PHS 398 (Rev. 11/07), Continuation Page

Transcription

Program Director/Principal Investigator (Last, First, Middle):
Weiner, Michael
Core: 1
Title of Core (not to exceed 81 spaces): Administrative Core
Core Leader: Weiner, Michael, W.
Position/Title: Director, Center for Imaging of Neurodegenerative Diseases; Professor, UCSF
Department, service, laboratory, or equivalent: Radiology
Mailing Address:
4150 Clement Street (114M)
San Francisco, CA 94121
Human Subjects (yes or no): No
If yes, state pages where a description of the plan for protection of human subjects can befound
and the pages where a description detailing the participation by both genders and all racial and
ethnic minorities can be found.
Vertebrate Animals Involved (yes or no): No
If "yes," identify by common names and underline primates. State pages where a description of
the plan for the protection of animals can be found. Also, if available, state the page number
where the IACUC approval can be found. Otherwise Just-in-Time procedures are applicable.
Dates of Proposed Project Period if different from that of the entire application:
PHS 398/2590 (Rev. 11/07)
Page
Continuation Format Page
Weiner, Michael
PROJECT SUMMARY (See instructions):
The overall goal of this project is to determine the relationships among the clinical, cognitive, imaging,
genetic and biochemical biomarker characteristics of the entire spectrum of Alzheimer’s disease (AD), as the
pathology evolves from normal aging through very mild symptoms, to mild cognitive impairment (MCI), to
dementia. ADNI will inform the neuroscience of AD, identify diagnostic and prognostic markers, identify
outcome measures that can be used in clinical trials, and help develop the most effective clinical trial
scenarios. ADNI2 continues the currently funded AD Neuroimaging Initiative (ADNI1), a public/private
collaboration between academia and industry to study biomarkers of AD as well as a recently funded Grand
Opportunities (GO) grant which supplements ADNI goals and activities. The Administrative Core, led by the
PI (MW Weiner) consists of his administrative staff, a statistician, and the Data and Publications Committee
administered by Robert Green at Boston University. Dr, Weiner has responsibility for all administrative,
financial, and scientific aspects of ADNI. The Administrative Core is reponsible for: scientific, administrative
and financial coordination of the entire project including non-compete and competitive renewals; all budgets
and subcontracts including tracking all subcontracts, reconciling budgets, tracking carryovers, maintaining an
updated financial accounting and projections of expenses that closely matches reconciliation data when final
status reports (FRS) are received; responsibility for interactions with NIH, FNIH, ISAB, SAB, and ADNI
projects in other countries; conferece calls with ADNI executive Committee twice/month; conference call with
leaders of the ISAB twice/month; monthly call with the ADNI Clinical Core at ADCS; conference call with MRI
and PET cores (twice/month); organizing Scientific Advisory Board meetings; Organizing the “ADNI
weekend” which consists of meetings of the Steering Committee, ISAB, Scientific Advisory Board, Excom,
and other meetings; interaction with all companies involved with/interest in ADNI; facilitating development of
ADNI projects in other countries. Finally, Dr. Weiner tracks details of subjects flow and procedures,
processing of all scans, generation of scientific results, and manuscripts, abstracts, and publications.
RELEVANCE (See instructions):
This ADNI project will provide new information which will greatly facilitate design of clinical treatment trials
and will help develop new diagnostic techniques which identify AD at an early stage, ultimately leading to
effective treatment and prevention of AD. The Administrative Core has overall responsibility for the
coordination and success of this entire highly complex project.
PROJECT/PERFORMANCE SITE(S) (if additional space is needed, use Project/Performance Site Format Page)
Project/Performance Site Primary Location
Organizational Name:
Northern California Institute for Research and Education
DUNS:
613338789
Street 1:
4150 Clement Street
City:
Street 2:
San Francisco
Province:
Project/Performance Site Congressional Districts:
County:
VAMC Building 13
San Francisco
USA
CA-008
Country:
State:
Zip/Postal Code:
CA
94121
Additional Project/Performance Site Location
DUNS:
Street 1:
Street 2:
City:
Province:
County:
Country:
State:
Zip/Postal Code:
PHS 398 (Rev. 11/07)
Page 2
Form Page 2
Weiner, Michael
SENIOR/KEY PERSONNEL. See instructions. Use continuation pages as needed to provide the required information in the format shown below.
Start with Program Director(s)/Principal Investigator(s). List all other senior/key personnel in alphabetical order, last name first.
Name
eRA Commons User Name
Organization
Role on Project
Weiner, Michael W.
michaelw
NCIRE
PI
Boston University
Co-Investigator
Green, Robert
OTHER SIGNIFICANT CONTRIBUTORS
Name
Organization
Role on Project
Human Embryonic Stem Cells
No
Yes
If the proposed project involves human embryonic stem cells, list below the registration number of the specific cell line(s) from the following list:
http://stemcells.nih.gov/research/registry/. Use continuation pages as needed.
If a specific line cannot be referenced at this time, include a statement that one from the Registry will be used.
Cell Line
PHS 398 (Rev. 11/07)
Page 3
Form Page 2-continued
Number the following pages consecutively throughout
the application. Do not use suffixes such as 4a, 4b.
Weiner, Michael
1. OVERALL INTRODUCTION AND ADMINISTRATIVE CORE
1.1. SPECIFIC AIMS: The goal of this project is to determine the relationships among the clinical, cognitive,
imaging, genetic and biochemical biomarker characteristics of the entire spectrum of Alzheimer’s disease (AD),
as the pathology evolves from normal aging through very mild symptoms, to mild cognitive impairment (MCI,
generally accepted as a precursor to dementia), to dementia. The Alzheimer’s Disease Neuroimaging Initiative
(ADNI) will inform the neuroscience of AD, identify diagnostic and prognostic markers, identify outcome
measures which can be used in clinical trials and will help develop the most effective clinical trial scenarios [1].
The project continues the currently funded ADNI (ADNI1), a public/private collaboration between academia and
industry to study biomarkers of AD as well as a recently funded Grand Opportunities (GO) grant (funded from
Sept 2009-Sept 2011, see more details below) which supplements ADNI goals and activities. We term the
renewal of ADNI1, for which this application requests funding, ADNI2.
1.1.1. Overall Goal: Predictors
and Outcomes: The overall goal
of ADNI is to validate biomarkers
of AD for several key reasons.
Initially ADNI1 was conceived as
a study that would validate
biomarkers for use as outcome
measures (measures of change)
in clinical trials. Furthermore,
biomarkers that directly or
indirectly measure AD pathology
may be used as predictors of
cognitive decline/dementia. Such
predictors will assist in the
enrichment and selection of
subjects with mild impairment
and normal elderly subjects for
treatment
trials
and
even
prevention trials. Thus, ADNI2 is
focused
on
a
broader
understanding of biomarkers in a
wider range of subjects in order
to
understand
how
these
Figure 1: Overall model of changes in the progression from normal aging to MCI to AD.
biomarkers can be used as both
predictors
and
outcomes.
Different biomarkers will be effective predictors or outcomes at different stages across the continuum from
normal cognition to AD dementia. Understanding the sequential change of biomarkers and their relative value
as predictors and outcomes at the presymptomatic, mild symptoms/mild cognitive impairments, and dementia
stages of the disease will add to our understanding of the neuroscience of AD, lead to improved diagnostic
tests, and facilitate design and power calculations of clinical trials for disease modifying agents.
1.1.2. Our Model: Our model (Figure 1) posits that AD begins with amyloid β (Aβ) accumulation in the brain,
which leads ultimately to synaptic dysfunction, neurodegeneration, and cognitive/ functional decline. This
predicts that the earliest detectable changes (measured in the GO/ADNI projects) are those related to Aβ
(Cerebrospinal fluid (CSF) and PET amyloid imaging). Subsequently neurodegeneration is detected by a rise
of CSF tau species, synaptic dysfunction by FDG-PET, and neuron loss indicated by atrophy most notably in
medial temporal lobe (measured with MRI). The temporal sequence of changes of Aβ deposition, CSF tau,
FDG-PET, and MRI remain to be determined. These changes ultimately lead to memory loss, general cognitive
decline and eventually dementia. Expression of each element of AD pathology (e.g. Aβ and tau deposits,
atrophy) is influenced by many modifying factors including age, APOE genotype, and cerebrovascular disease
(white matter lesions detected by FLAIR MRI) and microbleeds (detected by T2* MRI) and there are expected
to be wide differences among individuals.
We note that this model represents a simplification of current knowledge of the neurobiology of aging and
AD, and we understand that the relationships among aging, tau phosphorylation and conformational change,
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
amyloid peptide accumulation and conformational change, synaptic dysfunction and neuronal loss are complex
and not adequately conveyed by our simple model (Figure 1). Nonetheless, the model is useful in a heuristic
way in the interpretation of biomarker, cognitive and clinical data from ADNI and other studies, and in the
incorporation of biomarker measures into trial designs. We realize that this model may change with
incorporation of more data, and we will re-evaluate it over the course of the study.
Our model suggests that different imaging modalities, measurements, and different biochemical markers will
usefully serve as “predictors” (measurements which predict future change) and outcomes (measurements that
detect change) at different stages in the transition from normal aging, to MCI, to dementia. Further, the model
suggests that hat the imaging/biomarker measurements most likely to predict decline in normal subjects will be
measures of Aβ (CSF and PET), perhaps in combination with measures of CSF tau, FDG-PET, and MRI.
While amyloid biomarkers may be useful predictors of decline in early MCI (EMCI), CSF tau, FDG-PET and
MRI measures of regional atrophy, which likely change after amyloid markers change, may be more predictive.
In late MCI (LMCI) and AD, we hypothesize that the most effective biomarkers for prediction of further decline
will be FDG-PET, MRI, and cognition. Biomarkers that are most likely to correlate with, and augment the utility
of, cognitive and clinical measures as outcomes in clinical trials are FDG PET and possibly MRI measures of
volume (especially of hippocampus and temporal cortex) at early stages and atrophy throughout the brain at
later stages. However, it is recognized that the performance of the various imaging and CSF/blood
measurements depends both on the biological sequence of events as well as the sensitivity, accuracy, and
precision of the various measurements. Thus, for example, a test which best predicts future cognitive decline
in normal subjects may not necessarily represent the earliest biological change, but rather the earliest change
which is detected by a sensitive and robust test.
1.1.3. Specific Activities of ADNI2: The goals of ADNI2 will be accomplished by: 1) continuing annual
clinical/cognitive/MRI followup of the 476 normal controls and LMCI subjects previously enrolled in ADNI1 and
who will be followed in GO for 2 yrs; 2) following the 200 EMCI subjects enrolled in the GO grant after GO
ends; 3) additional enrollment of new healthy controls (n=150), EMCI (n=100 which adds to the 200 subjects
enrolled in GO), LMCI (n=150), and AD (n=150) subjects; 4) performance of F18 amyloid PET imaging (using
F18 AV-45 from Avid Radiopharmaceuticals) and FDG-PET imaging on all new subjects enrolled in ADNI2,
and obtain a 2nd F18 amyloid PET image on all remaining ADNI1, GO, and ADNI2 subjects 2 years after the
baseline scan. Obtain FDG PET scans at about the same time as F 18 amyloid scans 5) continue to obtain
annual clinical/cognitive/blood draw and MRI on all subjects including ADNI1 and GO . Obtain lumbar
puncture for CSF at baseline an every other year..All collected data will be processed and analyzed by ADNI
investigators including the Biostatistical Core, and will made available on the UCLA/LONI/ADNI website to all
qualified scientists in the world who request a password, without embargo.
1.1.4. The specific aims are defined below (see each core for specifics). These are focused on predictors,
outcomes and clinical trial design, but fulfillment of these aims with add considerably to what is known about
the basic neuroscience of AD in terms of progression and underlying mechanisms.
Aim 1: Predictors: Determine and define those biomarkers which best predict future cognitive decline and
conversion to MCI/dementia at the various stages of the progression from normal cognition to dementia. These
biomarkers may serve as predictive or early diagnostic markers, and could be used for selection of subjects or
as covariates in future treatment or prevention trials.
1. Replicate the important findings from ADNI1.
2. Test hypotheses concerning the EMCI subjects and F18 amyloid imaging:
a. Test the hypothesis that most but not all AD subjects will be amyloid +;approx 60% LMCI subjects will
be amyloid positive and amyloid positivity will predict conversion from LMCI to AD; a lower % of EMCI subjects
and of controls will be amyloid+ and amyloid positivity will predict cognitive decline and conversion to LMCI
and EMCI respectively.
b. Test the hypothesis that CSF Aβ amyloid and tau, taken together, will have a higher power to predict
cognitive decline than F18 amyloid imaging
c. Test the hypothesis that FDG PET and structural MRI will have superior power to predict cognitive
decline in LMCI and AD than amyloid imaging
3. Perform genome wide association analyses and targeted sequencing. Identify genetic markers associated
with AD, with brain amyloid deposition, and with rate of change of imaging/biomarker/clinical/cognitive data
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Aim 2: Outcomes: Determine and define those biomarkers that best serve as outcome measures to quantify
the rate of progress at the various stages from controls to dementia. These biomarkers may serve as outcome
measures in future treatment or prevention trials
1. Replicate results from ADNI1.
12Test hypotheses that:
a. Changes in structural MRI, especially of hippocampus and temporal lobe, have higher power than
changes in other brain areas to assess rates of progression; and
b. Changes in structural MRI and FDG PET in specific brain regions have much higher statistical power to
detect change than clinical measures. Such MRI and PET changes correlate with changes in clinical
measures.
2. With regard to the EMCI subjects,. Test the hypotheses that changes in structural MRI in median temporal
lobe and other temporal lobe regions, as well as FDG PET, have the highest power of all measures to detect
longitudinal change and that these measures correlate with ongoing cognitive change and also predict future
cognitive change
Aim 3: Clinical Trial Design: To improve clinical trials by developing various clinical trial protocol scenarios
which use clinical, cognitive, and biomarker measures as selection criteria, as covariates, and as outcome
measures, with maximum statistical power to detect treatment effects. Such scenarios would be developed for
subjects with dementia, with MCI, with mild symptomatology, and normal healthy controls. Further descriptions
of this aim are in the Clinical Core.
Finally the Neuropathology Core will perform pathological examination on brains obtain by autopsy to validate
the antemortum diagnoses.
1.1.5. Summary: Taken together, the overall impact of this ADNI2 renewal will be the following: 1) increased
knowledge concerning the sequence of events leading to AD dementia; 2) development of improved clinical
and biomarker methods for early detection of AD; 3) improved imaging and chemical biomarker methods for
monitoring progression of AD; 4) facilitation of clinical trials of treatments to slow disease progression,
ultimately contributing to the prevention of AD dementia. No other large multisite study in the world addresses
these complex issues with the sample size and statistical power of this study. The innovation of this proposal
lies in the longitudinal assessment of the spectrum from normal aging to AD (including the newly enrolled
EMCI) using an integrated combination of clinical/cognitive, CSF/plasma biomarker, MRI, amyloid/FDG PET,
and genetic measures. ADNI2 will also facilitate the development of a national network for F18 amyloid
imaging. As in ADNI1 and GO, ADNI2 will continue release of all data as acquired, without embargo, to any
qualified scientist who requests a password to our website.
1.2. Background and Significance:
1.2.1. Neuroscience of AD: Space does not allow much review of the current state of knowledge concerning
the pathophysiological events in AD or discussion of the neuroscience questions in the AD field. The model
shown in Figure 1 is widely viewed as the most probable model of AD progression, but the ADNI project is not
built around, and does not depend upon, the amyloid hypothesis. Despite the evidence in favor of this
hypothesis [2], other evidence does not necessarily support all aspects of it including 1) the early Braak stage
consists of tau tangles and synapse loss in the entorhinal cortex and hippocampus without amyloid
accumulation [3-5]; 2) a follow-up study of subjects in the Wyeth Elan 1792 vaccine trial showed amyloid
removal (at pathology) in some subjects, while they continued to decline cognitively [6]; and 3) there is poor
correlation between brain amyloid level and cognitive impairment. One possibility is that subjects with dementia
have such severe brain damage that amyloid removal does not slow progression of symptoms; this is one
reason for performing AD treatment trials on subjects with less severe cognitive impairments, such as MCI
(and a reason why the number of MCI subjects in ADNI is much greater than AD or control subjects). However,
the failure of anti amyloid clinical trials could be due to many reasons, especially that the treatments did not
sufficiently reduce brain amyloid. An important point to emphasize is that we have limited information
concerning the pathophysiological sequence of events of AD in humans from autopsy studies and from studies
measuring only cognition. Use of imaging and CSF biomarkers facilitates probing of the sequence of events in
AD. For example it has been recently reported (in 1 subject) that CSF amyloid falls prior to development of C11 PIB positivity which precedes cognitive impairment [7].Replication and extension of this sequence of events
in a multisite study with large sample size will provide critical information concerning the neuroscience of AD.
1.2.2. Overall Goals of the Field: We consider that the 3 major overall goals of the AD research field are:
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
1.) To develop a comprehensive understanding of the sequence of pathophysiological events which cause AD
and lead to dementia, at the molecular, cellular, brain, and clinical levels.
2.) To develop measurements which identify the various elements of AD pathology, and the factors which
influence AD pathology, in living human subjects. These measurements can then be used for early diagnosis,
as risk factors/predictors for cognitive decline/dementia. These measurements could ultimately have utility in
clinical trials and in clinical practice
3.) To develop treatments which slow the progression of AD and ultimately prevent the development of AD
(either secondary prevention or primary prevention).
AD currently affects more 5 million patients in the USA and will rise to 16 million by 2050 [8], costing the US
economy more than $140 billion/yr [8, 9] Globally, an estimated 35.6 million people have dementia (largely due
to AD) and that number is expected to reach 65.7 million in 2030 and 115.4 million in 2050 [10]. It is generally
accepted that there is a pressing need to develop effective disease modifying treatments to slow or halt
progression of AD pathology to be used in subjects with dementia, mild cognitive impairment, and in control
subjects at risk for development of cognitive decline and dementia. At the current time no treatments have
been convincingly demonstrated to slow the progression of AD pathology. Many current trials are forced to rely
on relatively small samples of convenience using those predictors and outcome measures which are of interest
and available.
1.2.3. Predictors: Imaging and CSF/blood biomarkers to detect AD pathology: It is generally accepted
that AD pathology (amyloid plaques, tau tangles, synapse loss, gross neuron loss and brain shrinkage) begins
many years prior to dementia and often exists with no evidence of cognitive impairment. The cognitive
impairment caused by AD pathology is thought to occur within the context of the cognitive changes which
occur in normal aging, and is characterized initially by problems with memory functioning. This progesses to
deficits in other cognitive domains, functional abilities, and frank dementia. Evidence exists that the
pathological and cognitive changes are non-linear in that there is a gradual acceleration of pathological and
cognitive changes. There is a compelling need to identify measurements whichidentify the presence and extent
of AD pathology in the living brain, thus characterizing the stage of disease. Because of the non-linear nature
of the process, knowledge of the stage of progression could potentiallybe used to predict the future rate of
cognitive decline and the future occurrence of dementia (the further advanced the progression, the greater the
rate of future change). As amyloid plaques develop, considerable evidence suggests that CSF Aβ amyloid falls
[11, 12]. Thus CSF Aβ is a putative measure of brain amyloid deposition. Brain amyloid is directly detected by
PET amyloid ligands. CSF tau increases in the progression of controls to MCI to AD [11, 12], and is a putative
measure of the deposition of tau tangles and neurodegeneration. No direct measures of brain synaptic density
exist in humans, but brain activity is reduced as synaptic density falls, and FDG PET is a quantitative measure
of brain activity which appears to identify early AD. Structural MRI detects brain atrophy, and hippocampal
volume shrinkage has been correlated with neuronal loss [13] and NFTs. Thus each of these measures has
predictive value, but their relative values at the different stages across the continuum have not been
established. Several investigators have proposed that imaging and CSF biomarkers could be used to identify
AD pathology in subjects who are not demented, and could thus be used for diagnosis of AD [14]. Several
pharmaceutical companies, and the Alzheimer’s Disease Cooperative Study have proposed performing AD
treatment studies employing subjects with “early AD” meaning non-demented subjects with cognitive
impairments who have imaging/CSF biomarker evidence of AD pathology (especially low CSF Aβ and/or C-11
PIB positivity), but the value of this approach has not been established. Genetics may also be considered a
predictor in AD. Aside from ApoE-4 measurements, genetic analysis was not an initial goal of ADNI1, but
supplemental funds were provided for the Genome Wide Association study (GWAS) of all 812 ADNI subjects
(data available on the LONI website, see below).
1.2.4. Outcomes: Imaging and CSF/blood biomarkers to detect progression of AD: Measures of rates of
change serve as outcomes in clinical trials. A problem with AD clinical trials is the length of time and large
sample sizes required, because of the high variability of clinical and cognitive measures. Numerous reports
suggested that changes in brain structure (detected by MRI) or brain glucose metabolism (detected by FDG
PET) had higher statistical power to detect change than clinical/cognitive measures because of their low
variability. Interest in biomarkers was further increased because measures of function and cognition are
affected by many things (e.g. depression, other illnesses) in addition to features of AD; are potentially affected
by drugs such as cognitive enhancers; have low statistical power to determine effects of disease modifying
treatments; and only indirectly reflect disease progression.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
1.2.5. Rationale for ADNI: The original impetus and rationale for ADNI was that many academic investigators,
pharmaceutical companies and biotech companies were beginning to develop treatments aimed at slowing the
progression of AD, but measurements of cognition or conversion from MCI to AD would not convincingly
demonstrate that a putative treatment was modifying the progression of AD pathology, because impaired
cognition in AD and MCI can be improved with symptomatic treatments such as acetylcholinesterase
inhibitors,. Also, numerous investigators were performing studies measuring various imaging and CSF/blood
biomarkers. However, 7 years ago there were no standards for obtaining or measuring imaging/biomarkers for
AD. Further, there were not sufficient data to determine the relative value of biomarker measures to detect
progression of AD in treatment trials.
1.2.6. Goals and design of ADNI1: Therefore the original goal of ADNI1 was to ascertain the relative value of
various imaging, and CSF/blood biomarkers as outcome measures in trials of AD and MCI subjects. Other
goals of ADNI1 were to develop improved standardized methods for performing AD trials, development of
clinical trial scenarios, and the acquisition of a data repository for academics and industry for a multiplicity of
purposes. It should be emphasized that the original ADNI1 grant had limited funding for analysis of data; funds
were only provided for analysis of the baseline, 6 month and 12 month data; and statistical support was limited.
The rationale of ADNI has been to study the value of imaging, blood/CSF biomarkers in controls, subjects with
early and late MCI, and demented subjects with AD in a large population where all subjects have as many
measurements as possible (discussed further below). Subjects were selected using inclusion/exclusion criteria
similar to that of a clinical trial, and the subjects are enrolled at sites participating in clinical AD treatment trials,
coordinated by our Clinical Core the AD Cooperative Study. In ADNI1 all subjects had 1.5 T MRI; 25% had 3T
MRI; 60% had FDG PET at baseline, 6, 12, 18, 24, and 36 months (36 month data not yet available); and 60%
had lumbar punctures at baseline, 1 yr and then annually. Furthermore the images were processed by multiple
groups (6 groups analyzed MRI, 3 groups analyzed FDG PET. The Biostatistical core has performed a rigorous
analysis to compare the results of the various groups. Furthermore, to “validate” the imaging/biomarker results,
the rates of change of these measurements have been correlated with change of clinical/cognitive
measurements. In retrospect, one limitation of ADNI1 was that FDG PET was performed on only 50% of the
subjects, lumbar puncture was performed on only 60% of the subjects and these two groups only partially
overlapped. The failure to get all essential data on all subjects has led to smaller samples sizes in some
analyses. This limitation will be overcome in GO and ADNI2 by obtaining clinical/cognitive/MRI/amyloid
PET/FDG PET/lumbar puncture on all subjects at baseline.
In the 5 years since the funding of ADNI1 there has been increased interest in the use of imaging and
CSF/blood biomarkers to identify AD pathology in subjects prior to dementia, and to develop diagnostic criteria
which employ these measurements [14]. Data from ADNI have proved to be a valuable resource to address
these issues, and thus the development of imaging and CSF/blood biomarkers as predictors has become an
important goal of ADNI.
1.2.6.1. Administrative History: ADNI, a cooperative agreement (U01) grant, was funded as a public/private
partnership with $40 million from NIA and $20 million from 13 companies in the pharmaceutical industry and
two Foundations for a total of $60 million. Since then, additional companies have joined, bringing the total to
22. An additional $7 million has been provided in the form of supplements for 1) the C-11 PIB sub-study 2) the
lumbar puncture extension (beyond the original 1 year of funding), and 3) the genome wide association
analysis of the DNA of all ADNI subjects. All funds from industry are provided to the Foundation for NIH which
then provides the combined funds to NIA who awards funds in the form of a UO1 grant to ADNI (The Northern
California Foundation for Research and Education (NCIRE) at the Veterans Administration Medical Center,
San Francisco, affiliated with the University of California San Francisco, is the recipient organization and the
location of the Administrative Core).
The tables are provided to show the schedule of events and scope of work performed for ADNI1, and the
proposed work for the funded GO grant and the current application, ADNI2. Table 1 shows the years for
ADNI1, GO, and ADNI2 and how Year 1 of the GO grant overlaps with Year 6 of ADNI1, and how Year 2 of the
GO grant overlaps with Year 1 of ADNI2. More explanatory material is in the Budget Justification
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Table 2 shows the schedule of events for ADNI1. Year 1 was the Preparatory phase with little enrollment.
Year 6 (no cost extension year) just began, and thus the actual number of subjects, scans. is not known.
1.2.7. Grand Opportunities (GO) grant: A GO grant was submitted by the identical team of investigators
involved with ADNI (M Weiner PI). Notice of Award for $24 million was announced on Sept 24 2009. The GO
grant was explicitly written to supplement the goals of ADNI1. The GO grant will provide funds for: 1) enrolling
200 EMCI subjects, some of whom will have early biomarker signals of AD pathology. This category of
subjects was not enrolled in ADNI1, thus this subject group bridges the gap between normal elderly and LMCI
who are more amnestic than EMCI. These GO subjects will have clinical/cognitive, blood/CSF/genetic, FDG
and amyloid PET, and MRI measurements during the 2 year period of GO. 2) Performing F18 amyloid PET
imaging on all normal control and LMCI subjects from ADNI1, and newly enrolled EMCI subjects, which will
allow correlation and comparison of this modality with all of the other clinical/cognitive, neuroimaging, genetic,
and biomarker data collected in ADNI1 and GO. 3) Extending the follow-up of LMCI and normal subjects who
were enrolled in ADNI1 and are being carried forward in GO. 4) Analysis of all of the ADNI1 data that could not
be done in the ADNI1 grant (since ADNI1 was a data collection grant, and few funds provided for analysis) as
well as analysis of the data from this GO project, to test hypotheses and perform data explorations. Table 1
shows that Year 1 of GO directly overlaps with Year 6 of ADNI1 (Year 6 is a no-cost extension year of ADNI1
and the last funded year of ADNI1) and that Year 2 of GO overlaps with Year 1 of ADNI2 (the application under
review). The very low budget for Year 1 of ADNI2 is explained by the fact that GO pays for the cost of
continuing the ADNI subjects during this period, and for the costs of the F 18 scans on ADNI1 and GO
subjects. Finally it should be emphasized that all data from GO will be part of the same data archive with
ADNI1 and ADNI2, and that all analyses will use data from all projects. Table 3 shows the schedule of events
for the GO grant, which just began 9/30/09.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
1.2.8. Rationale and Justification for ADNI2, which is a renewal of ADNI1: As previously stated, the major
goals of ADNI2 are: 1) to obtain new information concerning the pathophysiological sequence of changes in
the brain which occur across the continuum from normal aging to MCI to AD-dementia; 2) to identify and
validate imaging and blood/CSF biomarker predictors of cognitive decline/dementia; 3) to identify and validate
imaging and blood/CSF biomarker outcomes which reflect progression of AD pathology; and 4) to develop
information leading to improved clinical trials. These goals will be accomplished first by following the 476
normal controls and LMCI subjects who are continuing from ADNI1 and GO. During this prolonged follow-up
period it is expected that more MCI subjects will convert to AD, while other MCI will remain stable and not
convert. Some normal controls will begin to demonstrate cognitive decline, and convert to EMCI or LMCI. Thus
the relative predictive value of the various cognitive, imaging, and CSF/blood biomarkers can be assessed with
greater statistical power. Important new findings from ADNI1 will be replicated using ADN2 data. Second,
follow-up of EMCI subjects from GO and new enrollment of EMCI subjects in ADNI2 will provide new
information concerning this population, thus bridging the gap between the healthy controls and LMCI enrolled
in ADNI1. Of particular interest will be the relative value of clinical/cognitive tests and imaging and blood/CSF
biomarker as predictors of cognitive decline/dementia. Third, F18 amyloid PET, FDG PET and lumbar puncture
for CSF will be performed on all subjects at baseline. This will also establish a nationwide network for F18
amyloid PET which can be used for research and clinical trials. Importantly, the information obtained will allow
cost effectiveness comparisons of the predictive value of CSF Aβ/tau measurements versus amyloid PET .
Taken together ADNI is the only multisite longitudinal observational clinical/imaging/biomarker study being
performed in the US. ADNI data are widely available to all scientists throughout the world without embargo
through the UCLA/LONI/ADNI website. ADNI has already demonstrated its high value by providing a great deal
of scientific information (as evidenced by the numerous publications, see list below), and providing information
for development of clinical trial protocols which are being used in several current phase 3 studies. ADNI also
serves as a model of ADNI-like efforts in other countries. Thus the very practical deliverables of ADNI2 will be:
1) Advance the neuroscience of AD by the generation of new data, especially on normal elderly, EMCI, and
LMCI subjects concerning the sequence of events which occur across the continuum of AD. Addition of the
EMCI subjects and F18 amyloid imaging is expected to provide new insights concerning the neuroscience of
AD, in particular the changes which occur in the various imaging/biomarker measurements during the transition
from normal controls to EMCI to LMCI to AD. 2) Identification and validation of imaging/biomarker predictors of
cognitive decline/dementia which can be used in clinical trials and as diagnostic tests to identify AD pathology
in non-demented individuals. 3) Identification and validation of the best outcome measures at different stages
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
across the continuum from normal cognition to dementia. 4) Development of clinical trial scenarios which use
clinical/cognitive measures and imaging/CSF/blood biomarkers as both predictors and outcomes, to obtain
maximum statistical power for treatment studies across the continuum from normals (especially normal elders
with evidence of AD pathology) to dementia. In summary, the renewal of ADNI will contribute considerably to
the development of new diagnostic approaches, improved clinical trials, and to the identification of effective
treatments which slow the progression of AD pathology in demented and non demented subjects. Ultimately,
the results from ADNI will contribute considerably to the development of AD treatment trials and to effective
measures which prevent the development of AD.
Table 4 shows the schedule of events for the project proposed in this application, and if funded would begin
9/1/10. A detailed explanation of the schedule of events is provided in the Budget Justification.
Table 4 above shows first that in year 1 the only studies supported by ADNI2 are on newly enrolled subjects
including 150 normals, 100 EMCI, 150 L MCI and 150 AD. All these subjects receive
clinical/cognitive/MRI/amyloid PET/FDG PET and lumbar puncture for CSF at baseline. Because our ADNI1
data shows that a 6 month interval detects change on structural MRI, there will be a MRI at 3 months (to
examine the value of such a short between scan interval) and 6 months after baseline for these subjects. After
baseline measurements all subjects have annual clinical/cognitive measurements, and annual MRI exams.
Amyloid PET and FDG PET and lumbar puncture for CSF are performed after 2 yrs, and lumbar puncture
again after 4 years In year 2, ADNI2 begins to follow those normal and L MCI subjects from ADNI1 carried
forward by GO and those EMCI subjects enrolled in GO. Everyone will have annual clinical/cognitive/MRI
studies. Amyloid PET/FDG PET will be performed 2 years after baseline. Lumbar puncture for CSF will be
obtained on alternate years.
1.3. Progress Report:
1.3.1. Specific Aims of ADNI1:
Original Specific Aims of ADNI1: (shown in italic, copied from original application)
1.) Create uniform standards for MRI and PET acquisition, which could be used as surrogate measures in
future AD/MCI treatment and prevention trials.
2.) Create a publicly accessible data repository to provide further information concerning the longitudinal
changes in brain structure, function, cognition, blood, urine and CSF biomarkers which occur in normal aging,
MCI, and AD as well as transitions from one of these states to another.
3.) Validate MRI and PET imaging by examining their relationship with cognitive and functional measures.
4.) Validate blood and CSF biomarker measures by examining their relationship with cognitive and functional
measures.
5.) Identify the most effective measures for monitoring treatment effects in different stages in the progression of
normal aging, through MCI to AD.
6.) Develop improved methods of acquiring and processing multisite longitudinal data, which will increase costeffectiveness and power of future treatment trials by:
a. Improving MRI and PET acquisition methods to: Reduce between-manufacturer differences and
variability with time; Reduce problems caused by inhomogeneity of B and B , gradient non linearity, and
0
PHS 398/2590 (Rev. 11/07)
Page
1
Weiner, Michael
motion; Minimize instrumentation effects on PET data due to scattered photons, random coincidences,
attenuation correction, and resolution differences; Develop methods which account for hardware upgrades;
Determine advantages/disadvantages of 3 T vs. 1.5 T MRI.
b. Improving MRI and PET imaging processing methods for AD trials to: Reduce between site and between
manufacturer differences and upgrade effects; Improve processing methods, robust to random and nonrandom noise including motion; Determine methods for measuring change in brain volumes and metabolism,
which reduce propagation of error and have improved power to detect changes within groups, and transitions.
Develop a probabilistic brain map showing longitudinal differences of regional structure and metabolism, and
effects of age, diagnostic group, transitions, and other factors.
7.) Develop statistical models of cross sectional and longitudinal clinical, imaging, and biomarker data, which
can be used for future hypothesis generation and testing.
8.) Develop a harmonized network of sites, which may participate in future treatment trials.
This Initiative will have several aspects, two of which are general data collection to form a repository, and
specific hypothesis testing. The major specific a priori hypotheses to be statistically tested were:
1.) Prediction of cognitive decline: Baseline MRI and FDG PET, as well as specific biomarkers, will
correlate with baseline cognitive data and stage (normal, MCI, AD) and will predict both more rapid change in
quantitative measures and greater likelihood of conversion from MCI to AD. Short-term longitudinal change in
MRI and FDG PET will be correlated both with more rapid change in cognitive function and with greater rates
of subsequent cognitive decline and conversion.
2.) Specific regional hypotheses: The associations between imaging measures and clinical measures of
cognition will be region-specific: e.g. volumes of medial temporal lobe structures (hippocampus and entorhinal
cortex) will correlate more strongly with memory, and frontal lobe volume and metabolism will correlate more
strongly with executive function. These regional associations may be stage specific.
3.) Added power for clinical trials: The addition of imaging measures as markers of decline, as well as
biomarker data, to the standard clinical assessment will increase the precision of estimates of rates of change
in cognition and likelihood of conversion from MCI to AD. The added precision will improve the design of
clinical trials by allowing researchers to use shorter follow-up, smaller sample sizes, or greater power to detect
smaller effect sizes. Study design modifications will lead to greater cost-effectiveness in future trials.
In addition to the above hypotheses, it is expected that during the course of this project additional a priori
hypotheses and exploratory analyses will be performed. All such analyses will be carried out under the
supervision of the statisticians involved, and appropriate attention will be given to the problem of multiple
comparisons when analyzing such large volumes of data (This ends “original specific aims of ADNI1”).
1.3.2. Summary of the Cores: This section very briefly summarizes the accomplishments of each Core during
the past 5 years. Much more detailed information is provided in each Core. It is important to emphasize that
each Core did its own analyses, and then the Biostatistics Core has performed overall analyses. Therefore, in
some cases there are differences between results from different Cores, because of differences in sample size
and in methods.
1.3.2.1. Administrative Core: This core has overall responsibility for the entire project and oversaw activities
of all sites and cores, monitored funding and expenditures, submitted non competitive renewals, coordinated a
wide variety of meetings and teleconferences, facilitated development of other ADNI-like projects world-wide,
reviewed a large number of publications, successfully coordinated the writing of the GO grant, and has
coordinated the writing of this competitive renewal.
1.3.2.2. Clinical Core: Successfully facilitated recruitment of 229 normal, 380 LMCI, 210 AD subjects,
including retention of subjects with an attrition rate of only 6%. The core set up systems for electronic data
capture, quality control and reporting, of FDG-PET, C11 PIB PET, 1.5T MRI, 3T MRI and CSF biomarkers.
This core developed a flexible, powerful electronic data capture system for data entry at each participating
ADNI site. Protocols, template informed consent documents and procedures manuals have been developed
and modified for each amendment to ADNI (for example, the addition of PIB imaging, the extension of the
follow-up period, and the incorporation of additional lumbar punctures). Clinical Monitors, under the
supervision of the ADCS Medical Core, regularly visit each ADNI performance site to ensure compliance with
regulatory requirements and protocol procedures, and accurate data entry. Concerning performance of various
measures in clinical trials, ADNI results show that the CDR-SB is a powerful outcome measure in mild AD and
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
MCI. In LMCI, the ADAS13 (The ADAS-cog with 13 items) is superior to the ADAS11 or ADAS12 (these are
ADAS-cog with 11 or 12 items) in a 24 month trial. Imaging and CSFamyloid/tau covariates reduce samples
sizes by 5-10 % in LMCI and mild AD. Selection of LMCI subjects using low values of CSF Aβ reduces sample
sizes (presumably by selecting subjects with AD pathology). Subjects with LMCI and low CSF Aβ would meet
recently-proposed research criteria for AD, and consensus meetings suggest that standard AD-type cognitive
and clinical co-primary outcome measures will be appropriate for pivotal trials, and a single clinical measure
may be appropriate for a Phase II proof of concept trial. Using ADNI data, we have shown that a two-year
treatment period in LMCI, with appropriate covariates, has reasonable power to demonstrate effects on primary
measures such as ADAS Cog and CDR Sum of Boxes. A design based on these principles has recently been
launched as a Phase II proof of concept trial of a secretase inhibitor.
1.3.2.3. MRI Core: This core successfully implemented standardized procedures for multisite MRI across 89
MRI scanners at 59 sites using 38 different vendor and platform specific protocols; compared 5 different
analytical methods as predictors and outcomes. It demonstrated that longitudinal consistency is improved with
correction of scaling gradient non-linearity and with intensity non-uniformity correction. Some MRI analysis
methods had greater longitudinal power than others. The best performing measures (defined as smallest
sample sizes needed to detect a 25% rate reduction in AD and MCI subjects) were FreeSurfer hippocampal
volume and selected voxels in the temporal lobe using tensor-based morphometry. The longitudinal measures
with the greatest correlations with longitudinal change in general cognition (ADAS cog) were the whole brain
and ventricular Boundary Shift Integral (BSI) measures and also FreeSurfer ventricular volume. MRI has much
better longitudinal power to detect change than clinical instruments, resulting in substantially smaller sample
sizes needed for clinical trials in both MCI and AD patients. The results demonstrated no difference between
3T vs. 1.5T MRI in group-wise discrimination or sample sizes needed to power trials and found that
hippocampal rates of change in MCI and AD accelerate over time. Associations were also found between white
matter hyperintensity load and cognition/cognitive decline and associations between levels of atrophy and
genetic differences in the glutamate receptor, the FTO obesity gene, and several other candidate genes.
1.3.2.4. PET Core: This core successfully implemented standardized procedures for multisite FDG PET and
C-11 PIB PET and compared 3 different analytical methods as predictors and outcomes. It found high
correlations between CSF Aβ and C-11 PIB. Longitudinal measures of brain glucose metabolism using both
statistically defined ROIs and a priori defined ROIs are associated with cognitive decline. Using statistically
driven ROIs (i.e. Reiman lab), calculations suggest that impressively small sample sizes would be needed for
clinical trials. CSF and PET measures of Aβ are highly congruent, show evidence of Aβ deposition in a high
proportion of normal individuals, and appear to be reasonably good predictors of cognitive decline in patients
with MCI or mild AD. The results showed that that hippocampal volume mediates cognitive impairment
associated with reduction of CSF Aβ.
1.3.2.5. Biomarker Core: This Core has 1) Established biofluid collection, shipping and storage SOPs, and an
archive of ADNI biofluids; The Table below summarizes the applications for ADNI samples from outside users
which were submitted to Dr Montine’s Resource Allocation Review Committee (RARC) and their distribution.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Summary of Applications and RARC Reviews (As Of 24 August, 2009)
PI (site)
Date
Decision
Biospecimen
Primiano (Clonex)
31 October, 2006
Clarifications
requested
Burns (University of 1 April, 2007
Clarifications
Kansas)
requested
9 May, 2007
Release pending
CSF
Plasma
Lymphocytes
Hendrickson
3 August 2007
Decline - too
None
(Merck)
preliminary for ADNI
CSF samples
Iqbal
1 April, 2007
Clarifications
requested
9 May, 2007
Release pending
CSF
Soares (Pfizer)
17 August, 2007
Release pending
CSF
Dechairo (Pfizer)
9 May, 2007
Defer to NIA
DNA
officials
Power 3
9 May, 2007
Decline - too
CSF samples
Goate (Wash U)
20 August, 20007
Release pending
DNA
Loy (U of
22 September, 2007 Decline - too
Rochester)
CSF samples
McIntyre (St. Fancis 31 May, 2009
Phased release
Serum then CSF
Hospital, Albany,
NY)
Restrepo (UCLA)
31 May, 2009
Limited release for
Plasma
pilot study
2) Implemented measurements of CSF by quantifying total tau, P-tau181p and Aβ42 using the Luminex
platform and Innogentics reagents; 3) Validated methodology using a “round robin” study with other
laboratories; 4) Performed measurements of CSF Aβ and tau on ADNI samples from baseline, Yrs 1 and 2; 5)
Measured homocysteine in 813 baseline plasma samples and found differences between controls and AD, and
controls and MCI but no difference in AD vs. MCI; 6) Developed and validated a semi-automated highthroughput HPLC tandem mass spectrometry assay for plasma and urine 8-iso-PGF2a; Pilot results are
encouraging. Proteomic studies by Rules Based Medicine Inc are in progress using plasma and CSF samples
from the entire ADNI cohort (supported by ISAB funds)
1.3.2.6. Genetics Core: Genotyping for 620,901 single nucleotide polymorphism (SNP) and copy number
variation (CNV) markers was completed on all participants. One case-control analysis identified APOE and a
new risk gene, TOMM40 (translocase of outer mitochondrial membrane 40), at a genome-wide significance
level of 10-6 (10-11 for a haplotype). TOMM40 risk alleles were approximately twice as frequent in AD subjects
as controls. A second analysis using a more stringent criterion for quality control confirmed that SNP
rs2075650 is strongly associated with an increasing risk of AD in both asymptotic P values (p = 1.043 × 10-10)
and corrected empirical P values (p = 1× 10-4). This SNP is located in TOMM40 on chromosome 19, roughly 13
kilobase pairs distal to APOE, and exhibits strong linkage disequilibrium with APOE at D' = 93. A third study
developed a genome-wide, whole brain approach using voxel-based morphometry (VBM) and FreeSurfer
parcellation followed by GWAS. SNPs in the APOE and TOMM40 genes were confirmed as markers strongly
associated with multiple brain regions. Other top SNPs were proximal to the EPHA4, TP63 and NXPH1 genes.
Detailed image analyses of rs6463843 (flanking NXPH1) revealed reduced global and regional gray matter
density across diagnostic groups in TT relative to GG homozygotes. Interaction analysis indicated that AD
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
patients homozygous for the T allele showed differential vulnerability to right hippocampal GM density loss.
NXPH1 codes for a protein implicated in promotion of adhesion between dendrites and axons, a key factor in
synaptic integrity, the loss of which is a hallmark of AD.
1.3.2.7. Neuropathology Core: The major accomplishment was implementing a protocol to solicit permission
for brain autopsy in ADNl participants who die and to send appropriate brain tissue from the decedents to the
Core for a standardized, uniform, and state-of-the-art neuropathological assessment. The benefit to ADNl of
the implementation of the NPC is very clear. Prior to the establishment of the NPC in September 2007, there
were 6 deaths but no autopsies in ADNl participants. Subsequent to the establishment of the Core there have
been 17 deaths of ADNl participants and 10 autopsies. Hence, the autopsy rate has gone from 0% to 59%.
Another accomplishment is the detection of co-existent pathologies with AD in the autopsied cases. It is
possible that these comorbidities may contribute to any variance in ADNl data.
1.3.2.8. Biostatistics Core: The Biostatistical Core found differences in results when using univariate or
multivariate models, for the numerous measures of cognition, brain volumes and FDG brain glucose
metabolism: In general, ApoE-4 was associated with lower CSF Aβ, smaller brain volumes, reduced FDG PET
uptake and higher rates of decline. In the healthy controls 38% had CSF Aβ below the threshold of 192 which
separates AD from normal controls, 18% had CSF tau above the threshold of 93, and 47% had abnormally
high PIB retention, suggesting early AD pathology.
Outcomes: A major goal of ADNI1 was to determine which imaging and CSF markers had high statistical
power to quantify rates of change in AD and MCI for use in clinical trials. The results from these calculations
are expressed as the number of subjects/arm required to detect a 25% slowing of the rate of change in a 12
month 2 arm clinical trial. In AD, MRI measures of hippocampus, hippocampal region, or ventricles required
between 54-130 subjects/arm and a data driven PET measure required 96 subjects/arm. Other PET measures
required higher sample sizes. In MCI, MRI measures of volume of hippocampus, hippocampal regions or
ventricles required 83-300 subjects/arm and PET measures required 280-7700. In general MRI measures had
higher statistical power to detect change than FDG PET. Since there was little longitudinal change of CSF Aβ
and tau, these measures had very low power to detect change.
Predictors: Another goal of ADNI has been to identify those biomarkers which predict future change of
cognition, function, and measures of neurodegeneration and AD pathology. Even though the controls as a
group showed no cognitive decline over the first two years, low baseline CSF Aβ was associated with declining
performance on MMSE and increased rate of hippocampal atrophy. Reduced hippocampal volume and FDG
brain glucose metabolism also predicted cognitive decline. Furthermore elevated CSF tau was associated with
lower FDG PET uptake. These results are consistent with the hypothesis that early stage AD pathology can be
detected in healthy controls and that even though there were no changes in cognition, rates of change of brain
atrophy and FDG uptake increased as CSF Aβ declined, consistent with greater AD pathology causing greater
change.
In MCI, 72% were PIB positive and 74% had CSF Aβ below the threshold of 192 which separates AD from
normal control. PIB positivity, lower CSF Aβ, and high tau were associated with more rapid cognitive decline
and greater rates of brain atrophy. Furthermore, elevated tau and higher baseline PIB retention predicted
reduced FDG brain glucose metabolism and greater rates of change of FDG brain glucose metabolism.
Multivariate models supported the hypothesis that greater AD pathology was associated with more rapid
change of brain glucose metabolism and volumes.
Evidence of AD pathology at baseline was also predictive of time-to-conversion from MCI to AD. Smaller
hippocampal and whole brain volumes, FDG brain glucose metabolism, ApoE-4, and lower cognitive and
functional performance at baseline were all significant univariate predictors of time-to-conversion and
hippocampal volume was more predictive than whole-brain volume. Univariate models suggested that a
number of baseline fluid and imaging biomarkers were associated with shorter time to conversion from MCI to
AD, including hippocampal and ventricular volume and brain size; three complex summaries of FDG PET
hypometabolism and the P-tau/Aβ142 ratio. In addition, baseline cognitive function measures were predictive
of time to conversion to AD. In AD, 89% were PIB+ and this group had the lowest levels of CSF Aβ and highest
levels of CSF tau. Smaller hippocampi, reduced FDG metabolism and higher tau predicted greater rise of
ADAS Cog and change in other cognitive measures.
Clinical trial design: Estimates of rates of clinical change (ADAScog or CDR-SOB) based on linear mixedeffects models appear to be about 40% more efficient (in terms of required sample size) than time-toconversion estimates of a hypothesized 25% treatment effect in MCI. Furthermore, using baseline covariates
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
such as hippocampal volume can improve efficiency by about 5%, and coupled with sample enrichment (based
on CSF Aβ) can improve efficiency by about 40% compared to a linear mixed-effects model analysis with no
covariates or enrichment.
1.3.2.9. Informatics Core: A major and unique contribution of ADNI has been the sharing of all raw and
processed data, through a central website (UCLA/LONI/ADNI) without any embargo of raw and processed data
(except for QC purposes). Any scientist in the world can readily obtain a password to this website and
download any or all of the images (including raw and processed images) and the entire clinical/imaging data
base (which contains all de-identified clinical, cognitive, fluid biomarker, and genetic (GWAS data)). This Core
provided a secure and reliable environment for storing and sharing neuroimaging and related data and a
supportive, responsive team dedicated to meeting the evolving needs of the ADNI community. Beginning with
the ADNI preparatory phase, the informatics core provided an environment for sharing experimental imaging
protocols for evaluation by imaging core members. After finalization of the protocols, the informatics core has
continuously provided a reliable, long-term repository for imaging, clinical and related data storage and
distribution. This effort has involved providing infrastructure and support to the 58 ADNI acquisition sites, to
the MRI and PET imaging cores that perform image quality control, and to the more than 1,000 users
worldwide who have used this infrastructure to obtain imaging, clinical, biomarker and genetic data. To date,
more than 81,000 ADNI MRI and PET images have been stored in the archive with more than 646,886 images
downloaded by approved ADNI data users. Additionally, the clinical and genetic data have been provided to
hundreds of users. Finally, subsystems for project management, data user application and review, and a
comprehensive web site have been implemented. It should be emphasized that the unembargoed release of all
data in an observational study is completely unprecedented, and to our knowledge has never been done in any
previous study. The demonstration that this can be done is impacting the overall attitude towards data sharing,
and other studies are planning similar data release. One unanticipated result has been that many investigators
not funded by ADNI are downloading the data and publishing abstracts and papers on ADNI data (see
Informatics Core and Publications Committee). This will be continued in ADNI2.
1.3.3. Overall impact of ADNI on the field: World Wide ADNI: At the time ADNI1 was funded, there were no
plans for similar efforts in other countries. However, the establishment of ADNI stimulated many such efforts
resulting in: 1) Australian ADNI (PI Colin Masters) [15], otherwise known as AIBL, which is a 2 site longitudinal
study of 1100 subjects with MRI (using ADNI protocol), a subset with C-11 PIB, and cognitive measures
(similar to ADNI). The Alzheimer’s Association has funded data sharing between the Australian study and
ADNI2) Japanese ADNI (PI Takeshi Iwatsubo) [16], which studies 220 subjects using identical methods to
ADNI in all respects except for language 3) European ADNI (PI Giovanni Frisoni) is enrolling 150 subjects.
There are also several large longitudinal projects beginning in China using imaging/CSF biomarkers and a
Korean ADNI is being planned. The Alzheimer’s Association has organized a quarterly teleconference of all
worldwide ADNI PIs, is working to fund more data sharing efforts among the projects, and Dr Iwatsubo is
hosting the first World Wide ADNI meeting in Sendai, Japan in November 2009. The impact of these numerous
projects, and the value of the information gained to academic scientists and to the pharmaceutical industry is
huge. To our knowledge, ADNI is the only neuroscience project in the world which is having such a worldwide
impact in the AD field.
1.3.4. Limitations of ADNI: One limitation of ADNI is that our population represents a clinical trial population
and not an epidemiologically selected “real life” population. Our subjects do not include those with cortical
strokes, cancer, heart failure, substance abuse etc. Therefore the extent to which the results from ADNI can be
generalized to the entire population is limited. Future population based studies will be required to determine if
the information derived from ADNI is relevant to the greater population. One approach has been for ADNI
investigators to develop collaborations with investigators who are conducting population based studies, so that
ADNI methods can be used in such studies; such discussions are underway. A second limitation is that ADNI
only studies subjects over the age of 65, and there is considerable evidence that AD pathology may begin to
occur in the human brain well prior to this age. Autopsy studies and amyloid imaging have suggested that a
substantial fraction of cognitively normal subjects in their 70s have AD pathology. A full understanding of the
pathophysiological sequence of events which occur in AD will require longitudinal studies of subjects beginning
at a young age. A third limitation of ADNI is the types of data that are not being collected including:
computerized neuropsychological testing, EEG, MEG, MRS, metabolic and inflammatory markers, and lifestyle
information. The decision concerning which measures to include was reached by consensus among the Site
PIs, Core leaders, and NIA. Although many of these measures might provide useful information, they are not
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
included because: 1) the measures have not yet been demonstrated to have high value as either predictors or
outcomes, and are not currently being incorporated into clinical trials; 2) the subject burden of ADNI is already
quite great (clinical/cognitive battery, MRI, FDG/Amyloid PET, lumbar puncture) and there are concerns that
adding additional tests will impair enrollment and increase dropout; 3) the additional cost of these measures is
not supported by evidence for inclusion. A fourth limitation was that not all measurements (like FDG PET and
lumbar puncture) were obtained on all subjects, limiting the ability to compare methods. This is being
overcome in the current study in which all subjects will have (at least) baseline lumbar puncture and AV-45
amyloid imaging as well as the other measurements.
1.3.5. Publications arising from ADNI1
1) Trojanowski JQ. Searching for the biomarkers of Alzheimer’s. Practical Neurol., 3:30-34, 2004.
2) Mueller SG, Weiner MW, Thal LJ, Petersen RC, Jack CR, Jagust W, Trojanowski JQ, Toga AW, Beckett L.:
Ways toward an early diagnosis in Alzheimer’s disease: The Alzheimer’s Disease Neuroimaging Initiative
(ADNI), Alzheimer’s & Dementia 1:55-66, 2005. PMC1864941
3) Mueller SG, Weiner MW, Thal LJ, Petersen RC, Jack C, Jagust W, Trojanowski JQ, Toga AW, Beckett L.
The Alzheimer’s Disease Neuroimaging Initiative. Neuroimaging Clin N Am, 15(4):869-77, 2005.
PMC2376747
4) Fukuyama H. Neuroimaging in mild cognitive impairment. Rinsho Shinkeigaku, 46(11):791-4, 2006.
5) Iwatsubo T. Beta-and gamma-secretases. Rinsho Shinkeigaku, 46(11):925-6, 2006.
6) Leow AD, Klunder AD, Jack CR Jr, Toga AW, Dale AM, Bernstein MA, Britson PJ, Gunter JL, Ward CP,
Whitwell JL, Borowski BJ, Fleisher AS, Fox NC, Harvey D, Kornak J, Schuff N, Studholme C, Alexander
GE, Weiner MW, Thompson PM; ADNI Preparatory Phase Study.: Longitudinal stability of MRI for mapping
brain change using tensor-based morphometry. Neuroimage, 31(2):627-40, 2006. PMC1941663
7) Mueller SG, Weiner MW, Thal LJ, Petersen RC, Jack C, Jagust W, Trojanowski JQ, Toga AW, Beckett LA.
Ways toward an early diagnosis in Alzheimer’s disease: The Alzheimer’s Disease Neuroimaging Initiative.
Cognition and Dementia, 5(4):56-62, 2006.
8) Arai H. Alzheimer’s Disease Neuroimaging Initiative and mild cognitive impairment. RinshoShinkeigaku,
47(11):905-7, 2007.
9) Fletcher PT, Powell S, Foster NL, Joshi SC. Quantifying metabolic asymmetry modulo structure in
Alzheimer’s disease. Inf Process Med Imaging, 20:446-57, 2007.
10) Ihara Y. Overview on Alzheimer’s disease. Rinsho Shinkeigaku, 47(11):902-4, 2007. PMID 18210830
11) Murayam S, Saito Y. Neuropathology of mild cognitive impairment Alzheimer’s disease. Rinsho
Shinkeigaku, 47(11): 912-4, 2007.
12) Haschke M, Zhang YL, Kahle C, Klawitter J, Korecka M, Shaw LM, Christians U. HPLC-atmospheric
pressure chemical ionization MS/MS for quantification of 15-F2t-isoprostane in human urine and plasma.
Clinical Chemestry, 53:489-497, 2007.
http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=17259
231
13) Shaw LM, Korecka M, Clark CM, Lee VM-Y, and Trojanowski JQ. Biomarkers of neurodegenertaion for
diagnosis and monitoring therapeutics. Nat. Rev. Drug Discovery, 6:295-303, 2007.
655
14) Fan Y, Batmanghelich N, Clark CM, Davatzikos C, the Alzheimer’s Disease Neuroimaging Initiative. Spatial
Patterns of Brain Atrophy in MCI Patients, Identified via High-Dimensional Pattern Classification, Predict
Subsequent Cognitive Decline. NeuroImage, 39(4): 1731-1743, 2008.
15) Hampel H, Burger K, Teipel SJ, Bokde ALW, Zetterberg H, Blennow K. Core Candidate Neurochemical
and Imaging Biomarkers of Alzheimer’s Disease. Alzheimer’s & Dementia, 4(1):38-48, 2008.
16) Nestor SM, Rupsingh R, Borrie M, Smith M, Accomazzi V, Wells JL, Fogarty J, Bartha R, and the
Alzheimer’s Disease Neuroimaging Initiative. Ventricular Enlargement as a Possible Measure of
Alzheimer’s Disease Progression Validated Using the Alzheimer’s Disease Neuroimaging Initiative
Database. Brain 131(Pt 9):2443-54, 2008. PMC2724905
17) Shaw LM. PENN biomarker core of the Alzheimer’s Disease Neuroimaging Initiative. Neurosignals,
16(1):19-23, 2008. PMC2696349
18) Visser PJ, Verhey FRJ, Boada M, Bullock R, De Deyn PP, Frisoni GB, Frolich L, Hampel H, Jolles J, Jones
R, Minthon L, Nobili F, Olde Rikkert M, Ousset P-J, Rigaurd A-S, Scheltens P, Soininen H, Spiru L,
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Touchon J, Tsolaki M, Vellas B, Wahlund LO, Wilcock G, Winblad B, DESCRIPA study group.
Development of Screening Guidelines and Clinical Criteria for Predementia Alzheimer’s Disease.
Neuroepidemiology, 30:254-265, 2008.
http://www.ncbi.nlm.nih.gov/pubmed/18515975?ordinalpos=&itool=EntrezSystem2.PEntrez.Pubmed.Pubm
ed_ResultsPanel.SmartSearch&log$=citationsensor
19) Boyes RG, Gunter JL, Frost C, Janke AL, Yeatman T, Hill DL, Bernstein MA, Thompson PM, Weiner MW,
Schuff N, Alexander GE, Killiany RJ, Decarli C, Jack CR, Fox NC, for the ADNI study.: Intensity nonuniformity corrections using N3 on 3-T scanners with multichannel phased array coils. Neuroimage,
39(4):1752-62, 2008. PMC2562663
20) Jack CR Jr, Bernstein MA, Fox NC, Thompson P, Alexander G, Harvey D, Borowski B, Britson PJ, Whitwell
J, Ward C, Dale AM, Felmlee JP, Gunter JL, Hill DL, Killiany R, Schuff N, Fox-Bosetti, S, Lin C, Studholme
C, Decarli CS, Gunnar Krueger, Ward HA, Metzger GJ, Scott KT, Mallozzi R, Blezek D, Levy J, Debbins
JP, Fleisher AS, Albert M, Green R, Bartzokis G, Glover G, Mugler J, Weiner MW, ADNI Study. The
Alzheimer’s disease neuroimaging initiative (ADNI): MRI methods. J Magn Reson Imaging, 27(4):685-91,
2008. PMC2544629
21) Hua X, Leow AD, Lee S, Klunder AD, Toga A, Lepore N, Chou Y-Y, Brun C, Chiang M-C, Barysheva M,
Jack Jr. CR, Bernstein MA, Britson PJ, Ward CP, Whitwell JL, Borowski B, Fleisher AS, Fox NC, Boyes
RG, Barnes J, Harvey D, Kornak J, Schuff N, Boreta L, Alexander GE, Weiner MW, Thompson PM, the
Alzheimer’s Disease Neuroimaging Initiative. 3D characterization of brain atrophy in Alzheimer’s disease
and mild cognitive impairment using tensor-based morphometry. NeuroImage, 41(1):19-34, 2008.
PMC2556222
22) Frisoni GB, Henneman WJP, Weiner MW, Scheltens P, Vellas B, Reynish E, Hudecova J, Hampel H,
Burger K, Blennow K, Waldemar G, Johannsen P, Wahlund L-O, Zito G, Rossini PM, Winblad B, Barkhof
F, Alzheimer’s Disease Neuroimaging Initiative. The pilot European Alzheimer’s Disease Neuroimaging
Initiative of the European Alzheimer’s Disease Consortium. Alzheimer’s & Dementia, 4(4): 255-64, 2008.
PMC2657833
23) Morra JH, Tu Z, Apostolova LG, Green AE, Avedissian C, Madsen SK, Parikshak N, Hua X, Toga AW,
Jack CR Jr, Weiner MW, Thompson PM, the Alzheimer’s Disease Neuroimaging Initiative. Validation of a
fully automated 3D hippocampal segmentation method using subjects with Alzheimer’s disease mild
cognitive impairment , and elderly controls. Neuroimage, 43(1): 59-68, 2008. PMC2624575
24) Hua X, Leow AD, Parikshak N, Lee S, Chiang MC, Toga AW, Jack CR Jr, Weiner MW, Thompson PM, the
Alzheimer’s Disease Neuroimaging Initiative. Tensor-based morphometry as a neuroimaging biomarker for
Alzheimer’s disease: An MRI study of 676 AD, MCI, and normal subjects. NeuroImage, 43(3):458-69,
2008.
25) Becker RE, Greig NH. Alzheimer’s disease drug development: old problems require new priorities. CNS
Neurol Disord Drug Targets, 7(6):499-511, 2008.
26) Walhovd KB, Fjell AM, Dale AM, McEvoy LK, Brewer J, Karow DS, Salmon DP, Fennema-Notestine C; the
Alzheimer’s Disease Neuroimaging Initiative. Multi-modal imaging predicts memory performance in normal
aging and cognitive decline. Neurobiol Aging, Oct 5 [Epub ahead of print], 2008.
27) Clark CM, Davatzikos C, Borthakur A, Newberg A, Leight S, Lee VM-Y, Trojanowski JQ. Biomarkers for
early detection of Alzheimer pathology. NeuroSignals, 16:11-18, 2008.
155
28) Mormino EC, Kluth JT, Madison CM, Rabinovici GD, Baker SL, Miller BL, Koeppe RA, Mathis CA, Weiner
MW, Jagust WJ, and the Alzheimer’s Disease Neuroimaging Initiative. Episodic memory loss is related to
hippocampal-mediated beta-amyloid deposition in elderly subjects. Brain, 132(Pt 5):1310-23, 2009.
29) Morra JH, Tu Z, Apostolova LG, Green AE, Avedissian C, Madsen SK, Parikshak N, Toga AW, Jack CR Jr,
Schuff N, Weiner MW, Thompson PM, The Alzheimer’s Disease Neuroimaging Initiative. Automated
mapping of hippocampal atrophy in 1-year repeat MRI data from 490 subjects with Alzheimer’s disease,
mild cognitive impairment, and elderly controls. Neuroimage, 45(1 Suppl):S3-15, 2009. PMC2733354
30) Shaw LM, Vanderstichele H, Knapik-Czajka M, Clark CM, Aisen PS, Petersen RC, Blennow K, Soares H,
Simon A, Lewczuk P, Dean R, Siemers E, Potter W, Lee V, Trojanowski JQ, Alzheimer’s Disease
Neuroimaging Initiative. Cerebrospinal Fluid Biomarker Signature in Alzheimer’s Disease Neuroimaging
Initiative Subjects. Ann Neurol, 65(4):403-13, 2009. PMC2696350
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
31) Morra J, Tu Z, Apostolova LG, Green AE, Avedissian C, Madsen SK, Parikshak N, Hua X, Toga AW, Jack
CR, Schuff N, Weiner MW, Thompson PM (2008). Automated 3D Mapping of Hippocampal Atrophy and its
Clinical Correlates in 400 Subjects with Alzheimer's Disease, Mild Cognitive Impairment, and Elderly
Controls, Human Brain Mapping, 30(9):2766-88, 2009.
32) Misra C, Fan Y, Davatzikos C. Baseline and longitudinal patterns of brain atrophy in MCI patients, and their
use in prediction of short-term conversion to AD: Results from ADNI. NeuroImage, 44(4):1415-22, 2009.
33) Leow AD, Yanovsky I, Parikshak N, Hua X, Lee S, Toga AW, Jack CR Jr, Bernstein MA, Britson PJ, Gunter
JL, Ward CP, Borowski B, Shaw LM, Trojanowski JQ, Fleisher AS, Harvey D, Kornak J, Schuff N,
Alexander GE, Weiner MW, Thompson PM, Alzheimer’s Disease Neuroimaging Initiative. Alzheimer’s
Disease Neuroimaging Initiative: A One-Year Follow-up Study Using Tensor-Based Morphometry
Correlating Degenerative Rates, Biomarkers and Cognition. Neuroimage. 45(3):645-55, 2009.
PMC2696624
34) Weiner MW. Editorial: Imaging and Biomarkers Will be Used for Detection and Monitoring Progression of
Early Alzheimer’s Disease. J Nutr Health Aging, 13(4):332, 2009.
869
35) Schuff N, Woerner N, Boreta L, Kornfield T, Shaw LM, Trojanowski JQ, Thompson PM, Jack CR Jr, Weiner
MW, and the Alzheimer's Disease Neuroimaging Initiative. MRI of Hippocampal Volume Loss in Early
Alzheimer’s Disease in Relation to ApoE Genotype and Biomarkers. Brain. 132(Pt 4):1067-77, 2009.
PMC2668943
36) McEvoy LK, Fennema-Notestine C, Cooper JC, Hagler D Jr, Holland D, Karow DS, Pung CJ, Brewer JB,
Dale AM for the Alzheimer’s Disease Neuroimaging Initiative. Alzheimer’s Disease: Quantitative structural
neuroimaging for detection and prediction clinical and structural changes in mild cognitive impairment.
Radiology, 251(1):195-205, 2009.
37) Chupin M, Gerardin E, Cuingnet R, Boutet C, Lemieux L, Lehericy S, Benali H, Garnero L, Colliot O. Fully
Automatic Hippocampus Segmentation and Classification in Alzheimer’s Disease and Mild Cognitive
Impairment Applied on Data from ADNI. Hippocampus, 19(6):579-587, 2009.
497
38) Jack Jr. CR, Lowe VJ, Weigand SD, Wiste HJ, Senjem ML, Knopman DS, Shiung MM, Gunter JL, Boeve
BF, Kemp BJ, Weiner M, Petersen RC, and the Alzheimer’s Disease Neuroimaging Initiative. Serial PIB
and MRI in normal, mild cognitive impairment and Alzheimer’s disease: implications for sequence of
pathological events in Alzheimer’s disease. Brain, 132(Pt 5):1355-65, 2009. PMC2677798
39) Potkin SG, Guffanti G, Lakatos A, Turner JA, Kruggel F, Fallon JH, Saykin AJ, Orro A, Lupoli S, Salvi E,
Weiner M, Macciardi F. Hippocampal atrophy as a quantitative trait in a genome-wide association study
identifying novel susceptibility genes for Alzheimer’s disease. PLoS ONE, 4(8):e6501-15, 2009.
PMC2719581
40) Chou YY, Leporé N, Avedissian C, Madsen SK, Parikshak N, Hua X, Shaw LM, Trojanowski JQ, Weiner
MW, Toga AW, Thompson PM; The Alzheimer's Disease Neuroimaging Initiative. Mapping Correlations
Between Ventricular Expansion and CSF Amyloid and Tau Biomarkers in 240 Subjects with Alzheimer’s
Disease, Mild Cognitive Impairment and Elderly Controls.
NeuroImage, 46(2): 394-410, 2009.
PMC2696357
41) Kovacevic S, Rafii MS, Brewer BJ and the Alzheimer’s Disease Neuroimaging Initiative. High-throughput,
Fully-automated Volumetry for Prediction of MMSE and CDR Decline in Mild Cognitive Impairment.
Alzheimer Disease and Associated Disorders, 23(2):139-145, 2009.
42) Petersen RC, and Trojanowski JQ. Time for Alzheimer’s Disease biomarkers? Potentially yes for clinical
trials, but not yet for clinical practice. JAMA, 302(4):436-7, 2009.
43) Vemuri P, Wiste HJ, Weigand SD, Shaw LM, Trojanowski JQ, Weiner M, Knopman DS, Petersen RC, Jack
Jr CR, and the Alzheimer’s Disease Neuroimaging Initiative. MRI and CSF biomarkers in normal, MCI, AD:
Diagnostic discrimination and cognitive correlations. Neurol., 73:287-293, 2009a.
048
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
44) Vemuri P, Wiste HJ, Weigand SD, Shaw LM, Trojanowski JQ, Weiner M, Knopman DS, Petersen RC, Jack
Predicting future clinical change. Neurol., 73:294-301, 2009b.
049
45) Querbes O, Aubry F, Pariente J, Lotterie JA, Demonet JF, Duret V, Puel M, Berry I, Fort JC, Celsis P; the
Alzheimer's Disease Neuroimaging Initiative. Early Diagnosis of Alzheimer’s Disease Using Cortical
Thickness: Impact of Cognitive Reserve. Brain, 132(8):2036-47, 2009. PMC2714060
46) Risacher SL, Saykin AJ, West JD, Shen L, Firpi HA, McDonald BC, and the Alzheimer’s Disease
Neuroimaging Initiative (ADNI). Baseline MRI Predictors of Conversion from MCI to Probable AD in the
ADNI Cohort. Current Alzheimer Research, 6:347-361, 2009.
234
47) Petersen RC. Commentary on “A roadmap for the prevention of dementia II: Leon Thal Symposium 2008.
A National Registry on Aging.” Alzheimer’s & Dementia, 5(2):105-7, 2009. PMCID: PMC2544623
48) Petersen RC. Early Diagnosis of Alzheimer’s disease: is MCI too late? Curr Alzheimer Res, 6(4):324-30,
2009.
49) Jagust WJ, Landau SM, Shaw LM, Trojanowski JQ, Koeppe RA, Reiman EM, Foster NL, Petersen RC,
Weiner MW, Price JC, Mathis CA, and For the Alzheimer’s Disease Neuroimaging Initiative. Relationship
between biomarkers in aging and dementia. Neurology, 73:1193-9, 2009.
50) Petersen RC, Jack Jr, CR. Imaging and Biomarkers in Early Alzheimer’s Disease and Mild Cognitive
Impairment. Clinical Pharmacology and Therapeutics, 86(4):438-41, 2009.
51) Reiman E et. al.: Categorical and Correlational Analyses of Baseline Fluorodeoxyglucose Positron
Emission Tomography Images From the Alzheimer's Disease Neuroimaging Initiative (ADNI).
NeuroImage, 2009 Jul 23. [Epub ahead of print].
52) Landau SM, Harvey D, Madison CM, Koeppe RA, Reiman EM, Foster NL, Weiner MW, Jagust WJ, the
Alzheimer’s Disease Neuroimaging Initiative. Associations between cognitive, functional, and FDG-PET
measures of decline in AD and MCI. Neurobiology of Aging, 2009 Aug 4. [Epub ahead of print].
834
53) Fennema-Notestine C, Hagler DJ Jr, McEvoy LK, Fleisher AS, Wu EH, Karow DS, Dale AM; the
Alzheimer's Disease Neuroimaging Initiative. Structural MRI Biomarkers for Preclinical and Mild
Alzheimer’s Disease. Human Brain Mapping, 2009 Mar 10. [Epub ahead of print].
54) Hampel H, Shen Y, Walsh DM, Aisen P, Shaw LM, Zetterberg H, Trojanowski JQ, and Blennow K.
Biological markers of β-amyloid related mechanisms in Alzheimer’s disease. Exper. Neurol. In press, 2009.
55) Petersen RC, Aisen PS, Beckett LA, Donahue MJ, Gamst AC, Harvey DJ, Jack Jr CR, Jagust WJ, Shaw
LM, Toga AW, Trojanowski JQ, Weiner MW, and the Alzheimer’s Disease Neuroimaging Initiative.
Alzheimer’s Disease Neuroimaging Initiative (ADNI): Clinical characterization, Neurol., In Press, 2009.
56) Jagust WJ, Landau SM, Shaw LM, Trojanowski JQ, Koeppe RA, Reiman EM, Foster NL, Petersen RC,
Weiner MW, Price JC, Mathis CA, and the Alzheimer’s Disease Neuroimaging Initiative. Relationships
between biomarkers in aging and dementia. Neurology, In Press, 2009.
57) Petersen RC, Knopman DS, Boeve BF, Geda YE, Ivnik RJ, Smith GE, Roberts RO, Jack CR Jr. Mild
Cognitive Impairment Ten Years Later. Arch Neurol, In Press, 2009.
58) De Meyer G, Shapiro F, Vanderstichele H, Vanmechelen E, Engleborghs B, De Deyn P-P, Hanson O,
Minthon L, Zetterberg H, Blennow K, Shaw LM, Trojanowski JQ, and the Alzheimer’s Disease
Neuroimaging Initiative. A mixture modeling approach to biomarker assessment reveals an Alzheimer’s
disease signature in more than a third of cognitively normal elderly people. Arch. Neurol., Submitted, 2009.
59) Ewers M, Walsh C, Trojanowski JQ, Shaw LM, Petersen RC, Jack CR, Jr, Bokde AWL, Feldman H,
Alexander G, Sheltens P, Vellas B, Dubois B, Hampel H, and the Alzheimer’s Disease Neuroimaging
Initiative (ADNI). Multi-modal biological marker based signature and diagnosis of early Alzheimer’s disease.
Submitted, 2009.
60) Hu WT, Chen-Plotkin A, Arnold S, Grossman M, Clark CM, Shaw LM, Leight Sb, McCluskey L, Elman L,
Karlawish J, Hurting HI, Siderowf S, Soares S, Lee VM-Y, and Trojanowski JQ. CSF biomarkers for
Alzheimer’s disease and frontotemporal lobar degeneration, Submitted, 2009.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
61) Jack CR, Knopman DS, Jagust WJ, Shaw LM, Aisen PS, Petersen RC, Trojanowski JQ. Modeling dynamic
biomarkers of the Alzheimer's pathological cascade. Lancet Neurology, Submitted, 2009.
62) Korecka M, Clark CM, Lee VM-Y, Trojanowski JQ, Shaw LM. Quantification of 8-iso-PGF2α and 8,12-isoiPF2α-VI in CSF and brain tissue from patients with Alzheimer’s disease by HPLC with tandem mass
spectrometry. J. Neurochemistry, Submitted, 2009.
63) Okonkwo OC, Alosco ML, Griffith HR, Mielke MM, Shaw LM, Trojanowski JQ, Tremont G, and the
Alzheimer's Disease Neuroimaging Initiative. CSF abnormalities and rate of decline in everyday function.
Arch. Neurol., Submitted, 2009.
64) Hu WT, Chen-Plotkin A, Arnold S, Grossman M, Clark CM, Shaw LM, Leight Sb, McCluskey L, Elman L,
Karlawish J, Hurting HI, Siderowf S, Soares S, Lee VM-Y, and Trojanowski JQ. CSF biomarkers for
Alzheimer’s disease and frontotemporal lobar degeneration. Submitted, 2009.
65) Korecka M, Clark CM, Lee VM-Y, Trojanowski JQ, Shaw LM. Quantification of 8-iso-PGF2¦Á and 8,12-isoiPF2¦Á-VI in CSF and brain tissue from patients with Alzheimer’s disease by HPLC with tandem mass
spectrometry. J. Neurochem., Submitted, 2009.
1.3.6. Overall Summary and Impact of ADNI1: Taken together the major impact of ADNI has been: 1)
establishment of standardized methods for imaging/biomarker collection and analysis which are starting to be
used in clinical trials. ADNI results on LMCI subjects replicated rates of conversion in a similar group of MCI
subjects enrolled using the “Petersen” criteria in the ADCS Vitamin E/Donepezil trial. ADNI1 utilized a
standardized neuropsychological battery, which has been subsequently used by industry and ADCS trials. The
MRI Core developed a structural MRI protocol, identical across vendors, with an MRI phantom for calibration.
This protocol has been used in numerous phase 2 and 3 treatment trials since. The PET Core established
methods for multisite FDG PET, and the first multisite C-11 PIB study. The biomarker core established
standardized methods for measurements of CSF Aβ amyloid and species of tau. The importance of these
standardization efforts should not be underemphasized since the ADNI methods have now been adopted for
other ADNI-like studies outside of the U.S. and this will facilitate comparisons of results among countries,
cultures, and ethnicities, and provide an infrastructure for world-wide clinical trials by the pharmaceutical
industry. 2) Provision of a large data base of images, genetic, fluid biomarker, and clinical data which is
being used by many investigators and industry. 3) New results concerning the neuroscience of AD, evidence of
AD pathology in normal subjects associated with greater rates of change of brain structure and brain glucose
metabolism; demonstration of outcome measures with high power to detect treatment effects; evidence that
abnormal CSF biomarkers predict future rates of brain atrophy, brain glucose metabolism, and cognition in
MCI; evidence that amyloid imaging and CSF Aβ provide similar information. An important long term goal of
our field is to identify and validate imaging/biomarkers for AD progression which can be used as “surrogate
markers” in place of clinical/cognitive tests in clinical trials. This is a very long way off, because such surrogate
markers must be validated in the treatment setting, across various types of treatments. Nevertheless, the ADNI
results are providing an important first step towards this goal. FDA is represented on the ADNI Steering
Committee, and the ADNI Executive Committee met with Dr. Russell Katz of the FDA in the winter of 2009; he
expressed very strong FDA support for the goals, processes, and results of ADNI. Taking all of the above
together, we conclude that the original stated goals of ADNI1 have been accomplished and surpassed (e.g. C11 PIB sub-study, the GWAS analysis) evidenced by the large number of publications resulting from this
project, and the other ADNI projects around the world. In this ADNI2 renewal, we propose continued follow-up
of the control, LMCI, and EMCI subjects from ADNI1 and GO, and enrollment of new control, EMCI, LMCI and
AD subjects, with baseline 6 month and then annual follow-up with clinical/cognitive/MRI and blood/urine
sampling. In addition there will be a 2nd time point for amyloid and FDG PET, and lumbar puncture for CSF
every other year on all subjects. ADNI is the only multisite study which obtains comprehensive longitudinal
clinical, imaging, and biomarker information across the continuum from normal aging to dementia, and it is
expected that ADNI2 will substantially contribute to development of effective treatments and preventative
approaches to AD. The success of ADNI1 has encouraged a competitive renewal application. The overall
goals and methods of each Core in ADNI2 are summarized next (many details in the various Core sections
which make up the body of this application).
1.4. METHODS: Our overall strategy is to 1) continue to follow all previously-enrolled EMCI, LMCI, and normal
control subjects who consent to continue in our study; 2) enroll new subjects who are normal controls EMCI,
LMCI, and AD; and 3) continue previous methods while introducing new methods including amyloid imaging
with AV-45, 3 Tesla MRI on all subjects, MRI sub-studies (diffusion tensor imaging, resting bold, and arterial
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
spin labeled perfusion MRI), new analyses of CSF and plasma, and targeted DNA sequencing. Major goals of
ADNI2, will be to: 1) replicate selected findings of ADNI1; 2) determine the most effect predictor of future rate
of change and outcome measures of change in EMCI subjects 3) compare the predictive value of F18 amyloid
imaging with CSF measurements of Aβ/tau at all stages from normals to AD. The blood and CSF samples
obtained from ADNI will be available for genetic studies, exploratory biomarkers studies, and replications of
new predictive methods developed by investigators. Taken together ADNI2 will continue to elucidate disease
mechanisms, identify the most useful predictors and outcomes at each stage along the AD continuum,
improve the efficiency of trial designs, and characterize a very early stage of disease, EMCI, that may be
optimal for assessing disease-modification interventions.
A brief summary of work to be done by each Core follows:
1.4.1. Administrative Core (see details below)
1.4.2. Clinical Core: Will continue all Clinical Core functions including development of protocols, consent
forms, training and procedure manuals, host Steering Committee meetings of Site PIs and Study coordinators,
assist in recruitment efforts including press releases and media contacts, monitor enrollment, receive all ADNI
clinical/cognitive data as well as numerical results from biomarker, MRI, PET, and genetics cores.
1.4.3. MRI Core: Will maintain MRI methodological consistency in previously enrolled ADNI1 subjects and
modernize the MRI protocol for newly enrolled subjects at 3T with 3D T1 volume, FLAIR, and a long TE
gradient echo volumetric acquisition (GRE) for micro hemorrhage detection. Additionally, it will perform pilot
sub-studies of arterial spin labeling (ASL) perfusion (Siemens), resting state functional connectivity (RSFC)
(GE) and diffusion tensor imaging (DTI) (Philips). The service aims of the central MRI core lab at Mayo Clinic
needed to generate high quality MRI data. The 5 funded ADNI MRI core analysis labs will generate numeric
summary MRI data which will also be made available to the general scientific community.
1.4.4. PET Core: Will maintain PET methodological consistency for FDG PET, and expand the PET protocol
for AV-45 and new PET scanners from sites that joined the PET project for the AV45 imaging. In addition to the
PET QC core, the 3 PET analysis labs will replicate selected findings from ADNI1 by applying the same ROIs
to a new sample of AD patients. Hypotheses to be tested are that FDG PET is a strong predictor, along with
memory function, in a multivariate model of conversion in MCI. And that the best predictors of ADAS-Cog
change are the CSF tau/Aβ levels and FDG PET. Additionally, results from AV-45 will be compared with C-11
PIB on those subjects who have both. The most important new work to be done by the PET Core in ADNI2 will
be longitudinal amyloid imaging, especially in EMCI
1.4.5. Biomarker Core: Will bank and curate biofluids from all subjects, distribute AD samples to investigators
and study promising AD biomarkers. It will continue measurements of cerebrospinal fluid (CSF) Aβ and tau
and perform plasma Aβ studies. We will study BACE and 20 new promising AD biomarkers in CSF and
plasma. It will determine whether a panel of CSF/ plasma biomarkers (rather than any single analyte) will: (1)
Predict conversion from normal to MCI or to AD and conversion from MCI to AD as well as identify MCI
subjects who have stable MCI and do not convert to AD; (2) Reflect the progression of AD from its prodromal
phase through to early/moderate stages of AD; (3) Predict the likelihood of maintaining healthy brain aging or
resistance to AD in the normal control population.
1.4.6. Genetics Core: Will receive and bank blood samples, extract DNA and RNA, store immortalized cell
lines, and perform APOE genotyping for new samples. It will perform genotyping using an updated Illumina
Human BeadChip compatible with the array used in ADNI1. It will perform quality control, sample verification
and organization of samples. It will collaborate with ADNI cores (clinical, biomarker, informatics, biostatistics,
MRI, PET) and with, National Cell Repository for Alzheimer's Disease (NCRAD) , the NIA-funded AD Genetics
Consortium, outside experts and other relevant entities, to facilitate integrative analyses for hypothesis testing
and novel discovery. It will perform comprehensive analysis of genetic influences on baseline data including
identifying novel risk markers and validating established markers that predict conversion and progression.
Emphasis will be on continuous phenotypic measures from structural, functional, and molecular neuroimaging,
other biomarkers such as CSF, blood, and urine, and clinical and neuropsychological variables. This core will
perform comprehensive analysis of genetic influences on longitudinal data with the goal of identifying genetic
features associated with rate and characteristics of progression.
1.4.7. Neuropathology Core: Will continue to perform pathological analysis on autopsy samples
1.4.8. Biostatistics Core: Will continue to perform statistical analysis on all data from ADNI1, GO and ADNI2
and to assist the other cores, and outside users concerning such analyses.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
1.4.9. Informatics Core: Will continue to receive and distribute data, and develop enhanced and user friendly
methods for querying the data base.
1.5. ADNI Governance and Administrative Core:
1.5.1. Overall description of organization and governance. ADNI is a U01 cooperative agreement grant,
and the NIA requires that this project be governed by a Steering Committee which consists of: the PI and all
funded Core leaders, all Site PIs, representatives from NIH and FDA, representatives from each of the
contributing companies as observers only. More than 100 people attended the Steering Committee in the
Spring of 2009.
The Figure above describes the governance and organization of ADNI1, GO and ADNI2
Concerning governance, the ADNI project closely follows the study design and methodology laid out in the
original ADNI1 grant application (and this will be true for ADNI2) but changes in scope are permitted. The day
to day decisions are made by the ADNI Executive Committee (Excom) which includes the PI, the Core leaders,
a representative of the NIA (Dr. Buckholtz), the current, past, and future Chairs of the Industry Scientific
Advisory Board (as observers), and David Lee of the Foundation for NIH (as observer). The ADNI Excom holds
two teleconference calls/month and has in-person meetings at the annual meeting of the Steering Committee
and at the International Conference on AD (ICAD). In addition, the Clinical Core (ADCS) has an ADNI
conference call at least monthly. The MRI, PET, Genetics, and Biostatistics Cores have conference calls
twice/month. Dr Weiner attends almost all teleconference calls.
All sites are managed by the ADNI Clinical Core at the ADCS, UC San Diego (Paul Aisen, PI). The
publications Committee vets all publications using ADNI data (see description below). The Industry Scientific
Advisory Board (ISAB) is composed of all companies which provide funds to ADNI and is managed by the
Foundation for NIH. The ISAB is chaired on a rotating basis. Chairs have included: William Potter (Merck), Eric
Siemers (Lilly), Patricia Cole (Eisai), Holly Soares (Pfizer), and in 2010 Mark Schmidt (Novartis).
All requests for specimens (blood, plasma, CSF, DNA, immortalized cell lines) go directly to the Resource
Allocation Review Committee consisting of members independent of ADNI and approved by NIA and chaired
by Dr. Tom Montine. Following approval by the RARC, NIA must approve release of all specimens.
1.5.2. Administrative Core: The Administrative Core, located at the VA Medical Center/University of
California San Francisco/Northern California Institute for Research and Education (the nonprofit foundation to
which all ADNI funds are awarded). It consists of the Principal Investigator of ADNI (M.W. Weiner), his
administrative staff, statistical support , and the Data and Publications Committee administered by Robert
Green at Boston University. Dr Weiner has responsibility for all administrative, financial, and scientific aspects
of ADNI. The following lists specific functions of the Administrative Core:
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
1) Responsibility for all activities of ADNI including strategic direction and scientific, administrative and
financial coordination of the entire project including non-competing and competitive renewals.
2) Responsibility for all budgets and subcontracts. Tracking all subcontracts, reconciling budgets, tracking
carryovers, maintaining an updated financial accounting and projections of expenses which closely
matches reconciliation data when financial status reports (FRS) are received.
3) Responsibility for interactions with NIH, and the ADNI Scientific Advisory Board (SAB)and with ADNI
projects in other countries.
4) Conference call with ADNI Executive Committee twice/month
5) Conference call with leaders of the ISAB twice/month
6) Monthly call with ADNI Clinical Core at ADCS
7) Conference call with MRI and PET cores (twice/month)
8) Organizing the meeting with the Scientific Advisory Board: Members are Drs. Zaven Khachaturian
(President, Prevent Alzheimer’s Disease 2020, Inc. [PAD2020], Editor in Chief Alzheimer’s & Dementia:
Journal of Alzheimer’s Association), William Thies (Alzheimer’s Association), Peter Snyder (University of
Connecticut), Howard Fillit (The Institute for the Study of Aging and the Alzheimer's Drug Discovery
Foundation), Gregory Sorenson (Harvard, MGH), Lewis Kuller (U Pittsburgh), Dennis Choi (Emory U.), and
William Potter (retiring from Merck).
9) Organizing the “ADNI weekend”, (together with the Clinical Core) which consists of meetings of the
Steering Committee, ISAB, Scientific Advisory Board, Excom, and other meetings.
10) Tracking of all scientific activity of ADNI
11) Interaction with all companies involved with/interest in ADNI e.g. amyloid imaging, and blood/CSF
biomarker companies, GE, Siemens, Philips, Alzheimer’s Association, other ADNI projects around the
world and to facilitate development of ADNI projects in countries where they don’t exist (e.g. Korea and
China now starting up such projects)
Dr Weiner is also PI of the GO grant, which closely relates to ADNI and separately administered with its own
account/fund and separate subcontracts. Dr Weiner supervises Enrique Menendez who is the lead grants
administrator for ADNI1/ADNI2 and GO. The financial tracking is highly complex (e.g. ADNI1 received over $65
million total costs) and involves the awarding of many different subcontracts to the various Core leaders,
tracking work performed and unspent funds etc. Finally, some image analysis of ADNI data, using FreeSurfer
(from Massachusetts General Hospital, Bruce Fischl PI) is performed at San Francisco VA, overseen by Dr
Schuff. This work is part of the MRI Core.
1.5.2.1. Data and Publications Committee (DPC. PI Robert Green): The DPC performs three tasks: (1)
develops and proposes policy to the Executive and Steering Committees with regard to data access and
publication; (2) screens all applications for access to ADNI data; and (3) reviews all publications for adherence
to ADNI publication policy guidelines. The DPC helped develop policies for open data access such that virtually
all requests for data access are granted. Persons requesting access to the data fill out a brief online
application form in which they indicate their academic affiliation and reason for requesting access or statement
about the project area they are interested in. Each of these applications is individually reviewed by the DPC
Chair. A table of individuals with access to the data and the projects they are pursuing is publically available
so that data users can be aware of the interests of others and reach out to other data users to form
collaborations if they wish. The DPC Administrator reviews manuscripts and requires all scientists who are
developing manuscripts using ADNI data to adhere to ADNI publication guidelines. These guidelines request
that authorship be stated in the “modified corporate authorship” format, in which the particular writing team is
named, and the authorship list is followed by the words “for the ADNI Study*” The asterisk then refers to a web
page where the ADNI leadership and individual site directors and co-investigators at each ADNI site are
named. In this manner, the ADNI leadership and ADNI site investigators can obtain “group authorship credit”
that provides at least modest academic credit for the work they are doing toward all ADNI publications.
A member of the DPC also reviews each manuscript for any that may have egregiously poor quality, but
importantly, does not attempt to review manuscripts for scientific quality or for duplication. It has been our
conscious policy to avoid practices that would inhibit or slow the utilization of ADNI data by the worldwide
scientific community. Therefore, we have decided that scientific review should occur at the level of publication
review, and that we will tolerate, and even encourage, multiple examinations of ADNI data by multiple
investigators. While this philosophy raises the possibility that two papers could present conflicting analyses or
interpretations, we have elected to let such potential conflicts play out in the “marketplace of ideas”.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
1.6. Human Subjects (see clinical Core)
1.7. References
1.
Miller, G, Alzheimer's Biomarker Initiative Hits Its Stride. Science, 2009. 326: p. 386-9.
2.
Hardy, J,Selkoe, DJ, The amyloid hypothesis of Alzheimer's disease: progress and problems on the
road to therapeutics. Science, 2002. 297(5580): p. 353-6.
3.
Braak, H,Braak, E, Frequency of stages of Alzheimer-related lesions in different age categories.
Neurobiol Aging, 1997. 18(4): p. 351-7.
4.
Thal, DR, Rub, U, Orantes, M,Braak, H, Phases of A beta-deposition in the human brain and its
relevance for the development of AD. Neurology, 2002. 58(12): p. 1791-800.
5.
Thal, DR, Rub, U, Schultz, C, Sassin, I, Ghebremedhin, E, Del Tredici, K, Braak, E,Braak, H, Sequence
of Abeta-protein deposition in the human medial temporal lobe. J Neuropathol Exp Neurol, 2000. 59(8):
p. 733-48.
6.
Holmes, C, Boche, D, Wilkinson, D, Yadegarfar, G, Hopkins, V, Bayer, A, Jones, RW, Bullock, R, Love,
S, Neal, JW, Zotova, E,Nicoll, JA, Long-term effects of Abeta42 immunisation in Alzheimer's disease:
follow-up of a randomised, placebo-controlled phase I trial. Lancet, 2008. 372(9634): p. 216-23.
7.
Cairns, NJ, Ikonomovic, MD, Benzinger, T, Storandt, M, Fagan, AM, Shah, AR, Reinwald, LT, Carter,
D, Felton, A, Holtzman, DM, Mintun, MA, Klunk, WE,Morris, JC, PiB-PET Detection of Cerebral Abeta
May Lag Clinical, Cognitive, and CSF Markers of Alzheimer's Disease: A Case Report. Archives of
Neurology, 2009. In Press.
8.
2009 Alzheimer's disease facts and figures. Alzheimers Dement, 2009. 5(3): p. 234-70.
9.
Katzman, R,Fox, P, The World-Wide Impact of Dementia. Projections of Prevalence and Costs, in
Epidemiology of Alzheimer's Disease: From Gene to Prevention, R. Mayeaux and Y. Christen, Editors.
1999, Springer-Verlag: Berlin, Germany. p. 1-17.
10.
Acosta, D,Wortmann, M, Alzheimer's Disease International World Alzheimer Report 2009, M. Prince
and J. Jackson, Editors. 2009, Alzheimer's Disease International: London, UK. p. 1-92.
11.
Visser, PJ, Verhey, F, Knol, DL, Scheltens, P, Wahlund, LO, Freund-Levi, Y, Tsolaki, M, Minthon, L,
Wallin, AK, Hampel, H, Burger, K, Pirttila, T, Soininen, H, Rikkert, MO, Verbeek, MM, Spiru, L,Blennow,
K, Prevalence and prognostic value of CSF markers of Alzheimer's disease pathology in patients with
subjective cognitive impairment or mild cognitive impairment in the DESCRIPA study: a prospective
cohort study. Lancet Neurol, 2009. 8(7): p. 619-27.
12.
Mattsson, N, Zetterberg, H, Hansson, O, Andreasen, N, Parnetti, L, Jonsson, M, Herukka, SK, van der
Flier, WM, Blankenstein, MA, Ewers, M, Rich, K, Kaiser, E, Verbeek, M, Tsolaki, M, Mulugeta, E,
Rosen, E, Aarsland, D, Visser, PJ, Schroder, J, Marcusson, J, de Leon, M, Hampel, H, Scheltens, P,
Pirttila, T, Wallin, A, Jonhagen, ME, Minthon, L, Winblad, B,Blennow, K, CSF biomarkers and incipient
Alzheimer disease in patients with mild cognitive impairment. JAMA, 2009. 302(4): p. 385-93.
13.
Bobinski, M, de Leon, MJ, Wegiel, J, Desanti, S, Convit, A, Saint Louis, LA, Rusinek, H,Wisniewski,
HM, The histological validation of post mortem magnetic resonance imaging-determined hippocampal
volume in Alzheimer's disease. Neuroscience, 2000. 95(3): p. 721-5.
14.
Gauthier, S, Dubois, B, Feldman, H,Scheltens, P, Revised research diagnostic criteria for Alzheimer's
disease. Lancet Neurol, 2008. 7(8): p. 668-70.
15.
Ellis, KA, Bush, AI, Darby, D, De Fazio, D, Foster, J, Hudson, P, Lautenschlager, NT, Lenzo, N,
Martins, RN, Maruff, P, Masters, C, Milner, A, Pike, K, Rowe, C, Savage, G, Szoeke, C, Taddei, K,
Villemagne, V, Woodward, M,Ames, D, The Australian Imaging, Biomarkers and Lifestyle (AIBL) study
of aging: methodology and baseline characteristics of 1112 individuals recruited for a longitudinal study
of Alzheimer's disease. Int Psychogeriatr, 2009. 21(4): p. 672-87.
16.
Arai, H, [Alzheimer's disease neuroimaging initiative and mild cognitive impairment]. Rinsho
Shinkeigaku, 2007. 47(11): p. 905-7.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Core: 2
Title of Core (not to exceed 81 spaces): Clinical Core
Core Leader: Aisen, Paul S.
Position/Title: Professor, University of California, San Diego
Department, service, laboratory, or equivalent: Neuroscience
Mailing Address:
9500 Gilman Drive #0949
La Jolla, CA 92093-0949
Human Subjects (yes or no): Yes – Pages 335-337
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael/Aisen, Paul S
The major goals of this Alzheimer’s Disease Neuroimaging Initiative (ADNI) are: 1) Develop improved
methods, which will lead to uniform standards for acquiring longitudinal multisite MRI and PET data on
patients with Alzheimer’s disease (AD), mild cognitive impairment (MCI), and elderly controls. 2) Acquire a
generally accessible data repository, which describes longitudinal changes in brain structure and
metabolism. In parallel, acquire clinical, cognitive and biomarker data for validation of imaging surrogates. 3)
Determine those methods, which provide maximum power to determine treatment effects in trials involving
these patient populations. In order to continue to retain the active population of NL and MCI subjects
enrolled in ADNI all identified subjects will have clinical/cognitive assessments and 1.5 T structural MRI
(eMRI) at an additional yearly visit. All scans will be transferred to central sites (MRI to Mayo Clinic and
PET to U Michigan) for immediate quality assurance so that subjects may be rescanned if necessary. All
clinical data will be collected, monitored, and stored by the Coordinating Center at UCSD. All biomarker
samples will be collected by UPenn. All raw scan data will be electronically transferred to the Laboratory of
Neuroimaging (LONI) at UCLA. All data will be monitored and analyzed by project statisticians, and data
base queries will be performed on request. Clinical, cognitive, imaging, and biomarker data bases will be
linked and raw, processed, and statistically analyzed data will rapidly be accessible to the public through
the Internet. Key UCSD responsibilities are: serving as the Coordinating Center for the clinical data,
monitoring enrollment of subjects, serving as a back-up repository for the entire dataset.
The overall goal of ADNI is to define the rate of progress of mild cognitive impairment and Alzheimer's
disease, to develop improved methods for clinical trials in this area, and to provide a large database which
will improve design of treatment trials. Recruiting new subjects and gathering expanded data in order to
collect information related to long term progression of cognitive function is vital to this task.
University of California, San Diego
80-435-5790
Street 1: Alzheimer's Disease Coop. Study
DUNS:
City:
La Jolla
Province:
Street 2:
County:
Country:
8950 Villa La Jolla Drive, Ste C227
State:
US
Zip/Postal Code:
CA
92037
53
DUNS:
Street 1:
Street 2:
City:
Province:
County:
Country:
State:
Zip/Postal Code:
PHS 398 (Rev. 11/07)
Page 2
Form Page 2
Weiner, Michael/Aisen, Paul S
SCIENTIFIC/KEY PERSONNEL. See instructions. Use continuation pages as needed to provide the required information in the format shown below.
Start with Program Director(s)/Principal Investigator(s). List all other key personnel in alphabetical order, last name first.
Name
Organization
Role on Project
Aisen, Paul
paisen
rgthomas
agamst
reraman
mrafii
peters8
UCSD
Clin Core Director
UCSD
Data Unit Director
UCSD
DataCore Faculty Coll
UCSD
DataCore Asst Director
UCSD
Medical Unit Assoc Dir
Mayo Clinic Rochester
Co Dir Clin Core
Thomas, Ronald
Gamst, Anthony
Raman, Rema
Rafii, Michael
Petersen, Ronald
Name
Organization
Role on Project
No
Yes
Cell Line
PHS 398 (Rev. 11/07)
Page 3
Weiner, Michael
Use only if additional space is needed to list additional project/performance sites.
DUNS:
ͺͲǦͶ͵ͷǦͷ͹ͻͲ
Street 1:
̵
City:

Province:
Street 2:
County:
Country:
ͺͻͷͲǡʹʹ͹

USA
State:
Zip/Postal Code:
ͻʹͲ͵͹
53

DUNS:
ͲͷǦͳͳͳ͵͵͵Ͳ
Street 1:
Department of Neurology
City:
Street 2:

Province:
County:
Country:
͸ͷͷͲǡ͓ͳͺͲͳ
State:
Zip/Postal Code:
͹͹Ͳ͵Ͳ

DUNS:
ͲͶͻͶ͵ͷʹ͸͸
Street 1:
̵
City:

Street 2:
County:
Province:
Country:
͹ͳͷǡǦ͵ʹͲ
State:
Zip/Postal Code:
Ͳʹͳͳͺ
ǯ
DUNS:
Ͳ͵Ͳͺͳͳʹ͸ͻ
Street 1:

City:

Province:
Street 2:
County:
Country:
ʹʹͳ
State:
Zip/Postal Code:
Ͳʹͳͳͷ

DUNS:
Ͳ͹Ǧ͹͹ͷǦͺͶͲ͹
Street 1:
12200 Fairhill Road
City:
Street 2:

Province:
County:
Country:
State:
Zip/Postal Code:

ͶͶͳʹͲ
Columbia University
DUNS:
049179401 Street 1:
:WK6W
City:
Street 2:
1HZ<RUN&LW\
Province:
County:
Country:
36%R[
State:
86$
Zip/Postal Code:
1<
Emory University
DUNS:
Street 1:
'HSDUWPHQWRI1HXURORJ\
City:
Street 2: :HVOH\:RRGV1HXUREHKDYLRUDO3URJUDP&OLIWRQ5G1(
$WODQWD
Province:
County:
Country:
State:
86$
Zip/Postal Code:
*$
Georgetown University
DUNS:
Street 1:
'HSDUWPHQWRI1HXURORJ\2QH%OHV%XLOGLQJ
City:
:DVKLQJWRQ
Province:
Street 2:
5HVHUYRLU5G1:
County:
Country:
State:
86$
Zip/Postal Code:
'&
Indiana University
DUNS:
Street 1:
18QLYHUVLW\%OYG6XLWH
City:
,QGLDQDSROLV
Province:
Street 2:
County:
Country:
State:
86$
Zip/Postal Code:
,1
Johns Hopkins University
DUNS:
Street 1:
1:ROI6WUHHW
City:
Street 2:
%DOWLPRUH
Province:
County:
Country:
State:
86$
Zip/Postal Code:
Mayo Clinic, Jacksonville
DUNS:
Street 1:
6DQ3DEOR5RDG
0H\HU
Street 2:
0'
City:
-DFNVRQYLOOH
Province:
County:
Country:
State:
86$
Zip/Postal Code:
)/
Mayo Clinic, Rochester
DUNS:
Street 1:
City:
Street 2:
5RFKHVWHU
Province:
County:
Country:
)LUVW6WUHHW6:
State:
86$
Zip/Postal Code:
01
Medical University of South Carolina
DUNS:
Street 1:
&RUH5RDG6XLWH
City:
Street 2:
1RUWK&KDUOHVWRQ
Province:
County:
Country:
State:
86$
Zip/Postal Code:
SC
Mt. Sinai Medical Center
DUNS:
Street 1:
'HSDUWPHQWRI3V\FKLDWU\%R[
City:
1HZ<RUN&LW\
Province:
Street 2:
County:
Country:
*XVWDYH/HY\3ODFH
State:
USA
Zip/Postal Code:
1<
New York University
DUNS:
Street 1:
6LOEHUVWHLQ$JLQJ'HPHQWLD5HVHDUFK
City:
1HZ<RUN&LW\
Province:
County:
Country:
Street 2: 'HSDUWPHQWRI3V\FKLDWU\)LUVW$YHQXH7+1
State:
86$
Zip/Postal Code:
1<

DUNS:
Street 1:
&RJQLWLYH1HXURORJ\$'&HQWHU
City:
&KLFDJR
Province:
Street 2:
County:
Country:
86$
(6XSHULRU6WUHHW
State:
Zip/Postal Code:
,/
60611
Oregon Health & Science University
DUNS:
Street 1:
6:6DP-DFNVRQ3DUN5G&5
City:
3RUWODQG
Province:
Street 2:
'HSDUWPHQWRI3V\FKLDWU\/
County:
Country:
State:
86$
Zip/Postal Code:
25
Rhode Island Hospital, Brown University
DUNS:
Street 1:
5KRGH,VODQG+RVSLWDO$3&
City:
Street 2:
3URYLGHQFH
Province:
(GG\6WUHHW
County:
Country:
State:
86$
Zip/Postal Code:
5,
Rush University Medical Center
DUNS:
Street 1:
5XVK$O]KHLPHU
V'LVHDVH&HQWHU
City:
&KLFDJR
Province:
Street 2:
County:
Country:
63DXOLQD1
State:
86$
Zip/Postal Code:
,/
Stanford University
DUNS:
Street 1:
$JLQJ&OLQLFDO5HVHDUFK&HQWHU
City:
Street 2:
6WDQIRUG
Province:
County:
Country:
'HSDUWPHQWRI3V\FKLDWU\&4XDUU\
State:
86$
Zip/Postal Code:
&$
Sun Health Research Institute
DUNS:
Street 1:
&OLQLFDO5HVHDUFK&HQWHU
City:
Street 2:
6XQ&LW\
Province:
County:
Country:
:6DQWD)H'ULYH
State:
86$
Zip/Postal Code:
$=
University of Alabama at Birmingham
DUNS:
Street 1:
WK$YHQXH6RXWK6XLWH
City:
%LUPLQJKDP
Province:
Street 2:
County:
Country:
86$
State:
Zip/Postal Code:
$/
University of California, Davis
DUNS:
047120084
Street 1:
Alzheimer's Disease Center
City:
Street 2:
Sacramento
Province:
County:
Country:
150 Muir Road (127A)
State:
86$
Zip/Postal Code:
CA
95817
University of California, Irvine
DUNS:
Street 1:
,QVWLWXWHIRU%UDLQ$JLQJDQG'HPHQWLD
City:
,UYLQH
Province:
Street 2: *LOOHVSLH1HXURVFLHQFH5HVHDUFK)DFLOLW\5RRP
County:
Country:
State:
86$
Zip/Postal Code:
&$
University of Nevada
DUNS:
Street 1:
'LYLVLRQRI1HXURORJ\
City:
Street 2:
/DV9HJDV
Province:
County:
Country:
:&KDUOHVWRQ6XLWH
State:
86$
Zip/Postal Code:
19
University of Pennsylvania
DUNS:
Street 1:
7KH5DOVWRQ+RXVH0HPRU\'LVRUGHUV&OLQLF
City:
3KLODGHOSKLD
Province:
Street 2:
County:
Country:
&KHVWQXW6W6XLWH
State:
86$
Zip/Postal Code:
3$
University of Pittsburgh
DUNS:
Street 1:
$'5&
City:
Street 2:
3LWWVEXUJK
Province:
County:
Country:
86$
University of Texas, Southwestern
:HVW830&0RQWHILRUH/RWKURS6WUHHW
State:
Zip/Postal Code:
3$
DUNS:
Street 1:
+DUU\+LQHV%OYG
City:
Street 2:
'DOODV
Province:
County:
Country:
State:
86$
Zip/Postal Code:
7;
University of Southern California
DUNS:
Street 1:
'HSDUWPHQWRI3V\FKLDWU\
City:
Street 2:
/RV$QJHOHV
Province:
County:
Country:
=RQDO$YHQXH.$0
State:
86$
Zip/Postal Code:
&$
University of Rochester
DUNS:
Street 1:
0RQURH&RPPXQLW\+RVSLWDO'HSDUWPHQWRI
City:
5RFKHVWHU
Province:
Street 2:
County:
Country:
3URJUDPLQ1HXUREHKDYLRUDO7KHUDSHXWLFV
State:
86$
Zip/Postal Code:
1<
University of Michigan
DUNS:
Street 1:
City:
Street 2:
$QQ$UERU
Province:
County:
Country:
(0HGLFDO&HQWHU'ULYH7DXEPDQ
State:
86$
Zip/Postal Code:
0,
Washington University, St. Louis
DUNS:
Street 1:
City:
Street 2:
6W/RXLV
Province:
County:
Country:
)RUHVW3DUN$YHQXH6XLWH
State:
86$
Zip/Postal Code:
02
Yale University
DUNS:
Street 1:
&HGDU6WUHHW&%
City:
Street 2:
1HZ+DYHQ
Province:
County:
Country:
86$
State:
Zip/Postal Code:
&7
DUNS:
University of Kansas
016060860
Street 1:
City:
Street 2:
KANSAS CITY
Province:
County:
Country:
State:
USA
Zip/Postal Code:
KS
66160
DUNS:
046705849
Street 1:
101 The City Drive South
City:
Street 2:
Orange
Province:
Bldg 54, Rm 203
County:
Country:
State:
USA
Zip/Postal Code:
CA
92868-4280
Duke University Medical
DUNS:
044387793
Street 1:
200 Trent Drive
City:
Street 2:
Durham
Province:
County:
Country:
State:
USA
Zip/Postal Code:
NC
27705
Wein Center
DUNS:
046025144
Street 1:
Mt. Sinai Medical Center/Miami Heart
City:
Miami Beach
Province:
County:
Country:
Street 2:
4300 Alton Road
State:
USA
Zip/Postal Code:
FL
33140
DUNS:
Thomas Jefferson University
053284659
Street 1:
City:
Street 2:
PHILADELPHIA
Province:
County:
Country:
State:
USA
Zip/Postal Code:
PA
19107-5587
DUNS:
University of Wisconsin
161202122
Street 1:
City:
Street 2:
MADISON
Province:
County:
Country:
USA
State:
Zip/Postal Code:
WI
53715-1218
Neurological Care of Central New York
DUNS:
796072135
Street 1:
1000 East Genessee St. Suite 405
City:
Syracuse
Province:
Street 2:
County:
Country:
State:
USA
Zip/Postal Code:
NY
13210
DUNS:
Howard University
056282296
Street 1:
City:
Street 2:
Washington
Province:
County:
Country:
State:
USA
Zip/Postal Code:
DC
20059
Olin Neuropsychiatry Research Center
DUNS:
065533796
Street 1:
Institute of Living
City:
Street 2:
Hartford
Province:
County:
Country:
200 Retreat Avenue, Whitehall Bdlg.
State:
USA
Zip/Postal Code:
CT
06106
DUNS:
University of California, Irvine - BIC
046705849
Street 1:
City:
Street 2:
Irvine
Province:
County:
Country:
State:
USA
Zip/Postal Code:
CA
92697-7600
University of California, San Francisco
DUNS:
094878337
Street 1:
San Francisco, CA 94143
City:
Street 2:
San Francisco
Province:
County:
Country:
State:
USA
Zip/Postal Code:
CA
94143-0962
Dartmouth Medical School and NRC
DUNS:
041027822
Street 1:
1 Rope Ferry Road
City:
Street 2:
Hanover
Province:
County:
Country:
USA
State:
Zip/Postal Code:
NY
03755-1404
Ohio State University
DUNS:
071650709
Street 1:
Enarson Hall 154 W 12th Avenue
City:
Street 2:
Columbus
Province:
County:
Country:
State:
USA
Zip/Postal Code:
OH
43210
DUNS:
University of Kentucky
939017877
Street 1:
City:
Street 2:
LEXINGTON
Province:
County:
Country:
State:
USA
Zip/Postal Code:
KY
40506-0057
Banner Alzheimer's Institute
DUNS:
788240674
Street 1:
901 East Willetta Street
City:
Street 2:
PHOENIX
Province:
County:
Country:
State:
USA
Zip/Postal Code:
AZ
85006
DUNS:
MD Clinical
621232482
Street 1:
City:
Street 2:
HALLANDALE BEACH
Province:
County:
Country:
State:
USA
Zip/Postal Code:
FL
33009
Wake Forest University
DUNS:
041418799
Street 1:
1834 Wake Forest Road
City:
Street 2:
WINSTON-SALEM
Province:
County:
Country:
State:
USA
Zip/Postal Code:
NC
27106
Albany Medical College
DUNS:
039486923
Street 1:
43 New Scotland Avenue
City:
Street 2:
ALBANY
Province:
County:
Country:
USA
State:
Zip/Postal Code:
GA
12208
DUNS:
Street 1:
City:
University of California Los Angeles
092530369
405 HILGARD AVE
Street 2:
Los Angeles
Province:
County:
Country:
LA
USA
State:
Zip/Postal Code:
CA
90095
DUNS:
190592162
Street 1:
43 New Scotland Avenue
City:
Street 2:
ALBANY
Province:
County:
Country:
State:
USA
Zip/Postal Code:
NY
12208
Banner Alzheimer’s Institute (BAI)
DUNS:
788240674
Street 1:
901 East Willetta Street
City:
Street 2:
PHOENIX
Province:
County:
Country:
State:
USA
Zip/Postal Code:
AZ
85006
Byrd Institute
DUNS:
150180060
Street 1:
4001 East Fletcher Avenue
City:
Street 2:
TAMPA
Province:
County:
Country:
State:
USA
Zip/Postal Code:
FL
33613
Dartmouth Medical School and NRC
DUNS:
041027822
Street 1:
1 Rope Ferry Road
City:
Street 2:
HANOVER
Province:
County:
Country:
State:
USA
Zip/Postal Code:
NH
03755
Dent Neurologic Institute
DUNS:
020413619
Street 1:
200 Sterling Drive
City:
Street 2:
Orchard Park
Province:
County:
Country:
USA
Duke University Medical Ctr
State:
Zip/Postal Code:
New York
14127
DUNS:
044387793
Street 1:
City:
Street 2:
DURHAM
Province:
County:
Country:
State:
USA
Zip/Postal Code:
NC
27705
Howard University
DUNS:
056282296
Street 1:
2400 Sixth Street
City:
Street 2:
WASHINGTON
Province:
County:
Country:
State:
USA
Zip/Postal Code:
DC
20059
MD Clinical
DUNS:
621232482
Street 1:
2500 E Hallandale Bch Blvd, Suite 505
City:
HALLANDALE BEACH
Province:
Street 2:
County:
Country:
State:
USA
Zip/Postal Code:
FL
33009
DUNS:
071650709
Street 1:
Enarson Hall 154 W 12th Avenue
City:
Street 2:
COLUMBUS
Province:
County:
Country:
State:
USA
OH
Zip/Postal Code:
43210
State:
Florida
Premiere Neurological Group
DUNS:
Street 1:
City:
4631 Congress Ave., Ste. 200
Street 2:
West Palm Beach
Province:
County:
Country:
USA
Zip/Postal Code:
33407
Thomas Jefferson University
DUNS:
053284659
Street 1:
1020 Walnut Street
City:
Street 2:
PHILADELPHIA
Province:
County:
Country:
DUNS:
016060860
USA
State:
Zip/Postal Code:
PA
19107-5587
Street 1:
City:
Lawrence
Street 2:
KANSAS CITY
Province:
County:
Country:
State:
USA
Zip/Postal Code:
KS
66160
DUNS:
939017877
Street 1:
City:
Street 2:
LEXINGTON
Province:
County:
Country:
State:
USA
Zip/Postal Code:
KY
40506-0057
DUNS:
161202122
Street 1:
N. Lake Street
City:
Street 2:
Madison
Province:
County:
Country:
State:
USA
Wisconsin
Zip/Postal Code:
53706
State:
Texas
UT, Southwestern Med Ctr at Dallas
DUNS:
800771545
Street 1:
5323 Harry Hines Boulevard
City:
Street 2:
Dallas
Province:
County:
Country:
USA
Zip/Postal Code:
75390- 9105
Wake Forest University SOM
DUNS:
937727907
Street 1:
1834 Wake Forest Road
City:
Street 2:
Winston-Salem
Province:
County:
Country:
State:
USA
Zip/Postal Code:
North Carolina,
27157
Wein Ctr for Clinical Research
DUNS:
046025144
Street 1:
4300 Alton Road
City:
Street 2:
Miami Beach
Province:
PHS 398/2590 (Rev. 11/07)
County:
Country:
USA
State:
Zip/Postal Code:
Florida
33140
Page
Project/Performance Site Format Page
Weiner, Michael
2. CLINICAL CORE:
A primary focus of the Alzheimer’s Disease Neuroimaging Initiative 2 (ADNI2) is to recruit subjects in the predementia state of cognitive impairment. ADNI1 recruited subjects with amnestic mild cognitive impairment
(aMCI) to represent this state. We will continue to pursue this strategy in ADNI2 but will add a group of
subjects with a milder degree of memory impairment and will call these subjects early MCI (EMCI). To contrast
these subjects with those recruited in ADNI1, the ADNI1 aMCI subjects will now be labeled as late MCI (LMCI).
Consequently, we will be following four groups of subjects in the Clinical Core, cognitively normal individuals
(CN), EMCI, LMCI and mild Alzheimer’s disease (AD) subjects. In addition to clinical and cognitive
assessments, these subjects will undergo MRI, FDG-PET and amyloid imaging, as well as cerebrospinal fluid
(CSF), blood and urine biomarkers. The trajectories of these imaging and chemical biomarkers will be followed
longitudinally in these four groups of subjects.
[Note: A list of abbreviations used is included at the end of this Core.]
2.1. Specific Aims:
The Specific Aims and hypotheses of the ADNI2 Clinical Core are divided into three groups: supportive aims
related to the management of ADNI activities, clinical assessment aims and hypotheses regarding the
cognitive and clinical data collected by ADNI, and clinical trial design aims and hypotheses utilizing ADNI data
to develop feasible scenarios for AD drug development.
2.1.1. Supportive Aims:
a. Develop/update protocol, procedures manual, technical procedures, consent forms.
b. Maintain the ADNI network of performance sites. Monitor site staffing, recruitment, compliance and data
entry, maintain financial support, replace or supplement sites as needed.
c. Manage supplies to sites.
d. Develop/modify the ADNI data system in accordance with ongoing and new ADNI aims.
e. Recruit and follow 550 new subjects (150 normal, 100 EMCI, 150 LMCI, 150 mild AD), and continue to
follow normal and LMCI subjects from ADNI1, and 200 EMCI subjects from GO.
f. Maximize lumbar punctures in new subjects.
g. Manage adjudication for conversion from CN to MCI, CN to AD and MCI to AD.
h. Provide help desk and reporting support to sites and investigators.
i. Maintain up-to-date clinical datasets on the ADNI data-sharing site at LONI.
j. Assist the Administrative Core in maintaining communication among ADNI investigators, sites, cores and
committees
2.1.2. Clinical Assessment Aims:
a. Confirmation of initial ADNI1 findings such as a rate of progression from LMCI to AD of 16-20% per year,
and the expected cognitive profile and rate of decline of CN, LMCI and mild AD subjects. The long-term followup of ADNI subjects is essential to characterize the longitudinal trajectories of cognitive, clinical and biomarker
assessments in each cohort, and to relate early changes to later clinical decline and dementia.
b. Characterization of EMCI
b1. EMCI is distinct from CN and LMCI in terms of cognition (ADAScog13, MoCA, and other measures),
clinical staging (CDR-SB), function (FAQ), brain volumes (hippocampus), brain activity (posterior cingulate by
FDG-PET), brain amyloid by PET, APOE4 prevalence, and CSF Aβ42, tau and p-tau, but not behavior (NPI).
b2. For tracking EMCI longitudinal decline (in terms of annual change divided by SD), the assessments are
ordered as follows: hippocampal volume>FDG-PET posterior cingulate>CDR-SB>FAQ>ADAScog13>CSF
Aβ42, tau, p-tau.
b3. There is greater decline, and longitudinal performance of these measures is substantially better, in the
APOE4 carrier subgroup in EMCI.
b4. In EMCI, there is a significant association among ADAScog13, CDR-SB, hippocampal volume and
FDG-PET posterior cingulate activity.
b5. In EMCI, there is an association among longitudinal decline in each of these measures.
c. Progression of EMCI to LMCI and dementia
c1. EMCI progresses to LMCI and dementia; the rate of progression to dementia is slower in EMCI than in
LMCI. Within EMCI, APOE4 carriers progress more rapidly to dementia than non-carriers.
c2. APOE genotype is strongly associated with progression of EMCI to LMCI and dementia
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
c3. Brain amyloid, as measured by low CSF Aβ42 or amyloid PET imaging, is strongly associated with
progression of EMCI to LMCI and dementia. The proportion of subjects with positive amyloid scans will
increase as follows: CN<EMCI<LMCI<AD.
2.1.3. Clinical Trial Design Aims:
a. Confirmation of the ADNI1 observations regarding the power of neuroimaging measures as outcomes for
trials in AD, LMCI and CN, and extension of these findings to EMCI.
a1. In the EMCI cohort, the order of power of assessments to demonstrate longitudinal decline will be: MRI
hippocampal volume>FDG-PET posterior cingulate>CDR-SB >ADAS-cog13 >ADAS-cog11.
a2. Three month changes in volumetric MRI measures will have good performance as outcomes,
correlating with later volumetric change and with long-term cognitive and clinical change.
b. Confirm the ADNI1 observations regarding the value of neuroimaging, cognitive and clinical measures as
covariates in longitudinal analyses. Extend these analyses to the EMCI population.
b1. Incorporation of covariates that reflect disease progression, such as hippocampal volume and posterior
cingulate activity by FDG-PET, will reduce sample sizes for EMCI clinical trials by 10%.
c. Confirm the ADNI1 observations regarding the relative power of longitudinal change versus survival to
diagnosis designs. Confirm the ADNI1 findings regarding the power of longitudinal change designs,
incorporating biomarkers for selection and covariates, in LMCI. Extend these analyses to the EMCI population.
c1. In EMCI, longitudinal change in CDR-SB and ADAScog-13 will provide greater statistical power than
survival to AD diagnosis. Power will be substantially improved by selecting subjects using amyloid biomarkers.
AD may be one of the most pressing problems facing all countries around the world as the population ages.
Currently, no therapies are available for AD that alter the underlying nature of the disease process, and while
symptomatic treatments provide some benefit, they are not the answer for the looming crisis.
Fortunately, there are more than 100 compounds under investigation by various pharmaceutical companies
and university medical centers around the world. Most of these therapies are designed to have an impact on
the underlying disease process itself. The earlier the intervention takes place, presumably, the greater the
protection against further neuronal damage will be appreciated.
ADNI has been a groundbreaking project, establishing pre-competitive collaboration and real-time data
sharing among academia and industry investigators to clarify the relationships among demographic, genetic,
clinical, cognitive, neuroimaging and biochemical measures throughout the course of AD neurobiology, in order
to facilitate the development of effective therapeutics. The project has exceeded expectations, providing
insights into disease mechanisms as well as hugely valuable advances, based primarily on the use of
standardized biomarkers, to drug development programs. ADNI has increased the rate of drug development;
disease-modifying therapies will arrive in the clinic sooner. A number of the leading disease-modifying drug
development programs are now employing ADNI methodology toward more efficient trial design, particularly in
the critically important early (pre-dementia) AD population.
AD can be diagnosed quite accurately at the dementia stage. In fact, a recent evidence-based medicine
review of the literature by the American Academy of Neurology documented that clinicians were quite accurate
when the clinical diagnoses were subsequently compared to neuropathological findings [1]. However, as one
identifies the disease process at an earlier point in the clinical continuum, the precision of the diagnosis is
reduced. Nevertheless, the challenge is to try to identify the process at the pre-dementia stage and enhance
the specificity of the clinical diagnosis through the use of imaging and other biomarkers. This approach
assumes an underlying cascade of pathological events that lend themselves to intervention. Biochemical and
neuroimaging biomarkers can provide a window on the underlying neurobiology, facilitating early identification
and intervention. Figure 1 presents a hypothetical model of the trajectories of biomarkers that have been
studied in ADNI1 and will continue to be followed longitudinally during the continuation of ADNI. To test this
model, it is essential to acquire very long-term longitudinal follow-up; this proposal covers up to a decade of
follow-up of the original ADNI1 subjects.
As is indicated in the Figure 1, there is evidence that the
accumulation of Aβ42 (presumed by some investigators to be the molecular trigger in AD neurodegeneration)
occurs early in the process. This can be detected by molecular imaging techniques such as PET scanning
using an amyloid-specific ligand or through measurement of cerebrospinal fluid (CSF) Aβ42. The next event
involves neuronal injury and dysfunction which may be detected by FDG-PET measures of regional metabolic
activity, elevated levels of CSF tau indicating neuronal damage, phospho-tau indicating accumulating tangle
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
pathology, and MRI volumetric changes. When a threshold of neuronal dysfunction is reached, cognitive and
then functional manifestations of AD accelerate. However, as mentioned above, by this point in the continuum,
it is likely that considerable damage has occurred in the central nervous system, and some of this may be
irreversible. Consequently, most investigators believe that early intervention is preferable to waiting to treat
individuals when the full dementia syndrome is present.
Figure 1. Trajectories of biomarkers during
the progression of AD. This hypothetical
graph is designed to capture the following
points: 1. CSF Aβ42 and amyloid PET,
reflecting amyloid accumulation in brain,
move in tandem. 2. Amyloid accumulation
precedes cognitive and functional decline by
years and changes only gradually once
symptoms develop. 3. Compared to CSF
Aβ42 and amyloid PET, CSF tau, MRI
volumes and FDG-PET are more dynamic
biomarkers of disease progression across
the spectrum of AD neurobiology. 4.
Cognitive decline becomes evident at the
onset of EMCI, and accelerates as the
disease progresses. 5. Functional decline
becomes evident at the onset of dementia,
and accelerates as the disease progresses.
6. All of these points are conjectural to
varying degrees; they require confirmation
with long-term longitudinal follow-up in
ADNI2 and beyond.
As shown in the Figure 1, amyloid
biomarker abnormalities are present during the asymptomatic stage. Ideally, one would like to intervene at this
point to have the greatest impact at preventing subsequent neuronal damage. However, there is a tremendous
challenge to identifying subjects at risk with sufficient sensitivity and specificity in the asymptomatic stage to
allow intervention at this point. ADNI2 will address the role of neuroimaging and other biomarkers at the stage
of early clinical symptom presentation (early MCI, or EMCI). Ultimately, it may be feasible to move diagnosis
and intervention into the asymptomatic stage.
In the current grant cycle, ADNI1 has focused on subjects with amnestic mild cognitive impairment (aMCI).
The construct of MCI has been extensively evaluated around the world and thousands of studies have been
completed in the past decade [2]. While all these studies are not entirely consistent, the wealth of the data
coalesce to indicate that aMCI is an identifiable entity with
a predictable progression to clinical dementia [3].
However, data to be reviewed below from ADNI1 indicate
that significant structural and functional imaging changes
as well as chemical biomarker profiles are evident at the
MCI stage as defined in ADNI1 [4]. For ADNI2, we will
evaluate a group of subjects who are earlier in the MCI
spectrum than was the case in ADNI1.
Figure 2. MCI construct.
In ADNI2, we plan to study a group of subjects with less severe memory impairment than found in the MCI
cohort enrolled in ADNI1. It is important to emphasize that these subjects will still meet criteria for aMCI, but
they will be at an earlier point in the clinical spectrum, as is shown in Figure 2. To accomplish this goal, with
funding from the National Institute on Aging through a GO grant, we will be recruiting 200 early MCI (EMCI)
from 2009 to 2011. In ADNI2, we propose to continue to follow these EMCI subjects along with 202 subjects
who are cognitively normal and 274 subjects who have late MCI (LMCI) (as defined in ADNI1) going forward.
In addition, we will recruit new cohorts of 150 cognitively normal subjects, 100 additional EMCI subjects, 150
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
LMCI subjects, and 150 subjects with mild AD. ADNI2 thus combines these newly recruited subjects with
those recruited in ADNI1 and GO (as shown in Table 8, below).
There is an increasing amount of data indicating that subjects with aMCI, similar to those characterized in
ADNI1, progress to clinical AD at a rate of 10 to 15% per year [5]. In addition, there are supporting data
indicating that adding structural MRI measures such as atrophy of medial temporal lobe structures,
hypometabolism patterns on FDG-PET or the CSF profile of decreased Aβ42 or elevated tau or phospho-tau
enhance the likelihood of predicting the rate of progression from MCI to AD [6]. Data from ADNI1 support
these contentions and imply that the clinical characterization of subjects with MCI can be enhanced with
imaging and chemical biomarkers to predict which subjects are more likely to progress more rapidly. This
premise will be extended in ADNI2 by studying a group of milder subjects with MCI hereby characterized as
EMCI.
The present application for continued funding aims to support this effort, building on all of the experience of
ADNI1. ADNI1 clearly established the utility of C11 PIB amyloid imaging; ADNI2 will examine the more widely
feasible F18 amyloid imaging (AV-45). ADNI2 will confirm and extend the striking findings linking regional
brain volumes (eg, hippocampal volume, regional cortical thickness) to AD progression, providing a potential
selection criterion, covariate or outcome measure for trials, as well as the utility of functional brain measures by
FDG-PET. The striking correspondence between CSF Aβ42 and amyloid imaging will be confirmed and
extended to the EMCI population. The utility of measures appropriate to primary care settings for the
screening and selection of mildly impaired subjects will be assessed.
The primary purpose of this ongoing work is to continue to elucidate disease mechanisms, improve the
efficiency of trial designs, and characterize a very early stage of disease, EMCI, that may be optimal for
disease-modification interventions.
2.3.1. Structural/supportive progress: The Clinical
Core successfully facilitated the accomplishment of
the ADNI aims, including the recruitment and
retention of over 800 subjects (229 normals, 380
LMCI, 210 AD), retention of subjects with an attrition
rate of only 6%, electronic data capture, quality
control and reporting, and coordination of FDG-PET,
C11 PIB PET, 1.5T MRI, 3T MRI and CSF
biomarkers. It has also initiated the GO project,
which will include the recruitment and follow-up of
an additional 200 subjects with EMCI. The
relationship of the Clinical Core to the flow of ADNI
data is shown in Figure 3.
Figure 3. Flow of data in ADNI.
The ADNI Clinical Core Infrastructure utilizes the Alzheimer’s Disease Cooperative Study (ADCS)
Administrative, Clinical Operations, Medical and Data Cores, all located at UCSD. The ADCS, a clinical trials
consortium continuously funded by NIA since 1991, consists of 35 primary sites (the site Principal Investigators
of which constitute the ADCS Steering Committee), 50 additional affiliated sites, 9 full-time clinical monitors,
and experienced informatics and biostatistical teams. The ADCS has completed over 25 clinical trials,
including over 6000 subjects.
To provide support to ADNI, the ADCS Data Core developed a flexible, powerful electronic data capture
system for data entry at each participating ADNI site. Investigator meetings, protocols, template informed
consent documents and procedures manuals have been developed, and modified for each amendment to
ADNI (for example, the addition of PIB imaging, the extension of the follow-up period, and the incorporation of
additional lumbar punctures). Clinical Monitors, under the supervision of the ADCS Medical Core, regularly
visit each ADNI performance site to ensure compliance with regulatory requirements and protocol procedures,
and accurate data entry. The ADCS Administrative Core manages the subcontracts for ADNI, tracking data
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
entry to guide the quarterly site payments, as well as all of the ADNI Investigator meetings. Day-to-day
management of ADNI is directed by the ADCS Clinical Operations Core; this work includes preparation and
maintenance of ADNI-related documents (such as procedures manuals), “help-desk” support to ADNI
investigators and staff and consumers of ADNI data, management of study supplies and specimens,
management of ADNI Executive Committee and site investigator communications, and tracking of overall study
progress.
The continuation of ADNI has been greatly facilitated by the award of funds through the Grand Opportunity
program within the ARRA NIH funding. The GO funding award was announced just one month prior to
submission of the current ADNI2 application. It provides support to ADNI activities over a two year period,
overlapping with and supplementing the first year of this proposed ADNI2 project. Specifically, the GO award
is supporting the following Clinical Core activities:
1. Two years of longitudinal follow-up for the original ADNI late MCI (LMCI) and cognitively normal (CN)
cohorts.
2. The recruitment of a new cohort of early MCI (EMCI) subjects with milder episodic memory impairment
than classical LMCI subjects enrolled in ADNI1.
3. The addition of new cognitive and clinical measures.
4. Additional biomarker and imaging data collection (as described in the corresponding sections of this
application).
2.3.2. Clinical Progress:
2.3.2.1. Subjects: ADNI1 began enrollment in late 2005 and completed enrollment of 819 subjects in 2007.
Fifty-seven sites in the U.S. and Canada participated in the enrollment. The original plan was to enroll 200
cognitively normal (CN), 400 aMCI and 200 mild AD subjects, and the final enrollment figures were as follows:
CN 229, aMCI 398 and mild AD 192.
The subjects enrolled were ages 55 to 90 (inclusive) and were required to have a study partner for an
independent evaluation of function. The general inclusion-exclusion criteria were as follows: All subjects had
to have a Hachinski Ischemic Score of less than or equal to 4, permitted medications had to be stable for four
weeks prior to screening, a Geriatric Depression Scale score of less than 6 and a study partner with 10+ hours
per week of contact either in person or on telephone, and the study partner had to accompany the participants
to the clinical visits. Subjects had to be in good health with no disease precluding enrollment, and women had
to be sterile for at least two years or past child-bearing potential. All subjects had to have six grades of
education or an equivalent work history and speak English or Spanish fluently. The subjects could not be
enrolled in other trials or studies concurrently. The study was approved by the local Institutional Review
Boards of the participating institutions.
The criteria for classification of subjects were as follows: With respect to memory complaints, the normal
subjects had none while the MCI and AD subjects both had to have complaints. On the MMSE, the range for
the normal subjects was 24-30 and for AD 20-26, all inclusive. The global CDR score for the CN group was 0
and for aMCI was 0.5 with a mandatory requirement of the memory box score being 0.5 or greater, and for the
AD subjects the CDR had to be 0.5 or 1. For the memory criterion, delayed recall of the Logical Memory II
subscale of the Wechsler Memory Scale-Revised was used with cutoff scores as follows based on education.
CN: greater than or equal to 9 for 16 years of education, greater than or equal to 5 for 8-15 years of education
and greater than or equal to 3 for 0-7 years of education. MCI and AD: less than or equal to 8 for 16 years of
education, less than or equal to 4 for 8-15 years of education and less than or equal to 2 for 0-7 years of
education.
No subjects could be taking antidepressant medications with anticholinergic properties and cholinesterase
inhibitors and memantine were permitted if the dose had been stable for three months prior to screening for
subjects with MCI and AD. All subjects provided demographics, family history and medical history.
At baseline, subjects were given the American National Adult Reading Test and the following cognitive
measures: digit span, category fluency, Trail Making A and B, Digit Symbol Substitution of the Wechsler Adult
Intelligence Scale-Revised, Boston Naming Test, Auditory Verbal Learning Test, clock drawing,
Neuropsychiatric Inventory Q, Alzheimer’s Disease Assessment Scale-Cognitive Subscale (ADAS-cog) and
Functional Activities Questionnaire (FAQ). All subjects had an MRI scan at 1.5 Tesla signal strength while
25% of the subjects also had an MRI scan at 3 Tesla, 50% had an FDG-PET scan, and approximately 50% of
the subjects at each site had a lumbar puncture.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
The original ADNI1 grant supported follow-up of the CN and MCI groups for 3 years, and the AD group for 2
years. Additional funding (notably the GO grant) supports ongoing annual evaluations of the CN and MCI
cohorts.
Characteristic
Age, mean ± SD, y
Education, mean ± SD, y
Years from symptom onset, y
Sex (% Female)
Marital status, %
Married
Widowed
Divorced
Never Married
Unknown
Apolipoprotein E є4, %
Carriers
Non-Carriers
Ethnicity
American Indian
Asian American
African American
Hispanic
White
Other
Controls
(n = 229)
MCI Group
(n = 398)
Mild AD Group
(n = 192)
P Value
75.8±5.0
16.0±2.9
NA
48.0
74.7±7.4
15.7±3.0
NA
35.4
75.3±7.5
14.7±3.1
3.9±2.5
47.4
0.137
<0.001
NA
0.002
0.002
a, c
a
<0.001
a, b, c
68.1
17.5
7.4
6.6
0.4
80.2
12.1
6.3
1.5
0
81.2
10.4
4.7
3.6
0
26.6
73.4
53.3
46.7
66.1
33.9
0
1.3
7.0
0.9
90.8
0
0.3
2.3
3.5
3.5
90.5
0
0
1.0
4.2
2.1
92.2
0.5
P<0.05 *
b, c
0.174
Table 1. Demographic Characteristics of the Participant Groups in ADNI1. * Multiple comparisons abbreviated as a: controls differ from
MCI, b: controls differ from AD, c: MCI differ from AD.
Table 1 shows the mean age and demographic features of the enrolled subjects. There were approximately
an equal number of men and women in the CN and AD subject groups, but there were more men than women
in the MCI group. The estimated premorbid verbal IQ of these subjects was almost 120 for the normal subjects
and 116 for the MCI and 114 for the AD subjects. Most subjects were white, and this was equivalent across
groups.
Assortment Variable
MMSE Score
CDR Global Score
CDR Sum of Boxes
Memory
Orientation
Judgment
Community Affairs
Hobbies
Personal Care
Hachinski Score
GDS Score
ADCS MCI-ADL (FAQ) Score
ADAS-Cog total
ADAS word list imm. recall
ADAS word list recognition
ADAS-cog without word list
PHS 398/2590 (Rev. 11/07)
Controls
MCI
Mean±SD
Mean±SD
29.1± 1.0 0.0± 0.0 0.0± 0.1 0.0± 0.0 0.0± 0.0 0.0± 0.1 0.0± 0.0 0.0± 0.0 0.0± 0.0 0.6± 0.7 0.8± 1.1 0.1± 0.6 6.2± 2.9 2.9± 1.1 2.6± 2.3 0.8± 0.9 27.0± 1.8
0.5± 0.0
1.6± 0.9
0.6± 0.2
0.2± 0.3
0.4± 0.3
0.2± 0.2
0.2± 0.3
0.1± 0.2
0.6± 0.7
1.6± 1.4
3.9± 4.5
11.5± 4.4
4.6± 1.4
4.6± 2.7
2.3± 2.0
Page
Mild AD
Z-score
MCI-Ctrl
-18.8
397
34.9
61.3
17.5
21.8
13.1
15.1
4.4
0.8
7.3
16.2
18.1
16.8
10.1
12.9
Mean±SD
23.3± 2.1
0.7± 0.3
4.3± 1.6
1.0± 0.3
0.8± 0.4
0.8± 0.4
0.7± 0.4
0.8± 0.5
0.2± 0.4
0.7± 0.7
1.7± 1.4
13.0± 6.9
18.6± 6.3
6.1± 1.5
6.6± 2.8
5.9± 4.1
Z-score
AD-MCI
-21.3 13.4 21.3 16.5 17.7 14.6 16.9 15.9 4.3 0.7 0.6 16.8 14.0 12.2 8.2 11.4 P-value
P<0.05
*
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
0.418
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
a, b, c
a, b, c
a, b, c
a, b, c
a, b, c
a, b, c
a, b, c
a, b, c
a, b, c
NA
a, b
a, b, c
a, b, c
a, b, c
a, b, c
a, b, c
ADAS word list delayed recall
AVLT Trials 1-5
AVLT delayed recall
AVLT DR/Trial 5 %
Trails A (seconds)
Trails B (seconds)
Category fluency (animal)
Category fluency (vegetable)
Number cancellation*
Boston naming test
Digits backwards
Clock drawing
CSF Biomarkers (pg/mL)
Tau A-beta 142 P-tau 181P 2.9± 1.7 43.3± 9.1 7.4± 3.7 65.8±27.6 36.5±13.2 89.2±44.3 19.9± 5.6 14.7± 3.9 0.4± 0.7 27.9± 2.3 7.2± 2.2 4.7± 0.7 (N=114)
69.7±30.4 205.6±55.1 24.9±14.6 Weiner, Michael
6.2± 2.3
30.7± 9.0
2.8± 3.3
32.1±31.3
44.9±22.8
130.7±73.5
15.9± 4.9
10.7± 3.5
1.0± 0.9
25.5± 4.1
6.2± 2.0
4.2± 1.0
(N=199)
101.4±62.2
162.8±56.0
35.5±18.0
20.8
-16.7
-15.6
-13.9
5.9
8.8
-9.1
-12.7
8.0
-9.4
-6.0
-7.6
6.0
-6.6
5.7
8.6± 1.6
23.2± 7.7
0.7± 1.6
11.2±22.0
68.0±36.9
198.9±87.2
12.4± 4.9
7.8± 3.3
1.8± 1.3
22.4± 6.2
5.0± 1.8
3.4± 1.3
(N=102)
119.1±59.6
143.0±40.8
41.6±19.8
15.0 -10.4 -10.3 -9.3 8.0 9.2 -8.1 -9.8 7.6 -6.2 -7.2 -7.5 <0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
a, b, c
a, b, c
a, b, c
a, b, c
a, b, c
a, b, c
a, b, c
a, b, c
a, b, c
a, b, c
a, b, c
a, b, c
2.4 -3.5 2.6 <0.001
<0.001
<0.001
a, b
a, b, c
a, b, c
Table 2. Baseline assessments of ADNI1 cohorts.
The mean scores for screening measures are shown in Table 2. In all cases except the Hachinski and the
GDS, the mean scores for the three groups differed significantly at p < 0.001 with CN performing best, AD the
worst and MCI in the middle. Overall, subjects with MCI had a mean ADAS-cog score of 11.5 while the normal
subjects were 6.2 and the mild AD subjects were 18.6 (p < 0.001). The three groups also differed significantly
on each of the subscales with the MCI mean scores worse than those for normal participants but not as low as
those for participants with AD. The neuropsychological battery indicated that, in general, subjects with MCI
were more impaired than the normal subjects on memory items and were only mildly impaired in the nonmemory domains.
2.3.2.2. Change Data: The annual change scores on cognitive and clinical assessments of subjects in the
three clinical groups over the course of 12 months are shown in Table 3.
As is apparent, the cognitively normal subjects did not decline on the MMSE while the MCI subjects
demonstrated a decline of approximately 0.7 points over 12 months. The mild AD subjects also declined but
somewhat more slowly than would be expected in more moderately impaired subjects.
The change on the ADAS-cog demonstrates similar findings. The normal control subjects are not declining to
any significant extent on any of the cognitive measures as a group, but the MCI and AD subjects are
worsening as would be expected. Projected sample sizes for detecting various interventions of therapeutics
are discussed below.
Assortment Variable
MMSE Score
CDR Global Score
CDR Sum of Boxes
Memory
Orientation
Judgment
Community Affairs
Hobbies
Personal Care
GDS Score
ADCS MCI-ADL (FAQ)
Score
ADAS-Cog total
ADAS word list imm.
recall
PHS 398/2590 (Rev. 11/07)
Controls
MCI
AD
Pvalue
P<0.0
5*
Mean±SD (N)
Mean±SD (N)
Zscore
Mean±SD (N)
Z-score
AD-MCI
0.0±1.4 (211)
0.0±0.1 (207)
0.1±0.3 (207)
0.0±0.1 (207)
0.0±0.1 (207)
0.0±0.2 (207)
0.0±0.1 (207)
0.0±0.1 (207)
0.0±0.0 (207)
0.2±1.2 (211)
0.1±1.0 (210)
-0.7±2.5 (358)
0.0±0.2 (358)
0.6±1.2 (358)
0.1±0.3 (358)
0.1±0.3 (358)
0.1±0.3 (358)
0.1±0.3 (358)
0.2±0.4 (358)
0.0±0.3 (358)
0.4±1.8 (358)
1.9±4.0 (354)
-4.1
-0.6
8.6
2.8
7.4
4.2
7.6
6.7
1.7
1.5
8.1
-2.4±4.1 (162)
0.3±0.5 (160)
1.6±2.2 (160)
0.2±0.4 (160)
0.3±0.5 (160)
0.2±0.5 (160)
0.3±0.5 (160)
0.3±0.6 (160)
0.2±0.6 (160)
0.3±1.8 (159)
4.6±5.6 (161)
-5.0
5.7
5.2
4.0
2.8
3.5
3.8
3.3
4.5
-0.9
5.5
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
0.340
<0.001
a, b, c
b, c
a, b, c
a, b, c
a, b, c
a, b, c
a, b, c
a, b, c
b, c
NA
a, b, c
-0.5±3.0 (210)
0.1±1.0 (210)
1.1±4.4 (357)
0.3±1.2 (358)
5.1
2.7
4.3±6.6 (161)
0.4±1.1 (162)
5.7
0.3
<0.001
0.020
a, b, c
a, b
Page
ADAS word list
recognition
ADAS-cog without word
list
ADAS word list delayed
recall
AVLT Trials 1-5
AVLT delayed recall
AVLT DR/Trial 5 %
Trails A (seconds)
Trails B (seconds)
Category fluency
(animal)
Category fluency
(vegetable)
Number cancellation*
Boston naming test
Digits backwards
Clock drawing
Weiner, Michael
-0.6±2.7 (210)
0.1±3.2 (358)
2.7
1.0±3.2 (161)
3.0
<0.001
a, b, c
-0.1±1.0 (210)
0.7±2.3 (357)
5.2
2.9±4.9 (161)
5.6
<0.001
a, b, c
0.1±1.6 (210)
0.5±1.9 (358)
2.6
0.4±1.2 (161)
-0.5
0.030
a
0.2±7.8 (209)
0.4±3.4 (210)
4.9±43.7 (209)
-2.3±11.2 (211)
-6.6±38.0 (210)
0.5±4.6 (211)
-1.3±6.3 (357)
-0.4±2.4 (358)
-5.0±28.3 (353)
1.2±16.1 (358)
9.0±56.1 (352)
-0.7±4.4 (358)
-2.4
-3.1
-2.9
3.0
3.9
-2.9
-3.6±5.7 (156)
-0.5±1.6 (155)
-7.1±21.0 (147)
3.9±21.7 (157)
20.0±85.6 (133)
-1.5±4.0 (160)
-4.2
-0.4
-0.9
1.4
1.4
-2.1
<0.001
<0.001
<0.001
0.001
<0.001
<0.001
b, c
a, b
a, b
a, b
a, b
a, b
-0.1±3.5 (211)
-0.6±3.1 (358)
-1.6
-1.0±2.8 (160)
-1.4
0.030
b
0.0±0.7 (209)
0.5±1.7 (210)
0.1±1.9 (211)
0.0±0.8 (211)
-0.1±0.9 (354)
-0.2±2.9 (356)
-0.3±1.7 (356)
-0.1±1.0 (357)
-0.9
-4.0
-2.2
-1.1
0.3±1.4 (155)
-1.5±3.7 (159)
-0.2±1.7 (152)
-0.4±1.3 (162)
3.0
-3.8
0.4
-2.8
<0.001
<0.001
0.070
<0.001
b, c
abc
NA
bc
Table 3. 12 month change in assessments in ADNI1 cohorts.
2.3.2.3. Early MCI: The EMCI group is a newly characterized set of subjects to be recruited in the GO grant
funding period. To assess the clinical characteristics of the EMCI and LMCI subject groups, we interrogated
the database from the NIA-sponsored Alzheimer’s Disease Center Program through the National Alzheimer’s
Coordinating Center
(NACC) under the direction of Dr. Walter
EMCI (N = 181)
LMCI (N = 369)
Kukull. We used the NACC database since
12%
27%
Progression to
it represents subjects who were classified
Dementia (Year 1)
as MCI using essentially the same criteria
Progression to CDR 1
proposed in ADNI. However, since there
(Year 1)
7%
15%
was no sub-categorization of EMCI or LMCI
MMSE
in the NACC database, we imposed the
Baseline
28.0 (1.6)
27.3 (1.8)
proposed criteria for GO and ADNI2 on
Year 1
27.6 (2.1)
26.2 (2.8)
previously collected aMCI subjects. The
CDR – SB
summary of the cognitive characteristics of
Baseline
1.2 (0.9)
1.5 (1.0)
LMCI subjects from ADNI1 have been
Year 1
1.6 (1.3)
2.1 (1.6)
described above, but the features of EMCI
FAQ
subjects have not been characterized, and
Baseline
3.7 (5.3)
4.7 (5.6)
Year 1
5.5 (6.9)
6.5 (6.4)
consequently, a comparison of these two
clinical groups at
Table 4. MCI data from NACC; annual rates of progression
and instrument means.
baseline and with respect to rates of progression from the NACC database is shown in Table 4. As can be
seen, the EMCI subjects represent individuals with milder degrees of cognitive and functional impairment than
the LMCI subjects and their rate of progression is slower. We anticipate the subjects recruited in ADNI2 will
conform to these general clinical characteristics.
2.3.3. Clinical Trial Design Progress: The real-time, public sharing of ADNI demographic, clinical, cognitive
and biomarker data has facilitated clinical trial design in academic and industry programs world-wide.
The great majority of AD drug development programs focus on symptomatic and disease-slowing effects in
subjects with AD dementia [7]. The most widely-used co-primary outcome measures for such trials are the
ADAS-cog for cognition and the CDR-SB for clinical status. The Neuropsychiatric Inventory (NPI) is the
standard measure of behavioral symptoms. Each of these measures is part of the assessment battery of
ADNI.
Therefore, trialists can and do use the shared ADNI data to explore the relationships among demographic
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
parameters and performance on these measures, can explore analytical methods and covariance structures,
and can estimate the power of various trial designs. The value of genetic, biochemical and neuroimaging
biomarkers for subject selection, reduction of explained variance and supportive outcome measures can
likewise be explored. There is a growing consensus that the optimal population for disease-modification programs is not AD
dementia, but rather pre-dementia individuals on the spectrum of AD neurobiology [7]. It is reasonable to
assume that interventions targeting the pathophysiological mechanisms underlying AD will have the greatest
effect before the pathology is at the advanced stage that corresponds to dementia. Many efforts are under
way by academic, industry, foundation and government groups to facilitate this direction; these efforts include
proposed revisions to diagnostic criteria, and various meetings and task forces to explore trial design issues.
ADNI data have provided the basis for much of this work. ADNI has focused on amnestic MCI, and has
included the leading candidate outcome measures and biomarkers, allowing assessment of proposed trial
designs. Importantly, ADNI data has revealed that appropriate use of standard outcome measures and
biomarkers can yield powerful and feasible trial designs for the pre-dementia (MCI) population.
Specifically, ADNI1 data [8] indicate:
a. The CDR-SB is a powerful outcome measure in mild AD and MCI.
b. In MCI, the ADAS-cog13 is superior to the ADAS-cog11 or ADAS-cog12 in a 24 month trial.
c. Covariates reduce samples sizes by 10-15 % in LMCI and mild AD
d. Selection of LMCI subjects using CSF Aβ42 reduces sample sizes
2.3.3.1. Rate of change designs in comparison to survival to dementia designs, and the impact of
biomarker selection and covariates: We have used ADNI data to propose study designs (Table 5) for
disease-modifying interventions in the pre-dementia population [8]. For such a trial, it is rational to select
subjects with evidence of amyloid accumulation in brain; ADNI data suggests that amyloid PET imaging and
low CSF Aβ42 are equivalent indicators of amyloid accumulation. Subjects meeting the ADNI criteria for
amnestic MCI and selected based on an abnormal amyloid biomarker would meet the proposed research
criteria for AD [9], and consensus meetings suggest that standard AD-type cognitive and clinical co-primary
outcome measures will be appropriate for pivotal trials, and a single clinical measure may be appropriate for a
Phase II proof of concept trial. Using ADNI data, we have shown that a two-year treatment period in this
population, with appropriate covariates, has reasonable power to demonstrate effects on primary measures
(Table 6, below). A design similar to this has recently been launched as a Phase II proof of concept trial of a
secretase inhibitor (ClinicalTrials.gov identifier NCT00890890).
Cognitive Status
Clinical Dementia Rating
global score
MMSE range
Biomarker for subject
selection
Biomarker for subject
stratification
Primary cognitive
outcome measure
Primary global/functional
outcome measure
Analysis covariates
Biomarker outcome
Duration of treatment
Primary analysis
Mild AD Trial
Mild dementia
Early AD Trial
Mild cognitive
impairment
0.5
Very Early AD Trial
Cognitively normal
25-30
Amyloid imaging and/or
CSF Aβ42
APOE genotype
28-30
Amyloid imaging and/or
CSF Aβ42
APOE genotype
CDR-SB
ADAScog12 (includes
delayed recall)
CDR-SB
Sensitive memory
and/or exec. fxn test
none
Baseline cognition and
regional brain volume
Regional brain atrophy
Baseline cognition and
regional brain volume
Regional brain volume
18months
Change score or slope
of cp-primaries:
ADAScog11, CDR SB
24months
Change score or slope
of co-primaries:
ADAScog12, CDR-SB
0.5-1
16-26
None
None or APOE
genotype
ADAScog11
0
(as surrogate endpoint)
24-36months
rate and cognitive
decline
Table 5. Clinical trial design scenarios utilizing change in continuous outcomes being explored in ADNI.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Randomized, placebo controlled trials of AD therapies for MCI have historically used time-to-conversion to
AD as the primary endpoint [10-12]. This approach faces practical, technical and analytic difficulties. Using
simulations and analytic methods based on data from ADNI1, we have compared the power to detect a
treatment effect using two endpoints: rate of cognitive change as estimated by linear mixed models (LMM) and
time-to-conversion as estimated by Cox [13] proportional hazards models. We have also evaluated the impact
of using hippocampal volume as a baseline covariate and selecting subjects on the basis of biomarker
evidence of amyloid dysregulation. We find that linear mixed models, as well as the baseline covariate and
subject selection strategies, can improve the power to detect a treatment effect.
Rate of change on continuous measures such as the ADAS-Cog and CDR-SB can be estimated by LMM
[14]. An important disadvantage of the rate-of-change endpoint, relative to time-to-dementia, is its clinical
interpretation. That is, the clinical significance of a percent change on a cognitive assessment score is not as
apparent as the clinical significance of a percent change in the two year conversion-free survival rate.
However, one might expect statistical inference based directly on continuous measures of disease severity to
be more efficient than inference based on the dementia endpoint subjectively derived from the same
continuous measures.
A direct analytic comparison based on standard power calculations under the two models (Cox vs. LMM) is
not straightforward because treatment effect sizes are not easily translated from one model to the other. In
other words, one cannot assume that a 25% change in rate of ADAS-Cog decline is equivalent to a 25%
change in the progression-free survival rate. To ensure we were comparing like treatment effects, we opted to
simulate randomized clinical trials and analyze the simulated data with both models to compare the efficiency
with which they detect a simulated treatment effect.
Simulated continuous outcomes can be easily generated from linear mixed models fit to the ADNI data.
However, simulating conversion events associated with the same continuous data requires an algorithmic
definition of conversion. To derive such an algorithm, we regressed observed ADNI conversions on baseline
and follow-up ADAS-Cog, CDR-SB, and the Functional Activities Questionnaire (FAQ) [15] z-scores using a
repeated binary outcome Generalized Estimating Equation (GEE) logistic regression model [16]. The resulting
model provides a continuous linear predictor of conversion, to which a conversion threshold can be applied.
The conversion threshold was tuned to produce about a 40% conversion rate over two years in simulated
placebo group data, comparable to observed conversion rate observed in ADNI. Our estimated conversion rule
was in agreement with actual clinical decisions for 327/391=83.6% of MCI subjects. The sensitivity and
specificity of the algorithm for detecting clinical conversion decisions was 111/134=82.8% and 216/257=84.0%.
Note that in the multivariate simulations, the Cox models are utilizing information from two assessments that is
not available to the LMM model.
To test the relative efficiency of a pre-specified enrichment strategy, simulations were conducted using
estimates from the ADNI MCI population as a whole, as well as the subgroup exhibiting beta amyloid
dysregulation (MCI-Aβ). This was defined using a CSF Aβ42 cutpoint of 192 pg/ml, independently estimated by
Shaw et al [6]. We also used baseline FreeSurfer hippocampal volumes provided by UCSF, and serial ADASCog, CDR-SB, and FAQ assessed every six months for two years. The available sample size for estimating the
model parameters was N=331 for MCI and N=126 for MCI-Aβ. Treatment group data is simulated assuming
various percent improvements in rate of change in the outcome variables compared to the placebo rates
estimated from ADNI data. We also simulate dropout resulting in about 15% attrition. We simulated data from
1000 clinical trials, analyzed using LMM and Cox models with and without baseline hippocampal volumes, and
estimated statistical power by the proportion of trials that rejected the null hypothesis of no treatment effect
(p<0.05).
A byproduct of the simulations were estimated conversion rates associated with a given treatment effect on
the rate of change of the continuous measures. Given these estimated conversion rates, we were able to do a
more direct analytical comparison of LMM and Cox. For the LMM sample size calculations we used the formula
of Liu and Liang [17] and for the Cox sample sizes we used a simulation based sample size calculation
provided by the Hmisc statistical software package [18]. The LMM calculations use the standard conservative
inflation to account for missing data, while Cox estimates use simulated missingness.
Table 6 summarizes the simulation-based comparisons of the LMM v.Cox methods of primary analysis; we
found there to be a clear advantage to using LMM versus the Cox model. For instance, to attain 80% power
(α=5%) to detect a δ=25% treatment effect would require N=998 per group with Cox versus N=558 with LMM
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
(CDR-SB). In the MCI sub-population with amyloid dysregulation, power was similar using Cox compared to
ADAScog, but improved for CDR-SB (N=427 Cox vs. N=338 CDR-SB for δ=25%). Baseline hippocampal
volumes consistently improved the efficiency of LMM.
Population
MCI
Treatment
Effect (%)
40
25
MCI-Aβ
40
25
Outcome
ADAS13
CDR-SB
ADAS13
CDR-SB
ADAS13
CDR-SB
ADAS13
CDR-SB
Mixed Model N (per
group)
No cov
Hipp vol
301
219
769
558
173
132
441
338
292
202
745
516
156
118
399
299
Cox Model N (per
group)
No covariate
Assumed Conversions (%)
Control
Active
440
35.9
27.5
998
36.0
30.3
200
46.2
33.0
427
46.4
37.4
Table 6. Analytic sample size calculations (power=80%) assuming the simulated conversion rates for the assumed treatment effect on
the continuous outcome.
2.3.3.2. The potential value of neuroimaging measures as surrogate outcomes: ADNI data has confirmed
that the annual change and variance for neuroimaging measures provides much better power to detect
disease-slowing effects than do standard cognitive and clinical measures.
For example, Table 7 provides group sizes for a study aiming to demonstrate a 25% slowing of disease
progression as indicated by one year change in various outcome measures (analyzed by linear mixed effects
models), with 80% power and an alpha of 0.05.
MCI
AD
Hipp.vol.(avg, L&R, Dale)
208
99
FDG-PET (Jagust)
3360
255
ADAS-cog 11
4099
407
MMSE
4162
632
CDR-SB
954
465
Table 7. Comparison of imaging and cognitive/clinical outcome measures to power trials.
In general, we observe that hippocampal volumetric change has excellent power in AD and MCI. This
suggests that for an intervention expected to slow clinical progression and brain atrophy, a Phase II a one year
proof of concept trial in AD might be conducted with reasonable sizes. The sample sizes in the table above
can be substantially reduced by the incorporation of biomarker selection criteria and covariates, as shown for
ADAS-cog and CDR-SB in Table 6.
To extend this idea further, we have found that 6 month change in volumetric MRI measures provides good
power to demonstrate slowing of progression in MCI. For example, for a 6 month trial to demonstrate 25%
slowing in AD and MCI would require 1055 and 13074 subjects using the ADAScog, 1084 and 3388 using the
CDR-SB, but only 216 and 528 for hippocampal volume. If we optimize biomarker selection and covariates, a
6 month proof of concept study is feasible. We note that the 6 month change in hippocampal volume is highly
correlated with later interval changes, and is correlated to later decline in cognitive and clinical measures.
In ADNI2, we will extend this still further, by adding 3 month volumetric MRI scans for all newly enrolled
subjects, to explore the feasibility of this measure for brief proof of concept trials, and for interim
analysis/adaptive designs for longer trials.
2.3.4. ADNI-Related Publications by Clinical Core Investigators:
1. Mueller SG, Weiner MW, Thal LJ, Petersen RC, Jack C, Jagust W, Trojanowski JQ, Toga AW, Beckett L.
The Alzheimer’s Disease Neuroimaging Initiative. Neuroimaging Clin N Am, 15(4):869-77, 2005.
PMC2376747
2. Mueller SG, Weiner MW, Thal LJ, Petersen RC, Jack C, Jagust W, Trojanowski JQ, Toga AW, Beckett LA.
Ways toward an early diagnosis in Alzheimer’s disease: The Alzheimer’s Disease Neuroimaging Initiative.
Cognition and Dementia, 5(4):56-62, 2006.
3. Shaw LM, Vanderstichele H, KnapikCzajka M, Clark CM, Aisen PS, Petersen RC, Blennow K, Soares H,
Simon A, Lewczuk P, Dean R, Siemers E, Potter W, Lee V, Trojanowski JQ and the Alzheimer's Disease
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Neuroimaging Initiative. Cerebrospinal Fluid Biomarker Signature in Alzheimer's Disease Neuroimaging
Initiative Subjects. Annals of Neurology, 65: 403-13, 2009. PMC2696350
4. Jack CR Jr, Lowe VJ, Weigand SD, Wiste HJ, Senjem ML, Knopman DS, Shiung MM, Gunter JL, Boeve
BF, Kemp BJ, Weiner M, Petersen RC; Alzheimer's Disease Neuroimaging Initiative. Serial PIB and MRI in
normal, mild cognitive impairment and Alzheimer's disease: implications for sequence of pathological
events in Alzheimer's disease. Brain,132(Pt 5):1355-65, 2009. PMC2677798.
5. Khachaturian ZS, Petersen RC, Gauthier S, Buckholtz N, Corey-Bloom JP, Evans B, Fillit H, Foster N,
Greenberg B, Grundman M, Sano M, Simpkins J, Schneider LS, Weiner MW, Galasko D, Hyman B, Kuller
L, Schenk D, Snyder S, Thomas RG, Tuszynski MH, Vellas B, Wurtman RJ, Snyder PJ, Frank RA, Albert
M, Doody R, Ferris S, Kaye J, Koo E, Morrison-Bogorad M, Reisberg B, Salmon DP, Gilman S, Mohs R,
Aisen PS, Breitner JC, Cummings JL, Kawas C, Phelps C, Poirier J, Sabbagh M, Touchon J, Khachaturian
AS, Bain LJ. A roadmap for the prevention of dementia: the inaugural Leon Thal Symposium. Alzheimers
Dement, 4:156-63, 2008.
6. Andrieu S, Coley N, Aisen PS, Carrillo M, Dekosky S, Durga J, Fillit H, Frisoni G, Froelich L, Gauthier S,
Jones R, Jonsson L, Khachaturian Z, Morris J, Orgogozo J-M, Ousset P-J, Robert P, Salmon E, Sampaio
C, Verhey F, Wilcock G, Vellas B. Methodological issues in primary prevention trials for neurodegenerative
dementia. J Alzheimer’s Disease, 16: 235-270, 2009.
7. Aisen PS. Interpreting Biomarker Data in Therapeutic Trials. Journal of Health Nutrition and Aging, 13:
370-2, 2009.
8. Aisen PS. Facilitating Alzheimer’s disease drug development in the United States. Alzheimer’s &
Dementia, 5: 125-7, 2009. PMC2750898.
9. Rafii MS, Aisen PS. Recent Developments in Alzheimer’s Disease Therapeutics
10. BioMed Central Medicine, BMC Med 7: 7, 2009. PMC2719107.
11. Khachaturian ZS, Snyder PJ, Doody R, Aisen P, Comer M, Dwyer J, Frank RA, Holzapfel A, Khachaturian
AS, Korczyn AD, Roses A, Simpkins JW, Schneider LS, Albert MS, Egge R, Deves A, Ferris S, Greenberg
BD, Johnson C, Kukull WA, Poirier J, Schenk D, Thies W, Gauthier S, Gilman S, Bernick C, Cummings JL,
Fillit H, Grundman M, Kaye J, Mucke L, Reisberg B, Sano M, Pickeral O, Petersen RC, Mohs RC, Carrillo
M, Corey-Bloom JP, Foster NL, Jacobsen S, Lee V, Potter WZ, Sabbagh MN, Salmon D, Trojanowski JQ,
Wexler N, Bain LJ. A roadmap for the prevention of dementia II: Leon Thal Symposium 2008. Alzheimer’s
& Dementia, 5: 85-92, 2009.
12. Aisen PS. Alzheimer’s Disease Therapeutics: The Path Forward. Alzheimer’s Research and Therapy, 1:
2, 2009. PMC2719107.
13. Petersen RC, Trojanowski JQ. Use of Alzheimer disease biomarkers: potentially yes for clinical trials but
not yet for clinical practice. JAMA, 302(4):436-7, 2009.
14. Petersen RC. Commentary on "A roadmap for the prevention of dementia II: Leon Thal Symposium 2008.
A national registry on aging. Alzheimer's & Dementia, 5(2):105-7, 2009.
15. Petersen RC. Early diagnosis of Alzheimer's disease: is MCI too late? Curr Alzheimer Res, 6(4):324-30,
2009.
16. Petersen RC, Jack CR Jr. Imaging and biomarkers in early Alzheimer's disease and mild cognitive
impairment. Clin Pharmacol Ther, 86(4):438-41, 2009.
17. Petersen RC, Aisen PS, Beckett LA, Donohue MJ, Gamst AC, Harvey DJ, Jack CR, Jagust WJ, Shaw LM,
Toga AW, Trojanowski JQ, and Weiner MW. Alzheimer’s Disease Neuroimaging Initiative (ADNI): Clinical
Characterization. Neurology 2009, in press.
18. Petersen RC, Knopman DS, Boeve BF, Geda YE, Ivnik RJ, Smith GE, Roberts RO, Jack CR Jr. Mild
Cognitive Impairment Ten Years Later. Arch Neurol, 2009, in press.
19. McEvoy LK, Edland SD, Holland D, Hagler, Jr. DJ, Roddey JC, Fennema-Notestine C, Salmon D, Koyama
AK, Aisen PS, Brewer JB, Dale AM, for the Alzheimer’s Disease Neuroimaging Initiative. Neuroimaging
Enrichment Strategy for Secondary Prevention Trials in Alzheimer’s Disease. Alzheimer’s Disease and
Associated Disorders, 2009, in press.
20. Vemuri P, Wiste HJ, Weigand SD, Shaw LM, Trojanowski JQ, Weiner M, Knopman DS, Petersen RC, Jack
Diagnostic discrimination and cognitive correlations. Neurology, 73:287-293, 2009.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
21. Vemuri P, Wiste HJ, Weigand SD, Shaw LM, Trojanowski JQ, Weiner M, Knopman DS, Petersen RC, Jack
Predicting future clinical change. Neurology, 73:294-301, 2009.
22. Hampel H, Shen Y, Walsh DM, Aisen P, Shaw LM, Zetterberg H, Trojanowski JQ, and Blennow K.
Biological markers of β-amyloid related mechanisms in Alzheimer’s disease. Exper. Neurol. In press, 2009.
23. Jagust WJ, Landau SM, Shaw LM, Trojanowski JQ, Koeppe RA, Reiman EM, Foster NL, Petersen RC,
Weiner MW, Price JC, Mathis CA, and the Alzheimer’s Disease Neuroimaging Initiative. Relationships
between biomarkers in aging and dementia. Neurology, In Press, 2009.
24. Ewers M, Walsh C, Trojanowski JQ, Shaw LM, Petersen RC, Jack CR, Jr, Bokde AWL, Feldman H,
Alexander G, Sheltens P, Vellas B, Dubois B, Hampel H, and the Alzheimer’s Disease Neuroimaging
Initiative (ADNI). Multi-modal biological marker based signature and diagnosis of early Alzheimer’s disease.
Submitted, 2009.
25. Jack CR, Knopman DS, Jagust WJ, Shaw LM, Aisen PS, Petersen RC, Trojanowski JQ. Modeling dynamic
biomarkers of the Alzheimer's pathological cascade. Lancet Neurology, Submitted, 2009.
2.4. Research Design and Methods
The Clinical Core will continue to be responsible for providing the operational infrastructure for this project,
including the recruitment of 150 additional subjects in each of the three original ADNI cohorts: normal, LMCI
and mild AD, and 100 additional subjects with EMCI, as well as the longitudinal follow-up of all remaining
ADNI1 normal and LMCI subjects, and GO EMCI subjects. As in ADNI1 and GO, this infrastructure will be
provided by the ADCS Administrative, Clinical Operations, Medical and Data Cores at UCSD. The ADCS
Cores occupy approximately 25,000 square feet of space across the street from the UCSD Medical School
Campus.
For ADNI2, the ADCS Administrative Core will:
a. Maintain subcontracts with all performance sites, monitoring work accomplished as indicated by data entry,
and providing quarterly payments
b. Providing regulatory oversight, including development of template consent forms for protocol and
amendments, reviewing and approving site consent forms, tracking overall site readiness (eg, regulatory
documents, IRB and radiation committee approvals, imaging certifications, rater certifications)
c. Organize investigator meetings
d. Work closely with the sites to reach their recruitment goals and effectively retain their subjects
ADNI has a multi-faceted recruitment plan in place; the overall goal being to raise awareness of ADNI trials
among targeted populations. The ADNI will partner with NIA and coordinate with its ADEAR Center to take
advantage of existing resources. The ADEAR center will also serve as the call center. In addition, a public
relations/advertising firm will be consulted for broader coverage, and for seeking celebrity spokespersons and
testimonies from other study participants or family members. The ADNI will determine the special requirements
of each site and pattern their individual public relations support around those needs. In that context the ADNI
will develop targeted messages in flyers, brochures, press releases, and presentations. Reference cards and
online access to recruitment materials for the sites will also be available. Paid advertisements, direct mail and
the Internet will be used as needed to supplement recruitment. A separate plan for minority recruitment is
being developed. Enrollment will be monitored and tracked and additional support provided where appropriate.
Additionally, the ADNI will provide background to sites on how to reach target audiences as well as assist in
identifying them. Technical assistance will be offered to the sites on an ongoing basis.
Several steps will be taken to assure the high follow up rate that is essential to the validity of the study
results. All staff members will be carefully instructed regarding the need for an expectation of full follow up
participation and the process of removing barriers to participation. At entry, each participant, and a significant
other informant will be queried regarding plans to change residence or leave the area. Frequent contact by
telephone will be maintained by participants at a minimum of six month intervals. Each participant will receive a
thank you note following the clinical evaluation and a personalized greeting card on his or her birthday or on a
major holiday. Progress of the study will be placed in a newsletter, distributed to the sites, for distribution to
subject participants.
The goal of ADNI2 is to obtain as close to 100% participation in lumbar puncture as is feasible. ADNI1
aimed for 25% participation and achieved more than 50% participation. For GO and ADNI2, we aim to limit
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
recruitment to those that consent to LP. Exceptions will be made to meet other goals such as minority
recruitment.
Efforts to maximize LP participation will include training of site personnel at a meeting scheduled for January
2010. At this meeting, the ADCS Recruitment Core assisted by the PIs and staff at the most successful LP
sites in ADNI1 will share recruitment, consent and procedural methods. Training videos for both the
recruitment/consent process and atraumatic LP techniques will be shared with all sites.
The Administrative and Clinical Operations Core maintain and review spreadsheets documenting the
performance of each ADNI site, including time to site readiness, enrollment rate, retention rate, lumbar
puncture rate, quality issues with imaging and data entry and minority enrollment. This data is shared with the
ADCS Recruitment Core (led by Mary Sano at Mount Sinai), which assists in site performance assessment,
and the evaluation of potential new sites.
At the launch of ADNI2, site performance will be reviewed. Balancing recruitment goals and site
performance quality, the enrollment of new ADNI2 subjects will be initiated at those sites with acceptable
records. Sites that are dropped from new enrollment as a result of quality concerns (all sites will maintain
follow-up of previously enrolled subjects) will be replaced by qualified applicant sites. A list of such applicant
sites is maintained by the Recruitment Core.
For ADNI2, the ADCS Clinical Operations Core will:
a. Manage the day-to day operations of ADNI
b. Provide progress reports to the Executive Committee
c. Manage supplies for all of the performance sites
d. Maintain the data system and provide Help-Desk support
e. Prepare and update protocol, procedures manual and other study documents
The Clinical Operations group is highly experienced, responsible for the conduct of all ADCS therapeutic
trials. The group includes seasoned project managers, each of whom has supervised multiple multicenter
trials, as well as informatics and computer specialists with many years of experience with the ADCS data
system. The group has provided Protocols and Procedures Manuals for all ADCS trials.
For ADNI2, the ADCS Medical Core will:
a. Provide on-site monitoring of all ADNI performance sites
b. Review and code all safety data, and facilitate interactions with the DSMB
The Medical Core includes two physicians and several support staff who have managed safety data analysis
and reporting for all ADCS trials; in addition, the Core includes 9 full-time clinical monitors split between the
East and West coasts. Core staff are responsible for the coding of all adverse events (using MedDRA) and
medications (with WHO Drug).The Medical Core directs the clinical monitoring procedures for all ADCS clinical
trials, as well as multi-center protocols to develop novel instruments for use in AD clinical trials. Clinical
monitors are responsible for ensuring that protocols are conducted properly at each center, and that center
staff are appropriately trained in protocol procedures. Additionally, the clinical monitors review
neuropsychological and behavioral ratings in detail, to verify that the instruments are being properly
administered and rated. Such oversight is conducted via remote auditing of case report forms, as well as onsite visits three times per year to review source documents and to meet with study personnel.
For ADNI2, the ADCS Data Core will:
a. Modify the ADNI electronic data capture system to accommodate the adjustments to the assessments for
existing cohorts, including real-time electronic data field checks and cross-form checks for quality control, and
provision of data to the Administrative Core for financial management
b. Provide real-time web-based reporting on data flow
c. Assure optimal data security and redundant data back-ups
d. Design and perform biostatistical analyses (in collaboration with the UC Davis biostatistical group) in
support of the Clinical Core Aims, particularly the trial design work
e. Perform tabulation and analysis of safety data for review by the Clinical Core leaders, Medical Core and
DSMB
f. Support the Clinical Operations and Administrative Core maintenance of the Data System
The computing environment currently deploys a Linux centric environment providing a multitude of services.
The core system is a complement of Postgresql and MySQL databases with logic and user interfaces drivened
by various modern technologies (Yahoo! User Interface Library and Google API). The web applications are
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
served up by highly available web servers deploying Intel Virtualization architecture. An assortment of file
system, backup, security and data management services round out the list of services.
The computing systems are linked to the UCSD WAN via fiber lines providing bandwidth exceeding 1 Gigabit
per second. Backups are performed through a combination of local hard disks and remote online storage
repositories. The CC LAN employs a firewall server between the UCSD feed and the rest of the systems.
Secondary firewalling services are enabled locally on the individual servers.
The Data Core staff provides support and training for computer users. In addition, computing support is
available through the UCSD Academic Computing and Telecommunication Department. This resource is
available on an as needed basis to provide specialized hardware, network and software consulting services.
Data management is conducted through a combination of Postgresql and MySQL relational data base
management systems. Each is powerful, reliable, well-supported, multi-user, high capacity, relational clientserver database system. Each has its own strengths. For example, Postgresql is a flexible security framework
allowing select, modify and delete privileges to be set at the table or even field level. The CC makes extensive
use of these security tools setting specific access rules for each individual staff member. This allows all CC
staff to access the database without compromising the security of the critical control tables or other sensitive
tables.
Application servers distribute the data framework and applications to the database servers. Both application
and database systems are hosted on Dell Poweredge Virtualization hardware. The application servers as a
whole are load balanced and high availability systems.
Extensive backup systems for the CC databases are in place. The database systems utilize a redundant
backup scheme involving local and remote repositories. The databases are backed up onto locally housed
hard disks daily. The local hard disk copies are then propagated redundantly to independent local hard disk
repositories. Additional database copies are electronically transferred to an offsite commercial file repository
over SSL encrypted channels. These off-site copies provide safeguards against theft, or loss, such as fire,
flood, or earthquake.
Audit trail: Changes to the database record after initial entry are automatically recorded in an audit trail. Each
audit trail record includes the name of the field being changed, the reason for the change, the old and new
values, change date, and the name of the person making the change. Only selected staff members are
permitted to make changes. The audit trail enables us to construct the complete change history for any record.
Quality assurance (QA): The QA program of the CC is extensive. It includes development of clear and
complete documentation of procedures and databases, cleaning of data and locking of completed databases
prior to analysis, clinical monitoring of data both in-house and at sites, as well as computerized data editing. All
procedures for performing QC checks are fully documented and updated as needed.
Data is cleaned by Quality Control (QC) on an on-going basis, during the data collection phase of the
protocol as well as after closure of the protocol. During each protocol, computerized data checks are used to
confirm that subjects meet inclusion and exclusion criteria at the time of entry, identify missing or out-of-range
items, identify missing forms from visit packets, identify any duplicate entries into the database, evaluate
longitudinal consistency between visits, and track subject status in the protocol (active vs. discontinued).
Requests for corrections are sent to sites and site monitor. The sites are asked to provide missing information
or to clarify contradictory responses, and to make the appropriate changes on the copies of the CRFs stored
on-site (or the electronic record if distributed entry is employed). Corrections received from sites are entered
into the database by QC staff, and an audit trail completed.
2.4.2. Clinical Assessment Aims
2.4.2.1. Methods: We propose to enroll subjects in four clinical groups representing a combination of subjects
followed from ADNI1, the GO grant and the newly recruited subjects as part of ADNI2. We will recruit subjects
in four clinical categories: cognitively normal (CN), early MCI (EMCI), late MCI (LMCI) equivalent to the aMCI
subjects in ADNI1 and mild AD subjects. For clarification, as discussed above, the EMCI and LMCI subjects
will meet criteria for aMCI of a degenerative etiology, but they will differ as to their degree of cognitive
impairment within the aMCI designation. Table 8 below demonstrates the four subject groups being proposed
and their relationships to ADNI1, the GO grant and ADNI2.
As specified in the first section of this Core, the aims include confirmation of the cross-sectional, longitudinal
and predictive findings of ADNI1, and extension of analyses to the EMCI cohort. Of critical importance, ADNI2
will extend the follow-up of original ADNI1 subjects for up to 10 years total, allowing the evaluation of
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
hypothetical trajectories of cognitive, clinical and biomarker assessments (Figure 1) and the relationship of
early changes to later clinical decline.
These aims will be addressed using the statistical approaches outlined in the Biostatistics Core. Specifically,
the statistical approach to group comparisons (i.e., across the CN, EMCI, LMCI and AD cohorts) is described
in Section 8.4.2.1. The approach to prediction of trajectories of change is described in Section 8.4.2.2. The
analysis of prediction of conversion to dementia is described in Section 8.4.2.3.
2.4.2.2. Recruitment: Subjects followed in ADNI2 are shown in Table 8. As mentioned, approximately 474
subjects (202 CN and 274 LMCI) will be carried forward from ADNI1. All of the participating sites have been
surveyed with respect to the subjects’ willingness to continue to participate in ADNI2, and the sites have
confirmed these numbers. We will also carry forward the 200 EMCI subjects from the GO grant, and we
anticipate a very low discontinuation rate based on ADNI1 continued participation rates of approximately 94%
per year (6% annual discontinuation rate).
ADNI1
202
EMCI
LMCI
AD
GO
200
274
ADNI2
150
100
150
150
Cumulative
352
350
424
150
Table 8. Number of subjects followed in ADNI2, including those
recruited in ADNI1 and GO.
2.4.2.3. New Subjects: New subjects will have similar recruitment procedures, cognitive and behavioral
instruments as performed in ADNI [4]; they are summarized in Table 9.
Visit name
Visit Type
Explain study
Obtain consent
Demographics, Family History, Inclusion and
Exclusion Criteria
Medical History, Physical Exam, Neurological
Exam, Hachinski
Vital Signs
Height
Screening Labs
ApoE Genotyping
Cell Immortalization Sample Collection
American National Adult Reading Test
Mini Mental State Examination
Logical Memory I and II
Everyday Cognition (ECog)
Montreal Cognitive Assessment (MoCA)
Category Fluency (Animals)
Trails A & B
Boston Naming Test (30-item)
Auditory Verbal Learning Test
Geriatric Depression Scale
Clock drawing
Neuropsychiatric Inventory Q
ADAS-Cog 11 (with Delayed Word Recall)
PHS 398/2590 (Rev. 11/07)
Screen
In-Clinic
Baseline
In-Clinic
Month 3
Imaging
Month 6
In-Clinic
Annual
In-Clinic
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
Biannual
Imaging/
LP
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
Page
X
Clinical Dementia Rating Scale
Activities of Daily Living (FAQ)
Collect and process biomarkers
Concomitant Medications
Subject payments
Adverse events
Diagnostic Summary
3T MRI Imaging (100%)
FDG-PET Imaging (100%)
F18-Amyloid PET Imaging (100%)
Lumbar Puncture (LP) (100%)
Weiner, Michael
X
X
X
X
X
X
X
X
X
X
X
X
X
X*
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
Table 9. Schedule of Events for ADNI2. *Month 3 MRI is timed from Screening MRI **Subjects will be followed for 2 years, after which
subjects will be asked to consent to annual follow-up under a separate grant and protocol
Enrolled subjects will be 55-90 years of age, have a reliable informant and will speak either English or Spanish.
Certain psychoactive medications will be excluded. Identical training procedures for the administration of the
CDR, AD Assessment Scale Cognitive Subscale, and the Neuropsychiatric Inventory-Q will be carried out as in
ADNI1.Inclusion / exclusion criteria will be the same as in ADNI1. Enrollment will occur at most of the 50-60
clinical sites currently used in ADNI. To reach the goal of 200 subjects with EMCI, approximately four subjects
will be enrolled per site during the first year of the GO grant and will be re-evaluated in year 2. Thus this
enrollment is feasible. The specific inclusion criteria for EMCI will be as follows: MMSE score of ≥ 24, CDR-SB
summary score 0.5 (memory box score ≥ 0.5), non-demented, in the judgment of the site clinician, delayed
recall of one paragraph from the Logical Memory subtest of the Wechsler Memory Scale-Revised with cutoff
scores (derived from the National Alzheimer's Coordinating Center (NACC) database)by education as
follows:≥16 years: 9-11; 8-15 years: 5-9; 0-7 years: 3-6. The scores represent recall performance
approximately 0.5 - 1.5 SD below education adjusted means for one paragraph of the Logical Memory subtest.
Please note that the LM cutoff scores for CN and EMCI subjects overlap. Since the EMCI subjects have only a
mild impairment in memory, they are differentiated from the CN subjects by having a CDR of 0.5. EMCI
subjects will have clinical and cognitive assessments and a 1.5 T-structural MRI at enrollment, 6 months and
12 months. In addition to the EMCI subjects, we will continue following ~ 476 subjects (202 normals and 274
MCI subjects) currently active in ADNI1 as part of year 6 of ADNI1 (at no cost to the GO grant in year In GO
we will continue to see ~ 506 of these subjects in year 2, (assuming 5% dropout). The 200 EMCI subjects will
undergo MRI scans in year 1 (at baseline and 6 months) and year 2 of GO, one F18 amyloid scan and FDG
PET at baseline, and FDG PET at 1 year. The 474 ADNI carry-forward subjects (including those who have had
C-11 PIB) will undergo one F 18 amyloid scan and the GO grant will also fund one FDG PET scan/ subject
(50% of ADNI1 subjects receive FDG PET funded by ADNI1 in Year 6 of ADNI which is Year 1 of GO). Table 8
above shows the distribution and budgetary allocation of subjects between ADNI1, ADNI GO and ADNI2. The
clinical evaluations for the EMCI and the normal and LMCI subjects followed in the GO project will include the
ADNI1 assessments, as well as brief instruments (such as the MoCA and the ECog) that may be useful in the
identification of very early stage AD in a general practice setting (see below). The investigators at each site will
be asked to adjudicate each subject according to the specified procedures in ADNI as to the clinical diagnoses
of normal, MCI (EMCI or LMCI) and dementia. All conversion decisions are reviewed by one of the Clinical
Core directors (RCP) and a central review committee.
The ADNI2 participating sites are listed in Table 10; as described above, some sites may be replaced as
recruitment sites (though all will continue to follow existing subjects) based on ongoing performance. We have
surveyed the sites for continued participation as indicated above, and those participating are noted. Since we
will be recruiting 550 new subjects (150 CN, 100 EMCI, 150 LMCI and 150 AD), the sites have indicated that
they will be able to recruit approximately 11 new subjects, 3 in each of the three clinical categories CN, LMCI,
AD and 2 in the EMCI noted above, in one year. The sites have extensive experience in recruiting these
subjects, and we do not anticipate a problem.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Site Name
Baylor College of Medicine
Name
Earl Zimmerman
Pierre Tariot
Rachelle Doody
Site Name
Premiere Research Institute
Rhode Island Hospital
Rush University Medical Center
Boston University
Ron Killiany
Brigham and Women's Hospital
Butler Hospital Memory and Aging Program
Case Western Reserve University
Cleveland Clinic Lou Ruvo Center for Brain
Health
Columbia University
Dent Neurologic Institute
Duke University Medical Center
Emory University
Georgetown University
Howard University
Dartmouth Hitchcock Medical Center
Indiana University
Johns Hopkins University
Mayo Clinic, Jacksonville
Mayo Clinic, Rochester
McGill University / Jewish General Hospital
Memory Clinic
Medical University of South Carolina
Mount Sinai School of Medicine
Neurological Care of CNY
New York University Medical Center
Northwestern University
Reisa Sperling
Stephen Salloway
Alan Lerner
Charles Bernick
St. Joseph's Health Center - Cognitive
Neurology
Stanford / PAIRE
Sun Health Research Institute
Sunnybrook Health Sciences Centre
University of Alabama, Birmingham
Jerome Yesavage
Marwan Sabbagh
Sandra Black
Daniel Marson
Yaakov Stern
Vernice Bates
Murali Doraiswamy
Allan Levey
Brigid Reynolds
Thomas Obisesan
Andrew Saykin
Martin Farlow
Marilyn Albert
Neill Graff-Radford
Ronald Petersen
Howard Chertkow
University of British Columbia
University of California, Davis
University of California, Irvine (BIC)
University of California, Los Angeles
University of California, San Francisco
University of Michigan, Ann Arbor
University of Pennsylvania
Robin Hsiung
Charles DeCarli
Ruth Mulnard
Steven Potkin
Liana Apostovola
James Brewer
Howard Rosen
Jeffrey Burns
Charles Smith
Judith Heidebrink
Steven Arnold
Oscar Lopez
Jacobo Mintzer
Hillel Grossman
Smita Kittur
Henry Rusinek
John (Chuang-Kuo)
Wu
Douglas Scharre
Godfrey Pearlson
Jeff Kaye
University of Rochester Medical Center
University of Southern California
University of Texas, Southwestern MC
Wake Forest University Health
Sciences
Washington University, St. Louis
Wien Center for Clinical Research
Yale University School of Medicine
M. Saleem Ismail
Lon Schneider
Kyle Womack
Sterling Johnson
Jeff Williamson
Olin Neuropsychiatry Research Center
Oregon Health and Science University
Parkwood Hospital
Table 10. ADNI2 Participating Sites.
Name
Carl Sadowsky
Brian Ott
Leyla deToledoMorrell
Andrew Kertesz
John Morris
Ranjan Duara
Christopher van
Dyck
Michael Borrie
2.4.2.4. Screening Visit: As is shown in Table 11, all new subjects will be screened for participation using the
MMSE, single paragraph delayed recall from Logical
CN
EMCI
LMCI
AD
Memory Subtest of the Wechsler Memory ScaleCDR
0
0.5
0.5
0.5-1
Revised, and the CDR. If the subjects meet criteria for
MMSE
24-30
24-30 24-30 20-26
inclusion in the various clinical groups at the screening
LM-DR (cutoffs)
visit, they will then be asked to return for a baseline visit.
Education: 0-7
≥3
3-6
≤2
≤2
The baseline evaluations are also outlined in Table 9,
8-15
≥5
5-9
≤4
≤4
above. At baseline, DNA for ApoE, blood for laboratory
≥16
≥9
9-11
≤8
≤8
studies and the first set of biomarkers will be obtained.
Dementia
No
No
No
Yes
All of the new subjects will undergo a lumbar puncture as
CN
EMCI
LMCI
AD
part of the baseline visits, and soon after the baseline
CDR
0
0.5
0.5
0.5-1
visit, all subjects will undergo a 1.5 T MRI scan. All
MMSE
24-30
24-30 24-30 20-26
subjects will also have an FDG-PET scan and an
LM-DR (cutoffs)
amyloid imaging scan.
Education: 0-7
≥3
3-6
≤2
≤2
8-15
≥16
Dementia
≥5
≥9
No
5-9
9-11
No
≤4
≤8
No
≤4
≤8
Yes
Table 11. Characteristics of four ADNI2 cohorts and Logical
Memory delayed recall cutoff scores (see text).
2.4.2.5. Follow-up Visits: Full follow-up visits will be conducted on an annual basis (plus a Month 6
assessment first year only). A complete battery of clinical and neuropsychological measures will be collected
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
at each time point, and imaging studies will be done (MRI, annually, FDG-PET and amyloid PET biannually).
Biomarkers will be collected at each clinic visit, and lumbar puncture will be repeated at two-year intervals.
The subjects will be re-evaluated clinically at each time point, and a diagnostic decision will be made as to
whether the subjects are cognitively normal, meet criteria for EMCI, LMCI or AD. The AD criteria will be
compatible with the NINDS/ADRDA criteria for AD. If a subject changes his or her clinical classification, the
criteria for the new diagnosis will be reviewed by one of the Clinical
Core directors (RCP), and a random sample subset of cases will be reviewed by a central review committee,
as is done in ADNI1.
2.4.2.6. Primary Care Physician Instruments: Since MCI is a sufficiently mild condition, it is most likely that
individuals with this degree of cognitive complaint will present initially to primary care physicians (PCP). As
such, there is a need to develop instruments that can be administered efficiently and inexpensively in the PCP
setting to allow the clinicians to determine which patients might be candidates for further evaluation and
possibly therapies. Toward this end, we have selected a brief cognitive instrument to allow detection of the
cognitive aspects of MCI and a functional instrument to determine the degree of functional impairment.
The primary screening tools for ADNI to determine if a person is eligible for the trial involve delayed recall of
one paragraph from the Logical Memory subtest of the Wechsler Memory Scale-Revised to provide a metric of
memory function to corroborate the individual’s cognitive complaint, and the CDR to be certain that the degree
of memory function represents a change from the previous level of performance. The combination of these
instruments is designed to determine that a person is cognitively and functionally impaired but not to the extent
that they would constitute criteria for dementia. However, using the Logical Memory paragraph and the CDR
would be too time consuming in the PCP setting, and consequently, brief instruments need to be evaluated to
see if they may be appropriate for the PCP.
The brief cognitive instrument that has been selected is the Montreal Cognitive Assessment test (MoCA)
which is designed to detect subjects at the MCI stage of cognitive dysfunction [19]. This instrument has been
shown to have adequate sensitivity and specificity in clinical settings to detect suspected MCI. The MoCA is
believed to be more sensitive than general screening instruments such as the MMSE or the Short Test of
Mental Status. The MoCA can be administered in approximately ten minutes.
For a functional assessment, we have selected the Measurement of Everyday Cognition (ECog) [20]. This
instrument has been developed to assess functional impairment of a very mild nature as can be seen in MCI.
The ECog is an informant-rated questionnaire comprised of multiple subscales and takes approximately ten
minutes to administer. Previous research on this instrument indicates that ECog correlates well with
established measures of functional status and global cognition but only weakly with age and education. ECog
was able to differentiate among cognitively normal, MCI and AD subjects. Results of ECog suggest that it is a
useful tool for the measurement of general and domain-specific everyday functions in the elderly.
The MoCA and ECog will be administered to all participants in ADNI2 but will not be used for any screening
or diagnostic decisions themselves. Rather, the traditional screening measures involving the single paragraph
from the Logical Memory subtest and the CDR will be used to determine the appropriate level of function for
subjects in ADNI2 just as it was in ADNI1. However, the performances of the MoCA and ECog will be followed
to determine their ability to differentiate among the four groups.
2.4.3. Clinical Trial Design Aims: ADNI1 has allowed us to propose novel trial designs in the pre-dementia
AD population; these ideas have been incorporated into one industry trial in progress, and they are likely to
influence many other industry and academic trials in the near future. We propose to continue and extend this
work utilizing the data collected in ADNI2.
2.4.3.1.
Statistical Methods: The statistical methods for the hypotheses of this section are described in the
Biostatistics Core (particularly Section 8.4.2.6.), and in the Preliminary Data section (2.3.3.1.) of this Core.
2.4.3.2.
Confirm and extend the trial design results from ADNI1: As described above, ADNI1 data
suggests the feasibility of longitudinal change designs in selected LMCI subjects selected for amyloid
accumulation using CSF Aβ42. In ADNI2 we will enroll 150 additional LMCI subjects, and all (or nearly all) will
have both CSF Aβ42 and F18 AV-45 amyloid imaging. We will thus have an independent sample of similar size
to that used for the ADNI1 power estimates. We will confirm the utility of CSF Aβ42 selection, the equivalent
value of AV-45 amyloid imaging, the utility of genotype and MRI volumetric (and other disease stage)
covariates, and confirm the group size estimates.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
We hypothesize that it will be feasible to extend these ideas to the milder EMCI population. That is, we
propose that selection of EMCI subjects by amyloid markers, and utilizing disease stage and APOE covariates,
with CDR-SB alone or with ADAS1 as continuous outcomes, we will have reasonable power to demonstrate
disease modifying effects. We expect that clinical effects of modifying disease mechanisms will be greater at
the earlier EMCI stage, offsetting the slower rates of decline. Thus we may have similar group sizes to power
EMCI studies to demonstrate a 40% effect as we do to power LMCI studies to see a 30% effect.
We also hypothesize that longitudinal change designs will be advantageous compared to survival to AD in
the EMCI population, as it is in LMCI.
2.4.3.3.
Extend the evaluation of imaging biomarkers as potential surrogate outcome measures:
ADNI data has demonstrated the effect size (ration of longitudinal change to variance) advantages of imaging
measures to cognitive and clinical measures in AD, LMCI and normals. We will confirm these findings in the
new ADNI2 subjects.
In ADNI1, we demonstrated that 6 month change in volumetric MRI measures had excellent properties for
potential surrogates, with strong effect sizes and correlation to cognitive and clinical measures. For example, a
6 month trial powered to demonstrate a 25% slowing of decline in hippocampal atrophy would require 353
subjects per group in AD, and 1235 in MCI (without biomarker selection or covariates); by contrast, 25%
slowing of decline on the ADAScog would require 1055 subjects in AD and 13074 subjects in MCI. Six month
change in hippocampal atrophy is correlated to cognitive and clinical change over the same period, confirming
its clinical relevance. In ADNI2, we will obtain a volumetric MRI 3 months after enrollment on all new subjects,
allowing evaluation of 3 month atrophy as an endpoint. If longitudinal change over variance is favorable in such
a short interval, this will present the possibility of conducting proof-of-concept trials of disease modifiers with as
short as a three month treatment period. Further, it will allow 3 month MRI change to be used for interim
analyses for futility or adaptive allocation.
We will explore 3, 6, 12 and 24 month change in MRI volumes (eg, entorhinal cortex, hippocampus, regional
cortical thickness, 12 and 24 month change in FDG-PET in subgroups of EMCI and normal controls. These
subgroups will be defined by one or more genotypic, imaging or cognitive measures.
Note: The validation of a surrogate outcome for pivotal trials requires demonstration, usually in multiple drug
development programs, that drug effects on a biomarker reflect eventual clinical benefits. Since ADNI does
not involve interventions, ADNI alone cannot establish a surrogate. But ADNI lays the groundwork, facilitating
the incorporation of candidate surrogates into therapeutic trials. Our expectation is that such biomarkers may
be validates in mild AD trials (though clinical benefits will probably be modest), and then the biomarkers can be
used as surrogates at early stages of disease (eg, pre-symptomatic AD) when cognitive and clinical outcome
measures have limited use, but when the ultimate impact of disease-modification may be large. The long-term
(up to 10 years) of longitudinal follow-up of multiple biomarker domains will be critical to the effort to link early
changes to later clinical decline.
2.4.3.4.
Extend clinical trial design evaluation to the EMCI population: The evaluation of the EMCI
cohort will include analysis of longitudinal trajectories of cognitive and clinical assessments, MRI volumetric
measures, FDG-PET as well as amyloid imaging and CSF markers. The impact of APOE genotype (and
potentially other genetic markers) on these trajectories will be assessed. Subgroups selected by amyloid
biomarkers, as well as cutoff values of hippocampal volume, FDG-PET activity and cognitive and clinical
assessments will likewise be examined. The value of biomarker covariates in reducing unexplained variance
of longitudinal change will be analyzed.
A major purpose of examining this new cohort in this manner will be to inform trial design. We hypothesize
that we will be able to extend similar design features that have yielded exciting findings in the LMCI cohort to
this more mildly impaired population. We aim to provide feasible trial designs for studies enrolling EMCI
subjects.
Our specific hypothesis is that we can select subjects using amyloid biomarkers (AV-45 imaging or CSF
Aβ42), perhaps also using APOE genotype selection, to define an EMCI subpopulation with longitudinal decline
on standard or supplemented measures that will allow proof of concept and pivotal testing of disease-modifying
agents. While amyloid biomarkers represent primary selection candidates, we will also evaluate other
biomarkers including CSF tau and P-tau (noting that the latter may be particularly appropriate for
neuroprotection and kinase inhibitor studies, respectively).
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
2.4.3.5.
Extend to the asymptomatic population: As indicated in Table 2 above, we expect that cognitive
and clinical measures will not be feasible outcomes in symptomatic individuals, even those selected using
biomarkers. While we expect to see some decline in standard measures, at this stage the efficiency of such
measures is unlikely to yield feasible trial sizes. In this population, we hypothesize, however, that potential
surrogate measure will allow proof of concept trial design. That is, as we increase our recruitment and longterm follow-up of asymptomatic subjects during ADNI2, we will explore the trajectories on imaging measures
(such as entorhinal cortex atrophy, regional cortical thickness measures, and FDG-PET measures) in subjects
selected on the basis of genotype and/or markers of amyloid or tau pathology.
An essential aim of ADNI2 is also to examine the relationship between longitudinal change and later
conversion to dementia in asymptomatic, EMCI and LMCI subjects. In addition to the comparison of
longitudinal change and survival to diagnosis trial designs, this will strengthen the link between (i.e., establish
the predictive value of) early change in cognitive, clinical and biomarker measures to later clinical progression.
While the experience of interventional studies will be essential, the ADNI2 analyses can support the validation
of potential surrogates by establishing predictive value.
2.5. References
[1] Knopman DS, DeKosky ST, Cummings JL, Chui H, Corey-Bloom J, Relkin N, Small GW, Miller B, Stevens
JC. Practice parameter: Diagnosis of dementia (an evidence-based review): Report of the Quality Standards
Subcommittee of the American Academy of Neurology. Neurology 56: 1143-1153, 2001.
[2] Petersen RC, Knopman D, Boeve B, Geda Y, Ivnik R, Smith G, Roberts R, Jack C. Mild Cognitive
Impairment: Ten Years Later. Neurology, in press, 2009.
[3] Gauthier S, Reisberg B, Zaudig M, Petersen RC, Ritchie K, Broich K, Belleville S, Brodaty H, Bennett D,
Chertkow H, Cummings J, deLeon M, Feldman H, Ganguli M, Hampel H, Scheltens P, Tierney MC,
Whitehouse P, Winblad B, Mild cognitive impairment. Lancet, 367:1262-1270, 2006
[4] Petersen RC, Aisen PS, Beckett LA, Donahue MJ, Gamst AC, Harvey DJ, Jack CR, Jr, Jagust WJ, Shaw
LM, Toga AW, Torjanowski JQ, Weiner MW for the Alzheimer's Disease Neuroimaging Initiative (ADNI):
Clinical Characterization. Neurology, in press, 2009.
[5] Farias S T, Mungas D, Reed BR, Harvey D, DeCarli C. Progression of mild cognitive impairment to
dementia in clinic- vs community-based cohorts. Arch Neurol, 66(9):1151-1157, 2009.
[6] Shaw, L. M. Vanderstichele, H. Knapik-Czajka, M. Clark, C. M. Aisen, P. S. Petersen, R. C. Blennow, K.
Soares, H., Simon, A. Lewczuk, P. Dean, R. Siemers, E. Potter, W. Lee, V. M. Trojanowski, J. Q.
Cerebrospinal fluid biomarker signature in Alzheimer's disease neuroimaging initiative subjects. Ann Neurol,
65(4):403-413, 2009.
[7] Rafii MS, Aisen PS. Recent Developments in Alzheimer’s Disease Therapeutics. BioMed Central Medicine,
BMC Med 7: 7, 2009.
[8] Aisen PS. Alzheimer’s Disease Therapeutics: The Path Forward. Alzheimer’s Research and Therapy, 1: 2,
2009.
[9] Dubois B, Feldman HH, Jacova C, Dekosky ST, Barberger-Gateau P, Cummings J, Delacourte A, Galasko
D, Gauthier S, Jicha G, Meguro K, O'brien J, Pasquier F, Robert P, Rossor M, Salloway S, Stern Y, Visser PJ,
Scheltens P. Research criteria for the diagnosis of Alzheimer's disease: revising the NINCDS-ADRDA criteria.
Lancet Neurol. 6(8):734-46, 2007.
[10] Donohue M, Aisen P, Gamst A, Weiner M. Using the Alzheimer’s disease neuroimaging initiative (ADNI)
data to improve power for clinical trials. International Conference on Alzheimer's Disease, Chicago, IL, 2008.
[11] Petersen RC, Thomas RG, Grundman M, Bennett D, Doody R, Ferris S, et al. Vitamin E and donepezil for
the treatment of mild cognitive impairment. The New England Journal of Medicine, 352(23), 2379-88, 2005.
[12] Thal LJ, Ferris SH, Kirby L, Block GA, Lines CR, Yuen E, et al. A randomized, double-blind, study of
rofecoxib in patients with mild cognitive impairment. Neuropsychopharmacology : Official Publication of the
American College of Neuropsychopharmacology, 30(6), 1204-15, 2005.
[13] Cox DR. Regression models and life-tables. Journal of the Royal Statistical Society.Series B
(Methodological), 34(2), 187-220, 1972.
[14] Laird NM & Ware JH. Random-effects models for longitudinal data. Biometrics, 38(4), 963-74, 1982.
[15] Pfeffer RI, Kurosaki TT, Harrah CH, Jr, Chance JM, Filos S. Measurement of functional activities in older
adults in the community. Journal of Gerontology, 37(3), 323-9, 1982.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
[16] Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuous outcomes. Biometrics, 42(1),
121-30, 1986.
[17] Liu G, Liang KY. Sample size calculations for studies with correlated observations. Biometrics, 53(3), 93747, 1997.
[18] Harrell FE, Jr. Hmisc: Harrell miscellaneous. R package version 3.6-0, 2009.
[19] Nasreddine ZS, Phillips NA, Bedirian V, Charbonneau S, Whitehead V, Collin I, Cummings JL, Chertkow
H, The Montreal Cognitive Assessment (MoCA): a brief screening tool for mild cognitive impairment. JAGS.
53:695-699, 2005.
[20] Farias ST, Mungas D, Reed BR, Cahn-Weiner D, Jagust W, Baynes K, Decarli C. The measurement of
everyday cognition (ECog): scale development and psychometric properties. Neuropsychology. 22(4):531-544,
2008.
2.6. Abbreviations:
AD: Alzheimer's disease
ADAScog 11,12, 13: Alzheimer's Disease Assessment Scale- cognitive subscale versions with 11, 12 or 13
items (ADAScog11 is the standard cognitive outcome in AD trials)
ADCS: Alzheimer’s Disease Cooperative Study (NIA-funded clinical trials consortium)
ADNI: Alzheimer’s Disease Neuroimaging Initiative
aMCI: amnestic mild cognitive impairment
ApoE: apolipoprotein E
CDR: Clinical Dementia Rating
CDR-SB: Clinical Dementia Rating sum of boxes
CN: cognitively normal
CSF: cerebrospinal fluid
CSF Aβ42: level of amyloid beta protein of 42 amino acids in the CSF
CSF tau, p-tau: level of tau protein or phosphorylated tau in the CSF
ECog: Everyday Cognition instrument
EMCI: early amnestic mild cognitive impairment
FAQ: Functional Assessment Questionnaire
FDG-PET: fluorodeoxyglucose positron emission tomography
GO: Grand Opportunity (NIH funding mechanism)
LMCI: late amnestic mild cognitive impairment
LMM: linear mixed model
LP: lumbar puncture
MoCA: Montreal Cognitive Assessment battery
NPI: Neuropsychiatric Inventory
SD: standard deviation
2.7. Human Subjects:
2.7.1. Conduct of Study: Ethical and Regulatory Considerations: This study will be conducted according to
Good Clinical Practice guidelines, US 21CFR Part 50 – Protection of Human Subjects, and Part 56 –
Institutional Review Boards (IRBs) / Research Ethics Boards (REBs), and pursuant to state and federal HIPAA
regulations. Written informed consent for the study must be obtained from all participants and/or authorized
representatives and the study partners before protocol-specific procedures are carried out.
2.7.2. Institutional Review Board / Research Ethics Boards: Institutional Review Boards and Research
Ethics Boards must be constituted and their authority delegated through the institution's normal process of
governance according to applicable State and Federal requirements for each participating location. The
protocol will be submitted to appropriate Boards and their written unconditional approval obtained and
submitted to Regulatory Affairs at the Coordinating Center (CC) prior to commencement of the study. The CC
will supply relevant data for investigators to submit to their IRBs/REBs for protocol review and approval.
Verification of IRB/REB unconditional approval of the protocol and the written informed consent statement with
written information to be given to the participants and/or their authorized representatives and study partners
and will be transmitted and validated by the CC in order to obtain approval for shipment of study supplies and
CRF’s to study sites. Sites’ approval must refer to the study by exact protocol title and number, identify
documents reviewed, and state the date of review. IRBs/REBs must be informed by investigators of all
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
subsequent protocol amendments and of serious or unexpected adverse experiences occurring during the
study that are likely to affect the safety of the participants or the conduct of the study. IRB approval for such
changes must be transmitted in writing to the ADCS.
2.7.3. Informed Consent and HIPAA Compliance: Informed consent will be obtained in accordance with US
21 CFR 50.25, the Tri-Council Policy Statement: Ethical Conduct of Research Involving Humans and the
Health Canada, ICH Good Clinical Practice Practices and applicable HIPAA privacy notifications will be
implemented before protocol procedures are carried out. Information should be given in both oral and written
form as deemed appropriate by the Site’s IRB.
Participants, their relatives, guardians or authorized representatives and study partners must be given ample
opportunity to inquire about details of the study. The consent form generated by the investigator with the
assistance of the CC must be approved, along with the protocol, and HIPAA privacy notifications by the
IRB/REB and be acceptable to the CC. Consent forms must be in a language fully comprehensible to the
prospective participants and/or their authorized representatives and study partners. Informed consent will be
documented by the use of a written consent form approved by the IRB/REB and signed by the participant
and/or an authorized representative and study partner. Consent must be documented by the dated signature of
the participant and/or authorized representative pursuant to local regulations. Each participant’s signed
informed consent and/or HIPAA research authorization must be kept on file by the investigator for possible
review by regulatory authorities and/or ADCS monitors. HIPAA privacy requirements will be met by either
inclusion of required HIPAA text within the IRB-approved consent document or by separate HIPAA research
authorization, pursuant to local regulations.
2.7.4. Informed consent for biomarkers, genetic material, and imaging data: The informed consent will not
only cover consent for the trial itself, but for the genetic research, biomarker studies, biological sample storage
and imaging scans as well. The consent for storage will include consent to access stored data, biological
samples, and imaging data for secondary analyses. Consent forms will specify that DNA and biomarker
samples are for research purposes only; the tests on the DNA and biomarker samples are not diagnostic in
nature and participants will never receive results. MRI scan findings of clinical significance, determined by the
site radiologist, will be shared with participants The informed consent and/or HIPAA notification will specify that
the ADCS will receive and store all research data; that Mayo Clinic Rochester will receive MRI images, the
University of Michigan will receive and store PET images, the University of Pennsylvania Alzheimer’s Disease
Biomarker Fluid Bank Laboratory will receive and store biomarker samples while the University of Los Angeles
Laboratory of Neuroimaging will house a full set of all the data. All data will be made available to: the
pharmaceutical industry, academic investigators and other interested parties in the public domain. A policy for
distribution of data will be developed.
2.7.5. Procedures to maintain confidentiality of genetic material and biomarkers:
2.7.5.1. Genetic research and storage of genetic material: The de-linking of the sample from the participant
occurs at the time the blood is sent to the University of Pennsylvania. All samples will be inventoried and
tracked using commercially available software. A database will be created and used for the inventory of stored
samples in conjunction with a bar code reading system. Bar code labels affixed to each sample vial will contain
the following information: sample ID# (to preserve confidentiality), date of collection and processing, total initial
volume collected, sample type (urine, plasma, serum, CSF), volume, aliquot number, freezer, shelf, rack, box,
location in the box. A bar code label will be used on the sample tracking form. Immortalized cell lines and DNA
will be prepared at U Pennsylvania. However, neither the CC nor the University of Pennsylvania will have
information regarding the participant’s name and thus are unable to link the DNA analysis results to the person.
Also, since the results are not ever transmitted to the site that enrolled and followed the participant, the site will
be equally unable to link the results to the participant. To gain the maximum utility for research on genetic
material and biological markers, the CC will be able to analyze clinical research data collected on each
participant in relation to biological specimens from that participant. However, there will be no link to research
done on these specimens with participants’ names. It is important to note that the linkage is between DNA
research data and study research data, and the linkage will take place only one of our data centers. The data
centers (UCSD, UCLA) do not have any record of the names of the study participants, or of specific medical
identifiers such as clinical medical record numbers. The participating sites do not receive APO-E results or any
DNA results, and do not have access to the database in which these results are stored. Therefore, even
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
though DNA results can be linked to clinical research data for purposes of analyses, there is no way to achieve
linkage of DNA test results to names of participants.
The procedures for patient confidentiality will be approved by the IRB of the University of Pennsylvania. The
protection of patient confidentiality and the use of stored DNA specimens will be in accordance with the rules
and procedures established by the U of Pennsylvania IRB. The DNA is banked in a locked freezer at U Penn
dedicated to the ADNI. The samples are without a link to identity of the participant from whom the sample
came. All samples are bar coded and identified by a bar code.
Specific procedures for requesting and accessing DNA will be created by the Resource Allocation Review
Committee (RARC) of the ADNI in accordance with recommendations proposed in the NBAC Human
Biological Materials Report. These DNA guidelines have also been developed in accordance with the American
Society for Human Genetics’ position paper on the NBAC report and the Ad Hoc Committee on Stored Tissue
of the College of American Pathologists.
2.7.5.2. Biomarker research and sample storage: Blood samples will be labeled by bar coding samples.
Participant’s names will not be provided to the University of Pennsylvania. Samples will be stored by bar code
number and no other identifying information will be provided.
2.7.6. MRI and PET imaging and data storage: MRI scans will be labeled according to each site’s imaging
machine capabilities. In most cases, this will only include research site number, participant ADCS code, and
date of scan. All efforts will be exerted to have scans sent only with this information. In cases where the
machine is not capable of editing information sent, the informed consent at those sites will specify what
additional information will be transmitted. MRI scan findings of clinical significance, determined by the site
radiologist, will be shared with participants
2.7.7. Data and Safety Monitoring Board: The CC currently has an extremely active Data and Safety
Monitoring Board (DSMB) that reviews the safety of all patients enrolled in trials on an ongoing basis. Even
though no drugs are involved, there is a potential for adverse events related to participation in this study. Thus,
our DSMB will review safety data collected on a quarterly basis including adverse events and laboratry
surveillance. After reviewing emerging safety data, the DSMB can make recommendations regarding the
conduct/continuation of this trial.
2.7.8. Inclusion of Women and Minorities: Women and members of minority groups will be actively recruited
during this protocol. Based on the participating sites data regarding enrollment of minorities, we expect 12% of
subjects enrolled will be minorities. This is close to the aged minority population in the U.S. which is 14%. A
more detailed projection is provided in the following Targeted/Planned Enrollment Table.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
2.8. ENROLLMENT TABLE
TARGETED/PLANNED ENROLLMENT: Number of Subjects
Ethnic Category
Sex/Gender
Females
Males
Total
Hispanic or Latino
13
9
22
Not Hispanic or Latino
306
222
528
Ethnic Category: Total of All Subjects *
319
231
550
American Indian/Alaska Native
1
1
2
Asian
6
5
11
Native Hawaiian or Other Pacific Islander
0
0
0
Black or African American
28
19
47
White
285
207
492
Racial Categories: Total of All Subjects *
319
231
550
Racial Categories
* The “Ethnic Category: Total of All Subjects” must be equal to the “Racial Categories: Total of All Subjects.”
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Core: 3
Title of Core (not to exceed 81 spaces): MRI Core
Core Leader: Jack Jr., Clifford, R.
Position/Title: Professor, Mayo Clinic
Mailing Address:
200 First Street SW
Rochester, MN 55905
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
The broad objectives of ADNI 2 fall into two general categories – improving methods for clinical trials and
thoroughly evaluating biomarkers in AD and its preclinical stages. Our approach to the ADNI 2 MRI protocol
will be to maintain MRI methodological consistency in previously enrolled ADNI 1 subjects. We will also
modernize and expand the MRI protocol for newly enrolled subjects in order to remain technically current in
MRI. These principles point to a multi track MRI acquisition approach with the following features; 1) ADNI 1
subjects: continue to follow existing ADNI 1 subjects with serial MRI studies on the same 1.5T scanner on
which they have been scanned, using the ADNI 1 1.5T protocol. 2) ADNI 2 core protocol: scan newly
enrolled ADNI 2 subjects at 3T with a core set of three sequence types - 3D T1 volume, FLAIR, and a long
TE gradient echo volumetric acquisition (GRE) for micro hemorrhage detection. 3) ADNI 2 experimental sub
studies: in addition to the core ADNI 2 protocol described above, we will perform pilot sub-studies of arterial
spin labeling (ASL) perfusion, resting state functional connectivity (RSFC) and diffusion tensor imaging (DTI).
One of these sequences may be added to the core protocol on each of the systems belonging to a single
MRI vendor. The MRI core of ADNI has two components; the central lab at the Mayo Clinic and, the five
funded image analysis co investigators. In addition, the MRI core will form small consulting advisory
committees in 3 areas – DTI, ASL, and RSFC analysis.
MRI core methods fall into two categories: (1) Service aims of the central MRI core lab at Mayo Clinic
needed to generate high quality MRI data in all subjects at each time point. (2) The 5 funded ADNI MRI core
analysis labs will generate numeric summary MRI data. All data will be made available to the general
scientific community.
Hypotheses testing Aims of the MRI core fall into three categories: 1) replication of important results from
ADNI 1, 2) testing new hypotheses related to the three core MRI protocol sequences, and 3) testing new
hypotheses related to experimental sequences in protocol.
Based on results from ADNI 1, we can conclude that MRI is one of the most effective ways to measures
longitudinal progression of disease. MRI is therefore an essential component of a comprehensive study of
biomarkers of AD at various clinical disease stages.
Mayo Clinic Rochester d/b/a Mayo Clinic College of Medicine
DUNS:
006471700
Street 1:
200 First Street SW
City:
Street 2:
Rochester
County:
Province:
Olmsted
USA
MN-001
Country:
State:
Zip/Postal Code:
MN
55905
Northern California Institute for Research and Education
DUNS:
613338789
Street 1:
4150 Clement Street
City:
Street 2:
San Francisco
Province:
PHS 398 (Rev. 11/07)
County:
Country:
VAMC Building 13
San Francisco
USA
State:
Zip/Postal Code:
CA
94121
CA-053
Page 2
Form Page 2
Weiner, Michael
Name
Organization
Role on Project
Jack, Clifford
Jack12
Mayo Clinic
PI
Schuff, Norbert
Nschuff
NCIRE
Co-Investigator
University College
Co-Investigator
UC-Davis
Co-Investigator
UCLA
Co-Investigator
Fox, Nick
DeCarli, Charles
cdecarli
Thompson, Paul
Name
Organization
Role on Project
No
Yes
Cell Line
PHS 398 (Rev. 11/07)
Page 3
Weiner, Michael W.
Dementia Research Centre, UCL Institute of Neurology
DUNS:
Street 1:
8-11 Charles Symonds House
City:
Street 2:
London
Province:
County:
Country:
State:
United Kingdom
Zip/Postal Code:
WC1N 3BG
REGENTS OF THE UNIVERSITY OF CALIFORNIA‐DAVIS
DUNS:
04‐712‐0084
Street 1:
DEPARTMENT OF NEUROLOGY
City:
Street 2:
Sacramento
Province:
County:
Country:
4860 Y STREET, SUITE 3700
Sacramento
USA
State:
Zip/Postal Code:
CA
95817
CA‐001
Regents of the University of California ‐ Los Angeles
DUNS:
09‐253‐0369
Street 1:
Laboratory of Neuro Imaging
City:
Street 2:
Los Angeles
Province:
County:
Country:
635 Charles E Young Dr South Ste 225
Los Angeles
USA
State:
Zip/Postal Code:
CA
90095
CA‐030
DUNS:
Street 1:
City:
Street 2:
Province:
County:
Country:
State:
Zip/Postal Code:
State:
Zip/Postal Code:
DUNS:
Street 1:
City:
Street 2:
Province:
PHS 398/2590 (Rev. 11/07)
County:
Country:
Page 1
Weiner, Michael
3. MRI CORE:
3.1. Specific Aims:
The primary objective of ADNI 1 was to improve methods for clinical trials. This is widely regarded to have
been a successful effort and therefore the objectives of ADNI 2 will include steps to further evaluate
methodological improvements in applications of MRI for clinical trials. In addition however, the data generated
by ADNI 1 has had a substantial impact on advancing understanding the nature of the AD disease process.
The broad objectives of ADNI 2 therefore fall into two general categories – improving methods for clinical trials
and thoroughly evaluating biomarkers in AD and its preclinical stages. These two broad thematic objectives
point to a series of specific aims that can be divided into three categories: 1) Generation of high quality multicenter MRI data, 2) Replication of important ADNI 1 results, and 3) Testing new hypotheses. Note that we
designate ADNI 1 as the current ADNI grant for which we are seeking renewal, and ADNI 2 as the competitive
renewal, should it be funded.
ADNI 2 will follow 3 cohorts of subjects. 1) Cognitively Normal (CN) and late MCI (LMCI) subjects carried
forward from ADNI 1, 2) EMCI (early MCI) enrolled in GO and carried forward into ADNI 2, 3)CN, LMCI, and
Alzheimer’s Disease (AD) subjects newly enrolled in ADNI 2. Our approach to the ADNI 2 MRI protocol will be
to maintain MRI methodological consistency to the greatest extent possible in previously enrolled ADNI 1
subjects, to maximize the value of the longitudinal MRI data in these subjects. We also believe however that
the ADNI MRI protocol should evolve to reflect recent technical progress in MRI. Therefore we will modernize
and expand the MRI protocol in order to remain technically current in MRI, within the constraints imposed by
having to operate in a reasonably consistent manner across an estimated 110 individual scanners. Finally,
given the tight linkage between GO and ADNI 2, we intend to employ the same MRI methods for subjects
newly enrolled in GO and those newly enrolled in ADNI 2. These principles point to a multi-track MRI
acquisition approach with the following features.
1. ADNI 1 subjects: Continue to follow existing ADNI 1 subjects with serial MRI studies on the same 1.5T
scanner on which they have been scanned, using the ADNI 1 1.5T protocol.
2. ADNI 2 core protocol: Scan newly enrolled GO and ADNI 2 subjects at 3T with a core set of three types of
sequences - 3D T1 volume, FLAIR, and a long TE (e.g., TE = 20 ms) gradient echo volumetric acquisition
(referred to from here forward simply as GRE) for micro hemorrhage detection. Each MRI exam will contain
both an accelerated and a non-accelerated 3D T1 acquisition.
3. ADNI 2 experimental sub-studies: In addition to the core ADNI 2 core protocol described above, we will
perform pilot sub-studies of arterial spin labeling (ASL) perfusion, resting state functional connectivity
(RSFC), and diffusion tensor imaging (DTI). One of these sequences may be added to the core protocol on
each of the systems belonging to a single MRI vendor (Siemens, GE, or Philips). The purpose of these
experimental sub-studies is to demonstrate the feasibility of acquiring useful data in a multi-center (but
single vendor) setting. We plan to assess the feasibility of this approach early in ADNI 2 and perform those
sub-studies that we determine can be reasonably powered based on enrollment and the specific system
configurations present across the ADNI 2 enrollment sites.
The multi-track protocol approach optimally addresses two desirable goals; 1) to maintain longitudinal
technical consistency in currently enrolled ADNI 1 subjects carried forward into GO and ADNI 2), 2) to
modernize the protocol in newly enrolled subjects (for whom consistency with past scans is not an issue).
Functions of the MRI core can be divided into two categories: 1) service functions carried out by the central
MRI core lab at the Mayo Clinic, 2) analyses carried out by each of the five funded MRI analysis labs.
3.1.1. Service Aims of the Central MRI Core Lab at the Mayo Clinic
1. Obtain high quality multi-center data that is consistent over time, and to the greatest extent possible
consistent across the different systems in the study, while minimizing the burden (i.e., scan time) on
participating subjects across multiple scanning sessions.
2. Create and deliver vendor and platform specific MRI protocols electronically to all scanners in the study at
baseline and again each time a system upgrade is performed at the site.
3. Qualify (and re-qualify after upgrades) each scanner on the MRI protocol.
4. Verify the MRI protocol is properly executed by checking the DICOM image headers and perform
appropriate image quality control throughout the study.
5. Identify appropriate quality control methods for non-anatomic imaging sequences, e.g. DTI, ASL, RSFC.
6. Correct specific classes of image artifacts in each 3D T1 image acquired; intensity non-uniformity, image
warping due to gradient non-linearity, and scaling changes over time.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
7. Monitor each scanner longitudinally in the study using the ADNI phantom.
3.1.2. Data Production Aims of MRI Analysis Labs: Perform quantitative measurements of all images
acquired using state-of-the-art image analysis methods. Methods will include both numeric summary data and
voxel-wise analyses.
3.1.3. Hypotheses Testing Aims of MRI Core:
3.1.3.1. Hypothesis Replication of ADNI 1:
1. Some structural MRI analysis methods have greater power to detect change over time than others which is
reflected in estimates of smaller sample sizes needed to detect a theoretical treatment effect.
2. Structural MRI has better (or at least equivalent) longitudinal measurement properties than FDG PET, CSF
biomarkers, and clinical indices, resulting in generally smaller (or at least equivalent) sample sizes for
clinical trials in both MCI and AD patients.
3. Longitudinal change on structural MRI generally has better (or at least equivalent) correlation with change
in cognitive measures compared with change in FDG PET or CSF biomarkers.
4. Structural MRI has better prediction of risk of conversion from MCI to AD than CSF biomarkers.
5. White matter hyperintensity (WMH) load differs by clinical group and is ordered AD > MCI > CN.
6. Greater WMH load at baseline is associated with faster rates of cognitive decline within each clinical group.
3.1.3.2. New Hypotheses Related to Core MRI Protocol Sequences
1. Structural MRI atrophy rates will accelerate as subjects progress from CN to EMCI, from EMCI to LMCI,
and from LMCI to AD.
2. Micro hemorrhages will be more prevalent at baseline in subjects with CSF or amyloid imaging evidence of
brain amyloid deposition than in subjects without evidence of brain amyloid.
3. The incidence of new micro hemorrhages will be greater in subjects with CSF or amyloid imaging evidence
of brain amyloid deposition than in subjects without evidence of brain amyloid.
4. WMH load differs by clinical group and is ordered AD > EMCI > MCI > CN.
5. Due to reduced subject motion, compared to non-accelerated acquisitions, accelerated 3D T1 scans on
average will have greater power to detect change over time, reflected in estimates of smaller sample sizes
needed to detect a theoretical treatment effect.
6. Due to reduced subject motion, compared to non-accelerated acquisitions, accelerated 3D T1 scans will
have better correlation with change in cognitive measures.
3.1.3.3. New Hypotheses Related to Experimental Sequences in Protocol
1. Cross-sectional measures of ASL, DTI, and RSFC will correlate with clinical indices (e.g., CDR or clinical
diagnostic groups) of disease severity cross-sectionally and will do so at least as well as existing MR, FDG
PET, and CSF measures.
2. Cross-sectional measures of ASL, DTI, and RSFC will correlate with other imaging and CSF biomarkers.
3. Serial measures of ASL, DTI, and RSFC will correlate with change in clinical indices of disease severity
over time and will do so at least as well as existing imaging and CSF measures.
4. Serial measures of ASL, DTI, and RSFC will correlate with change in other imaging and CSF biomarkers.
5. The experimental sequences (ASL, DTI, and RSFC) add predictive power for conversion from normal to
EMCI, EMCI to LMCI, and LMCI to AD over and above the predictive power from other imaging and CSF
biomarkers alone.
3.2.1. Use of Manufacturer-available Pulse Sequences: Much of the background section is devoted to
describing the rationale for the selection of MRI acquisition methods for ADNI 2. A manufacturer-available
pulse sequence (sometimes called a “product” pulse sequence) is one that is provided as part of the stock
software load on a commercial scanner, or alternatively is offered as a purchasable option on commercial
scanners. In distinction, a “work-in-progress” (WIPS) or “research pulse” sequence is not routinely available
from the vendor. A WIPS pulse sequence may have been created by the vendor for pre-production beta testing
purposes, or it may be a sequence created by an academic MR physicist for a special application.
Manufacturer-available pulse sequences do not require a formal research agreement to be in place between
the vendor and the MRI site. In contrast, WIPS sequences do require that a formal research agreement exists
between the vendor and MRI site. WIPS pulse sequences also require special attention (e.g., conversion,
recompilation, and redistribution) each time the software revision of the MRI system is upgraded. One of the
most important lessons learned in ADNI 1 was the difficulties with using WIPS sequences in a large multicenter study. Consequently, ADNI 2 will be limited exclusively to manufacturer-available pulse sequences.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
ADNI 1 included an extensive pre-study testing and evaluation phase (i.e., preparatory phase) of the primary
morphometric sequence – 3D T1. In May 2005, at the ISMRM meeting in Miami, we held an in-person meeting
of the MRI core along with our external advisors (G Glover, J Gore, J Mugler). Data generated during the ADNI
1 preparatory phase were reviewed and various options for the ADNI 1 MRI protocol were evaluated. The
assembled group selected a sagittal MPRAGE as the primary morphometric sequence. In ADNI 1 we placed
highest priority on cross-platform uniformity. One of the three major MRI vendors (GE) did not offer MPRAGE
as a manufacturer-distributed pulse sequence. Moreover, the analogous GE manufacturer-available pulse
sequence at the time (IR-FSPGR) was unable to acquire more than 128 slices on older (i.e., 9.x) GE scanners.
Even on the newest GE scanners in 2005, IR-FSPGR suffered from several severe image artifacts, including
un-modifiable cephalad chemical shift direction for acquisitions in the preferred sagittal plane, which caused
the fat signal to overlap the signal from the base of the brain. For all of these reasons, a WIPS 3D T1
sequence was needed for GE systems in ADNI 1. The ADNI MRI core therefore worked with NIA, GE, and the
MPRAGE patent holder (Univ. of Virginia) to come to an agreement allowing a customized version of MPRAGE
to be distributed to GE ADNI sites. Matt Bernstein, Ph.D. created this customized MPRAGE sequence for all
GE systems in ADNI 1, which addressed all the problems delineated earlier. While distribution of this WIPS
sequence was very successful in terms of image quality, consistency and technical performance, it also came
with undesirable consequences. These included:
1. Administrative overhead: A research agreement with GE was required with each site in order to run the
scanner in research mode. Wherever this was not in place prior to ADNI, such an agreement had to be
newly established. In addition, a unique letter agreement was required for each site in order to receive the
ADNI MRI protocol on CD. Changes in management at GE just prior to start up of ADNI 1 delayed
execution of this agreement, delaying the entire ADNI study.
2. General use for clinical trials: A letter agreement with each site, a research agreement with GE, and
installation of custom software on each scanner was necessary in order to run MPRAGE on a GE system.
Also, notably, approval from UVA to use MPRAGE is required for each individual research study. This
precluded many non-ADNI studies (especially industry partners supporting ADNI) from using the ADNI
MPRAGE sequence on GE scanners for their own clinical trials. This undermined a major objective of
ADNI 1, which was to create standardized methods that could be used generally in clinical trials.
3. Legal: A patent infringement lawsuit was filed in late 2008 by UVA against GE. This ongoing legal
proceeding makes seeking a use agreement between GE and UVA for new sites in ADNI 2 problematic.
For all the above reasons, we will use only manufacturer-available sequences in ADNI 2 (with the
possible exception of the ADNI 1 cohort, provided UVA grants permission to continue using MPRAGE in those
subjects for whom this agreement has been in place). For the foreseeable future, Siemens and Philips will
continue to offer MPRAGE as the inversion prepared 3D T1 manufacturer-available pulse sequence, while the
analogous manufacturer-available sequence for GE systems is IR-FSPGR (or variants, such as BRAVO).
Therefore, going forward in ADNI 2, 3D T1-weighted imaging will be performed with MPRAGE on Siemens and
Philips systems and with IR-FSPGR on GE systems. In addition to the 3D T1 sequence, the principle of using
only manufacturer-available sequences impacts other sequences considered for the ADNI 2 MRI protocol as
described later.
3.2.2. Rationale for a “Multi-track” Protocol for ADNI 2: ADNI 2 will follow 3 cohorts of subjects. 1) CN and
LMCIs carried forward from ADNI 1, 2) EMCI enrolled in GO and carried forward into ADNI 2, 3) CN, MCI,
LMCI, and AD subjects newly enrolled in ADNI 2. Our approach to the ADNI 2 MRI protocol will be, to the
greatest extent possible, maintain MRI methodological consistency in all previously enrolled subjects in order
to maximize the value of the longitudinal data in these subjects. We also believe, however, that the ADNI MRI
protocol should evolve to reflect recent technical progress in MRI. Therefore we will modernize and expand the
MRI protocol, to remain scientifically current in MRI. Finally, given the tight linkage between GO and ADNI 2, it
is logical to employ identical MRI methods for subjects newly enrolled in GO as those newly enrolled for ADNI
2. These principles imply a multi-track MRI acquisition approach, which we intend to pursue with the following
features.
1. ADNI 1 subjects: Continue to follow existing ADNI 1 subjects with serial MRI studies on the same 1.5T
scanner on which they have been scanned, using the ADNI 1 1.5T protocol which consists of: Two back-toback MPRAGE scans and a proton density/T2 dual fast spin echo sequence. (This assumes that UVA will
grant approval to continue use of MPRAGE on the ADNI 1 cohort.).
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
2. ADNI 2 core protocol: Scan newly enrolled GO and ADNI 2 subjects at 3T with a core set of three
sequences - 3D T1 volume, FLAIR, and GRE to detect micro hemorrhages. The 3D, T1-weighted volume
scan will run twice: first without acceleration and next with a parallel imaging acceleration factor of 2. This
procedure will enable the testing of new hypotheses 5) and 6).
3. ADNI 2 sub-studies: In addition to the core ADNI 2 core protocol we will perform "pilot sub-studies": ASL,
RSFC and DTI. One of these sequences will be added to the core protocol on each of the systems
belonging to a single MRI vendor (Siemens, GE, and Philips). The purpose of these sub-studies is to
demonstrate the feasibility of acquiring useful data in a multi-center (but single vendor) setting.
3.2.3. Rationale for Selection of Field Strength for the ADNI 2 Protocol: In considering field strength (1.5T
vs. 3T) for ADNI 2, data from the funded ADNI 1 image analysis labs indicated that effect sizes to detect rate of
change in AD and MCI subjects were not consistently different between 3T vs. 1.5T, nor was the magnitude of
the correlations between change in MRI and change in general cognition [1]. Consequently we determined that
scanning newly enrolled subjects in the ADNI 2 at 3T was advisable. Moving to 3T offers several advantages
for the core protocol sequences including: (1) enabling us to accelerate the FLAIR acquisitions on systems
(where this is a manufacturer-available feature), and do so without a significant loss in SNR due to the higher
field strength; (2) use of 3T will increase the sensitivity of GRE to micro bleeds compared to 1.5T; and (3) it will
enable us to accelerate the 3D T1 acquisition in order to perform a head-to-head comparison of accelerated to
non-accelerated 3D T1-weighted imaging. It will also allow us to pilot experimental sequences (ASL, RSFC,
and DTI) more effectively than at 1.5T. Finally, most future clinical trials will likely include both 1.5T and 3T
scanners, so including both field strengths in ADNI 2 will appropriately reflect the design of future clinical trials.
Preliminary site survey results indicate that 3T is available at every ADNI enrollment site.
3.2.4. Duration of Exam: Considerations in designing the MR protocol involve optimizing tradeoffs among
competing objectives. A complex (long) MR exam in which many sequences are acquired would maximize the
amount of data collected per exam. However, a shorter less complex exam would be met with greater patient
acceptance and lower attrition. ADNI requires that subjects agree to multiple investigative procedures acquired
at multiple time points for at least 5 years. This is a major commitment, particularly for a study in which no
therapy is offered. The frequent imaging performed in this study alone imposes a greater scanning burden on
participants than in most natural history studies. We therefore settled on an upper time limit of approximately
30 minutes of scan time for the MRI protocol beyond which we would be concerned about excessive patient
burden. This imposes a hard limit on the number of different imaging sequences that can be included in the
protocol.
3.2.5. Rationale for Selection of Core Sequences for the ADNI 2 Protocol: The ADNI 2 protocol is divided
into a core set of sequences that will be performed on all scanners in all subjects, and a series of more
experimental vendor specific sub-studies. This approach was arrived at after extensive discussion within the
ADNI MRI core, aligned external investigators, the ADNI Steering committee, and the industry scientific
advisory board (ISAB).
The 3D T1 and FLAIR sequences were considered the minimum requirements of the protocol. The 3D T1 is
the primary sequence for structural MRI analysis, while FLAIR is the sequence most widely used in clinical
neuroradiology for general pathology detection – including cerebro vascular disease. We then considered
adding 1 or 2 additional sequences to this minimum core from the following: GRE micro hemorrhage imaging,
RSFC, ASL, and DTI. Criteria for selecting additional sequences to include in the core protocol were: relevance
to clinical trials, availability among the three major MR vendors (GE, Siemens, Philips), ability to standardize
across vendor platforms, reliability of quantitative measurements, evidence of test-retest precision of
longitudinal measures, availability of data (cross-sectional and longitudinal) indicating evidence of diagnostic
efficacy, and interest on the part of both the ISAB as well as associated ADNI scientific investigators.
We elected to include GRE imaging as part of the core protocol on all systems. GRE imaging has become a
routine component of the MRI protocol for all anti-amyloid clinical trials. Up to 20% of all AD subjects enrolled
in anti-amyloid clinical trials have an abnormality detectable on GRE images (M. Bednar, ICAD 2009).
Including GRE imaging in the core protocol will permit ADNI to measure prevalence and incidence of micro
hemorrhages in a typical clinical trial population. In addition, the GRE sequence will allow us to assess
relationships between micro bleeds and both amyloid deposition and vascular disease. We considered the
option of using a 2D vs. a 3D GRE sequence. Although a 3D GRE has been advocated as more sensitive to
micro hemorrhages [2], our investigation into implementation issues showed that there are practical difficulties
associated with the 3D GRE acquisition in a multi-center, multi-vendor setting. Specifically, not all vendors offer
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
parallel imaging for this particular 3D pulse sequence, so the acquisition time for 3D whole head coverage
sometimes exceeds 10 minutes. Because ADNI 2 will use only manufacturer-provided sequences, to keep
acquisition time near 4 minutes while obtaining whole brain coverage we that believe the most practical
solution is to employ the standard 2D GRE sequence. Compared to the 2D pulse sequence described in
Vernooij et al [2], we will increase sensitivity by using 3T instead of 1.5T and by decreasing slice thickness
from 5 mm to 4 mm.
In summary, the features of the ADNI 2 protocol for newly enrolled subjects are:
1. Only manufacturer-available pulse sequences will be employed.
2. Scanning will be done at 3T.
3. The GO and the ADNI 2 protocols will be identical.
4. We will use MPRAGE for the structural T1-weighted sequence for Siemens and Philips systems and a
closely-related manufacturer-available pulse sequence, 3D IR-FSPGR, sequence on GE systems.
5. Both an accelerated and a non-accelerated 3D T1 sequence will be acquired in each exam.
6. The dual echo fast spin echo sequence that was used for vascular pathology detection in ADNI 1 will be
replaced by a T2-weighted FLAIR, which will be accelerated using parallel imaging by a factor of 2 on
systems where this is possible.
7. A long TE gradient echo (GRE) sequence will be included for micro hemorrhage detection.
3.2.6. Rationale for Including Both an Accelerated and Non-accelerated 3D T1 Sequence in the Core
Protocol: ADNI 1 employed back-to-back identical MPRAGE acquisitions so that in the event of severe artifact
on the first MPRAGE, a second MPRAGE might salvage the exam and prevent the subject from being asked to
return for a re-scan. As noted later in the progress report, the yield for this was minimal and therefore this
approach is not justified in ADNI 2. A second rationale for back-to-back MPRAGE acquisitions was for
purposes of signal averaging to increase SNR; however, to our knowledge this has seen little application in the
ADNI user community. However, for completely different reasons, we do plan to include two 3D T1 volume
sequences in ADNI 2, one of which is accelerated 2X and the other non-accelerated. The reasoning for this is
as follows. Compared to non-accelerated scanning at 1.5T, the increased SNR at 3T can be traded off for
reduced scan time through acceleration with parallel imaging. This is an attractive approach for clinical trials
that should reduce scan times and reduce the overall prevalence of patient motion artifacts. However, there is
a well-known SNR penalty with scan acceleration, as well as associated image artifacts. Moreover, the ability
of accelerated scans to appropriately capture the atrophy associated with AD has not been validated in large
multi-center studies to our knowledge. Although the potential for reduced artifacts due to patient motion is very
appealing, using solely an accelerated 3D T1 acquisition for the entire ADNI 2 study seems quite risky. The
alternative we selected was to acquire both an accelerated and a non-accelerated 3D T1 scan in all subjects
as part of the core protocol. This would allow a head-to-head comparison of these two approaches without
incurring any risk to the integrity of the study. A thorough comparison will include not only cross-sectional
analyses but also must include a comparison of the ability of accelerated vs. non-accelerated scans to
accurately measure rates of change on serial studies and correlate rates with clinical group and clinical indices
of cognitive change. This type of rigorous longitudinal comparison cannot be accomplished in a 1 to 2 month
preparatory phase and hence must be built into the main study protocol as we have done. We anticipate that at
the end of roughly 2 years of data acquisition, enough longitudinal data will have been acquired to determine in
a rigorous manner where the benefit-risk line lies with respect to accelerated 3D T1 acquisitions for multicenter AD trials. If the data are clear cut at that time, then we can drop either the accelerated or the nonaccelerated acquisition from the protocol in years 3-5.
3.2.7. Rationale for Selection of Experimental Sub-study Sequences for ADNI 2: We elected not to
include ASL, RSFC or DTI in the core protocol for all sites for several reasons. These include absence of
longitudinal data documenting acceptable test-retest precision, minimal evidence of diagnostic efficacy in AD,
questionable relevance to clinical trials in the near term, absence of accepted phantoms to measure the
physical properties of perfusion, diffusion, or connectivity, and the need to keep the protocol to an acceptable
time for patient acceptance purposes. Perhaps most importantly, we note that due to rapidly developing
technology and differing design choices among vendors it is not practical to standardize DTI and especially
ASL across vendors using manufacturer-available sequences at the current time. And an iron clad guideline for
ADNI 2 is that only manufacturer-available sequences will be used. We believe that placing fundamentally
incompatible across-vendor MRI data into the public domain would be a disservice to the scientific community.
At the same time, we did think that it was important to consider including each of these important and emerging
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
sequence types in ADNI 2 in order to pilot these approaches for potential use in multi-center clinical trials.
These considerations led us to conclude that a viable approach would be to add a different one of each of the
three experimental sequence types to the core protocol of each of the three major MRI vendors – creating
vendor specific protocols. Including only one experimental sequence per vendor protocol will keep the over
duration of the protocol to an acceptable time (~ 30 minutes) for patient acceptability purposes.
In considering options to deploy one of the three experimental sequences per vendor protocol, the most
obvious driving factor centers on ASL. GE presently only offers a WIPS ASL sequence. Therefore only
Siemens or Philips can be considered for ASL sub-studies. The MRI core has strong technical expertise and
industry contacts with GE and Siemens. Therefore the most logical approach is to deploy a complex sequence
like ASL on Siemens (rather than Philips) systems. DTI, which is also complex, is a logical fit for GE systems.
The least complex sequence, resting state BOLD, is a logical fit for Philips systems. We believe that the
protocol described above is well justified, and will provide valuable information both for industry sponsors as
well as scientific investigators. Consensus was reached on this multi-track approach following extensive
discussions within the MRI core and industry supporters of ADNI.
Our plan is to evaluate the feasibility of including each of the three more experimental sequences in ADNI 2
in the manner outlined above. Factors to consider will include which systems are available for ADNI 2 and
which sites have purchased the relevant software keys on their own (ADNI cannot be expected to fund
purchase of optional software license keys for individual MRI sites). We will only know this information after the
in-depth equipment survey that the MRI core will conduct prior to start of ADNI 2. We will evaluate the
feasibility of adding the more experimental sequences to ADNI 2 and will do only that which will result in
statistically meaningful data based on enrollment and the specific system configurations throughout the ADNI 2
enrollment sites.
The MRI core can count many accomplishments from ADNI 1. These include:
1. We achieved consistent acquisition methods across 89 MRI scanners at 59 sites. A total of 38 different
vendor and platform specific ADNI 1 protocols were created and distributed to sites, and posted publically
in PDF from on the ADNI web site.
2. Demonstrated that longitudinal consistency is improved with correction of scaling [3] gradient non-linearity
[4], and with intensity non-uniformity correction [5, 6].
3. The source code for ADNI phantom analysis was made freely available on the ADNI-LONI web site.
4. Demonstrated that some MRI analysis methods had greater longitudinal power than others. The best
performing measures (defined as smallest sample sizes needed to detect a 25% rate reduction in AD and
MCI subjects) were FreeSurfer hippocampal volume and selected voxels in the temporal lobe using tensorbased morphometry. The longitudinal measures with the greatest correlations with longitudinal change in
general cognition (ADAS cog) were the whole brain and ventricular BSI measures and also FreeSurfer
ventricular volume (presented at ADNI Steering Committee Meeting, Seattle, 2009).
5. MRI has much better longitudinal power to detect change than clinical instruments, resulting in substantially
smaller sample sizes for clinical trials in both MCI and AD patients [1, 7].
6. MRI has equivalent or better longitudinal power to detect change than FDG PET, resulting in generally
smaller sample sizes for clinical trials in both MCI and AD patients (presented at ADNI Steering Committee
Meeting, Seattle, 2009).
7. Longitudinal change on MRI has at least equivalent correlation with change in cognitive measures
compared with change in FDG PET (presented at ADNI Steering Committee Meeting, Seattle, 2009).
8. MRI has better correlation with clinical group membership and with cognitive tests than CSF tau or Aβ 42
[8].
9. MRI has better prediction of risk of conversion from MCI to AD than CSF tau or Aβ 42 [9].
10. Provided sample size estimates for powering clinical trials for MCI and AD using different MRI
measurement methods (presented at ADNI Steering Committee Meeting, Seattle, 2009).
11. Demonstrated no superiority for 3T vs. 1.5T in group-wise discrimination or sample sizes needed to power
trials [1].
12. Demonstrated that MRI rates of change in cognitively normal subjects are greater in APOE ε4 carriers than
non-carriers [10, 11].
13. Demonstrated that hippocampal rates in MCI and AD accelerate [10].
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
14. Demonstrated greater white matter hyperintensity load in AD than control and in MCI than control subjects
that would be typically enrolled in therapeutic trials (ADNI Steering Committee Meeting, Seattle, 2009).
15. Found associations between levels of atrophy and genetic differences in the glutamate receptor [12], the
FTO obesity gene [13], and several other candidate genes [14].
16. The ADNI1 protocol included two back-to-back MPRAGE scans. As the source of morphometric data,
MPRAGE was the major MRI focus in ADNI1. In 13% or all exams the quality ranking of the second
MPRAGE was better than the first. However, a primary driving force behind the choice of including two
MPRAGEs was that in the event of severe artifact on the first MPRAGE, a second MPRAGE might salvage
the exam and prevent the subject from being asked to return for a re scan. Among 2473 1.5T scan pairs,
we found 89 instances where the first MPRAGE failed and the second MPRAGE was of acceptable quality.
The exam salvage rate obtained from incorporating a second MPRAGE in the protocol was therefore
89/2473 or 3.6%. This seems like a low return-on-investment and therefore we will not acquire back-toback identical 3D T1-weighted scans in each exam in ADNI2 (presented at ADNI Steering Committee
Meeting, Seattle, 2009). Note that the planned accelerated vs. non-accelerated 3D T1 acquisitions address
an entirely different issue.
3.3.1. Full Length Peer Reviewed Publications by Funded ADNI MRI Core Members:
1. Schuff N, Woerner N, Boreta L, Kornfield T, Shaw LM, Trojanowski JQ, Thompson PM, Jack CR Jr, Weiner
MW; Alzheimer's Disease Neuroimaging Initiative. MRI of hippocampal volume loss in early Alzheimer's
disease in relation to ApoE genotype and biomarkers. Brain. 2009:1067-77. PMID: 19251758.
2. Boyes RG, Gunter JL, Frost C, Janke AL, Yeatman T, Hill DLG, Bernstein MA, Thompson PM, Weiner
MW, Schuff N, Alexander GE, Killiany RJ, DeCarli C, Jack CR, Fox NC. Intensity non-uniformity correction
using N3 on 3-T scanners with multi-channel phased array coils. Neuroimage 2008 39(4):1752-62 PMID:
18063391.
3. Jack CR Jr, Bernstein MA, Fox NC, Thompson P, Alexander G, Harvey D, Borowski B, Britson PJ, L
Whitwell J, Ward C, Dale AM, Felmlee JP, Gunter JL, Hill DL, Killiany R, Schuff N, Fox-Bosetti S, Lin C,
Studholme C, Decarli CS; Gunnar Krueger, Ward HA, Metzger GJ, Scott KT, Mallozzi R, Blezek D, Levy J,
Debbins JP, Fleisher AS, Albert M, Green R, Bartzokis G, Glover G, Mugler J, Weiner MW; ADNI Study.
The Alzheimer's disease neuroimaging initiative (ADNI): MRI methods. J Magn Reson Imaging.
2008;27(4):685-91. PMID: 18302232.
4. Hua X, Leow AD, Lee S, Klunder AD, Toga AW, Lepore N, Chou YY, Brun C, Chiang MC, Barysheva M,
Jack CR Jr, Bernstein MA, Britson PJ, Ward CP, Whitwell JL, Borowski B, Fleisher AS, Fox NC, Boyes RG,
Barnes J, Harvey D, Kornak J, Schuff N, Boreta L, Alexander GE, Weiner MW, Thompson PM. Alzheimer's
Disease Neuroimaging Initiative. 3D characterization of brain atrophy in Alzheimer's disease and mild
cognitive impairment using tensor-based morphometry. Neuroimage. 2008; 41: 19-34. PMID: 18378167.
5. Hua X, Leow AD, Parikshak N, Lee S, Chiang M-C, Toga AW, Jack CR Jr., Weiner MW, Thompson P,.
Tensor-based morphometry as a neuroimaging biomarker for Alzheimer’s disease: An MRI study of 676
AD, MCI, and Normal Subjects. NeuroImage, 2008 Jul 22. Epub ahead of print. PMID: 18691658.
6. Morra JH, Tu Z, Apostolova L, Green A, Avedissian C, Madsen S, Parikshak N, Hua X, Toga A, Jack CR
Jr, Weiner M, Thompson P. Validation of a fully automated 3D hippocampal segmentation method using
subjects with Alzheimer’s Disease, mild cognitive impairment, and elderly controls. NeuroImage 2008 Oct
15; 43(1):59-68. PMID: 18675918.
7. Morra J, Thompson P, Tu Z, Apostolova L, Green A, Avedissian C, Madsen S, Parikshaik N, Hua X, Schuff
N, Weiner M, Jack CR Jr. Automated 3D mapping of hippocampal atrophy and its clinical correlates in 400
subjects with Alzheimer’s disease, mild cognitive impairment, and elderly controls. Neuroimage. 2009
Mar;45(1 Suppl):S3-15. Epub 2008 Nov 8. PMID: 19041724.
8. Leow AD, Yanovsky I, Parikshak N, Hua X, Lee S, Toga AW, Jack CR Jr, Bernstein MA, Britson PJ, Gunter
JL, Ward CP, Borowski B, Shaw LM, Trojanowski JQ, Fleisher AS, Harvey D, Kornak J, Schuff N,
Alexander GE, Weiner MW, Thompson PM, The Alzheimer’s Disease Neuroimaging Initiative. Alzheimer’s
Disease Neuroimaging Initiative: A One-Year follow-up study using tensor-based morphometry correlating
degenerative rates, biomarkers and cognition. NeuroImage 2009 Apr 15; 45(3):645-55. PMID: 19280686.
9. Leow AD, Klunder AD, Jack CR, Toga AW, Dale AM, Bernstein MA, Britson PJ, Gunter JL, Ward CP,
Whitwell JL, Borowski B, Fleisher A, Fox NC, Harvey D, Kornak J, Schuff N, Studholme C, Alexander GE,
Weiner MW, Thompson PM, ADNI Study (2006). Longitudinal Stability of MRI for Mapping Brain Change
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
using Tensor-Based Morphometry, Neuroimage. 2006 Jun;31(2):627-40. Epub 2006 Feb 15. PMID:
16480900.
10. Chou YY, Leporé N, Avedissian C, Madsen SK, Parikshak N, Hua X, Trojanowski JQ, Shaw L, Weiner
MW, Toga AW, Thompson PM (2009). Mapping Correlations between Ventricular Expansion, and CSF
Amyloid & Tau Biomarkers in 240 Subjects with Alzheimer’s Disease, Mild Cognitive Impairment and
Elderly Controls, NeuroImage. 46(2):394-410, June 2009; Epub 2009 Feb 21. PMID: 19236926.
11. Chou YY, Leporé N, Avedissian C, Madsen SK, Hua X, Jack CR, Weiner MW, Toga AW, Thompson PM
(2009). Mapping Ventricular Expansion and its Clinical Correlates in Alzheimer’s Disease and Mild
Cognitive Impairment using Multi-Atlas Fluid Image Alignment, SPIE Medical Imaging 2009, SPIE Paper
Number 7259-111, Feb 9 2009 (peer-reviewed, 12 pages).
12. Jack CR Jr, Lowe VJ, Weigand SD, Wiste HJ, Senjem ML, Knopman DS, Shiung MM, Gunter JL, Boever
BF, Kemp BJ, Weiner M, Petersen RC, Alzheimers Disese Neuroimaging Inititative. Serial PIB and MRI in
normal, MCI, and AD: implications for sequence of pathological events in AD. Brain 2009 May 132(Pt
5):1355-65 PMID: 19339253.
13. Evans MC, Barnes J, Nielsen C, Kim LG, Clegg SL, Blair M, Leung KK, Douiri A, Boyes RG, Ourselin S,
Fox NC; and the Alzheimer’s Disease Neuroimaging Initiative. Volume changes in Alzheimer's disease and
mild cognitive impairment: cognitive associations. Eur Radiol. 2009 Sep 16. PMID: 19760240
14. Clarkson MJ, Ourselin S, Nielsen C, Leung KK, Barnes J, Whitwell JL, Gunter JL, Hill DLG, Weiner MW,
Jack CR, Fox NC and The Alzheimer’s Disease Neuroimaging Initiative. Comparison of phantom and
registration scaling corrections using the ADNI cohort. Neuroimage. 2009 Oct 1;47(4):1506-13 PMID:
19477282.
15. Hua X, Lee S, Yanovsky I, Leow AD, Chou YY, Ho A, Gutman B, Toga A, Jack CR Jr, Bernstein MA,
Reiman EM, Havery D, Kornak J, Schuff N, Alexander GE, Weiner MW, Thompson P. Optimizing power to
track brain degeneration in Alzheimer’s Disease and mild cognitive impairment with tensor-based
morphometry: an ADNI study of 515 subjects. NeuroImage, 2009 Dec;48(4):668-81. Epub 2009 Jul 14.
PMID: 19615450.
16. Ho A, Hua X, Lee S, Leow AD, Yanovsky I, Gutman B, Dinov ID, Lepore N, Stein J, Toga AW, Jack CR Jr,
Bernstein MA, Reiman EM, Harvey DJ, Kornak J, Schuff N, Alexander GE, Weiner MW, Thompson PM.
Comparing 3 Tesla and 1.5 Tesla MRI for Tracking Alzheimer’s disease progression with tensor-based
morphometry. Hum Brain Mapp. 2009 Sep 24. Epub ahead of print. PMID: 19780044
17. Gunter JL, Bernstein MA, Borowski BJ, Ward CP, Britson PJ, Felmlee JP, Schuff N, Weiner M, Jack CR Jr,
and the Alzheimer’s Disease Neuroimaging Initiative. Measurement of MRI scanner performance with the
ADNI phantom. Med Phys 2009 Jun; 36(6):2193-205. PMID: 19610308.
3.4. MRI Core Methods:
3.4.1. Organization and Overview: The MRI core of ADNI has two components; the central lab at the Mayo
Clinic, and the five funded image analysis co-investigators. Key personnel at the central lab at Mayo include
Drs. Jack, Bernstein, and Gunter in addition to support staff such as Bret Borowski, RTR and Kaely Steinert,
RTR. The five funded image analysis co-investigators are Drs. Paul Thompson, Nick Fox, Charles DeCarli,
Clifford Jack, and Norbert Schuff. Each of these individuals has established a significant publication record in
MRI of AD. Dr. Jack will be responsible for overall day-to-day MR-related operations for ADNI throughout the
study. The MR core will communicate by teleconference monthly and email as needed to review progress,
identify problem areas, and arrive at appropriate solutions. This approach is already in place for ADNI1 and
has been successful.
We anticipate roughly 55 clinical enrollment sites. Because we will scan ADNI1 subjects who have been
carried forward at 1.5T and also scan newly enrolled GO and ADNI2 subjects at 3T, two ADNI scanners must
be supported at each enrollment site. This results in an estimated 110 MRI scanners that must be maintained
by the MRI central lab at the Mayo Clinic.
MRI core methods fall into two broad categories. The first category concerns service aims of the central MRI
core lab at Mayo Clinic needed to generate high quality MRI data in all subjects at each time point. This MRI
data will be made available to the general scientific community and will also be used by the five funded ADNI
MRI core analysis labs to generate numeric summary MRI data. This numeric summary data will also be made
available to the general scientific community. In addition, each of the five funded ADNI MRI investigators
themselves will conduct specific hypothesis testing.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
3.4.2. Service Aims of Central MRI Core Lab at Mayo Clinic
An overarching goal of the MRI core lab is optimization of MR imaging across all scanners with an emphasis
on methods that will be applicable to clinical trials, acceptable to the FDA for registration purposes, and will
also provide data that will advance the neuroscience of AD. An overview of the functions provided by the
central core lab at Mayo Clinic for ADNI is below.
3.4.2.1. MRI Protocol Creation, Distribution, Site Certification
The steps entailed are the following:
1. Perform a detailed equipment survey at each site – determine technical specifications for each potential
scanner in the study, select optimum 3T scanner at each site, and monitor upgrades throughout the study.
2. Create a generic, non-platform specific MRI protocols.
3. With the aid of the MRI vendor support network created in ADNI1, create a platform specific protocol for
each scanner – with roughly 110 scanners we anticipate 60 or more distinct compiled versions of the ADNI
protocol will be required at any given time. This will entail piloting the protocol on every platform prior to site
distribution.
4. To avoid inevitable errors that occur with transcription of paper protocols, the protocols will be distributed to
each scanner electronically, as was done in ADNI1. One of the major lessons learned in ADNI1 was that
protocols should be distributed electronically, and loaded onto the appropriate scanner at each site without
manual alterations by MRI technologists at the site. This approach dramatically increases consistency and
was one of the major strengths of ADNI1.
5. Create an ADNI2 MRI procedures manual and distribute to sites.
6. Certify each scanner at baseline using both phantom and human volunteer scans, and re-certify scanners
after upgrades.
3.4.2.2.
Quality Control: Image quality control is a labor intensive process, but as clearly demonstrated in
ADNI1, it is necessary to insure protocol adherence and image quality. The following quality control operations
will be performed.
1. The images are inspected manually for the presence of artifacts (e.g., patient motion), to evaluate overall
image quality, to ensure that all slices were transmitted, and that the entire head was included in the field of
view. This manual inspection takes place for each of the sequences in the MRI protocol and the results of
the inspection recorded in a database.
2. Adherence to the parameters specified in the protocol for each of the sequences is checked. A software
program searches the DICOM header fields of each incoming data set for relevant pulse sequence
parameters, for example, TR, TE, field of view, matrix, etc. The automatic parameter check program
compares incoming parameter data against the values established for each sequence and flag any
deviation from protocol.
3. For scans acquired after baseline, we will spatially register each subsequent scan for every subject to
his/her baseline scan and toggle between scans looking for change in contrast, linear and non-linear
scaling changes, artifacts, intensity inhomogeneity, etc. – and to ensure that the scan is actually the same
person (which surprisingly enough is not always the case).
4. We will need to identify appropriate methods of quality control for non-anatomic imaging sequences, i.e.,
DTI, ASL, RSFC. Because of limited time in the scanning session, and the current unavailability of
standardized phantoms for these emerging applications, we think that realistically these checks may be
limited to DICOM header parameter checks and visual QC of processed images.
5. If an image quality problem is identified (1 or 3 above), or a deviation of MR protocol is identified (2 above),
the site will be contacted immediately. If the deviations are significant, the site will be asked to re-scan the
patient (although we hold re-scanning to a minimum in order to not burden subjects).
6. Although each site is asked to inform the MR central lab at Mayo of any imminent hardware or software
upgrades, typically this does not happen. Therefore, the hardware and software versions of each incoming
MR dataset will be verified at Mayo.
7. The results of the MR QC inspection are cross-logged with the informatics core at LONI and with the
clinical coordinating center.
8. ADNI phantom images are analyzed to check the gradient calibration at the site. This ensures spatial
fidelity to a level much greater than the anatomical changes due to progression of AD [4].
3.4.2.3.
Accommodating MRI Upgrades: On average, we can expect that each MR system in the study
will be upgraded at least once per year. With 110 scanners in the study we anticipate that at least 550
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
significant system upgrades will occur during the course of the 5 year ADNI2 study. A significant effect which
must be accounted for is the possibility that when the sequences in the protocol are recompiled under an
updated version of the scanner’s operating system, changes in some of the sequence parameters may occur,
obviously an undesirable result in a longitudinal study that hinges on methodological consistency over time.
For example, changes in TE or bandwidth, etc., may be an unintended consequence of pulse sequence
conversion for the new software revision. Our approach to software and hardware upgrades will be to
anticipate effects of an impending upgrade at every participating site. We will test the pulse sequences in the
ADNI protocol on every new hardware/software platform revision prior to installation at an ADNI site. This will
allow us to effect appropriate corrective action before the fact. For example, parameters may need to be
changed for the recompiled sequence to be executed in a manner identical to the pre-upgrade sequence. A
second major problem with hardware/software upgrades is that various calibration files often become
corrupted. Our solution to this problem is continuous phantom-based measurements of gradient calibration at
all scanners throughout the initiative.
3.4.2.4.
MR Data Pre-processing Methods: Modern MR scanners are capable of acting as high precision
measurement devices, but have not always been engineered to maximize that objective. Instead, commercial
MR scanners are often optimized to provide the best possible qualitative data for visual interpretation of
medical images. In a sense therefore we will be asking the scanners to perform a function (accurate and
precise quantification) for which they were not explicitly designed.
One of the accomplishments of the MRI core in ADNI1 was explicit demonstration of benefits of three specific
classes of image pre-processing to correct image artifacts. These data were presented at the ADNI Steering
Committee Meeting in Seattle 2009: 1) Correcting image scaling changes reduces by 10-12% the sample size
needed to detect a 25% rate reduction in AD subjects when compared to the same scans in which gradient
scaling changes have not been corrected [3]. 2) Correction of gradient non-linearity reduces by 9-10% the
sample size needed to detect a 25% rate reduction in control, MCI, and AD subjects when compared to the
same scans in which gradient unwarping has not been performed [4]. 3) Off-line processing to reduce intensity
non-uniformity significantly improves the precision of longitudinal measurements, particularly at 3T and on
images acquired on multi-array coils [5, 6]. This is particularly relevant given the protocol design in ADNI2,
which will emphasize 3T and also modern scanners with multi-array receiver coils.
Therefore, all 3D T1-weighted images acquired in ADNI2 will undergo correction for image non-uniformity,
warping due to gradient non-linearity, and scaling change over time. These corrections will be performed using
an automated pipeline developed by Jeff Gunter, Ph.D. at the Mayo Clinic which has been thoroughly tested
and validated on several thousand MRI studies obtained in ADNI1 and at the Mayo Clinic.
3.4.2.5.
Scanner Monitoring with the ADNI Phantom: Every scanner in ADNI2 will be monitored by the
ADNI phantom (Fig. 3.4.1.). As in ADNI1, we will acquire a phantom image at the end of each patient study.
Given the 30-minute limit placed on the duration of the patient scan and a 45-minute average time purchased
with each ADNI MRI examination slot, coupling the phantom QC scan with each subject exam is a cost
effective approach to scanner monitoring QC. The alternative would be paying for independent time slots at
each site for phantom-only scans, which would impose additional costs on the study. Phantom images are
analyzed at Mayo Clinic using a software program created by Jeff Gunter, Ph.D. specifically for ADNI called
AQUAL. (Note that source code for AQUAL has been made freely available on the ADNI-LONI web site).
AQUAL measures and reports out longitudinal measures of SNR, contrast, and both linear scaling and image
warping for each scanner in the study.
Each current ADNI site now has an ADNI phantom and therefore only a relatively small number of additional
phantoms will be necessary for ADNI2. Our experience in ADNI1 has been that the ADNI phantom is
invaluable for monitoring image geometry. Specific examples of major problems in ADNI1 that would not have
been identified without the use of the ADNI phantom are:
1. Identifying correct gradient type: Reporting of gradient type in the
DICOM header is ambiguous for some scanner models. In addition,
several sites self-reported an incorrect gradient model for their own
systems in the detailed MRI survey in ADNI1. In every such case,
scanning the ADNI phantom uncovered the problem. Without this
phantom surveillance the MRI core would have applied incorrect
gradient unwarping coefficients to all scans acquired on each of the
affected scanners throughout the entire ADNI1 study.
Fig. 3.4.1. ADNI Phantom
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
2. Identifying incorrect patient table positioning: At one site, table positioning was erroneously offset by 5 cm
following a vendor upgrade due to a mis-aligned laser land-marking light. This resulted in patient scans
being acquired away from magnet iso-center and the wrong gradient unwarping values being applied. This
error was unknown to the site and to the vendor who performed the upgrade, and was only uncovered by
scanning with the ADNI phantom.
3. Protocol error: An error occurred in creating the protocols for Phillips 1.5T systems early in the study –
specifically, autoshim was not specified when it should have been. Autoshim enabling status is not
recorded in the DICOM header for Phillips systems, so there is no way to monitor this feature of the
acquisition based on information recorded in the image header by the vendor. Geometric instability for
these systems was uncovered by the phantom monitoring QC program and the protocol problem was
corrected. Without the phantom monitoring system in place, this protocol error would have never been
identified.
All told, we estimate that 15% of all ADNI1 scans would have been affected by the errors such as these.
Monitoring scanner performance with the
ADNI phantom is a relatively inexpensive
way to insure that these types of errors are
uncovered and appropriate corrective action
is taken. Note that patient images will not be
altered on the basis of phantom scans.
3.4.3. Image Analysis:
3.4.3.1. Analysis of the Core Protocol
Sequences by the Five Funded Analysis
Labs: In addition to the central lab at Mayo,
the MRI core of ADNI consists of five funded
image analysis laboratories. These five labs
are headed by individuals who have
established significant publication track
records in MR imaging of Alzheimer’s
disease. The number of funded MRI image
analysis groups has been decreased from
ADNI1 to ADNI2 primarily to include
laboratories that reported the most
promising results in ADNI1. PIs of the five
Fig. 3.4.2. Tensor-based Morphometry Maps How Fast the
funded image analysis groups and their
Brain is Losing Tissue. In our longitudinal studies using our new
responsibilities for analyses of the core
TBM processing pipeline, we mapped how fast the brain loses
protocol sequences are:
tissue in Alzheimer’s disease (AD; top left) and mild cognitive
1. Paul Thompson – responsible for
impairment (MCI; top right). 0.3% per year loss rates in controls
longitudinal TBM of the 3D T1-weighted
compared to 2-3% per year in AD. In MCI, the loss rate was
scans
predictive of who would deteriorate to AD within a 1-year follow2. Nick Fox – responsible for longitudinal
up interval. TBM gave sample size estimates 10 times lower than
BSI and hippocampal volume measures
the best clinical scores (bottom right).
of the 3D T1-weighted scans
3. Norbert Schuff - responsible for longitudinal FreeSurfer measures of the 3D T1-weighted scans
4. Clifford Jack - responsible for longitudinal STAND score measures of the 3D T1-weighted scans
5. Clifford Jack - responsible for longitudinal quantification of micro hemorrhages on the long TE GRE scans
6. Charlie DeCarli - responsible for baseline volume measures of WMH lacunar infarctions and stroke on
FLAIR images
3.4.3.2.
Morphometry - 3D T1 images
1. Paul Thompson will perform tensor-based morphometry of the 3D T1 images (Fig 3.4.2.). 3D maps of rates
of atrophy (as a percent per year) will be derived by fluid registration of each follow-up scan to the baseline
scan using an extensively validated non-linear image registration algorithm [5, 15-17]. After fluidly aligning
maps of annualized atrophic rates to a geometrically-centered mean anatomical template [7, 18], we will fit
general linear models and longitudinal mixed models to map voxel-by-voxel correlations with diagnosis,
cognitive decline, and pathological biomarkers, and FDR will be used for multiple comparisons correction
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
[15]. To boost power, numeric summaries of atrophic rates will be computed in a statistically pre-defined
region of interest within the temporal lobes [7, 19] showing greatest effect sizes in independent training
data.
2. Norbert Schuff will perform measurements of the 3D T1 images using the probabilistic-based FreeSurfer
(FS) software. The results obtained from ADNI thus far show that hippocampal volume loss over time,
measured by FS, has extremely high statistical power to detect potential treatment effects. These efforts
using FS will be assisted by Dr. Bruce Fischl of MGH, one of the primary developers of FS, who will be
available for consultations related to FS process optimization. In short, the FreeSurfer pipeline consists of
five stages: an affine registration with Talairach space, an initial volumetric labeling, bias field correction,
non-linear alignment to the Talairach space, and a final labeling of the volume. The fully automated labeling
of volumes is achieved by warping a population-based brain atlas to the target brain and by maximizing an
a-posteriori probability of the labels given specific constraints. A full description of the FreeSurfer
processing steps can be found in [20]. The procedures have been extensively validated. Volume
measurements of about 96 anatomical brain regions will be computed, including the hippocampus, which
showed the most prominent volume change in previous studies [10]. In addition to volumes, thickness,
curvature and other geometric measures will be computed for cortical regions. For longitudinal
measurements of change, a Markov chain-based protocol will be applied [10], in which past measurements
are used as priors for current measurements. We have demonstrated that a Markov chain approach can
drastically reduce within subject variability in longitudinal data, increasing the sensitivity to detect volume
change.
3. Nick Fox will perform brain boundary shift integral (BSI) measures on 3D T1-weighted (T1-w) images. Each
individual’s 3D T1-w image will have the brain and ventricular regions delineated using a validated semiautomated method to give an approximate volume for each time point. Follow-up scans will be affine
registered to baseline incorporating scaling and differential bias corrections [21]. Brain and ventricular
volume change will be measured using the BSI. The BSI calculates the 3D integral of change in the
boundary of the region of interest directly from the registered, normalized and subtracted images and has
been shown to be sensitive to early neurodegenerative loss [22] and applicable to intervals of only 6
months [23]. Recent developments in BSI normalization and calculation (in submission) suggest that
robustness to contrast change can be improved and power increased: reducing sample sizes by over 20%
without excluding scan pairs (in submission) – we will further assess this in ADNI2. In addition, we will
investigate hippocampal BSI [24] and non-linear registration (fluid) [25] to track change within the medial
temporal lobe. These exploratory analyses will assess change based upon baseline hippocampal regions
derived from other processing labs and also explore the use of other automated template-based region
definition [26]. Preliminary results suggest that fully automated and precise region of interest measures and
rates of change are possible and would be appropriate to the analysis of using multiple time point
acquisitions. Using these promising approaches the Fox lab will perform automated hippocampal volume
and atrophy rate measures.
4. Clifford Jack (lab) will perform measurements with an algorithm named Structural Abnormality iNDex
(STAND). STAND is an algorithm that extracts atrophy information from individual patient’s 3D T1 MRI
scans and assigns a continuous STAND-score to the scan based on the degree of atrophy in comparison
to patterns extracted from a large library of clinically well characterized subjects’ MRI scans [27]. Algorithm
training and voxel selection is based on a linear support vector machine classifier. In Mayo subjects who
underwent ante mortem MRI and went on to autopsy, the rank correlation between Braak NFT stage and
STAND-scores was 0.62 (p<0.0001) [28]. Analyzing ADNI data using this method, the correlation with
clinical group membership, general cognitive performance and functional performance was better with
STAND than CSF biomarkers (Aβ 42, t- tau, p- tau) [8]. The ability to predict future progression from MCI to
AD was better with STAND than CSF biomarkers [9].
3.4.3.3. Cerebro Vascular Disease:
Charles DeCarli will calculate measures of white matter disease burden using the same methods that were
applied to ADNI1. All baseline protocol sequences will be used for detection of white matter hyperintensities
(WMH) and for manual infarct ratings; i.e., the 3D T1 and DSE sequences at 1.5T and 3D T1 plus FLAIR at 3T.
The validated, fully-automated WMH detection method aligns the imaging data to an elderly template image,
where WMHs are identified on a per-voxel basis based on image intensities and prior knowledge of the
probability of WMH occurrence at each location in the brain [29]. His laboratory also has developed an
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
optimized method for evaluation
of regional DTI in association
with local gray matter and white
matter volumes and plans to
extend
this
multi-modal
approach to include ASL. These
methods for DTI and ASL
analysis will be evaluated early
in ADNI2 as described below
under “experimental sequences”.
Finally, a trained and validated
expert will determine the gross
locations, sizes, and etiologies of
MRI-evident infarcts at baseline
and longitudinally using the
same
reliable,
repeatable
protocol that has been used for
ADNI and a variety of other
Fig. 3.4.3. Our DTI Processing Pipeline (Thompson/UCLA) Performs Highstudies,
including
the
throughput Fluid Registration and Statistical Analysis of Populations of
Framingham Heart Study.
DTI Data, revealing genetic effects, and cognitive correlates of DTI – here,
3.4.3.4.
GRE Imaging of
fractional anisotropy, perhaps the commonest DTI measure of fiber integrity,
correlates highly with IQ in 90 subjects (bottom left; r ~ 0.3-0.4 in blue regions).
Micro Hemorrhages
These point-by-point maps will reveal where changes in fiber integrity are linked
Clifford Jack (lab)will visually
with clinical measures, outcomes, and other biomarker data.
grade GRE scans for abnormal
tissue iron deposits. The number
and location of micro hemorrhages and other abnormal iron deposits will be quantified by visual review and
entered into a data form. Preliminary data indicate that 20% of subjects entered into typical AD therapeutic
trials may have abnormal tissue iron deposits – primarily micro hemorrhages (M Bednar, ICAD 2009).
3.4.3.5.
Experimental Sequences and Image Analysis Expert Consulting Group: In addition to the five
funded investigators above, the MRI core will form small advisory committees in 3 areas – DTI, ASL, and
RSFC analysis. ADNI will seek out 2-3 of the most productive scientists in each of these 3 areas. These
scientists will be asked to consult to ADNI via formal consulting contracts. By bringing state-of-the-art expertise
in these new areas into ADNI, we will guarantee that ADNI2 is at the cutting edge of image analysis science in
each of the MRI methods included in the study. The precise analysis methods that will be used for the three
experimental sequences are not fixed at this time. The methods, and the groups who will be assigned specific
analysis tasks, will be determined early in ADNI2 through pilot studies and interactions among the MRI core
members and the ADNI consultants. Descriptions of possible approaches to analysis of DTI, ASL, and RSFC
appear below.
3.4.3.6.
Diffusion Tensor Imaging: An illustration of one possible approach to DTI analysis is provided by
Paul Thompson here (Figure 3.4.3.). Our primary statistical analysis will characterize (1) the longitudinal
pattern of changes in DTI-derived measures of fiber integrity, in normals, MCI and AD patients; (2) relate DTI
changes to measures of impairment severity, future decline and CSF biomarkers; (3) identify baseline DTI
predictors of future decline, and assess their predictive value and sample size requirements when used in
conjunction with, or independently of, other imaging modalities. Correction for motion artifacts, eddy currents,
susceptibility artifacts: Motion artifacts and eddy current artifacts cause spatial misalignment and geometric
distortion among DTI data volumes scanned with different diffusion gradient directions [30]. We use the FMRIB
software library (FSL, http://www.fmrib.ox.ac.uk/fsl/) to correct for geometric distortions due to motion, eddy
current, and susceptibility artifacts. DTI data are then registered by 9-parameter transformation to the ICBM
space. All DTI scans are denoised using Riemannian methods [31], and mutually-registered to a geometricallycentered mean tensor image, using our validated fluid registration method based on information theory, driven
by the full 6D diffusion tensor [32]. Group Statistical Analyses: We will perform a voxel-by-voxel analysis of the
following DTI-derived measures: fractional anisotropy (FA), geodesic anisotropy (GA), mean diffusivity (MD),
and parallel and transverse diffusivity (diffusion tensor eigenvalues). To improve power, we will correct all
these indices for fiber crossing/mixing using our tensor deconvolution methods [33]; [34]; [35]; [36]. We will
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
also perform statistical analysis of the full 6D
diffusion tensor, which can boost power in
group DTI analyses [37-40]. ROI-based
Analyses: To provide regional summaries, we
will also fluidly align the parcellated Mori
DTI81 atlas to our mean DTI template and
compute average values of all DTI indices in
Young Normal
Old Normal
Alzheimer
MCI
regions of interest [41]. To boost power, we
Fig. 3.4.4. Arterial Spin Labeling
will also use the training-testing method [1, 7]
to create statistically-defined ROIs for minimal
sample size analyses. Joint DTI-atrophic Rate Analyses: As the atrophic rate and DTI maps are co-registered,
voxel-based statistics, using data from both modalities, will reveal where these indices offer synergistic or
independent information for diagnosis or prognosis. We will assess whether atrophic rates are greater in brain
regions with fiber integrity deficits at baseline. By fluid registration of serial DTI data [32]], we will adjust
measures of fiber integrity change for atrophy or geometric changes in anatomy (see Jahanshad et al [41] for
an examination of structural versus DTI-guided registrations).
3.4.3.7.
Arterial Spin Labeling: An illustration of one possible approach to ASL analysis is provided by
Norbert Schuff here [42-44]. An increasing body of work has shown that ASL MRI can detect regional pattern
of reduced brain perfusion in AD [42, 44] and even MCI [44]. In addition, MCI and AD present a characteristic
pattern of reduced perfusion that is different from the perfusion pattern other dementias, such as FTD [42-44].
Moreover, we have shown that alterations in brain atrophy and perfusion in AD can be discordant [44], implying
that ASL-MRI provides information complementary to structural MRI. However, ASL-MRI studies have been
challenging to a large part because signal-to-noise is low and biological variations are difficult to control. To
reduce variability in ASL MRI data, the processing for this study will include: 1) co-registration of ASL MRI to
the corresponding structural MRI data by fluid registration algorithm to correct for brain atrophy and gray/white
matter partial volume effects; 2) intensity normalization to reduce distortions from the RF bias field, and 3)
intensity calibration to the global mean value of perfusion to account for global variations in cerebral blood flow
from scan to scan. In brief, the perfusion-weighted image is stepwise co-registered and interpolated to
MPRAGE using affine transformations by first registering separately the labeled and unlabeled ASL MRI
images (which are T2-weighted) to the GRE-weighted image and then registering the GRE image to MPRAGE.
Once the images are co-registered, a perfusion-weighted image is calculated by subtracting the labeled from
unlabeled ASL images. To reduce signal bias from potential subtraction errors in ASL, histograms are
computed and the distribution of negative signal components (which derive from subtraction errors and noise)
are used to estimate a Gaussian noise distribution which is then used to estimate the likelihood of a bias free
ASL signal. Furthermore, the perfusion-weighted image is normalized to the corresponding ASL reference
scan signal without background suppression to account for B1-field inhomogeneity and other instrumental
variations. The ASL signal is then further corrected for partial volume effects (PVE) to account for underlying
structural variations in brain atrophy and in gray/white matter voxel
composition using information from the probabilistic tissue
segmentation maps of MPRAGE co-registered to ASL maps. The
rationale for PVE correction is to obtain an ASL signal that is largely
independent of structural alterations.
The final outcome is a perfusion map per unit gray matter for each
individual in the native space of MPRAGE. In addition, the
corresponding uncorrected perfusion map will be made available also
to allow the application of other PVE correction schemes.
Representative individual perfusion maps (4 Tesla) of gray matter
derived with this processing scheme are shown in (Figure 3.4.4.).
3.4.3.8.
Combined Multi-Modality Analytic Methods: An
additional feature of image analysis will be the integrated analysis of
different imaging modalities. An example of this approach is
illustrated here (Figure 3.4.5.) from Norbert Schuff where ASL
Fig. 3.4.5. Multi-Modality Analysis of
perfusion (A) is mapped on the cortical surface (developed by Dr.
Perfusion and Structural MRI to
Duygu Tosun in the CIND group) and perfusion and structural MRI
Predict Cognitive Status
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
(B) are then directly compared in terms of their predictive value of cognitive status (MMSE).
The unifying feature across perfusion and structural MRI (and in principle also FDG PET) is that we expect
longitudinal changes in aging, MCI and AD to occur in the same regions in all these image modalities. Because
we have specific hypotheses about the regions of greatest change, region of interest (ROI)-based techniques
will be applied in addition to regionally unbiased voxel-wise methods. ROI-based analyses will be
accomplished by spatially transforming the images from different modalities into the same space as each
individual’s MPRAGE after FS processing so that the individual FS parcellation maps can be directly overlaid
on the images. Furthermore, the interpolation of the low resolution perfusion images to the same MPRAGE
resolution, which is part of the spatial normalization process, will ease complications of consistently tracing
ROIs on images with different resolutions. For voxel-wise analyses, perfusion, PET and structural MRI data will
be spatial normalization to MNI brain atlas space using non-linear transformations.
We will test separately ROI and voxel-wise change of perfusion over time using an in-house developed
software package written in R (http://www.r-project.org/) [45]. This program utilizes the general linear model
statistical libraries in R as well as mixed effects statistical libraries to account for within subject variations over
time and the AnalyzeFMRI fdrtool tools in R. Among other applications, our package provides for voxel-based
uni- and multi-variate analysis of covariance. While there are currently a number of approaches to addressing
the multiple comparison problem that arises when testing a large number of voxels, including Random Field
Theory-based and non-parametric methods, the most convenient in the context of uni-/multi-variate analysis in
the R statistical package is to use a false discovery rate (FDR) threshold via the fdrtool tool [46]. While the
different methods constitute statistical tests with different interpretations and should each be considered in the
particular context of a given study, it seems reasonable to assume that an experimental multi-center study of
ASL data would be relatively insensitive to this choice.
We will also perform exploratory tests comparing ASL perfusion, FDG PET and structural MRI in terms of
their magnitude and variance in longitudinal change. We will use for these tests a non-parametric statistical
method, which our group developed previously [43] to identify cross-modality relationships (e.g. between
perfusion and atrophy) without need to explicitly model the relations. In brief, this test performs a statistic on
the T-values of each image modality, for example on perfusion and gray matter loss. The analysis includes
first, the design of combining functions of concordance or discordance between two modalities, followed by
permutation tests to obtain the distribution of each combining function and to perform a statistic. This approach
will test regionally to which extent are the variations in two modalities, e.g. perfusion and PET or perfusion and
gray matter atrophy, concordant or discordant.
3.4.3.9.
Resting State Functional Connectivity: An illustration of one possible approach to ASL analysis
is provided by Drs. Prashanthi Vemuri and Clifford Jack here. Pre-processing: The main steps involved in preprocessing fMRI data are removal of time dependent drifts of the fMRI signal, elimination of the first three
slices to allow for acquisition signal to reach steady state and inter-frame motion correction within the time
series of each subject using a six-degrees-of-freedom co-registration. Each patient’s pre-processed fMRI scan
will be co-registered (based on their first frame) to their 3D T1 scan and then be co-registered to the Talairach
atlas. Additional steps that will be applied to remove spurious signals that might affect the neuronal activity are:
low pass filtering images from 0.01 Hz to 0.1 Hz, global mean signal removal, removing signal from CSF and
white matter and spatially smoothing the images at 8 mm FWHM. Resting State Functional Connectivity
(RSFC) Analysis: A sphere of 12 mm will be placed on each of the already published nodes of the main
networks. A recent paper [47] compiled all the published and well-established nodes into a single table (Table
3.4). To re-check the nodes, we will run a group ICA analysis (using the FSL MELODIC group ICA toolbox:
http://www.fmrib.ox.ac.uk/fsl/melodic/index.html) and look at each of the seven networks in the independent
components and re-center the published nodes if necessary for our data. Correlation coefficients will be
computed between each of the nodes in each network for all subjects and will be z-transformed.
3.4.4. Statistical Analysis Methods: As presented in the Specific Aims for this Core, the hypotheses are
grouped into three categories: 1) hypothesis replication of ADNI1; 2) new hypotheses related to core MRI
protocol sequences; and 3) new hypotheses related to experimental sequences in protocol. Most of the
hypotheses address comparisons across different structural MRI methods on sample size, correlation with
cognitive change, or association with conversion from MCI to AD (for example, hypotheses 1-3 in Category
(1)), comparisons of measures or change in measures across diagnostic groups (for example, hypothesis 5 in
Category (1) and hypothesis 4 in Category (2)), baseline associations between MRI measures and clinical
outcomes (for example, hypotheses 1-2 in Category (3)), associations between baseline levels of an MRI
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
measure and cognitive decline (for example, hypothesis 6 in Category (1)), correlations between change in
MRI measures and change in other imaging or cognitive function measures (for example, hypotheses 3-4 in
Category (3)), or associations with clinical progression (for example, hypothesis 5 in Category (3)). We briefly
state below the methods to be used for each of these types of analyses, but more details may be found in the
Biostatistics Core.
A standardized framework for comparing different fluid and imaging biomarkers on a set of criteria, including
precision to measure change (related to sample size calculations) as well as clinical validity (correlation with
cognitive decline or clinical progression), was developed for ADNI1 and will be used to address all hypotheses
related to comparing across MRI measures or across MRI measures and CSF biomarkers. Analysis of
variance (ANOVA) techniques will be used to compare levels or change in measures across diagnostic groups.
Linear regression models will be used to assess correlations between baseline levels of an MRI measure and
baseline levels of another measure, such as cognitive function, other imaging measures or CSF biomarkers.
Mixed effects regression models will be used to assess associations between baseline level of an MRI
measure and cognitive decline. An extension of mixed effects regression models, called simultaneous random
effects models, which allows for multiple types of longitudinal outcomes, will assess correlations between
change in two (or more) different markers. Finally, survival models that account for interval censoring will be
used to assess associations with clinical progression.
ADNI2 will follow three cohorts of subjects with serial longitudinal imaging exams. 1) roughly 473 CN and
LMCIs carried forward from ADNI1, 2) 200 EMCI enrolled in GO and carried forward into ADNI2, 3) 150
subjects newly enrolled in ADNI2 in each of the following groups CN, MCI, LMCI, AD. With these numbers of
subjects, we should have ample power to address all of our hypotheses (please see Biostatistics Core for the
specific power calculations.
3.5. Literature Cited:
1.
Ho, A.J., X. Hua, S. Lee, A.D. Leow, I. Yanovsky, B. Gutman, I.D. Dinov, N. Lepore, J.L. Stein, A.W.
Toga, C.R. Jack, Jr., M.A. Bernstein, E.M. Reiman, D.J. Harvey, J. Kornak, N. Schuff, G.E. Alexander,
M.W. Weiner, and P.M. Thompson, Comparing 3 T and 1.5 T MRI for tracking Alzheimer's disease
progression with tensor-based morphometry. Hum Brain Mapp, 2009.
2.
Vernooij, M.W., M.A. Ikram, P.A. Wielopolski, G.P. Krestin, M.M. Breteler, and A. van der Lugt,
Cerebral microbleeds: accelerated 3D T2*-weighted GRE MR imaging versus conventional 2D T2*weighted GRE MR imaging for detection. Radiology, 2008. 248(1): p. 272-7.
3.
Clarkson, M.J., S. Ourselin, C. Nielsen, K.K. Leung, J. Barnes, J.L. Whitwell, J.L. Gunter, D.L. Hill,
M.W. Weiner, C.R. Jack, Jr., and N.C. Fox, Comparison of phantom and registration scaling corrections
using the ADNI cohort. Neuroimage, 2009. 47(4): p. 1506-13.
4.
Gunter, J.L., M.A. Bernstein, B.J. Borowski, C.P. Ward, P.J. Britson, J.P. Felmlee, N. Schuff, M.
Weiner, and C.R. Jack, Measurement of MRI scanner performance with the ADNI phantom. Med Phys,
2009. 36(6): p. 2193-205.
5.
Leow, A.D., A.D. Klunder, C.R. Jack, Jr., A.W. Toga, A.M. Dale, M.A. Bernstein, P.J. Britson, J.L.
Gunter, C.P. Ward, J.L. Whitwell, B.J. Borowski, A.S. Fleisher, N.C. Fox, D. Harvey, J. Kornak, N.
Schuff, C. Studholme, G.E. Alexander, M.W. Weiner, and P.M. Thompson, Longitudinal stability of MRI
for mapping brain change using tensor-based morphometry. Neuroimage, 2006. 31(2): p. 627-40.
6.
Boyes, R.G., J.L. Gunter, C. Frost, A.L. Janke, T. Yeatman, D.L. Hill, M.A. Bernstein, P.M. Thompson,
M.W. Weiner, N. Schuff, G.E. Alexander, R.J. Killiany, C. DeCarli, C.R. Jack, and N.C. Fox, Intensity
non-uniformity correction using N3 on 3-T scanners with multichannel phased array coils. Neuroimage,
2008. 39(4): p. 1752-62.
7.
Hua, X., S. Lee, I. Yanovsky, A.D. Leow, Y.Y. Chou, A.J. Ho, B. Gutman, A.W. Toga, C.R. Jack, Jr.,
M.A. Bernstein, E.M. Reiman, D.J. Harvey, J. Kornak, N. Schuff, G.E. Alexander, M.W. Weiner, and
P.M. Thompson, Optimizing power to track brain degeneration in Alzheimer's disease and mild
cognitive impairment with tensor-based morphometry: an ADNI study of 515 subjects. Neuroimage,
2009. 48(4): p. 668-81.
8.
Vemuri, P., H.J. Wiste, S.D. Weigand, L.M. Shaw, J.Q. Trojanowski, M.W. Weiner, D.S. Knopman, R.C.
Petersen, and C.R. Jack, Jr., MRI and CSF biomarkers in normal, MCI, and AD subjects: Diagnostic
discrimination and cognitive correlations. Neurology, 2009. 73(4): p. 287-293.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
9.
Petersen, and C.R. Jack, Jr., MRI and CSF biomarkers in normal, MCI, and AD subjects: Predicting
future clinical change. Neurology, 2009. 73(4): p. 294-301.
10.
Schuff, N., N. Woerner, L. Boreta, T. Kornfield, L.M. Shaw, J.Q. Trojanowski, P.M. Thompson, C.R.
Jack, Jr., and M.W. Weiner, MRI of hippocampal volume loss in early Alzheimer's disease in relation to
ApoE genotype and biomarkers. Brain, 2009. 132(Pt 4): p. 1067-77.
11.
Morra, J.H., Z. Tu, L.G. Apostolova, A.E. Green, C. Avedissian, S.K. Madsen, N. Parikshak, X. Hua,
A.W. Toga, C.R. Jack, Jr., N. Schuff, M.W. Weiner, and P.M. Thompson, Automated 3D mapping of
hippocampal atrophy and its clinical correlates in 400 subjects with Alzheimer's disease, mild cognitive
impairment, and elderly controls. Hum Brain Mapp, 2009. 30(9): p. 2766-88.
12.
Stein, J.L., X. Hua, J.H. Morra, S. Lee, A.J. Ho, A.D. Leow, A.W. Toga, J. Sul, H.M. Kang, E. Eskin,
A.J. Saykin, L. Shen, T. Foroud, N. Pankratz, M.J. Huentelman, D.W. Craig, J.D. Gerber, A. Allen, J.
Corneveaux, D.A. Stephan, J. Webster, B.M. DeChairo, S.G. Potkin, C.R. Jack, Jr., and M.W. Weiner,
Genome-wide association study of temporal lobe structure identifies novel quantitative trait loci for
neurodegeneration in
alzheimer's disease. Neuron, 2009.
13.
Ho, A.J., J.L. Stein, X. Hua, S. Lee, D.P. Hibar, A.D. Leow, I.D. Dinov, A.W. Toga, A.J. Saykin, L. Shen,
T. Foroud, N. Pankratz, M.J. Huentelman, D.W. Craig, J.D. Gerber, A. Allen, J. Corneveaux, D.A.
Stephan, J. Webster, B.M. DeChairo, S.G. Potkin, C.R. Jack, Jr., M.W. Weiner, C.A. Raji, O.L. Lopez,
J.T. Becker, and P.M. Thompson, Commonly carried allele within FTO, an obesity-associated gene,
relates to accelerated brain degeneration in the elderly. PNAS, 2009.
14.
Stein, J.L., X. Hua, S. Lee, A.J. Ho, A.D. Leow, A.W. Toga, A.J. Saykin, L. Shen, T. Foroud, N.
Pankratz, M.J. Huentelman, D.W. Craig, J.D. Gerber, A. Allen, J. Corneveaux, D.A. Stephan, J.
Webster, B.M. DeChairo, S.G. Potkin, C.R. Jack, Jr., M.W. Weiner, and P.M. Thompson, Voxelwise
Genome-Wide Association Study (vGWAS). Neuroimage, 2009.
15.
Hua, X., A.D. Leow, N. Parikshak, S. Lee, M.C. Chiang, A.W. Toga, C.R. Jack, Jr., M.W. Weiner, and
P.M. Thompson, Tensor-based morphometry as a neuroimaging biomarker for Alzheimer's disease: an
MRI study of 676 AD, MCI, and normal subjects. Neuroimage, 2008. 43(3): p. 458-69.
16.
Leow, A.D., I. Yanovsky, M.C. Chiang, A.D. Lee, A.D. Klunder, A. Lu, J.T. Becker, S.W. Davis, A.W.
Toga, and P.M. Thompson, Statistical properties of Jacobian maps and the realization of unbiased
large-deformation nonlinear image registration. IEEE Trans Med Imaging, 2007. 26(6): p. 822-32.
17.
Leow, A.D., I. Yanovsky, N. Parikshak, X. Hua, S. Lee, A.W. Toga, C.R. Jack, Jr., M.A. Bernstein, P.J.
Britson, J.L. Gunter, C.P. Ward, B. Borowski, L.M. Shaw, J.Q. Trojanowski, A.S. Fleisher, D. Harvey, J.
Kornak, N. Schuff, G.E. Alexander, M.W. Weiner, and P.M. Thompson, Alzheimer's disease
neuroimaging initiative: a one-year follow up study using tensor-based morphometry correlating
degenerative rates, biomarkers and cognition. Neuroimage, 2009. 45(3): p. 645-55.
18.
Hua, X., A.D. Leow, S. Lee, A.D. Klunder, A.W. Toga, N. Lepore, Y.Y. Chou, C. Brun, M.C. Chiang, M.
Barysheva, C.R. Jack, Jr., M.A. Bernstein, P.J. Britson, C.P. Ward, J.L. Whitwell, B. Borowski, A.S.
Fleisher, N.C. Fox, R.G. Boyes, J. Barnes, D. Harvey, J. Kornak, N. Schuff, L. Boreta, G.E. Alexander,
M.W. Weiner, P.M. Thompson, and I. Alzheimer's Disease Neuroimaging, 3D characterization of brain
atrophy in Alzheimer's disease and mild cognitive impairment using tensor-based morphometry.
Neuroimage, 2008. 41(1): p. 19-34.
19.
Hua, X., I. Yanovsky, A.D. Leow, S. Lee, A.J. Ho, N. Parikshak, A.W. Toga, C.R. Jack, Jr., M.W.
Weiner, and P.M. Thompson, Tensor based morphometry as surrogate marker for Alzheimer's disease
and mild cognitive impairment: Optimizing Statistical Power, in Organization for Human Brain Mapping.
2009.
20.
Fischl, B., D.H. Salat, E. Busa, M. Albert, M. Dieterich, C. Haselgrove, A. van der Kouwe, R. Killiany, D.
Kennedy, S. Klaveness, A. Montillo, N. Makris, B. Rosen, and A.M. Dale, Whole brain segmentation:
automated labeling of neuroanatomical structures in the human brain. Neuron, 2002. 33(3): p. 341-55.
21.
Lewis, E.B. and N.C. Fox, Correction of differential intensity inhomogeneity in longitudinal MR images.
Neuroimage, 2004. 23(1): p. 75-83.
22.
Ridha, B.H., J. Barnes, J.W. Bartlett, A. Godbolt, T. Pepple, M.N. Rossor, and N.C. Fox, Tracking
atrophy progression in familial Alzheimer's disease: a serial MRI study. Lancet Neurol, 2006. 5(10): p.
828-34.
PHS 398/2590 (Rev. 11/07)
Page
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
Weiner, Michael
Schott, J.M., S.L. Price, C. Frost, J.L. Whitwell, M.N. Rossor, and N.C. Fox, Measuring atrophy in
Alzheimer disease: a serial MRI study over 6 and 12 months. Neurology, 2005. 65(1): p. 119-24.
Barnes, J., R.G. Boyes, E.B. Lewis, J.M. Schott, C. Frost, R.I. Scahill, and N.C. Fox, Automatic
calculation of hippocampal atrophy rates using a hippocampal template and the boundary shift integral.
Neurobiol Aging, 2007. 28(11): p. 1657-63.
Crum, W.R., R.I. Scahill, and N.C. Fox, Automated hippocampal segmentation by regional fluid
registration of serial MRI: validation and application in Alzheimer's disease. Neuroimage, 2001. 13(5):
p. 847-55.
Barnes, J., J. Foster, R.G. Boyes, T. Pepple, E.K. Moore, J.M. Schott, C. Frost, R.I. Scahill, and N.C.
Fox, A comparison of methods for the automated calculation of volumes and atrophy rates in the
hippocampus. Neuroimage, 2008. 40(4): p. 1655-71.
Vemuri, P., J.L. Gunter, M.L. Senjem, J.L. Whitwell, K. Kantarci, D.S. Knopman, B.F. Boeve, R.C.
Petersen, and C.R. Jack, Jr., Alzheimer's disease diagnosis in individual subjects using structural MR
images: validation studies. Neuroimage, 2008. 39(3): p. 1186-97.
Vemuri, P., J.L. Whitwell, K. Kantarci, K.A. Josephs, J.E. Parisi, M.S. Shiung, D.S. Knopman, B.F.
Boeve, R.C. Petersen, D.W. Dickson, and C.R. Jack, Jr., Antemortem MRI based STructural
Abnormality iNDex (STAND)-scores correlate with postmortem Braak neurofibrillary tangle stage.
Neuroimage, 2008. 42(2): p. 559-67.
Schwarz, C.G., E. Fletcher, C. DeCarli, and O.T. Carmichael. Fully-Automated White Matter
Hyperintensity Detection with Anotomical Prior Knowledge and without FLAIR. in Information
Processing in Medical Imaging (IPMI). 2009. Williamsburg, VA.
Lenglet, C., J.S. Campbell, M. Descoteaux, G. Haro, P. Savadjiev, D. Wassermann, A. Anwander, R.
Deriche, G.B. Pike, G. Sapiro, K. Siddiqi, and P.M. Thompson, Mathematical methods for diffusion MRI
processing. Neuroimage, 2009. 45(1 Suppl): p. S111-22.
Kim, Y., P.M. Thompson, A.W. Toga, L. Vese, and L. Zhan, HARDI denoising: variational regularization
of the spherical apparent diffusion coefficient sADC. Inf Process Med Imaging, 2009. 21: p. 515-27.
Chiang, M.C., A.D. Leow, A.D. Klunder, R.A. Dutton, M. Barysheva, S.E. Rose, K.L. McMahon, G.I. de
Zubicaray, A.W. Toga, and P.M. Thompson, Fluid registration of diffusion tensor images using
information theory. IEEE Trans Med Imaging, 2008. 27(4): p. 442-56.
Leow, A.D., S. Zhu, L. Zhan, K. McMahon, G.I. de Zubicaray, M. Meredith, M.J. Wright, A.W. Toga, and
P.M. Thompson, The tensor distribution function. Magn Reson Med, 2009. 61(1): p. 205-14.
Zhan, L., A.D. Leow, S. Zhu, M.C. Chiang, M. Barysheva, A.W. Toga, K. McMahon, G. De Zubicaray,
M.J. Wright, and P.M. Thompson, Analyzing Multi-Fiber Reconstruction in High Angular Resulution
Diffusion Imaging using the Tensor Distribution Function, in Imaginh Science and Biomedical Imaging
(ISBI2009). 2009: Boston, MA. p. 4.
Patel, V., Y. Shi, P.M. Thompson, and A.W. Toga, Mesh-Based Spherical Deconvolution for Physically
Valid Fiber Orientation Reconstruction via Diffusion Weighted MRI, in Imaaging Science and
Biomedical Imaging (ISBI2009). 2009: Boston, MA. p. 4.
Hageman, N., A.D. Leow, D.W. Shattuck, L. Zhan, and P.M. Thompson, Segmenting Crossing Fiber
Geometries using Fluid Mechanics Tensor Distribution Function Tractography, in Imaging Science and
Biomedical Imaging. 2009: Boston, MA. p. 4.
Lee, A.D., N. Lepore, C.C. Brun, M. Barysheva, Y.Y. Chou, M.C. Chiang, S.K. Madsen, K. McMahon,
G. De Zubicaray, M.J. Wright, A.W. Toga, and P.M. Thompson, The multivariate A/C/E model and the
genetics of fiber architecture, in Imaging Science and Biomedical Imaging (ISBI2009). 2009: Boston,
MA. p. 4.
Lee, A.D., N. Lepore, C.C. Brun, Y.Y. Chou, M. Barysheva, M.C. Chiang, S.K. Madsen, G. De
Zubicaray, K. McMahon, M.J. Wright, A.W. Toga, and P.M. Thompson, Tensor-Based Analysis of
Genetic Influences on Brain Integrity using DTI in 100 Twins, in Medical Image Computing and
Computer Assisted Intervention (MICCAI2009). 2009: London, UK. p. 8.
Goh, A., C. Lenglet, P.M. Thompson, and R. Vidal, A nonparametric reimannian framework for
processing high angular resolution diffusion images (HARDI), in Computer Vision and Pattern
Recognition (CVPR) 2009. 2009: Miami Beach, FL.
Goh, A., C. Lenglet, P.M. Thompson, and R. Vidal, Estimating orientation distributions with probability
density constraints and spatial regularity
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
in Medical Image Computer Assisted Intervention (MICCAI2009). 2009: London, UK. p. 8.
41.
Jahanshad, N., A.D. Lee, Y.Y. Chou, N. Lepore, C.C. Brun, M. Barysheva, A.W. Toga, K. McMahon, G.
De Zubicaray, M.J. Wright, G. Sapiro, C. Lenglet, and P.M. Thompson, Reducing structural variation to
determine the genetics of white matter integrity across hemispheres - a DTI study of 100 twins, in
Imaging Science and Biomedical Imaging (ISBI2009). 2009: Boston, MA. p. 4.
42.
Du, A.T., G.H. Jahng, S. Hayasaka, J.H. Kramer, H.J. Rosen, M.L. Gorno-Tempini, K.P. Rankin, B.L.
Miller, M.W. Weiner, and N. Schuff, Hypoperfusion in frontotemporal dementia and Alzheimer disease
by arterial spin labeling MRI. Neurology, 2006. 67(7): p. 1215-20.
43.
Hayasaka, S., A.T. Du, A. Duarte, J. Kornak, G.H. Jahng, M.W. Weiner, and N. Schuff, A nonparametric approach for co-analysis of multi-modal brain imaging data: application to Alzheimer's
disease. Neuroimage, 2006. 30(3): p. 768-79.
44.
Johnson, N.A., G.H. Jahng, M.W. Weiner, B.L. Miller, H.C. Chui, W.J. Jagust, M.L. Gorno-Tempini, and
N. Schuff, Pattern of cerebral hypoperfusion in Alzheimer disease and mild cognitive impairment
measured with arterial spin-labeling MR imaging: initial experience. Radiology, 2005. 234(3): p. 851-9.
45.
Young, K., V. Weber, K. Govindaraju, A.D. Sharma, C. Studholma, and L. Hall, Multivariate Statistical
Mapping of Spectroscopic Imaging Data in Annual Scientific Meeting of the International Society of
Magnetic Resonance in Medicine. 2008: Toronto.
46.
Strimmer, K., fdrtool: a versatile R package for estimating local and tail area-based false discovery
rates. Bioinformatics, 2008. 24(12): p. 1461-2.
47.
Johnston, J.M., S.N. Vaishnavi, M.D. Smyth, D. Zhang, B.J. He, J.M. Zempel, J.S. Shimony, A.Z.
Snyder, and M.E. Raichle, Loss of resting interhemispheric functional connectivity after complete
section of the corpus callosum. J Neurosci, 2008. 28(25): p. 6453-8.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Core: 4
Title of Core (not to exceed 81 spaces): Positron Emission Tomography (PET) Core
Core Leader: Jagust, William, J.
Position/Title: Professor, University of California, Berkeley
Department, service, laboratory, or equivalent: Neuroscience, Public Health
Mailing Address:
50 University Hall #7360
Berkeley, CA 94720
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
The PET core of the Alzheimer's Disease Neuroimaging Initiative proposes several new research goals
based on advances in biomarker research First, a primary advance in PET imaging has been the recent
development and growth of tracers to image fibrillar forms of the A-beta peptide. While there is considerable
experience with the tracer [11C]PIB (Pittsburgh compound B) both in ADNI and worldwide, the requirement
of a cyclotron and radiochemistry program has limited the clinical dissemination of this technology.
However, the development of new ligands labeled with 18F affords the possibility of large scale amyloid
imaging. While several potential ligands are available, the PET core has chosen to use [18F]AV45 for a
variety of reasons defined in the application. One new goal is therefore the acquisition of AV45-PET on all
ADNI participants. A second new direction results from the realization that early detection of AD with
biomarkers is required to select subjects who would benefit most from treatment and to enrich clinical trials
with asymptomatic or mildly symptomatic subjects at risk. Therefore the PET core will image all early MCI
subjects. Finally, it is increasingly clear that different biomarkers change differently at different points in the
disease, and that markers of A-beta may provide different information than biomarkers of
neurodegeneration. Therefore, all subjects will receive an FDG-PET scan in association with the AV45-PET
scan so that changes in glucose metabolism can be tracked concurrently with amyloid deposition. Thus, by
the end of this 5 year project all participants will have had amyloid and glucose metabolic imaging at 2 time
points. The core will modify existing techniques developed in the first phase of the project to standardize,
acquire, quality check, process, and analyze all PET data and will make images and numerical summary
data available to the scientific community. Specific aims to be evaluated include establishing the prevalence
of amyloid positivity in different groups, particularly early MCI; defining the relationships between PET
measures of amyloid and other biomarkers, especially those that track neurodegeneration; and exploring the
relationships between these biomarkers and clinical measures of disease severity and progression.
The use of biomarkers in AD has major importance both for the detection of disease and for the prediction of
decline in those at risk. This approach to the development of biomarkers may help speed clinical trials and
identify suitable subjects for therapy.
The Regents of the University of California
DUNS:
12-472-6725
Street 1:
118 Baker Hall
City:
Street 2:
Berkeley
County:
Province:
Alameda
USA
CA-009
Country:
State:
Zip/Postal Code:
CA
94720
The University of Utah
DUNS:
D&B # 009095365
Street 1:
75 South 2000 East, Rm 211
City:
Street 2:
Salt Lake City
Province:
PHS 398 (Rev. 11/07)
County:
Country:
Salt Lake County
USA
State:
Zip/Postal Code:
UT
84112
UT-002
Page 2
Form Page 2
Weiner, Michael
Name
Organization
Role on Project
Chen, Kewei
keweichen
Banner Alzheimer's
Co-Investigator
Foster, Norman
nlfoster
University of Utah
Co-Investigator
Jagust, William
wjjagust
UC-Berkeley
PI
Koeppe, Robert
koeppe
University of Michigan
Co-Investigator
Mathis, Chester
mathis
Co-Investigator
Price, Julie
pricejc
Co-Investigator
Reiman, Eric
ereiman
Banner Alzheimer's
Co-Investigator
Name
Organization
Role on Project
No
Yes
Cell Line
PHS 398 (Rev. 11/07)
Page 3
Weiner, Michael W.
Regents of the University of Michigan
DUNS:
073133571
Street 1:
3003 S State St
City:
Street 2:
Ann Arbor
Province:
County:
Country:
Washtenaw
USA
State:
Zip/Postal Code:
MI
48109
MI‐015
DUNS:
004514360
Street 1:
ADRC
City:
Street 2:
Pittsburgh
Province:
4 West UPMC Montefiore, 200 Lothrop Street
County:
Country:
State:
USA
Zip/Postal Code:
PA
15213
DUNS:
788240674
Street 1:
901 East Willetta St
City:
Street 2:
Phoenix Province:
County:
Country:
Maricopa
USA
State:
Zip/Postal Code:
AZ
85006
AZ‐004
DUNS:
Street 1:
City:
Street 2:
Province:
County:
Country:
State:
Zip/Postal Code:
State:
Zip/Postal Code:
DUNS:
Street 1:
City:
Street 2:
Province:
PHS 398/2590 (Rev. 11/07)
County:
Country:
Page 1
Weiner, MW
4.0 POSITRON EMISSION TOMOGRAPHY (PET) CORE
4.1 SPECIFIC AIMS
The initial phase of the PET core of ADNI (ADNI1) was primarily aimed at establishing methods for
standardizing image acquisition and processing in order to investigate FDG-PET measures of glucose
metabolism as a potential outcome measure in clinical trials. During ADNI1, ADNI investigators became
interested in several new lines of research, three of which are incorporated in this renewal (ADNI2). The first
was the idea that clinical trials required more mildly affected subjects, with the consequence that we began to
consider ADNI2 as a project to develop biomarkers for the selection of very mildly impaired individuals for a
clinical trial. This resulted in the concept of early MCI (EMCI) as articulated in the clinical core. The corollary
of this goal for the PET core of ADNI2 is thus the use of PET to predict decline in the “traditional” study groups
(Normals, MCI, which we will refer to as “late MCI” or LMCI, and AD) as well as in very mildly affected patients
with EMCI. The ability to use PET as a predictor of decline in clinical settings could help move clinical trials to
earlier disease stages by using PET to select subjects with a high risk of subsequent cognitive decline, and
who would benefit from early treatment and comprise reasonable subjects in whom to test new therapies.
A second major direction resulted from the growth of amyloid imaging during ADNI1. At the start of ADNI1
(in 2004), this imaging modality was limited to only two very new tracers with human application: [18F]FDDNP
and [11C]PIB. For a number of scientific and logistical reasons we proceeded with an “add on” study of PIB
that recruited 103 subjects at 14 centers. While progress has been encouraging, the need for a local
cyclotron and radiochemistry program limits the ability to study a large cohort with this ligand. Since the
initiation of ADNI, however, a number of ligands labeled with the longer half-life 18F have become available.
For reasons discussed below we have decided to move forward with the compound manufactured by Avid
Radiopharmaceuticals, AV45, to perform amyloid imaging in all participants. Thus, our work with PIB showing
the technical feasibility of multi-site amyloid imaging and a growing body of evidence from the scientific
community have motivated our aims for ADNI2 to evaluate amyloid imaging as a biomarker in AD. This work
has, in fact, already begun, as ADNI investigators have received a Grand Opportunity (GO) grant to perform
AV45 imaging in currently enrolled control and MCI subjects along with a new group of EMCI subjects.
Finally, the data that have been collected in ADNI1 and other studies in laboratories around the world
have convinced us that the selection of a biomarker for AD is a complex question. In essence, it is clear that
biomarkers change differently at different stages of the disease, and may therefore be differentially useful in
diagnosis, prediction of decline, or measurement of outcomes depending on where in the disease course one
looks. This fundamental observation has motivated some of the questions in ADNI2 aimed at relating
changes in biomarkers to one another and to clinical symptoms at different stages of the disease.
Thus, the new features of this application that represent overarching goals are:
• The recruitment of cases of early MCI (EMCI) by the clinical core
• Use of PET, especially amyloid imaging, to characterize subjects and predict decline
• Evaluation of the differential utility and complementarity of different biomarkers in AD
The specific aims of ADNI2 can be separated into the operational/technical features that will provide useful
data for investigators worldwide, and scientific aims that will be tested by core investigators.
Operational/technical specific aims are:
1. Acquire standardized AV45-PET and FDG-PET scans on all individuals in the cohort at 2 timepoints, so
that all individuals in ADNI have 2 sets of scans taken 2 years apart.
2. Apply standardized quality control and processing to assure quality and comparability of all PET data.
3. Perform basic data analysis on all AV45 and FDG-PET data for regional measurement of brain amyloid
burden and glucose metabolism using several different and complementary approaches.
4. Maintain an administrative infrastructure for these operational tasks as well as interactions with other cores
and site PIs, and provide all images and data to the scientific community.
Scientific specific aims are:
5. Define the proportion of subjects in each clinical diagnostic group with evidence of significant accumulation
of fibrillar Aβ.
Hypothesis: Amyloid positivity will occur in approximately 30% of normals, 45% of EMCI, 60% of LMCI
and 90% of AD subjects recruited in this clinical trial.
6. Evaluate the relationship between AV45, FDG, and disease severity at baseline in each diagnostic group
Hypothesis: Baseline measures of FDG, but not AV45 will be associated with measures of disease
severity within diagnostic groups. A greater proportion of subjects with mild or absent symptoms
PHS 398/2590 (Rev. 11/07)
Weiner, MW
(normals, EMCI, LMCI) will show evidence of Aβ accumulation with AV45 than will show evidence of
glucose hypometabolism.
7. Evaluate the relationship between AV45, FDG, and disease severity over time in each clinical group.
Hypothesis: Baseline measures of Aβ with AV45 will be predictive of subsequent clinical decline in mildly
affected subjects (Normals, EMCI and LMCI). FDG measures of glucose metabolism will be predictive of
cognitive decline in all subjects. FDG will be a better predictor of decline in AD than AV45, the reverse will
hold for normal subjects and MCI patients. The two tracers together will have greater power to predict
decline than either tracer alone. Longitudinal changes in glucose metabolism will be correlated with
change in clinical symptoms, while longitudinal change in AV45 uptake will not.
8. Compare FDG and AV45-PET as outcome measures in clinical trials by estimating sample sizes
necessary to detect different sized treatment effects.
Hypothesis: Power analyses for FDG-PET in newly recruited MCI and AD patients will replicate results
obtained in ADNI1 using both pre-specified and data-driven (statistical) ROIs. Estimates of sample sizes
required to detect changes in AV45 will be greater than those required for FDG-PET.
9. Evaluate the use of FDG and AV45-PET for selecting MCI subjects in clinical trials
Hypothesis: Selection of EMCI or LMCI subjects using FDG- or AV45-PET will reduce the sample size
and time of follow up necessary to detect treatment effects in clinical trials. Use of the two tracers
together will further reduce sample sizes.
10. Compare changes in PET biomarkers to changes in other biomarkers and cognition, and compare the
predictive values of PET biomarkers to MRI and CSF variables.
Hypothesis: Longitudinal change and predictive value of AV45 will be similar to CSF measures of Aβ42,
while longitudinal changes and predictive value of FDG-PET will be similar to changes in MRI variables
and cognitive variables.
11. Compare different methods for assessing amyloid positivity, and for processing AV45-PET images, across
different laboratories.
Hypothesis: Regional and statistically-defined ROI measures of AV45 tracer uptake will be superior to
global measures at differentiating groups and identifying decliners.
4.2 BACKGROUND AND SIGNIFICANCE
4.2.1 Why biomarkers?
The application of biomarkers to the study of AD has entered a field of rapid growth for several different
reasons. One major reason is the availability of new technology. This includes methods for accurately
measuring biomarkers in CSF[1], and new ways of acquiring and analyzing structural MRI data [2, 3] both of
which are discussed in greater detail in the relevant cores. With regard to PET, a major advance has included
both the development of data analytic approaches [4,5] and, more importantly, availability of new tracers such
as [11C]PIB (Pittsburgh Compound B) for the quantitation of brain beta-amyloid (Aβ)[6].
Another major reason for the interest in biomarkers is the potential for disease-modifying therapies.
Although the subject is too broad to review here, many new approaches aimed at either Aβ or downstream
effects of this molecule have been proposed that might slow or halt the progression of AD [7-9]. For a host of
reasons including the failure of several large potentially disase-modifying clinical trials, many have suggested
that treatment must be initiated very early in the disease to have an effect[10]. One solution to this problem is
the use of biomarkers to help identify subjects for clinical trials, especially those with very mild symptoms[11].
4.2.2 Which biomarker?
It has become apparent that a single biomarker is not adequate for all purposes. A biomarker that is
effective for diagnosis may not be useful in predicting decline, and a biomarker that can predict decline may
not be effective in monitoring progression. Thus the hypotheses we have articulated in the PET core
specifically distinguish amongst these different uses of a biomarker. Although diagnosis is of some interest in
ADNI, the study design is not really targeted to validation of PET as a diagnostic biomarker since subjects are
not selected with a full range of dementia presentations but rather more reflect the more restricted ranges of
cases recruited to clinical trials. Nevertheless, some information on diagnostic accuracy will be obtained
when neuropathological diagnoses become available.
As one example of how biomarkers differ, consider the fact that both CSF and PET measures of Aβ are
highly congruent [12, 13], show evidence of Aβ deposition in a high proportion of normal individuals[14], and
appear to be reasonably good predictors of cognitive decline in patients with MCI or mild AD[15-19]. In
PHS 398/2590 (Rev. 11/07)
Weiner, MW
contrast, the evidence that levels of Aβ in patients with dementia are related to either cross-sectional or
longitudinal measures of dementia severity, is weak or negative, while changes in other biomarkers such as
FDG or MRI measures of atrophy are better correlates of severity in established disease [20, 21]. Thus, a
marker for Aβ might turn out to be very useful in selecting subjects for a clinical trial or treatment, but not
necessarily useful for monitoring therapy, unless of course the therapy was designed to produce an Aβlowering effect. Very similar arguments obtain for other biomarkers, including FDG-PET, which clearly
provides information qualitatively different from that obtained with Aβ biomarkers and which may reflect
downstream, or neurodegenerative effects that are closely tied to symptoms. A model for the time course of
Aβ deposition and subsequent downstream changes is proposed and outlined in the overview section of this
application. This model produces specific hypotheses that we will test in the PET core, and essentially posits
that inciting events involve Aβ deposition, and that biomarkers reflecting this process will be abnormal earlier
than biomarkers reflecting neurodegeneration. Neurodegeneration, in contrast, will be better detected later in
disease using FDG-PET (and MRI). Exactly how Aβ and neurodegeneration biomarkers may be related or
complementary in “intermediate” disease stages (such as EMCI and LCMI) is a major question we hope to
answer.
4.2.3 Which amyloid imaging biomarker?
There are a number of possible approaches to amyloid imaging with PET. Although the collective
experience with [11C]PIB both in ADNI and around the world is great, the requirement of an onsite cyclotron
and radiochemistry program has limited the diffusion of this technology to the clinical community. The
maturity of the amyloid imaging field (as evidenced by over 200 PubMed references on the topic) demands
more widespread use. A key goal of ADNI2 is to make this technology available to a wide range of academic
medical centers and standardize the acquisition and processing of amyloid PET images.
The operational model for our
18
F
HO
S
approach
is fundamentally drawn
F
NH 1 1CH3
from
the
field
of clinical oncology,
O
N
O
O
where
it
is
common
for even
O
community
hospitals
to have a
O
O
N
NHCH
3
NHCH
PET scanner and purchase FDG
[11C]PIB
from a regional commercial
supplier. Because of the longer
[18F]AV-45
half-life of 18F compared to 11C
F
(109.8 vs 20.4 min),
NC CN
O
radiochemical production within
O
CH3
a radius of several hours
18
F
O
N
transportation to a local site is
NHCH
CH3
feasible. This basic approach
underlies the commercial
development of several 18F
18
[ F]BAY94-9172
[18F]FDDNP
amyloid imaging ligands. The
Structures of some of the amyloid imaging agents mentioned in the
potential candidates for a
text. [18F]3’-F-PIB is similar to PIB but with 18F labeling the
multicenter study of this sort are
those with substantial human data: [18F]FDDNP, [18F]BAY94-9172 (formerly AV1 and now also known as
Florbetaben), [18F]3’-F-PIB (also known as GE-067 or Flutemetamol), and [18F]AV45 (also known as
Florpiramine). FDDNP, well studied in humans with dementia[22] has several shortcomings for a study of this
sort. The tracer binds to more than one abnormal protein, in particular tau[23], and perhaps for this reason
has lower target-to-background binding than other agents like PIB[24, 25]. Its binding to a number of other
abnormal brain proteins might further complicate data interpretation in a study of this sort[26, 27]. For these
reasons, as well as the fact that a widespread distribution network for this tracer is not currently available, we
have focused on other tracers for ADNI2.
Amongst the remaining choices, the available data as well as logistical considerations have prompted us
to select [18F]AV45 as the most suitable agent for ADNI2. Preliminary data available for this compound in
both preclinical models and humans are substantial and are reviewed in the next section of this application.
We note that data available for other compounds are similarly favorable, and it appears that the three [18F]
18
3
18
3
PHS 398/2590 (Rev. 11/07)
Weiner, MW
amyloid tracers are likely to be more similar to one another than different. However, a major reason for the
selection of AV45 has been the availability of manufacturing centers, along with a plan for distribution, that
makes delivery of the tracer to virtually all participating ADNI sites possible at a reasonable cost. This factor,
along with the cooperation of Avid Radiopharmaceuticals in providing us preliminary data and regulatory
compliance (ie, use of the Avid IND) underlies the selection of the compound for ADNI2 and GO. Indeed, by
the time this grant is reviewed, AV45 studies should be underway at ADNI sites funded by the GO grant. A
letter from Daniel Skovronsky, President and CEO of Avid, accompanies this application and outlines the
commitment from Avid to supporting our efforts in ADNI.
4.3 PROGRESS REPORT/PRELIMINARY RESULTS
The ADNI PET core has accomplished all of the goals outlined in the original application as well as the PIB
add-on study. Below we summarize and highlight the significant accomplishments according to our original
specific aims:
4.3.1 Develop and implement methods for PET data acquisition
The ADNI PET core has pioneered the development of standards for acquisition of FDG-PET and PIBPET data that are now widely utilized in both academic and pharmaceutical studies worldwide. The key
design features of this protocol include: (1) simplicity, permitting multisite application (2) Compatible with all
commercially marketed PET cameras (3) Flexible data acquisition for post-processing to standardize formats,
intensity and resolution (4) low radiation dose (5) high patient acceptance. The protocol has been
successfully implemented at all ADNI PET sites in North America (N=50) and has been adopted by a number
of pharmaceutical clinical trials and the Japanese ADNI. The protocol was developed collaboratively amongst
ADNI investigators with input from the sites and an outside panel of experts. In addition, in the PIB “add-on”
we developed a standardized protocol for the acquisition of PIB-PET data that has also been widely adopted.
4.3.2 Develop and implement methods for quality control and data processing
During the first two years of the project, a total of 50 sites were approved for PET scanning by performing
a pair of phantom scans on the 3-D Hoffman brain phantom following a protocol that matched the acquisition
and reconstruction protocol to be used for the human phase of the ADNI project. Scans were passed through
a quality control process that checked for statistical noise, agreement with a digital version of that phantom
(the gold standard), and assessed for image resolution and image uniformity. This process also helped
develop a method for assessing and correcting for differences in PET images across sites[28]. There are a
total of 17 different scanner models from three vendors that have been used by participating ADNI sites.
Intrinsic resolution differs by more than a factor of two across scanners, from as high as 3-4mm to as low as
8mm effective resolution, thus requiring assessment of resolution and partial volume effects. Besides
resolution adjustments, we also are looking at low-frequency effects that are due to differences in
reconstruction algorithms, and in particular effects due to differences in the attenuation and scatter corrections
between vendor scanner models.
Human PET studies were first acquired at the start of the second year of ADNI, September 2005. All
human PET scans are run through a stringent quality control procedure to assess image quality. QC checks
include a statistical noise check, motion assessment across temporal frames, checking for full coverage of the
brain, visual checks of images to look for the most common PET artifacts (such as normalization problems or
motion between attenuation and
emission scans), as well as visual and
image header checks to make sure the
exact ADNI protocol has been followed.
Half the enrolled ADNI subjects
underwent FDG-PET, and we now
have 404 baseline scans, 368 6 month
scans, 336 12 month scans, 154 18
month scans (MCI subjects only), 283
Figure 1: Examples of ADNI processing stream. Left: FDG-PET in
24 month scans, and to-date 106 36
native format, averaged 6 x 5 min frames. Middle: same as left, but with
month scans (MCI and control subjects
standard orientation, voxel size. Right: same as middle smoothed to 8
only). Besides FDG, approximately
mm FWHM resolution.
one quarter of the PET subjects
(n=103) also received PIB scans to image amyloid deposition. Repeat scans at a one-year interval have been
performed on 81 of the 103 subjects, while 21 of these 81 subjects have had a 3rd annual PIB scan.
PHS 398/2590 (Rev. 11/07)
Weiner, MW
The original “raw” PET image sets are automatically downloaded for the LONI image repository. Routines
read and convert PET images sets to a standard file format. The different temporal frames are co-registered,
and both a dynamic image set, as well as a single-frame averaged image set are produced in the original
patient orientation and in the original intrinsic pixel size and plane spacing for that scanner. In addition, all
images sets are re-oriented to a common spatial orientation and interpolated onto a uniform image grid
(1.5mm3 voxels). In the final step, images are smoothed with a scanner-specific filter derived from each site’s
Hoffman phantom[28] to provide a common resolution of 8 mm FWHW isotropic resolution. These processed
PET images sets are then uploaded back to LONI in a consistent DICOM format (see Fig 1 for examples). To
date, the QC failure rate is 12%, which has declined slightly from the initial years of the project when it was
15%, but it has remained relatively stable. The majority of scan failures were related to problems with
reconstruction and were fixed with reprocessing; less than 1% of scans required rescanning.
4.3.3 Analysis of FDG-PET data and PIB-PET data
Considerable work in the analysis of all PET data is reported by the biostatistics core. These results compare
different methods of PET data analysis to one another and to other biomarkers such as MRI. In order to
conserve space, these results are presented in other parts of the application (see overview, biostatistics core).
Here we present the results of analyses performed by the major PET labs funded for ADNI data analysis.
Funding support for PET data analysis in ADNI1 was actually minimal and involved only funding of the initial 3
timepoints – baseline, 6 and 12 month studies. These analyses were performed in 3 laboratories (Jagust,
Reiman, Foster), each of which took a different approach to analysis in an effort to define the relative merits of
each approach. For the PIB add-on, a fourth lab was added (Mathis). Considerably more data has been
analyzed than was initially funded, and below we present highlights of each laboratory’s analysis results.
4.3.3.1 UC Berkeley/Jagust Laboratory: The methodological approach to data analysis in the Jagust
laboratory involved the pre-selection of specific regions of interest (ROIs). This was done by identifying
regions cited frequently in FDG-PET studies
of AD and MCI patients. We conducted a
meta-analysis in PubMed using all
permutations of the following search terms:
AD or Alzheimer’s; MCI or Mild Cognitive
Impairment; FDG-PET or FDG or glucose
metabolism. Within the studies identified by
these terms we isolated those that listed
coordinates representing results of crosssectional and/or longitudinal voxelwise
analyses in which FDG uptake differed
Figure 2: FDG-ROIs were constructed based on coordinates
cited in studies as described in the text and were used to
significantly between groups, changed in the
generate the Composite ROI for analyses from the Jagust Lab.
same individuals over time, or correlated with
cognitive performance. This resulted in a
total of 292 MNI or Talairach coordinates and (if available) their accompanying Z-scores or T-values, of which
209 were from cross-sectional or correlational studies and 31 were coordinates from longitudinal studies. The
list of studies used to generate these FDG-ROIs is available online as supplementary data to our report on
FDG-clinical correlations[29] (see Figure 2). All coordinates were transformed into MNI space. Then intensity
values were generated for coordinates that reflected a combination of the Z-score or t-value associated with
the coordinate and the degree to which coordinates within the same region overlapped (indicating repeated
citations of the same region across studies). All t-values were transformed to approximate Z scores. Then,
overlapping Z scores, when they occurred, were added. The volumes were smoothed with a 14mm FWHM
smoothing kernel. Finally, the volume was then intensity normalized using the maximum value, resulting in a
map with values between 0 and 1. The cross-sectional coordinate map was then thresholded at 0.50, and this
resulted in a set of 5 regions located in right and left angular gyri, bilateral posterior cingulate gyrus, and left
middle/inferior temporal gyrus (see Fig 2). Because these 5 ROIs were highly correlated, we generated a
composite ROI by averaging across each subject. This composite ROI was used in the majority of
subsequent analyses.
PET as an outcome measure: As noted, one of the original goals of ADNI was to investigate imaging as a
biomarker of disease progression. Initial results from our lab have recently been reported[29] and will be
briefly summarized. In essence, we found that in a large sample (approximately 400 subjects comprised of
PHS 398/2590 (Rev. 11/07)
Weiner, MW
normals, MCI and AD patients with multiple time points up to 2 years) low glucose metabolism at baseline
predicted decline on the ADAS-cog, and that longitudinal decline in glucose metabolism was associated with
longitudinal decline on the ADAS-cog. We also determined that in the setting of a clinical trial, longitudinal
metabolic measurements would have greater statistical power for the detection of change than would
longitudinal ADAS-cog measurements.
Relationships between biomarkers: Another goal of ADNI involves the investigation of the relationships
between biomarkers, and between biomarkers and cognitive function, and a paper on this topic is currently in
press[13]. In this report we investigated relationships between PIB-PET, FDG-PET and CSF biomarkers
(Aβ42 and tau). A unique aspect of the study was the dichotomization of these biomarkers as “+” and “-“ for
AD using cutoffs obtained in samples that were distinct from ADNI, and then application of these cutoffs to the
ADNI sample. In summary, we found high agreement between PIB and CSF Aβ42 (91% agreement, κ=0.74),
but generally modest to low agreement between PIB and other biomarkers. In contrast, FDG-PET was highly
correlated with scores on the MMSE (R=0.63, p <0.0001), while both PIB and CSF Aβ42 were not. Together
these findings indicate that different biomarker measures of Aβ are in agreement, but that non-Aβ biomarkers
are better indicators of disease status.
Prediction of decline: As noted in the introduction, ADNI is gradually transitioning to place increased
emphasis on the use of biomarkers in the prediction of change as well as outcome measures. We have
recently completed a project looking at multiple predictors of decline in ADNI MCI patients. Again, a feature of
this project was the use of set cutpoints for biomarkers, defined in the ADNI controls and AD patients, and
then applied to the MCI patients to predict either conversion (ie, a dichotomous outcome) or cognitive decline.
We used ADNI MCI participant data (N=85) to directly compare the predictive value of candidate markers.
Specifically, we examined genetic
(APOE 4 status), neuroimaging (FDG-PET,
hippocampal volume), and CSF biomarker
(Aβ42, t-tau, p-tau181p) measures, in
addition to episodic memory performance
(Auditory Verbal Learning Test, AVLT)
obtained from all participants at baseline.
A “cutoff” for each biomarker was defined
based upon an ROC analysis between AD
and control subjects (importantly, a
Figure 3: Predicted survival curves based on Cox proportional
different group of subjects than those
hazards model illustrating results for predictors of conversion from
tested for progression). MCI patients were
MCI to AD. Each curve reflects the proportion of patients
then defined as normal or abnormal for
remaining non-demented for subjects scoring above or below the
each biomarker, including FDG-PET. The
cutoff for FDG-PET (left) or AVLT (right). The joint hazard ratio for
subjects scoring in the AD range for both variables was 15.
goal of the study was the determination of
which marker or combination of markers is optimal for predicting of conversion to AD or cognitive decline
(longitudinal ADAS-cog measurements) over a variable followup time (mean = 1.9 +/- .4 yrs). We found that
baseline FDG-PET using our composite ROI and AVLT were significant predictors of conversion, both
independently and when accounting for status on the other variables, indicating that FDG-PET and AVLT
have sensitivity to detect significant a clinical transition. Individuals who were abnormal on both measures
(low glucose metabolism and poor recall performance) were 15 times more likely to convert to AD than
individuals who were normal on these measures. With respect to prediction of cognitive decline, all variables
had value in predicting ADAS-cog change when considered individually, but only the CSF ratio ptau181p/Aβ42 (and, marginally, FDG-PET) predicted decline when accounting for the other variables,
suggesting that it may be sensitive to more subtle, sub-clinical change. The fact that FDG-PET played a role
in predicting both conversion and cognitive decline is consistent with previous findings [30, 31] and indicates
that glucose hypometabolism may be sensitive to a broader range of pathological processes than the other
variables. Overall, our results suggest that these markers provide complementary information that may aid in
selecting patients for clinical trials or identifying patients likely to benefit from a therapeutic intervention.
4.3.3.2 Banner Alzheimer’s Institute/Reiman Laboratory: The group at Banner was responsible for the voxelbased analysis of FDG PET data using the automated brain mapping algorithm SPM5
(http://www.fil.ion.acl.ac.uk/spm/). Along the way, they developed several new image-analysis methods[32]
PHS 398/2590 (Rev. 11/07)
Weiner, MW
and conducted amyloid imaging studies of their own[33] supporting some of the work proposed during the
next funding period.
Cross-Sectional Brain Mapping Studies: SPM5 was used to generate statistical maps of between-group
differences in regional-to-whole brain measurements of the cerebral metabolic rate for glucose (CMRgl) in
probable AD, MCI and NC groups, correlations with categorical and continuous measurements of clinical
disease severity, and differences between APOE4 carriers and non-carriers in each of these groups[5]. It
also provided the foundation to develop a “weighted hypometabolic convergence analysis” technique to
automatically capture the AD-related pattern and magnitude of an individual’s CMRgl reductions in an “ADrelated hypometabolic convergence index” and use it to predict rates of clinical conversion in MCI patients
as described below. In comparison with NC subjects, AD and MCI patients each had significantly lower
CMRgl bilaterally in posterior cingulate, precuneus, parietotemporal and frontal cortex, as well as occipital
cortex, which were correlated with more severe dementia (measured with the CDR and MMSE). In
comparison with APOE4 non-carriers in the same subject group, APOE4 carriers in the NC group had
significantly lower CMRgl in several AD-affected regions, supporting previous findings from this group[34,
35] and others[36, 37]. Measurements and Tailarach brain atlas coordinates and measurements from the
maximally significant locations were uploaded to the LONI website as a shared resource. Among other
things, ADNI2 will provide an opportunity to extend this work to individuals in each subject group with
significant fibrillar amyloid burden. As noted, this work is published[5].
Longitudinal Brain Mapping Studies: SPM5 was also used to characterize 12-month regional-to-whole
brain CMRgl declines in the AD, MCI and NC groups and
provided the foundation to develop the statistical ROI (sROI)
strategy and estimate the number of AD and MCI patients
needed to evaluate AD-slowing treatments in multi-center
RCTs as described below. The AD and MCI groups each has
significant twelve-month CMRgl declines bilaterally in posterior
cingulate, medial and lateral parietal, medial and lateral
temporal, frontal and occipital cortex (figure 4), which were
significantly greater than those in the NC group and correlated
with measures of clinical decline. Among other things, ADNI2
will provide an opportunity to extend this work to individuals in
each subject group with significant fibrillar amyloid burden.
AD-Related Hypometabolic Convergence Index (HCI)
Analysis. Banner researchers have been developing voxelFigure 4: 12 Month CMRglc declines in
based image analysis strategies to relate the pattern of
patients with AD (A) and MCI (B)
hypometabolism in an individual subject to that in AD patients,
augment the power of FDG PET (and other imaging modalities) using voxel-based image-analysis
techniques to help in the differential diagnosis of dementia (which is beyond the scope of ADNI) and predict
subsequent rates of clinical decline in the earlier symptomatic and asymptomatic stages of AD, alone or in
combination with other risk factors or biomarkers (one of the principal aims of ADNI 2). These strategies
include use of multivariate statistical methods to capture complex
image patterns in a single subject score[32] and include an
automated AD-related “weighted hypometabolic convergence
analysis” technique, briefly considered here. This method
generates a single “hypometabolic convergence
index” (HCI) to characterize the extent to which both the
magnitude and spatial distribution of hypometabolism in an
individual subject is related to the magnitude and spatial
distribution of AD patients in comparison to the measurements in
NC subjects. The method generates a spatially normalized z-score
map of CMRgl reductions in a person’s FDG-PET image and in a
probable AD patient group (transformed from a t-score map) in
Figure 5: Subsequent rates of clinical
comparison with the images from a NC group. The AD-related
progression in HCI-positive and
hypometabolic index is computed by summing the product of
negative MCI patients.
individual subject and AD patients z-scores in every voxel and
PHS 398/2590 (Rev. 11/07)
Weiner, MW
dividing by 10,000 to provide an HCI between 0 and 20. HCI’s were strongly associated with the
classification of subjects into AD patients, MCI patients who converted to probable AD within 18 months,
MCI patients who did not convert to AD during that time,
and NC subjects (linear trend p=1.8e-25), and in an
independent group of AD cases and controls, suggesting
its potential as a measure of metabolic AD severity.
Using the mean HCI in AD patients and NCs as the
empirically pre-defined threshold to classify FDG PET
scans as HCI-positive or negative, MCI patients with
HCI-positive scans had a significantly higher rate of
clinical progression to AD than those who remained
stable, as reflected in the Kaplan-Meier curve (figure 5).
This HCI approach together with definition of amyloid
positivity with AV45, can be used as a predictor or
subject selection variable in ADNI2 to model sample
sizes with various cutoffs for clinical trial effects.
Figure 6: Statistical ROIs (sROIs) of 12-month
Tracking AD Progression and Evaluating AD-Slowing
CMRglc decline
Treatments in sROI’s: Banner researchers have also
developed an “empirically pre-defined statistical region-of-interest (sROI)” strategy, providing a single
imaging endpoint with improved statistical power to evaluate AD-slowing treatments in randomized clinical
trials (RCTs), freedom from multiple regional comparisons, and the ability to customize the endpoint to the
subject group and between-scan interval of interest using
voxel based image analysis techniques[38]. sROI’s were
empirically pre-defined for each subject group and
between-scan interval in ADNI’s training data set (figure 6)
and then used in ADNI’s independent testing data set to
demonstrate its power to track sROI-to-reference region
CMRgl declines and evaluate AD-slowing treatments with a
fraction of the number of AD and MCI patients needed
using clinical endpoints in 12-, 18- and 24-month parallelgroup, placebo-controlled RCTs. sROI measurements were
uploaded to the LONI website, permitting the ADNI
Statistics Core to relate the combined use of FDG PET and this sROI strategy to other image-analysis
techniques, other imaging modalities and clinical endpoints in terms of the number of AD and MCI patients
needed to detect significant AD-slowing treatment effects (see Biostatistics Core).
4.3.3.3 University of Utah/Foster Laboratory: The focus of data analysis in our laboratory has been the use of
3-dimensional stereotactic surface projection (3D-SSP) with Neurostat, a free, non-commercial analysis
program[39]. This approach has been influential in the decision to include specific aims in ADNI2 using
analysis of individual images to classify individual subjects and predict outcomes. Techniques developed for
FDG have been optimized for use with AV45-PET through a collaboration with Dr. Satoshi Minoshima who will
collaborate with this laboratory during ADNI2. Data from Utah were analyzed, and uploaded to LONI as
numerical analyses of all ADNI baseline, 6, and 12 month scans. Numerical summary measures used by the
Biostatistics Core to calculate sample size, compare with other biomarkers and to examine the relationship
with cognitive measures were 1) the number of surface pixels with values > 2 standard deviations (SD) below
the normal control population, 2) the number of surface pixels with values > 3 SD below the normal controls,
3) sum of Z score values > 2 SC below normal controls, 4) Z score values based upon a regression of pons
values in serial studies of an individual, 5) glucose metabolism relative to pons measured in a predetermined
surface region representing the frontal cortex, 6) glucose metabolism relative to pons averaged in frontal,
parietal and temporal association cortices. Numerical summary measures 1-4 are the PET Core’s only
measures of topographic extent of glucose metabolism, a novel outcome we proposed.
Classification of cases as AD or FTD: The Utah/Neurostat approach is especially suited for classification
of individual images, which we have previously demonstrated can differentiate AD and frontotemporal
dementia (FTD), showing specificity of the FTD pattern of hypometabolism using 3D-SSP of >97%[40]. Using
this approach in ADNI, we found that 10/93 (10.8%) of AD subjects in ADNI clearly have an FTD pattern of
PHS 398/2590 (Rev. 11/07)
Weiner, MW
hypometabolism. Similar analyses performed in ADNI normal and MCI subjects find occasional patients with
patterns suggesting FTD or AD. We have suggested that misdiagnosis may be contaminating many AD
research studies and FDG-PET could improve the diagnostic accuracy of subjects enrolled in clinical
trials[41]. Interestingly, the proportion of subjects with an AD diagnosis and FTD pattern is similar to the
proportion of PIB- AD patients.
Glucose metabolic patterns and MCI Conversion: Our observation that some MCI patients had scans with
clear abnormalities consistent with AD,
led us to examine the relationship between
the topographic extent of glucose
hypometabolism (measured with 3D-SSP)
and subsequent diagnosis. We found that the
topographic extent of hypometabolism at
baseline was greater in MCI subjects who
converted to AD by 1 year, than in MCI
subjects who continued to be classified on
clinical grounds as MCI. We found that
variation in the number of abnormal pixels
largely reflected the timing of conversion, with
patients who converted at 6 months having a
greater extent of hypometabolism than those
who converted later. Perhaps not
surprisingly, converters had an AD-like
pattern of hypometabolism. The pattern of
Figure 7: SSP images in ADNI MCI subjects by diagnosis at
average z-scores in non-converters shows
12 months
little abnormality, while 6 month converters
show more extensive hypometabolism than 12 month converters (figure 7).
Neurostat and AV45 data: We have adapted Neurostat to allow processing of AV45-PET data and the
generation of amyloid binding and statistical
maps using data provided by AVID
Radiopharmaceuticals. Pre-processing steps
followed the simplified analysis described for
PIB[42]. We used a single late 50-90 minutes
scan following AV45 injection obtained from the
set of dynamic AV45-PET images, similar to what
we propose for this project. T1 trans-axial MR
images were intensity normalized and then
spatially normalized and skull stripped with
Freesurfer[43], coregistered to PET, and
converted to a binary format compatible with
Neurostat analysis. We developed a normal
comparison data set as “proof of concept” based
upon 5 normal control subjects. Amyloid binding
Figure 8: 3D SSP of AV45 activity in controls and an AD
values relative to cerebellum and Z-scores relative
patient
to normal controls for all surface pixels are shown
graphically in a standard display for an individual AD patient (figure 8). The normalized uptake values from
both the reference set of control subjects and the individual subject are shown. This AD patient has increased
cortical uptake of AV45 compared to controls, as expected.
4.3.3.4 University of Pittsburgh/Mathis Laboratory: The PIB-PET add-on study began its third year of funding
in May 2009. To date, 103 baseline PIB-PET studies have been completed at 14 participating ADNI PET
centers in 19 elderly cognitively normal controls (Mean age=78, MMSE=29), 65 MCI subjects (mean age = 75,
MMSE=27), and 19 AD subjects (Mean age=73, MMSE=22).
Baseline PIB-PET data: Regional assessment of the PIB-PET data involved sampling 13 different brain
areas using an automated region of interest (ROI) template method. The template analysis method was
PHS 398/2590 (Rev. 11/07)
Weiner, MW
based on the MRI of an elderly MCI subject (79 years), who was purposely chosen to represent a moderate
degree of atrophy. Each subject’s MRI scan was warped into MCI-Template space and then the same
parameters were applied to the subject’s PIB-PET
scan, resulting in alignment of the MRI and PIB-PET to
the MCI-Template. The MCI-Template ROIs were then
sampled on the warped PIB-PET images. Figure 9
shows regional brain PIB-PET values in all 103 subjects
in 4 cortical ROIs. In addition, a four region average (4
Reg Avg) value of all 4 ROIs was calculated for each
subject. Positive amyloid deposition (PIB+) was defined
when the SUVR (standardized uptake value ratio in the
region relative to cerebellar gray matter) exceeded a
regional value of 1.50 (red line in Figure 9). This PIB+
cut-off value was based on studies in a large group of
cognitively normal control subjects studied at the
Figure 9: Regional PIB-PET SUVR outcome
University of Pittsburgh and was described in a recent
measure values in four brain regions (ACG, anterior
paper[44]. The PIB-PET data from these ADNI studies
cingulate; FC, frontal cortex; PAR, parital cortex;
were generally consistent with literature reports from a
and PRC, precuneus) and a 4 cortical regions
averaged value (4 Reg Avg) for all 103 PIB-PET
variety of laboratories throughout the world. The control
baseline subjects in the ADNI PIB add-on studies.
subjects were comprised of two sub-groups: one subThe cut-off value of 1.50 SUVR units dichotomizes
group with relatively high SUVR values exceeding the
all of the subjects into two groups: PIB+ and PIB-.
PIB+ cut-off value of 1.50 (9 of 19) and another subgroup with relatively low SUVR values below the cut-off value (10 of 19) (Figure 10). Most literature studies
using PIB have shown the frequency of amyloid-positivity in elderly cognitive control subjects (average age 80
years) to be in the range of 30%. In the ADNI PIB-PET cohort, we found a higher frequency of PIB-positivity
of 47%. Although the prevalence of PIB positivity is high, as shown in the previous work from the Jagust Lab
Figure 10: Four Region
in conjunction with the CSF
Average SUVR values
biomarker core13, there is high
for all 103 baseline PIBcongruence between PIB and CSF Aβ
PET subjects separated
measures. As these control subjects
into control, MCI, and
will be studied longitudinally over the
AD groups and each
next few years, it will be interesting to
group further subdetermine whether their PIB-positivity
divided into PIB(-) and
becomes more robust and whether this
PIB(+) sub-groups. n is
change correlates with changes in
the number of subjects
in each sub-group
cognition. The AD group showed PIBpositivity in 17 of 19 subjects (89%)
(Figure 10). This seems to be typical of many centers throughout the world, which find that 10-20% of their
clinically diagnosed AD subjects are PIB-negative. Whether this is a reflection of inaccurate clinical diagnosis
Figure 11: Four Region
or the failure of PIB to
Average SUVR values for
detect amyloid deposits in a
73 PIB-PET one year
small fraction of AD subjects
longitudinal follow-up
must await future post-mortem
subjects separated into
confirmation studies. Finally,
control, MCI, and AD
the MCI group was comprised
groups and each group
of an apparent continuum of
further sub-divided into
PIB values, ranging from very
PIB(-) and PIB(+) subPIB-negative to very PIBgroups. n is the number of
positive with a number of
subjects in each sub-group
and hatched boxes indicate
subjects in an intermediate
the one year follow-up subzone around the cut-off value of
group SUVR averages.
1.50 SUVR units. For the ADNI
MCI cohort, 18 of 65 MCI subjects were PIB- (28%) and 47 of 65 MCI subjects were PIB+ (72%) (Figure 10).
These values agree with the literature, which typically shows about 2/3 of MCI subjects studied to be PIB+.
PHS 398/2590 (Rev. 11/07)
Weiner, MW
Longitudinal PIB-PET data: As of September 2009, 79 PIB-PET one year and 18 PIB-PET two year
longitudinal ADNI studies have been completed. Data from 73 of the one year longitudinal studies have been
processed by the University of Pittsburgh PET Group and posted on the LONI website (Figure 11). These
include 16 cognitively normal control, 44 MCI, and 13 AD subjects. In all subject groups, there were small
and insignificant changes in group average SUVR values over one year of follow-up. These results are
consistent with literature reports of group average changes over 2 years [20, 45]. However if individual
subjects are followed longitudinally, significant changes over time can be discerned. Using reliable change
index (RCI) methods[46] to track changes in individual subjects has been helpful in this regard. For
application of the RCI method to the PIB-PET longitudinal cohort, measurement variability was determined as
the standard deviation from test-retest PIB-PET data from the University of Pittsburgh, comprised of 22
subjects (7 controls, 9 MCI, and 6 AD) who received two PIB-PET scans within 28 days. The 4 Region
Average test-retest SUVR standard deviation of these data was 0.13 SUVR units. Hence, a change cut-off
value of 0.215 SUVR units (delta-SUVR value) was determined to be a significant change in the 4 Region
Average SUVR value of any subject with a confidence level of 95% and a z-score of 1.645. The number of
PIB+ and PIB- subjects with significant increases in the 4 Region Average SUVR value (>0.215 SUVR units)
over one year of follow-up is shown in the Table. Across all subject groups, the number of significant PIB+
subjects showing increases was 10 of 49 subjects (20%), while only 2 of 24 (8%) PIB- subjects increased.
Table. Number of PIB(+) and PIB(-) Subjects Showing Significant Increases in 4 Region Average SUVR Values Over 1
Year of Follow-up
Controls
MCI
AD
# >0.215 SUVR
# >0.215 SUVR
# >0.215 SUVR
PIB(+)
8
3
29
4
12
3
PIB(-)
8
2
15
0
1
0
4.3.4 Preliminary Results: AV45
As noted, we have selected [18F]-AV45 as the agent for PET amyloid imaging in ADNI for a number of
reasons which we review here. Please also see the letter of collaboration from Daniel Skovronsky, President
and CEO of Avid Radiopharmaceuticals, who has indicated substantial support from the company for this
project. Also, by the time ADNI2 is funded, initiation of AV45 studies as part of the GO grant will be underway
so that the infrastructure and many of the proposed methods will already be implemented.
4.3.4.1 Preclinical data Results of preclinical studies are currently in press and are briefly summarized
here[47]. Initial experiments indicated favorable qualities of this ligand as an amyloid imaging agent. The
octanol/water partition coefficient is 56 (logP=1.74). AV45 has high affinity specific binding for amyloid
plaques. Equilibrium binding studies were performed in AD brain homogenates, with specific binding
determined by incubation of different concentrations of [18F]AV45 with and without BTA-1. After incubation,
samples were rapidly filtered,
washed 3 times with PBS at pH7.4, and the radioactivity retained on the
filter disks was determined in a gamma counter with 90% counting
efficiency. Kd and Bmax were calculated by Scatchard plot and Rosenthal
analysis. The dissociation constant was Kd = 3.7±0.3nM, Bmax was 8811
fmole/mg protein. These values are well within the range of other
effective compounds ([11C]PIB has a Kd of 2 nM).
Brain sections from AD patients were frozen and cut into 20 mm
thick sections, incubated with [18F]AV45 for 1 hour at room temperature,
18
Figure 12: [ F]AV45 in vitro
dipped in saturated Li2CO in 40% aqueous ethanol, washed with 40%
autoradiograms of postmortem brain aqeous ethanol, rinsed in water, dried, and exposed for 15 hours.
sections from an AD patient (left)
Autoradiography showed patterns of staining indicative of binding to Aβ
and normal control.
plaques (figure 12).
Screening of non-radiolabelled AV45 shows low or no affinity for all
other CNS and cardiovascular receptors tested. The compound has been tested in rats at doses up to 100x
the maximal human dose (MHD=50mg) in a PET study, and in beagles at up to 8.7x MHD with no adverse
reactions. The PET tracer studies will thus be conducted safely, well below a “no observed effect level”
(NOEL) dose.
Primate PET experiments indicated rapid brain penetration and washout. Metabolism of the compound
was studied in vitro by incubation of the tracer with human liver microsomes,indicating that the primary
PHS 398/2590 (Rev. 11/07)
Weiner, MW
metabolite is an 18F-labeled N-desmethylated amine derivative. In vivo studies in mice showed that after 60
min, slightly less than 4% of inected 18F was in the blood, and only 30% of the radioactivity was parent
compound.
4.3.4.2 Initial Clinical Studies: 19 clinical trials have been completed or are currently underway with AV45, and
613 humans have undergone AV45-PET imaging worldwide. In the US, two Phase I studies investigated
safety, brain uptake, pharmacokinetics, and amyloid imaging properties, as well as whole body distribution
and radiation dosimetry (41 subjects).
Human radiation dosage is reported in the
Human Subjects section of the application.
In the U.S., completed Phase II trials
included a dose ranging study (20 subjects),
a test-retest reliability study (34 subjects,
each imaged twice), and a cross-sectional
study (182 subjects). Ongoing Phase II and
Phase III trials in the US include a
longitudinal progression study (measuring
symptomatic progression in 120 subjects
followed for 18 months post AV45 imaging)
and an imaging to histopathology correlation
study (involving 150 subjects imaged and
followed to autopsy). Key findings from
these studies are summarized here.
Clearance of [18F]AV45 in humans is
Figure 13: Top: 10 min PET data from AV-45 in an AD patient (left)
rapid, as within 10 min of injection total
and healthy control (right) showing uptake in cortex in AD and
plasma radioactivity is reduced by 80%.
nonspecific retention in white matter in the control. Bottom:
Kinetics that demonstrate rapid uptake and washout with retention
Human metabolites include polar
in cortical regions in the AD patient, washout in the same regions in
compounds, the N-desmethylated amine,
controls, and similar washout in white matter and cerebellum in
and an N-acetylated desmethylated amine.
both subjects.
While the 2 latter compounds are capable of
penetrating the brain, they occur in very low concentrations and have low affinity for Aβ and are therefore
unlikely to contribute to the PET signal.
Studies in both normal controls and patients with AD demonstrate a variety of favorable characteristics for
18
Figure 14 Left: Pattern of [ F]AV45 uptake (50-60min, normalized to cerebellum) in controls, AD and MCI
18
patients. (right) graph showing dependence of [ F]AV45 uptake on age.
an amyloid imaging tracer. Test-retest reliability is high, with variation of <6% for subjects studied up to 2
weeks apart. The compound shows a moderate degree of non-specific binding in white matter, but washout
from cortical regions in normal controls is rapid, while washout is slower in AD patients, indicative of binding to
cortical Aβ. The kinetics of the tracer are favorable, reaching steady states relatively early (see figure 13) so
that images obtained relatively soon after injection can be used for quantitation. This factor combined with the
PHS 398/2590 (Rev. 11/07)
Weiner, MW
longer half life of 18F makes short imaging times at about an hour after injection a very practical method of
obtaining data for quantiation. Images obtained using equilibrium models, late images, or late standardized
uptake value ratios (with cerebellum as a denominator) all show a pattern of binding identical to that seen with
other amyloid tracers such as [11C]PIB – that is, retention in cortical regions in AD patients that is most notable
in precuneus/posterior cingulate, medial inferior frontal cortex, and other regions of association cortex. In a
small number of subjects who have undergone both [11C]PIB and [18F]AV45 imaging studies, the two tracers
show very similar patterns of cortical uptake and rentention. In addition to matching the topology of known
Aβ binding sites, the relationships between [18F]AV45 binding and clinical/biological features of AD match
what is known from pathological studies and other imaging studies using [11C]PIB (figure 14). For example,
binding in controls shows a proportion of subjects with positive scans, indicative of Aβ deposition. The
proportion of controls with positive scans increases with age, from ~ 3% at <60 years of age to ~30% at >80
years of age. Approximately 45% of MCI subjects (mixed amnestic and non-amnestic) show positive scans
while the remainder show scans that are negative for tracer retention., and MCI patients show scans that may
be either positive or negative for tracer retention. The frequency of scan positivity increases with both age
and ApoE4 genotype, similar to results recently reported for [11C]PIB[14].
4.3.4.3 Logistics: One of the considerable advantages of using [18F]AV45 for this study is that transportation
arrangements for the delivery of tracer to all current ADNI sites are in place because of the many ongoing
research clinical trials that are already using AV45. For example, AV45 is being used in research studies
sponsored by NIH grants, comparative trials sponsored by the American College of Radiology, therapeutic
monitoring trials sponsored by Eli Lilly and Pfizer, and therapeutic trials being conducted by the ADCS. Of 58
current sites, tracer can be delivered by ground transportation from manufacturing centers within a several
hour radius to 45 sites. An additional 8 sites are reachable by a short airflight; some of these sites will be
supplied by new manufacturing centers within driving distance by the time ADNI2 is initiated. The remaining 5
sites are Canadian – regulatory issues will govern how soon and when tracer can be supplied to them. Even
if these issues cannot be solved by the time of ADNI2, the vast majority of sites will be able to participate.
Publications
Hua X, Lee S, Yanovsky I, Leow AD, Chou YY, Ho AJ, Gutman B, Toga AW, Jack CR Jr, Bernstein MA,
Reiman EM, Harvey DJ, Kornak J, Schuff N, Alexander GE, Weiner MW, Thompson PM Alzheimer’s
Disease Neuroimaging Initiative: Optimizing Power to Track Brain Degeneration in Alzheimer's Disease
and Mild Cognitive Impairment with Tensor-Based Morphometry: An ADNI Study of 515 Subjects.
Neuroimage 2009;48:668-81. Epub 2009 Jul 14.
Ho A, Hua X, Lee S, Leow AD, Yanovsky I, Gutman B, Dinov ID, Lepore N, Stein J, Toga AW, Jack CR,
Bernstein MA, Reiman EM, Harvey DJ, Kornak J, Schuff N, Alexander GE, Weiner MW, Thompson PM.
Comparing 3 tesla and 1.5 tesla MRI for tracking Alzheimer's disease progression with tensor-based
morphometry. Hum Brain Mapp 2009 (in press)
Jagust WJ, Landau SM, Shaw LM, Trojanowski JQ, Koeppe RA, Reiman EM, Foster NL, Petersen RC,
Weiner MW, Price JC, Mathis CA. Relationships between biomarkers in aging and dementia. Neurology,
In press.
Joshi A, Koeppe RA, Fessler JA. Reducing between scanner differences in multi-center PET studies.
NeuroImage 46:154-159, 2009.
Langbaum, JBS, Chen K, Lee W, Reschke C, Bandy D, Fleisher AS, Alexander GE, Foster NL, Weiner MW,
Koeppe RA, Jagust WJ, Reiman EM, Alzheimer’s Disease Neuroimaging Initiative. Categorical and
correlational analyses of baseline fluorodeoxyglucose positron emission tomography images from the
Alzheimer's Disease Neuroimaging Initiative (ADNI). NeuroImage 2009;45:1107-16.
Landau SM, Harvey D, Madison CM, Koeppe RA, Reiman EM, Foster NL, Weiner MW, Jagust WJ.
Associations between cognitive, functional, and FDG-PET measures of decline in AD and MCI.
Neurobiology of Aging, epub Aug 4 2009 NIHMS 132130.
Mormino EC, Kluth JT, Madison CM, Rabinovici GD, Baker SL, Miller BL, Koeppe RA, Mathis CA, Weiner
MW, Jagust WJ. Episodic memory loss is related to hippocampal-mediated beta-amyloid deposition in
elderly subjects. Brain, 132:1310-1323, 2009 (epub Nov 28 2008).
Reiman EM, Langbaum JBS. Brain imaging in the evaluation of putative Alzheimer’s disease slowing, riskreducing and prevention therapies. In Jagust WJ, D’Esposito M, eds. Imaging the Aging Brain. New York:
Oxford University Press; 2009:319-350.
PHS 398/2590 (Rev. 11/07)
Weiner, MW
4.4 METHODS
The plans for ADNI2 draw upon our experience with ADNI to date, as we plan to utilize methods and
approaches developed for both FDG- and PIB-PET data acquisition and analysis, with relatively slight
modifications to accomplish goals specific to this funding cycle. Through the accomplishment of our specific
aims, we will assure that all ADNI subjects have 2 sets of FDG- and AV45-PET scans obtained at 2 time
points 2 years apart, using optimal approaches that will, judging from the success of ADNI1, become
standards for the imaging community. This will also bring amyloid imaging to over 50 academic sites
throughout North America, and will enable the large-scale testing of hypotheses, articulated in the specific
aims, concerning the use of biomarkers in dementia.
4.4.1 Administration/Management Plan
We will continue the successful management strategy designed and applied in ADNI1 to date. Dr. Jagust, the
Core PI, will participate in all executive committee conference calls and meetings and will thereby regularly
interact with the ADNI PI (Dr. Weiner), other core leaders, NIA representatives, and industrial participants. He
will also participate in meetings of the Steering Committee, Industry Scientific Advisory Board, External
Advisory Committee, and international ADNI meetings as scheduled and necessary.
Current management of the PET core relies on twice-monthly conference calls involving all of the major
PET laboratories as well as advisors and representatives from the ISAB. These calls are concerned with
planning, operations, logistics, data analysis, and publications. The PET group also meets in person in
conjunction with the Steering Committee meetings.
The PET core, under the leadership of the Core leader will also have the responsibility for working closely
with the Clinical Core and the ADCS in drafting procedures and startup. This includes developing a study
protocol, modifying the existing PET technologist’s manual (developed as part of ADNI1 for FDG and PIB
imaging, which will require modification for AV45 procedures), and assisting with IRB approvals and drafting
consent documents. We will also interact with Avid leadership in regulatory issues to be certain that the
current Avid IND supports all ADNI activities – this will not be a problem as similar protocols are currently
approved under this IND.
4.4.2 Site qualification
Initial steps, which are currently underway as part of GO, involve ascertaining which of the approximately 60
ADNI sites are interested in participating in this protocol – so far, interest is keen. We will employ the
methods utilized for site qualification in ADNI to date to select sites capable of performing the study. This
approach was implemented at the start of the PET Core activities with FDG and requires all sites to image an
18
F-filled (generally with FDG) Hoffman brain phantom on two sequential days using the protocol identical to
that required for human imaging. This enabled us to ascertain the characteristics of the scanner (particularly
resolution and uniformity) and assured that sites were capable of performing the protocol for acquisition and
image reconstruction. The vast majority of sites that will participate in the AV-45 protocol are already qualified
for PET imaging using this procedure. We did not require re-qualification when we instituted the PIB protocol
and we will not do so for the AV-45 protocol. Approximately 10 ADNI sites were not in the orginal PET
protocol some of which may be interested in participating in this protocol – these sites, as well as any sites
that change scanners during the protocol will be required to qualify based on a screening questionnaire that
well be sent to all sites at the startup. Any such site will be provided with a Hoffman phantom (we have
purchased 4 at the start of ADNI) and will be provided with a technical manual (developed as part of ADNI)
and advice from the Banner team that participated in ADNI1 startup. All phantom images will be forwarded to
the University of Michigan where they will be reviewed and, if not approved, repeated. Site will be qualified
based upon interest, phantom imaging, and performance during ADNI1 (ie, any site with unacceptable
scanner QC performance will not be included).
4.4.3 Data acquisition
All scans will be acquired in pairs of [18F]FDG and [18F]AV45 scans, performed on separate days, between 1
day and 2 weeks apart, with either scan performed first. Subjects will be scheduled by the clinical sites,
working in conjunction with the PET centers. Prior to initiation of the study, logistical arrangements for
ordering and transporting the tracer will be worked out with each site as part of the startup.
The AV-45 protocol will entail the injection of 10 mCi of tracer followed by an uptake phase of 50 min
during which time the state of the subject is not important. At 50 minutes subjects will be positioned in the
scanner and 2 x 5 min frames of emission data collected. PET/CT scans will precede this acquisition with a
CT scan for attenuation correction; PET-only scanners will perform a transmission scan following the emission
PHS 398/2590 (Rev. 11/07)
Weiner, MW
scan. As we have done to date in ADNI, sites will be required to use a single iterative reconstruction for all
scans that is optimized for the instrument and which cannot change during the protocol. The vast majority of
sites are experienced with this; new sites will be instructed as part of the qualification procedure.
FDG scans will be acquired as they have been in the ADNI protocol to date: subjects’ blood glucose is
checked prior to scanning and must be < 180 mg/dL. After the injection of 5 mCi of tracer, subjects are in a
quiet, dimly lit room with eyes and ears unoccluded for 30 min, after which they are placed in the scanner.
Data are acquired as 6 X 5 min frames preceded by a CT scan or followed by a positron transmission scan.
At the time of imaging, all sites will fill out a PET imaging form as we have instituted for ADNI1. This
provides data on important parameters not captured routinely in all image headers such as amount of tracer
injected, exact times of injection, blood glucose level, subject state, etc. These data are reviewed at the time
of the QC checks. These forms are filled out online by the PET technologist or study coordinator –
compliance has been high because they are required for reimbursement.
4.4.4 Data flow and QC
All data will be uploaded to the UCLA Laboratory of Neuroimaging (LONI) as we have done to date with ADNI.
Instruction in this protocol is provided as part of site qualification and all PET sites are currently familiar with
this. Data are de-identified as part of the upload and placed into quarantine until they pass QC. Dr Koeppe’s
laboratory at the University of Michigan is notified when new scans are uploaded, and QC is performed within
24 hours followed by pre-processing of the images.
Procedures for assuring scan quality involve the following and will be performed on all human PET scans.
1. Download all PET data sets from LONI
2. Convert raw Image data of any format to CTI format as needed and store on local computer.
3. Raw Image QC Process
a. Visual inspection of all images: including both frames (temporal) and planes (spatial).
b. Extract and inspect all header information, and check versus required scan protocol.
c. Co-register (six degrees of freedom, rigid-body) all frames to first frame of the raw dynamic image set.
Assess subject motion by magnitude of translation and rotatation parameters.
d. Recombine co-registered frames to create both registered dynamic and registered average (averaged
over all frames) image sets in native image geometry and orientation.
e. Determine image quality metrics (global correlation, global mean square error, global absolute error)
both between frames on raw dynamic image data sets both pre- and post-coregistration. This includes
comparisons between all frames pairs (e.g. 15 comparisons for a 6-frame study)
f. Inspect PET scan information form completed by site for each scan. Note errors and correct.
g. Complete PET QC form (e.g. pass, fail with reprocessing, fail with rescan, fail without rescan).
4. Pre-analysis processing steps:
a. Reorient and resample baseline FDG-PET images into a standard image matrix and image orientation
(160x160x96 voxels; 1.5mm voxel size in all three dimensions).
b. Co-register all PET scans on each subject (all AV45 scans including baseline, and all repeat FDG
scans) to the baseline FDG scan reoriented into the standard image matrix in step a).
c. As above, create dynamic and averaged image sets in standard image matrix
d. Perform image intensity normalization on all FDG data sets, to set global average of the normalized
and thresholded image set to 1.0 (iterative process that makes average of all voxels above 0.5 equal
to 1.0). Renormalization of FDG images to pons, vermis, or other mildly affected areas will be
performed for specific analyses, however global normalization is a convenient step in intensity scaling
images sets from all different scanners to a common value. AV45-PET scans will be normalized to
cerebellar gray matter, such that its value will be 1.0.
e. Smooth images from all PET scanner models/vendors by an amount determined from Hoffman
phantom scans, in order to achieve a uniform effective resolution of 8 mm FWHM.
f. Determine image quality metrics between scans, as above for within-scans. This step is key as it
identifies problems between scans, such as motion induced artifacts, etc. This is very important and
mandatory step for all within-subject analyses.
g. Upload all “pre-analysis” processed images to LONI (four image sets total).
As noted, these procedures have all been successfully employed as part of ADNI to date for both FDG and
PIB tracers and many of the procedures for standardization of images have been published[28].
PHS 398/2590 (Rev. 11/07)
Weiner, MW
4.4.5 Data Analysis
Data will be analyzed by 4 laboratories that have each participated in the analysis and reporting of PET data
in ADNI, and who have reviewed accomplishments in the progress report section of this grant. Each
laboratory will use an independent method of analyzing all data obtained as part of ADNI2, and in conjunction
with the biostatistics core, will use results to test the hypotheses outlined in the specific aims. Below we
summarize the major approaches to data analysis. As we have done for ADNI1, all results will be tabulated
and uploaded so that summary numerical measures will be available immediately on the web. Each site will
also upload any templates, regions, and final image results for free access by the scientific community.
4.4.5.1 UC Berkeley/Jagust Laboratory: This group will analyze all FDG and AV-45 data. In addition, the AV45 analyses will be performed in collaboration with the Mathis group at the University of Pittsburgh. The
approach taken to FDG data will be identical to that used in ADNI1. This is described in the Progress Report
section above, and also in a recently published paper[29]. The result of this analysis stream yields 5 separate
ROIs (right and left angular gyri, bilateral posterior cingulate gyrus, and left middle/inferior temporal gyrus)
that we have averaged together to yield a single composite ROI. In order to reduce type I error, we will use
this single ROI but can use the individual ROIs to explore findings that are significant or that require a more
specific topographical approach. While we recognize that our approach is somewhat simple, it has a number
of major advantages including automated implementation and standardization, as well as limiting the number
of potential statistical tests because of the a priori selection of a very limited number of ROIs.
AV45 data will be analyzed using a more anatomical approach. We will define ROIs using a standard
template in standard space – the AAL atla[48]. The AAL atlas will be “reverse normalized” to the spatial
dimensions of the AV45 images, and counts will be extracted from these ROIs in order to widely sample
cortex, using regions that include prefrontal cortex, lateral parietal, medial parietal (precuneus/posterior
cingulate), and lateral temporal cortex. A cerebellar ROI will be defined as a reference tissue. We have used
these methods in our laboratory in Berkeley with [11C]PIB, and note that a number of potential problems have
either been solved or do not occur[49, 50]. For example, despite the relatively low count rate in cortex of Aβnegative control subjects, [11C]PIB images are easily normalized to either MRI or PET templates. The
process of reverse-normalization, using a segmented PET or MRI mask if necessary has also been validated
in our laboratory and is widely used in other labs[4]. Some of the variables in PET analysis will be explored in
conjunction with the Mathis lab, and these include selection of appropriate templates, other approaches to
ROI selection (which could also utilize freesurfer or hand-drawn ROIs), and different methods of optimizing
ROI selection in longitudinal scan comparisons. The fundamental datum of these AV45 analyses will thus be
tracer uptake in an ROI normalized to cerebellum, or basically a standardized uptake value ratio (SUVR). We
will explore different combinations of SUVRs to capture both the topographical extent and amplitude of tracer
uptake, in a manner similar to that which we and many other laboratories have used for [11C]PIB. One
desirable approach is to define a global index of tracer uptake that will be accomplished by averaging a large
number of cortical ROIs in which Aβ is deposited – ie frontal, temporal, parietal, precuneus/posterior
cingulate. This Aβ index can then be used either as a continuous variable or dichotomized into “Aβ+” and
“Aβ-“. We will explore different approaches to the classification of subjects, but one promising method for this
was developed by our colleagues at Pittsburgh and uses an iterative approach[44]. The very large sample
size of normal subjects will provide a strong resource for characterizing subjects. Alternatively, we can use
data collected as part of the Avid phase III program, process the images to ADNI standards, and develop
cutoffs in a separate group of subjects for validation in ADNI. Regardless of the exact approach, the end
result of this data analysis stream will be both a single global value for AV45 uptake that can be used in
continuous and dichotomous analyses, and regional values that can be explored as potentially more sensitive
markers.
4.4.5.2 Banner Alzheimer’s Institute/Reiman Laboratory: This group is responsible for voxel-based analyses
of AV45 and FDG PET data using SPM5 along with its empirically predefined “statistical region-of-interest
(sROI) strategy” and its recently developed “weighted hypometabolic convergence index (HCI)’ strategy (see
proress report). Statistical brain mapping strategies will be used to analyze continuous PET measurements on
a voxel-by-voxel basis in the different subject groups, thus providing regional information about betweengroup differences in baseline AV45 and FDG PET measurements, the extent to which these baseline PET
measurements are correlated with clinical disease severity and subsequent rates of clinical decline, and 24month changes in AV45 and FDG PET measurements. An AV45 PET sROI will be empirically characterized
and an sROI-to-cerebellar AV45 SUVR threshold determined using an ROC-derived sensitivity and specificity
PHS 398/2590 (Rev. 11/07)
Weiner, MW
analysis using a previously acquired data set from AD patients and young adult APOE4 non-carriers, who are
least likely to have fibrillar amyloid deposition. The empirically pre-defined threshold will then be used in the
independent ADNI2 data set to categorize the proportion of AV45-positive subjects in each group; to
characterize and compare the extent to which AV45-positive and negative scans are related to clinical
measures of disease severity, FDG PET measures of AD-related hypometabolism (including the researchers’
proposed AD-related HCI) and subsequent rates of metabolic and clinical decline in each subject group; and
to estimate the number of AV45-positive subjects in each group needed to evaluate AD-slowing treatments in
24-month, multi-center, parallel group, placebo-controlled RCTs. The HCI values from all subjects as well as
the threshold value differentiationg AD patients from controls found in ADNI1 data will be applied to subjects
in ADNI2 to predict decline and estimate sample sizes for modeled clinical trials.
sROI Analyses: The sROI strategy will be used to track 24-month CMRgl declines and evaluate ADslowing treatment effects in each of relevant subject groups with optimal statistical power and freedom from
the Type 1 error associated with multiple comparisons. As reported in the Progress Report section (4.3
above), we will use SPM5 in batch mode on a training data set to empirically define the sROI (the set of
voxels consistently associated with 24-month CMRgl decline in the relevant patient group), reference region
(to normal FDG PET scans for the variation in absolute measurements) and Gaussian smoothing filter (if any)
associated with the most significant 24-month CMRgl decline[38]. For each FDG PET training data set
analysis, the following candidate reference regions will be evaluated and compared: SPM5-defined whole
brain and Automatic Anatomical Labeling (AAL)-defined cerebellar, sensorimotor, and pontine ROIs, as well
as spared sROIs defined using the set of voxels associated with AD-related or progressive regional-to-whole
brain increases (interpreted as relevant sparing and previously found primarily in white matter) and different
significance thresholds. SPM5 will then be performed using the predefined sROI, reference region and
smoothing filter on an independent ADNI2 data set to characterize 24-month sROI-to-reference region CMRgl
declines and estimate the number of people in the particular subject group (AD, EMCI, LMCI) needed to
detect a 25% treatment effect in a 24-month multi-center RCT with 80% power, two-tailed testing, and P=0.05.
In order to confirm the sample size estimates for AD and MCI generated in ADNI1, the same sROI and spared
reference region will be applied to newly recruited AD, EMCI, and LMCI patients in ADNI2.
The sROI strategy will also be used to provide an sROI-to-cerebellar AV45 SUVR measurement and
threshold for the categorization of scans as AV-45 positive or negative. SPM5 will be performed in batch
mode on AV45 PET scans previously acquired in probable AD patients and adult APOE4 carriers in Avid’s
registered multi-center study to find the sROI (the set of voxels consistently associated with AD-related
increases in cerebral-to-cerebellar AV45 SUVR) and Gaussian smoothing filter (if any) associated with the
most significant AD-related SUVR increases. The AAL defined bilateral cerebellar crus I regions 91–92 will
provide the cerebellar reference region for the assessment of cerebral-to-cerebellar SUVR. An ROC analysis
will be used to define the sROI-to-cerebellar SUVR with the best trade-off between specificity and sensitivity
in the classification of cases and controls.
Weighted HCI Analyses: As previously noted, the automated AD-related “weighted hypometabolic
convergence index analysis” technique generates a single “hypometabolic convergence index” (HCI) to
characterize the extent to which both the magnitude and spatial distribution of hypometabolism in an
individual subject is related to the magnitude and spatial distribution of AD patients in comparison to the
measurements in NC subjects. This HCI will be used as a measure of AD-related disease severity in the
analyses proposed. An ROC analysis of ADNI1 AD and control data will generate the threshold for subject
classification of MCI patients and thus as a predictor of decline. In ADNI2, HCI’s will be computed in every
scan and an empirically preselected index will be used to classify scans as HCI-positive or negative.
While all of the proposed analyses are important, the research group is most interested in determining
the extent to which AV45-positivity and HCI-positivity, alone, together or in combination with other
measurements, predicts subsequent rates of progression to a more impaired clinical category in MCI, EMCI
and normal older adult APOE4 carriers and non-carriers, as well as the estimated number of AV45-positive
MCI, EMCI and cognitively normal older-adult APOE4 carriers and non-carriers needed to evaluate ADslowing treatments in a 24-month multi-center RCT using FDG PET, and clinical endpoints.
4.4.5.3 University of Utah/Foster Laboratory: As noted in the progress report, this group will continue to apply
3D-SSP approaches to all data with modifications for AV45-PET. Numerical summary measures for baseline
FDG-PET scans will be: 1) the number of surface pixels with values > 2 standard deviations (SD) below the
normal control population, 2) the number of surface pixels with values > 3 SD below the normal controls, 3)
PHS 398/2590 (Rev. 11/07)
Weiner, MW
sum of Z score values > 2 SD below normal controls, and 4) sum of Z scores >2SD below normal based upon
a regression of pons values in serial studies of an individual Numerical summary measures for AV45 scans
will be derived from values normalized to the cerebellum (rather than pons as we use in FDG-PET studies).
Regional values will be 1) average cerebral cortical binding, 2) insula binding, 3) medial parietal binding and
4) putamen binding. In addition, we will provide 4 measures based upon 3D-SSP Z-score maps analogous to
those we are submitting for FDG-PET: 1) the number of surface pixels with values > 2 standard deviations
(SD) below the normal control population, 2) the number of surface pixels with values > 3 SD below the
normal controls, 3) sum of Z score values > 2 SD below normal controls, and 4) sum of Z scores >2SD below
normal controls based upon a regression of cerebellar values in serial studies of an individual. We recognize
that identifying the best outcome measures for AV45 scans is not certain and we will periodically re-evaluate
these selections as our experience grows.
Classifying Abnormalities in Individual AV45-PET and FDG-PET scans: A major focus of ADNI2 is the
identification of scans as “normal” or “abnormal”. The approach to this in our lab is under development, but
will involve 3 potential methods. The first is to use the existing reference data set collected by Avid of AV45
normals and define statistical significance between an individual scan and this reference database using Z
scores on a voxel by voxel basis. A second approach will be to use an iterative outlier approach as
suggested by AIzenstein et al[44]. A third approach will be to use a multivariate approach with partial least
squares (PLS) regression. The PLS analysis will use AV45 SSP maps regressed upon diagnostic group
using a training data set. Individual subjects scans can be characterized using the PLS model and defining a
score of amyloid distribution based on how well the latent variable (amyloid distribution) of the subject is
related to the AD or control latent variable.
Statistical hypothesis testing: Once SSP maps are made and subjects’ scans classified, these variables
can be used to test the specific hypotheses as outlined in the final section of the application. For example, we
will examine the relationship between clinical outcomes – as either change in diagnosis or rate of change on
cognitive tests – and baseline FDG-PET or AV45-PET. Similarly, to evaluate potential use of AV45-PET as
an outcome measure in clinical trials, we will evaluate the 8 numerical summary scores described above in
terms of rates of change and relationship to cognition, and we can use these analyses to model clinical trials
and define sample sizes necessary to detect effects of amyloid-lowering drugs. The classification of subjects
based on the approaches described above can also be used to select subjects in modeled clnical trials. For
example, we expect that patients with an FTD-like pattern of hypometabolism will have progressive decline,
but in different measures than typical for AD. We also expect that VMCI and MCI subjects with abnormal
metabolic and amyloid binding maps will have greater rates of conversion to AD and greater decline on
cognitive measures.
4.4.5.4 University of Pittsburgh PET Laboratory: Some of the variables in PET analysis will be explored in in
collaborative efforts between the Pittsburgh and Berkeley groups, and these include selection of appropriate
templates (such as the AAL template, the MCI-Template (see PET Core Progress Report), and the probability
template[51] relative to the “gold standard” of carefully hand drawn template regions on each subject’s MRI)
and the effects of atrophy and white matter contributions. An advantage of the use of MRI template-based
ROI sampling methods for PET data analysis is the rapid standardized regional analysis of data for large
subject cohorts such as that proposed for ADNI2. However, the quality of the normalization may be affected
by the amount and location of age- or disease-related atrophy that could greatly impact the extracted PET
values. The AAL template is based on an averaged single subject MRI of a young individual, and its
applicability and accuracy in assessing amyloid PET studies in elderly control, MCI, and AD subjects has not
been critically evaluated. A basic goal is to critically evaluate the regional SUVR measures for the AAL
template, MCI-Template, and probability template (PT) relative to hand drawn SUVRs. In contrast to AAL,
the PT approach utilizes MR images with delineated ROIs for multiple individual subjects to define ROI
locations in native subject space that represent the maximum overlap of the individual normalized template
ROIs. The Pittsburgh PET group has utilized MR images for typical elderly control (n=6, 78±8 yrs) and AD
(n=6, 78±3 yrs) subjects to produce an elderly control and an AD PT, each with 14 defined regions. MCI
subjects are expected to fall between these two groups, but it may be necessary to define an MCI probability
template as well. An MRI-based atrophy correction method will be used to correct the AV45-PET data for the
dilutional effect of expanded cerebrospinal fluid (CSF) spaces accompanying normal aging and diseaserelated cerebral atrophy[52]. It is important to correct for atrophy when comparing in vivo PET amyloid data to
postmortem data, as the latter is relative to tissue mass and not total brain volume measures [53]. Atrophy
PHS 398/2590 (Rev. 11/07)
Weiner, MW
corrections may have a significant impact in determining AV45-positive control and MCI subjects by removing
atrophy as a contributing variable across the subject groups. The widely used two-component (2C) approach
involves segmentation of the MR image into (1) gray and white matter and (2) CSF voxels. The 2C method
has been widely applied in quantitative PET binding investigations that include PIB PET studies of amyloid
deposition[42, 54, 55]. For ADNI2, tissue segmentation will be performed using the SPM5 unified approach
to generate 3 tissue maps (gray matter (GM), white matter (WM), and CSF). For the 2C correction, gray and
white matter maps will be combined, a binary map will be generated (gray+white =1 and CSF =0) and
smoothed to match the PET scanner resolution (~8 mm), and regional correction factors will be sampled.
The atrophy correction factors determined for the various template methods will be compared. These atrophy
corrections can also be applied to data analyzed in the other PET analysis laboratories, and results
compared. Finally, the white matter (WM) contribution to AV45 uptake will be examined using the WM maps
generated from the SPM5 segmentations. We will examine AV45 WM uptake through evaluation of SUVR
image intensity in areas that overlap those voxels corresponding to 80-100% probability of WM in the
smoothed WM tissue maps.
Statistical Analyses/Hypothesis testing
The hypotheses for this Core are listed in the Specific Aim section and generally fall into five categories: 1)
hypotheses related to comparison of measures across diagnostic groups (for example, hypothesis for Specific
Aim 5); 2) correlations between baseline PET measures and baseline clinical outcomes (for example,
hypothesis for Specific Aim 6); 3) associations between baseline PET measures and cognitive decline (for
example, hypothesis for Specific Aim 7); 4) correlations between change in PET measures and change in
other imaging, cognitive, or CSF biomarker measures (for example, hypothesis for Specific Aim 10); and 5)
comparisons across PET measures on sample size, correlation with cognitive decline, or association with
clinical progression (for example, hypotheses for Specific Aims 8 and 11). We briefly state below the methods
to be used for each of these types of analyses, but more details may be found in the Biostatistics Core,
Section 8.4.2. These analyses focus on the numerical summaries generated from the PET scans (including
the sROI from the voxel-based methods), but similar techniques may be used on the voxel-level data as
described in Sections 4.3.3.2 and 4.4.5.2).
Analysis of variance and chi-square tests will be used to test hypotheses on differences between
continuous and categorical measures by diagnostic group (Category 1). Linear regression methods will be
used to assess associations between PET and clinical measures at baseline (Category 2). Mixed effects
regression models will be used to assess associations with cognitive decline (Category 3). An extension of
mixed effects regression models, called simultaneous random effects models, which allows for multiple types
of longitudinal outcomes, will assess correlations between change in two (or more) different measures. For
the analyses in this category that focus on AV45 measures or are restricted to EMCI subjects, in which we will
only have two measurements per person, linear regression models using difference scores for both the
predictor of interest and the outcome will be used (Category 4). Finally, a standardized framework for
comparing different fluid and imaging biomarkers on a set of criteria, including precision to measure change
(related to sample size calculations) as well as clinical validity (correlation with cognitive decline or clinical
progression), was developed for ADNI 1 and will be used to address all hypotheses related to comparing
across MRI measures or across MRI measures and CSF biomarkers (Category 5).
For these analyses, we will have approximately 238 Normals and LMCI carried-over from ADNI 1, 200
EMCI from GO, and 150 Normals, LMCI, and AD subjects newly recruited in ADNI 2. An additional 100 EMCI
will also be recruited in ADNI 2. All of these subjects will receive FDG-PET and AV45-PET scans. With these
sample sizes, we should have ample power to address all of our hypotheses (please see Biostatistics Core,
Section 8.4.2.4 for the specific power calculations.)
References
1. Shaw LM, Vanderstichele H, Knapik-Czajka M, et al. Cerebrospinal fluid biomarker signature in
Alzheimer's disease neuroimaging initiative subjects. Ann Neurol;65:403-413, 2009.
2. Vemuri P, Whitwell JL, Kantarci K, et al. Antemortem MRI based STructural Abnormality iNDex (STAND)scores correlate with postmortem Braak neurofibrillary tangle stage. Neuroimage;42:559-567, 2008.
3. Vemuri P, Wiste HJ, Weigand SD, et al. MRI and CSF biomarkers in normal, MCI, and AD subjects:
predicting future clinical change. Neurology;73:294-301, 2009.
PHS 398/2590 (Rev. 11/07)
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
Weiner, MW
Sun FT, Schriber RA, Greenia JM, He J, Gitcho A, Jagust WJ. Automated template-based PET region of
interest analyses in the aging brain. Neuroimage;34:608-617, 2007.
Langbaum JB, Chen K, Lee W, et al. Categorical and correlational analyses of baseline
fluorodeoxyglucose positron emission tomography images from the Alzheimer's Disease Neuroimaging
Initiative (ADNI). Neuroimage;45:1107-1116, 2009.
Klunk WE, Engler H, Nordberg A, et al. Imaging brain amyloid in Alzheimer's disease with Pittsburgh
Compound-B. Ann Neurol;55:306-319, 2004.
Avila J, Hernandez F. GSK-3 inhibitors for Alzheimer's disease. Expert Rev Neurother;7:1527-1533,
2007.
Buttini M, Akeefe H, Lin C, et al. Dominant negative effects of apolipoprotein E4 revealed in transgenic
models of neurodegenerative disease. Neuroscience;97:207-210, 2000.
Tuszynski MH, Thal L, Pay M, et al. A phase 1 clinical trial of nerve growth factor gene therapy for
Alzheimer disease. Nat Med;11:551-555, 2005.
Aisen PS. Alzheimer's disease therapeutic research: the path forward. Alzheimers Res Ther;1:2, 2009.
Dubois B, Feldman HH, Jacova C, et al. Research criteria for the diagnosis of Alzheimer's disease:
revising the NINCDS-ADRDA criteria. Lancet Neurol;6:734-746, 2007.
Fagan AM, Mintun MA, Mach RH, et al. Inverse relation between in vivo amyloid imaging load and
cerebrospinal fluid Abeta42 in humans. Ann Neurol;59:512-519, 2006.
Jagust WJ, Landau SM, Shaw LM, et al. Relationships between biomarkers in aging and dementia.
Neurology;In press2009.
Morris JC, Roe CM, Xiong C, et al. APOE predicts Abeta but not Tau Alzheimer's pathology in cognitively
normal aging. Ann Neurol In press.
Okello A, Koivunen J, Edison P, et al. Conversion of amyloid positive and negative MCI to AD over 3
years. An 11C-PIB PET study. Neurology 2009.
Wolk DA, Price JC, Saxton JA, et al. Amyloid imaging in mild cognitive impairment subtypes. Ann
Neurol;65:557-568, 2009.
Mattsson N, Zetterberg H, Hansson O, et al. CSF biomarkers and incipient Alzheimer disease in patients
with mild cognitive impairment. Jama;302:385-393, 2009.
Visser PJ, Verhey F, Knol DL, et al. Prevalence and prognostic value of CSF markers of Alzheimer's
disease pathology in patients with subjective cognitive impairment or mild cognitive impairment in the
DESCRIPA study: a prospective cohort study. Lancet Neurol;8:619-627, 2009.
Snider BJ, Fagan AM, Roe C, et al. Cerebrospinal fluid biomarkers and rate of cognitive decline in very
mild dementia of the Alzheimer type. Arch Neurol;66:638-645, 2009.
Engler H, Forsberg A, Almkvist O, et al. Two-year follow-up of amyloid deposition in patients with
Alzheimer's disease. Brain;129:2856-2866, 2006.
Jack CR, Jr., Lowe VJ, Weigand SD, et al. Serial PIB and MRI in normal, mild cognitive impairment and
Alzheimer's disease: implications for sequence of pathological events in Alzheimer's disease.
Brain;132:1355-1365, 2009.
Small GW, Kepe V, Ercoli LM, et al. PET of brain amyloid and tau in mild cognitive impairment. N Engl J
Med;355:2652-2663, 2006.
Shoghi-Jadid K, Small GW, Agdeppa ED, et al. Localization of neurofibrillary tangles and beta-amyloid
plaques in the brains of living patients with Alzheimer disease. Am J Geriatr Psychiatry;10:24-35., 2002.
Tolboom N, Yaqub M, Van der Flier WM, et al. Imaging Alzheimer Pathology in vivo: quantitative
comparison of 11C PIB and 18FFDDNP. Alzheimer's and Dementia;4:T6, 2008.
Shin J, Lee SY, Kim SH, Kim YB, Cho SJ. Multitracer PET imaging of amyloid plaques and neurofibrillary
tangles in Alzheimer's disease. Neuroimage;43:236-244, 2008.
Agdeppa ED, Kepe V, Liu J, et al. Binding characteristics of radiofluorinated 6-dialkylamino-2naphthylethylidene derivatives as positron emission tomography imaging probes for beta-amyloid
plaques in Alzheimer's disease. J Neurosci;21:RC189, 2001.
Bresjanac M, Smid LM, Vovko TD, Petric A, Barrio JR, Popovic M. Molecular-imaging probe 2-(1-[6-[(2fluoroethyl)(methyl) amino]-2-naphthyl]ethylidene) malononitrile labels prion plaques in vitro. J
Neurosci;23:8029-8033, 2003.
Joshi A, Koeppe RA, Fessler JA. Reducing between scanner differences in multi-center PET studies.
Neuroimage;46:154-159, 2009.
PHS 398/2590 (Rev. 11/07)
Weiner, MW
29. Landau SM, Harvey D, Madison CM, et al. Associations between cognitive, functional, and FDG-PET
measures of decline in AD and MCI. Neurobiol Aging 2009.
30. Chetelat G, Desgranges B, De La Sayette V, Viader F, Eustache F, Baron JC. Mild cognitive impairment:
Can FDG-PET predict who is to rapidly convert to Alzheimer's disease? Neurology;60:1374-1377, 2003.
31. Drzezga A, Grimmer T, Riemenschneider M, et al. Prediction of individual clinical outcome in MCI by
means of genetic assessment and (18)F-FDG PET. J Nucl Med;46:1625-1632, 2005.
32. Chen K, Reiman EM, Huan Z, et al. Linking functional and structural brain images with multivariate
network analyses: a novel application of the partial least square method. Neuroimage;47:602-610, 2009.
33. Reiman EM, Chen K, Liu X, et al. Fibrillar amyloid-beta burden in cognitively normal people at 3 levels of
genetic risk for Alzheimer's disease. Proceedings of the National Academy of Sciences of the United
States of America;106:6820-6825, 2009.
34. Reiman EM, Caselli RJ, Yun LS, et al. Preclinical evidence of Alzheimer's disease in persons
homozygous for the e4 allele for apolipoprotein E. N Engl J Med;334:752-758, 1996.
35. Reiman EM, Chen K, Alexander GE, et al. Functional brain abnormalities in young adults at genetic risk
for late-onset Alzheimer's dementia. Proceedings of the National Academy of Sciences of the United
States of America;101:284-289, 2004.
36. Small GW, Mazziotta JC, Collins MT, et al. Apolipoprotein E type 4 allele and cerebral glucose
metabolism in relatives at risk for familial Alzheimer's disease. Jama;273:942-947, 1995.
37. Small GW, Ercoli LM, Silverman DH, et al. Cerebral metabolic and cognitive decline in persons at genetic
risk for Alzheimer's disease. Proceedings of the National Academy of Sciences of the United States of
America;97:6037-6042, 2000.
38. Chen K, Langbaum JBS, Fleisher AS, et al. Twelve-Month Metabolic Declines in Probable Alzheimer’s
Disease and Amnestic Mild Cognitive Impairment Using an Empirically Pre-Defined Statistical Region-ofInterest: Findings from the Alzheimer’s Disease Neuroimaging Initiative. Submitted.
39. Minoshima S, Frey KA, Koeppe RA, Foster NL, Kuhl DE. A diagnostic approach in Alzheimer's disease
using three-dimensional stereotactic surface projections of fluorine-18-FDG PET. J Nucl Med;36:12381248, 1995.
40. Foster NL, Heidebrink JL, Clark CM, et al. FDG-PET improves accuracy in distinguishing frontotemporal
dementia and Alzheimer's disease. Brain;130:2616-2635, 2007.
41. Foster NL, Wang AY, Tasdizen T, Fletcher PT, Hoffman JM, Koeppe RA. Realizing the potential of
positron emission tomography with 18F-fluorodeoxyglucose to improve the treatment of Alzheimer's
disease. Alzheimers Dement;4:S29-36, 2008.
42. Lopresti BJ, Klunk WE, Mathis CA, et al. Simplified quantification of Pittsburgh Compound B amyloid
imaging PET studies: a comparative analysis. J Nucl Med;46:1959-1972, 2005.
43. Segonne F, Dale AM, Busa E, et al. A hybrid approach to the skull stripping problem in MRI.
Neuroimage;22:1060-1075, 2004.
44. Aizenstein HJ, Nebes RD, Saxton JA, et al. Frequent amyloid deposition without significant cognitive
impairment among the elderly. Arch Neurol;65:1509-1517, 2008.
45. Mathis CA, Price JC, Klunk WE, et al. Longitudinal PIB measures in control, MCI, and AD subjects. J
Nucl Med;49:35P, 2008.
46. Jacobson NS, Truax P. Clinical significance: a statistical approach to defining meaningful change in
psychotherapy research. J Consult Clin Psychol;59:12-19, 1991.
47. Choi SR, Golding G, Zhuang Z, et al. Preclinical properties of 18F-AV-45: a PET imaging agent for A-beta
plaques in the brain. J Nucl Med;In press2009.
48. Tzourio-Mazoyer N, Landeau B, Papathanassiou D, et al. Automated anatomical labeling of activations in
SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain.
Neuroimage;15:273-289, 2002.
49. Rabinovici GD, Jagust WJ, Furst AJ, et al. Abeta amyloid and glucose metabolism in three variants of
primary progressive aphasia. Ann Neurol;64:388-401, 2008.
50. Mormino EC, Kluth JT, Madison CM, et al. Episodic memory loss is related to hippocampal-mediated
beta-amyloid deposition in elderly subjects. Brain;132:1310-1323, 2009.
51. Svarer C, Madsen K, Hasselbalch SG, et al. MR-based automatic delineation of volumes of interest in
human brain PET images using probability maps. Neuroimage;24:969-979, 2005.
PHS 398/2590 (Rev. 11/07)
Weiner, MW
52. Meltzer CC, Kinahan PE, Greer PJ, et al. Comparative evaluation of MR-based partial-volume correction
schemes for PET. Journal of Nuclear Medicine;40:2053-2065, 1999.
53. Ikonomovic MD, Klunk WE, Abrahamson EE, et al. Post-mortem correlates of in vivo PiB-PET amyloid
imaging in a typical case of Alzheimer's disease. Brain;131:1630-1645, 2008.
54. Price JC, Klunk WE, Lopresti BJ, et al. Kinetic modeling of amyloid binding in humans using PET imaging
and Pittsburgh Compound-B. J Cereb Blood Flow Metab;25:1528-1547, 2005.
55. Lowe VJ, Kemp BJ, Jack CR, Jr., et al. Comparison of 18F-FDG and PiB PET in cognitive impairment. J
Nucl Med;50:878-886, 2009.
PHS 398/2590 (Rev. 11/07)
Weiner, Michael
Core: 5
Title of Core (not to exceed 81 spaces): Biomarker Core
Core Leader: Trojanowski, John, Q.
Position/Title: Professor, University of Pennsylvania
Department, service, laboratory, or equivalent: Pathology and Laboratory Medicine
Mailing Address:
3400 Spruce Street
7 Malony Bldg.
Philadelphia, PA 19104
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
The Biomarker Core in ADNI-2 will bank and curate biofluids from ADNI-1, ADNI GO and ADNI-2 subjects,
distribute ADNI samples to investigators qualified by the ADNI Resource Allocation Review Committee
(RARC) and study promising AD biomarkers. For example, ADNI-1 data on cerebrospinal fluid (CSF) Aβ and
tau established the diagnostic and predictive importance of these CSF analytes, while our Aβ plasma data
enable us to determine if plasma Aβ levels predict conversion from mild cognitive impairment (MCI)
including early MCI (EMCI) and late MCI (LMCI) to AD and/or reflect AD progression. We also identified 20
new promising AD biomarkers in CSF and plasma that, together with BACE, are promising for further study
in ADNI-2. Thus, the Aims of the ADNI-2 Biomarker Core are to: (1) Receive, aliquot, store, curate and track
all samples collected from ADNI-1, ADNI GO and ADNI-2 subjects, (2) Continue studies of CSF Aβ42, total
tau and phosphorylated tau (taup181) as well as plasma Aβ42 and Aß40; (3) Validate promising new CSF
and plasma biomarkers including BACE that we identified as having potential diagnostic applications for AD;
(4) Partner with the ADNI Industrial Scientific Advisory Board (ISAB) and investigators outside ADNI in
RARC approved “add-on” biomarker studies; (5) Collaborate with all ADNI Cores in studies of biomarker,
clinical, imaging and autopsy data; (6) Collaborate with World-Wide ADNI (WW-ADNI) sites in Europe,
Japan, Korea, China and Australia in studies of previously collected and new biomarker data. By
implementing these aims, we will test the Biomarker Core hypotheses that a panel of CSF/plasma
biomarkers (rather than any single analyte) will: (1) Predict conversion from normal to MCI or to AD and
conversion from MCI to AD as well as identify MCI subjects who have stable MCI and do not convert to AD;
(2) Reflect the progression of AD from its prodromal phase through to early/moderate stages of AD; (3)
Predict the likelihood of healthy brain aging or resistance to AD in the normal control (NC) population.
By accomplishing these Aims of the Penn Biomarker Core in ADNI-2, we will advance understanding of the
applications of validated and new AD biomarkers as predictive, diagnostic and progression markers from NC
to eMCI/MCI and thence to AD thereby contributing to the mission of ADNI-2.
University of Pennsylvania, Dept of Pathology & Laboratory Medicine
DUNS:
04-225-0712
Street 1:
7 Malony Bldg
City:
Street 2:
Philadelphia
Province:
County:
3400 Spruce Street
Philadelphia
USA
PA-002
Country:
State:
Zip/Postal Code:
PA
19104
DUNS:
Street 1:
Street 2:
City:
Province:
County:
Country:
State:
Zip/Postal Code:
PHS 398 (Rev. 11/07)
Page 2
Form Page 2
Weiner, Michael
Name
Organization
Role on Project
Leslie Shaw
SHAWLMJ
University of PA
PI-core director
John Trojanowski
trojanowski
University of PA
PI-core director
Name
Organization
Role on Project
Virginia Lee
University of PA
Collaborator
No
Yes
Cell Line
PHS 398 (Rev. 11/07)
Page 3
Weiner, Michael
5. BIOMARKER CORE:
Co-Core Leaders: Leslie M. Shaw, Ph.D. and John Q. Trojanowski, M.D, Ph.D.
5.1. Specific Aims: 1
Building on our progress in ADNI1, the Specific Aims of the Biomarker Core in ADNI2 are to bank and curate
biofluids from ADNI1, GO and ADNI2 subjects, distribute ADNI samples to investigators qualified by the RARC
and study promising AD biomarkers. Selection of biomarkers for study in ADNI2 is based on: 1) ADNI1
Biomarker Core studies; 2) other AD biomarker research. For example, ADNI1 established the diagnostic and
predictive importance of CSF Aβ and tau, while our Aβ plasma data enable us to determine the utility of
measuring plasma Aβ to predict conversion from EMCI/LMCI to AD or to monitor AD progression. With support
from other sources, we identified promising new biomarkers in studies of ~1500 non-ADNI CSF and plasma
samples from AD and NC followed at Penn using ADNI standard operating procedures (SOPs). We
interrogated these samples with the Rules Based Medicine (RBM) DiscoveryMAP panel of 189 analytes, and
identified ~20 potential AD biomarkers. Finally, others showed that BACE is promising for study in ADNI2.
Thus, the Specific Aims of the ADNI2 Biomarker Core are designed to test the following hypotheses:
a) A panel of CSF/plasma biomarkers (rather than any single analyte) will predict conversion from
normal to EMCI/LMCI or to AD and conversion from LMCI to AD as well as identify LMCI subjects
who have stable LMCI and do not convert to AD.
b) A panel of CSF/plasma biomarkers will reflect the progression of AD from its prodromal phase
through to early and moderate stages of AD
c) A panel of CSF/plasma biomarkers will predict the likelihood of healthy brain aging or resistance to
AD in the normal control (NC) population.
To accomplish this, the Specific Aims of the Penn Biomarker Core in ADNI2 are to:
1) Continue to receive, aliquot, store, curate, and track all samples collected from subjects in
ADNI1, GO and ADNI2
2) Continue biomarker studies of CSF Aβ42, total tau and taup181 as well as plasma Aβ42 and Aβ40.
3) Validate promising new biomarkers including BACE, and analytes identified by the RBM
DiscoveryMAP 189 analyte panel.
4) Partner with the ADNI industrial scientific advisory board (ISAB) and investigators outside ADNI
in RARC approved “add-on” biomarker studies.
5) Collaborate with all ADNI Cores in analyses of biomarker, clinical, imaging and autopsy data.
6) Collaborate with World-Wide ADNI (WW-ADNI) Sites in Europe, Japan, Korea, China and
Australia in joint studies of previously collected and new biomarker data.
By implementing these Aims to test our hypotheses, we will advance understanding of the applications of
validated and new AD biomarkers as predictive, diagnostic and progression markers from NC to EMCI/LMCI
and thence to AD thereby contributing to the mission of ADNI2.
5.2.1. AD and its Sequential Phases: The most common dementia is AD [1, 2], and AD’s hallmark lesions are
Aβ plaques and tau neurofibrillary tangles (NFTs). Clinical symptoms relate to NFTs [3], neurodegeneration
and synapse loss [4, 5]. AD can be divided into t3 phases: 1) A pre-symptomatic phase in which subjects are
cognitively normal but have AD pathology [4, 6, 7]; 2) A prodromal phase known as EMCI/LMCI [8]; 3) A phase
when patients show dementia and functional losses. Although it has been suggested that diagnostic criteria for
early AD should be redefined by the presence of memory impairments plus biomarker evidence of AD [9], this
is controversial despite studies showing AD biomarkers predict conversion from LMCI to AD [10-19], so the
diagnosis of AD still requires the presence of dementia [20]. A key feature of the AD model that informs our
ADNI2 Biomarker Core studies (see Fig. 1) is the bi-phasic nature of AD [21]. This is based on the view that
AD begins with abnormal processing of amyloid precursor protein (APP) thereby increasing brain Aβ [22] which
leads to neuron dysfunction and death [23]. Our model also emerges from our ADNI1 studies and other
research, and it assumes a lag phase between Aβ deposition and neuron loss, as well as differences in brain
resiliency, cognitive reserve or other factors likely account for the variable duration of this lag phase [24].
1
Citations in the text identified by * are publications from the Penn Biomarker Core of ADNI1 that are listed at the end of
the Progress Report while those without * are listed in Literature Cited. Due to the explosion in publications on AD
biomarkers, primary citations within the last 4 years are emphasized and earlier citations are included in reviews.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Additional pathologies (e.g. alpha-synuclein, TDP-43 lesions) also may contribute to clinical variations in AD
patients [25].
Figure 1: Proposed model illustrating the
ordering of biomarkers of AD pathology relative
to stages in the clinical onset and progression
of AD. Clinical disease, on the horizontal axis,
is divided into three stages; cognitively normal
or pre-symptomatic, EMCI, LMCI and dementia.
The vertical axis indicates the range from
normal to abnormal for each of the biomarkers
as well as memory and functional impairments
(ADL = activities of daily living). Amyloid
imaging and CSF Aβ are biomarkers of brain Aβ
amyloidosis. CSF tau and FDG PET are
biomarkers of neuron injury and degeneration
while structural MRI is a biomarker of abnormal
brain morphology.
5.2.2. Biomarkers of AD Pathology: This Core focuses on plasma and CSF biomarkers, among which
reduced CSF Aβ42 and elevated CSF tau are the most well established, but BACE and analytes in the RBM
DiscoveryMAP panel show promise as discussed below. Further, FDG PET, PET amyloid imaging, and
structural MRI are established AD biomarkers, and collaborations between the Biomarker Core and other ADNI
Cores integrates these biomarkers (*2, *6, *7, *10, *12, *15, *18, *19). Both CSF Aβ42 and PIB amyloid PET
imaging are biomarkers of brain Aβ amyloidosis. Nearly all clinically diagnosed AD subjects are PIB positive
[26, 27], and there is excellent correspondence between PIB imaging and fibrillar deposits of Aβ in brain [28].
Abnormally low CSF Aβ42 correlates with the clinical diagnosis of AD and Aβ pathology at autopsy [29, 30]
while there is ~100% concordance between low CSF Aβ42 and PIB positivity [31, 32]. Thus, PIB imaging and
CSF Aβ42 are biomarkers of brain Aβ amyloidosis [18, 33]. CSF tau is an indicator of tau pathology as well as
neuronal injury and both phospho-tau and total tau CSF levels increase in AD [34], while higher CSF tau levels
correlate with greater cognitive impairment in individuals on the NC to LMCI to AD spectrum [16, 35]. Elevated
CSF tau also reflects neuronal damage in ischemic [36] and traumatic [37] brain injury, while elevated CSF tau
in AD appears to result from tau pathology that leads to neurodegeneration, and the release of tau into the
CSF [38, 39]. Elevated CSF tau in AD correlates with NFTs at autopsy [40], but high CSF tau levels are not
seen in tauopathies such as progressive supranuclear palsy (PSP) and corticobasal degeneration (CBD) [41,
42]. This could be due to the effects of extracellular Aβ plaques on the clearance of tau in AD brains,
differences in the species of pathological tau in AD versus CBD and PSP or other unknown factors. FDG PET
is an indicator of impaired synaptic function and AD patients show a specific topographic pattern of decreased
glucose uptake [43]. Greater decreases in FDG uptake correlate with greater cognitive impairment from NC to
LMCI and then to AD [44]. Imaging and autopsy studies show good correlation between FDG PET and
postmortem evidence of AD [45]. Structural MRI is a measure of the loss of synapses and neurons [46].
Volumetric or voxel based measures of brain atrophy correlate with measures of cognitive impairment in NC,
LMCI and AD. Thus, rates of neuron and synaptic loss reflect the progressive rate of brain atrophy and
correlate with rates of cognitive decline [47, 48], while the degree of atrophy correlates with NFTs [49-51] in
subjects who have undergone antemortem MRI and postmortem AD staging [52].
5.2.3. Temporal Ordering of Biomarkers in AD: Figure 1 relates AD stage to AD biomarkers, and it is based
on the following assumptions: 1) these biomarkers become abnormal before clinical symptoms appear; 2) Aβ
biomarkers become abnormal before tau and neurodegenerative biomarkers; 3) tau and neurodegenerative
biomarkers correlate with clinical disease severity; 4) these biomarkers are temporally ordered. Growing
evidence supports these assumptions. Thus, 20-40% of elderly NC subjects have evidence of brain
amyloidosis by PIB imaging or CSF Aβ42 [16, 53-55]. These data agree with autopsy studies [4, 6, 7] and
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
support the view that brain Aβ alone does not produce dementia, but precedes clinical symptoms [56, 57].
Indeed, Aβ deposition may begin years before AD symptoms [58]. Aβ and NFT pathology can be present in
NC subjects, although NFTs tend to be confined to entorhinal cortex in NC, but are more widespread in AD [4,
6, 7]. Studies show that MRI, FDG PET, and CSF tau are abnormal in the LMCI phase of AD [16, 44, 59], that
abnormal neurodegenerative AD biomarkers precede the first cognitive symptoms of AD, and that FDG PET
abnormalities precede cognitive symptoms in subjects who later progress to AD [60, 61]. FDG PET studies are
abnormal in elderly NC who carry the APOEε4 allele [62, 63], and rates of MRI atrophy become abnormal in
subjects who later progress to AD although some subjects are still cognitively normal [64-66]. Further, rates of
MRI atrophy are greatest in subjects with clinical AD, least in NC subjects and intermediate in LMCI subjects
while rates of PIB accumulation do not differ by clinical group [58]. Thus, MRI rates map to cognitive
deterioration in AD, but rates of PIB change do not [58, 67]. Rates of brain atrophy correlate well with NFT and
other neurodegenerative pathology, but not with severity of Aβ pathology at autopsy [68]. PIB positive NC may
have normal MRIs, implying that Aβ can accumulate without affecting brain volumes [69] and data suggested
that the direct substrate of memory impairment is hippocampal volume on MRI, not PIB imaging [70]. Thus, Aβ
deposition appears to be an early event that nears a plateau prior to the appearance of atrophy on MRI and
cognitive symptoms, while CSF tau and MRI biomarkers are more closely linked to cognitive symptoms than
Aβ deposits [70, 71], but correlations with cognition are stronger with structural MRI than CSF tau in LMCI or
AD subjects [72]. However, both MRI and CSF tau are predictive of future conversion from LMCI to AD [72],
but CSF tau does not change once subjects have progressed to clinical AD [73] while rates of atrophy on MRI
are significantly greater in AD patients than in elderly NC subjects [74, 75].
5.2.4. Conclusions: Based on our model (Fig. 1), we propose studies of specific AD biomarkers for diagnosis
and disease staging. The staging biomarkers and their temporal relationships with the phases of AD discussed
here present opportunities to test our Biomarker Core hypotheses. Further, because of the highly interactive
nature of ADNI, the Biomarker Core works with other ADNI and non-ADNI investigators to test the Biomarker
Core hypotheses and those of ADNI2 overall.
5.3. Progress:
The Penn Biomarker Core made considerable progress to implement its ADNI1 Aims including establishing
biofluid collection, shipping and storage SOPs, an archive of ADNI biofluids and conducting studies of these
fluids as summarized below. We also established the RARC, conducted meetings with ISAB members and
other biomarker scientists, established an international CSF quality control program for quality assessment of
biomarker assays in WW-ADNI, and support for ADNI “add-on” studies. Notably, all ADNI biomarker data are
posted on the ADNI website (http://www.adni-info.org/index) immediately after they are obtained. Finally,
ADNI1 Biomarker Core progress also is demonstrated by 19 published/in press papers listed below as well as
10 submitted papers.
5.3.1. Development of SOPs for Biofluid Collection, Shipping, Aliquoting, Storage and Curation: The
Penn Biomarker Core established ADNI biomarker SOPs at the outset of ADNI1 (see ADNI website for
details). This was done in consultation with ISAB and other biomarker scientists for collection, handling,
shipment, labeling, aliquoting, storage and tracking 24/7x365 days/year of DNA, CSF, plasma, serum and
urine samples. We also worked with the Clinical Core to develop biofluid tracking forms that provide a detailed
history for each sample. These SOPs are essential to assure: a) sample integrity; b) accurate identification of
samples received and aliquots prepared from them; c) sample stability. An example of the value of detailed
characterization of each collected biofluid is documenting the time to freezing on dry ice at study sites (see Fig.
2 below). Thus, for the RARC approved proteomic studies of ADNI plasma and CSF samples, data on the time
at room temp before freezing each sample could be informative for interpreting results on time/temp sensitive
AD biomarkers, and detailed sample history permits selection of samples using a specific time stipulation, e.g.
1-2 hrs at room temp., but not longer. Another aspect of sample timing involves CSF collection where knowing
the length of time from collection to time of transfer is important. For example, some early CSF samples were
mistakenly collected into polystyrene collection tubes at the sites. This was rapidly corrected, but this
information enabled us to understand the effects of this collection error, and simulations of the effects of brief
exposure to polystyrene alleviated concerns about this potential confound. Moreover, Figure 3 below illustrates
the fastidious tracking by ADNI sites of how samples are obtained and processed since the average time CSF
was in contact with any transfer tube was 25.7 min for ADNI Baseline CSF samples thereby precluding a
significant exposure time to any CSF collection tube.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
5.3.2. Current Status of the ADNI1 Biofluid Archive: Shortly after SOPs for the collection, processing, bar
code labeling, packaging and shipment of biofluids for APOE genotyping, cell immortalization and biomarker
studies were finalized, these SOPs were incorporated into the ADNI procedure manual and distributed to all
ADNI sites. Following infrastructure
Figure 2
Figure 3
CSF and Plasma Samples
CSF Samples
Time from Collection to Freezing
160
140
Sample
Sa m ple CCount
o un t
Time from Collection to Transfer
CS F BL
CS F M12
Plasm a BL
Plasm a M12
N
120
100
80
CSF Yr 1
407
312
Average time (min)
36.5
35.2
95% CI (min)
32.7-40.3
30.6-39.7
Plasma BL
Plasma Yr1
816
722
N
60
CSF BL
Average time (min)
70.2
68.2
95% CI (min)
67.3-72.9
65.2-71.1
160
V6- Month 24
V4- M12
V2-Baseline
140
Sample
Count
120
CSF BL
100
407
312
Mean
25.7 min
24.0 min
95% CI
22.2 -29.2 min
19.8 -28.2 min
80
60
CSF Yr 1
N
40
40
20
20
0
0
0
10
20
30
40
50
60
70
80
90
15
30
100 110 120 130 140 150 160 170 180 240 >320
45
60
75
90
105
120
135
150
165
180
195 >210
Minutes
Time (mins)
Color Codes For Bar Graphs: In Figure 2, purple = CSF at baseline, green = CSF at year 1, red = plasma at baseline,
tan = plasma at year 1. In Figure 3 for CSF, purple = baseline, green = year 1 and blue = year 2.)
building to set up and manage the ADNI Biomarker Core laboratory and biofluid repository at Penn, the first
ADNI biofluids started to arrive in August 2005. Through April 30, 2007 a total of 1108 blood samples were
collected at the Screening visit and all were received and rapidly processed for APOE genotyping at Penn so
results could be entered into the ADNI database within a week of receipt to balance the ADNI cohort for APOE
status. Residual blood samples were stored at -80 0C for DNA preparation and genetic studies. As of June,
2009 a total of 12,053 ADNI biofluid samples have been received and processed (see Fig. 4). To respond to
Figure 4
the request of the Clinical Core that
clinic visits take place 5 days/week, we
ADNI Biofluid Samples Received
arranged for receipt and freezer
Total Fluids
storage of biofluids 6 days/week
Received
Number of Samples Received
APOE
1080
including Saturday. Thus far, a total of
Plasma
3344
119,106 aliquots of serum, plasma,
Serum
3344
CSF and urine have been prepared,
Urine
3429
bar code labeled and stored in
CSF
856
dedicated ADNI freezers at -80 0C.
Temperature monitoring of each
freezer is done 24/7/365 with a
telephone alarm system and one Penn
Biomarker Core staff is always on call
to respond to an alarm. For each
primary biofluid sample collected, the
following information is maintained in
the ADNI Biomarker Core database at Penn: biofluid type (CSF, plasma, serum, urine), coded subject and visit
ID, 6 digit license plate number, visit date and time, date and time of receipt, condition of samples as received,
biofluid sample volume and number of aliquots. The database is backed up daily on an external “brick” hard
drive and on a DVD disk. The latter are stored outside the Biomarker Core laboratory in a secure location in a
different building to assure data security in the event of a catastrophic failure of the server on which the
database resides.
5.3.3. Round Robin Study to Validate and Standardize Methods to Measure CSF Tau and Aβ: To reliably
measure Aβ and tau in ADNI CSF samples, the Biomarker Core identified sources of variation in quantifying
total tau, P-tau181p and Aβ42 that were shown by Luminex or ELISA methods to have at least 85% sensitivity
and 80% specificity for diagnosing AD, predicting LMCI progression to AD, and identifying elderly Clinical
CSF
ApoE
Plasma
3500
Serum
Urine
3000
2500
2000
1500
1000
500
Urine
Serum
10-12/2008
01-03/2009
04-06/2009
07-09/2008
04-06/2008
07-09/2005
10-12/2005
01-03/2006
04-06/2006
07-09/2006
10-12/2006
01-03/2007
04-06/2007
07-09/2007
10-12/2007
01-03/2008
0
PHS 398/2590 (Rev. 11/07)
Plasma
ApoE
CSF
Page
Weiner, Michael
Dementia Rating (CDR) scale 0 individuals likely to progress to CDR>0 (*1,*3,*4,*13-*17). We also validated
the INNO-BIA AlzBio3 reagents in a 7 site study (including academic and ISAB site) in follow up to our prequalification study wherein a standardized protocol for test performance was implemented. Each of two
analytical runs included 6 calibrators plus a 7th (lowest calibrator diluted in half with the kit diluent), two
aqueous-based quality control samples, 4 aqueous-based run validation samples containing reference
materials (synthetic peptides Aβ1-42, p-tau181p) or recombinant protein (total tau), spiked in diluent and 3 CSF
pools prepared from routine clinic CSF samples, aliquoted (0.5 cc) into polypropylene tubes and frozen at -80
0
C. This provided the required experience for each participating site in the round robin to perform in the
qualification study. The qualification study included the same calibrator scheme and control samples, 3 of the 4
run validation samples used in the pre-qualification study, and aliquots from 5 CSF pools including the same 3
used in the pre-qualification study and 2 prepared from AD patient CSF samples. Three analytical runs were
completed using the standardized test protocol. Briefly, the precision study demonstrated the following:
1. Repeatability and reproducibility
a. The within-center %CV for measurement of CSF biomarkers was ~10% over the 3 runs for each
biomarker and Figure 5 shows within-center (combined within-run and between run precision)
and between-center precision for CSF Aβ1−42 measured in 5 CSF pools in 7 laboratories.
Figure 5: Repeatability and Reproducibility
30
25
20
20.2
18.5
18.9
%CV 15
15.8
16.1
10
5
7.1
7.5
8.2
9.8
8.9
0
b et w een- cent er
00040
00041
w i t hi n- cent er
00042
CSF ID
00043
00044
b. For the CSF pools, the between-center variability is greater than that for the aqueous-based
controls, and studies are underway to determine possible explanations this. We collaborate with
the WW-ADNI biomarker quality control program to improved performance of these tests.
c. The results using the pre-made series of calibrators are highly reproducible between the 7
centers (see full report entitled “ADNI Interlaboratory Study” posted on the ADNI website).
2. Sample stability
a. Stability of the 3 aqueous-based and 3 CSF pools was demonstrated for the 8 month interval
between the pre-qualification and qualification studies. Ongoing studies in the Biomarker Core
are evaluating longer term stability.
b. The effect of brief exposure of 20 freshly obtained CSF samples to polystyrene for 1 hr at room
temp., as compared to polypropylene, decreased the concentration of Aβ42, by 14.5%, and tau
by 11%, but there was no change in P-Tau181P concentration.
3. ADNI CSF sample analyses
Tolerance intervals were generated from the CSF pool data for the Biomarker Core as a guide
for the acceptability of analytical runs of ADNI CSF samples.
4. Publication of results
These data together with the pre-qualification study data are being submitted for publication
(Shaw, et al, Submitted, 2009b). This paper reports on data from individual participating
laboratories regarding other aspects of CSF biomarker measurements such as comparison of
calibration algorithms (e.g., Luminex algorithm, Bio-Plex algorithm, Innogenetics algorithm), long
and short term stability of CSF samples, CSF freeze-thaw, Bio-Plex vs. Luminex (data
generated in the pre-qualification study by Kaj Blennow of EU-ADNI for two different Luminex
systems), linear response of biomarkers on dilution, pre-analytical issues concerning possible
diurnal changes in Aβ42 levels, etc. All these data were posted on the ADNI website
immediately after completion and brief summaries were distributed at ADNI meetings as well as
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
on the ADNI website. Moreover, we published a detailed review (*17, see book chapter in
Appendix 1) of the procedures required to bring potential AD biomarkers from concept to their
implementation as useful biomarkers with particular emphasis on CSF Aβ and tau.
5.3.4. Baseline Studies of CSF Tau and Aβ in >400 ADNI Subjects: This study of ADNI baseline CSF
samples developed a pathological CSF biomarker signature for AD (*15). To do this, Aβ42, total tau (t-tau) and
tau phosphorylated at threonine 181 (p-tau181p) were measured in: a) CSF samples obtained at baseline for
100 mild AD, 196 LMCI and 114 elderly NC subjects in ADNI; b) an independent set of 56 autopsy-confirmed
AD subjects and 52 age-matched elderly NC followed in the Penn ADCC with ADNI SOPs using the Luminex
multiplex immunoassay. Detection of an AD CSF profile for t-tau, p-tau181 and Aβ42 in ADNI subjects was
achieved using receiver operating characteristic (ROC) cutpoints and logistic regression models derived from
the autopsy-confirmed CSF biomarker data. Our data showed that CSF Aβ42 was the most sensitive
biomarker for AD detection in CSF from non-ADNI autopsy-confirmed subjects with an ROC area under the
curve of 0.913 and sensitivity for AD detection of 96.4% (see Table 1). A unique bimodal characteristic of the
distribution of CSF Aβ42 was detected in each ADNI subgroup (Fig. 6), and a logistic regression model for
Aβ42, t-tau and APOε4 allele count provided the
Table 1: ROC curve parameters for non-ADNI autopsy-based AD cases vs NC
tau
ROC AUC
Threshold value
Sensitivity (%)
Specificity (%)
Test accuracy (%)
Positive predictive value (%)
Negative predictive value (%)
0.831
93 pg/mL
69.6
92.3
80.6
90.7
73.8
Aβ42
0.913
192 pg/mL
96.4
76.9
87.0
81.8
95.2
p-tau181p
0.753
23 pg/mL
67.9
73.1
70.4
73.1
67.9
tau/Aβ42
0.917
0.39
85.7
84.6
85.2
85.7
84.6
p-tau181p/Aβ42
0.856
0.10
91.1
71.2
81.5
77.3
88.1
LRTAA
model
0.942
0.340
98.2
79.5
89.9
85.7
97.2
Figure 6: CSF Aβ42 concentration distribution plots for ADNI AD, LMCI and NC subjects
A: ADNI AD subjects
B: ADNI LMCI subjects
C: ADNI NC subjects
best delineation of mild AD. Application of cut points for the 4 most sensitive parameters for AD CSF
pathology, i.e. Aβ42, t-tau/Aβ42, p-tau181/Aβ42 and the LRTAA (Logistic Regression of Tau, Aβ and APOEε4
alleles) model, showed the presence of an AD-like CSF profile in 89.1%, 91.8%, 94.6% and 89.1%,
respectively in the 37 LMCI subjects who converted to probable AD at 12 months. The potential robustness of
these AD indices of risk for LMCI subjects to convert to probable AD will only be realized after a sufficient
period of time has elapsed to fully determine all converters to probable AD or other dementias and to
document the clinical stability of the non-converters. Support for our hypothesis that these AD biomarker
indices will be reliable harbingers of probable AD comes from our finding that a comparable incidence of CSF
AD profiles was observed for LMCI→AD converters at 24 months (89.7%, 88.2%, 92.6% and 88.2%,
respectively). Thus, the pathological CSF biomarker signature of AD we defined effectively detects mild AD in
a large multisite prospective clinical investigation, and this signature appears to predict conversion from LMCI
to AD. The cutoff values established by Shaw et al (*15) were validated in a follow up study with EU-ADNI and
ISAB collaborators (Ewers et al, Submitted, 2009b) wherein we identified AD biomarker patterns in an
independent manner, without clinical diagnoses using a mixture modeling approach to analyze the ADNI CSF
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
data in Shaw et al. (*21). This analysis was validated on two additional data sets, one of which was an
autopsy-confirmed EU cohort. Using the ADNI data set, a CSF Aβ42/ p-tau181p biomarker mixture model
identified one feature linked to AD, while the other matched the NC status. The AD signature was found in
90%, 72%, and 36% of patients in the AD, LMCI, and NC groups, respectively. The NC group with the AD
signature was enriched in APOEε4 allele carriers. Results were validated on other data sets (Ewers et al.,
Submitted, 2009a and b). In one data set on 68 EU autopsy-confirmed AD cases, 64/68 patients (94%
sensitivity) were correctly classified with the CSF AD feature. In another data set with LMCI patients followed
for 5 years by EU-ADNI investigators, the model showed a sensitivity of 100% (57/57) in patients progressing
to AD. These data accurately classified AD patients independent of clinical diagnosis. Further substantiation of
these diagnostic cutpoints is illustrated in Figure 7 which shows that there is a significantly greater
hippocampal atrophy rate, and lateral ventricular volume increase rate (data not shown) during the first year
after entry into ADNI for subjects with CSF Aβ42 levels below the 192 pg/mL cutoff concentration compared to
those with concentrations above the cutoff.
Figure 7: Hippocampal (left+right) atrophy rates (% volume change below or above the Aβ42 cutpoint
concentration value of 192 pg/mL) in ADNI subjects who provided baseline CSF. Note that individuals below the
192 pg/ml Aβ threshold (tan boxes) show greater atrophy than individuals above this threshold (green boxes).
Aβ threshold = 192 pg/mL
Hippocampal atrophy rates (% volume change)
Aβ42
<192 pg/mL
- 5.6±4.7
All subjects
- 8.0±5.9
AD
- 4.8±3.6
LMCI
- 3.6±3.2
NC
Aβ42
>192 pg/mL
- 2.6±4.1
- 4.2±3.5
- 2.9±3.7
- 2.2±4.3
We extended these analyses in collaboration with EUADNI colleagues (Ewers et al., Submitted, 2009a and
b). Briefly, we developed a diagnostic classification
model combining core psychometric and biological
markers to detect AD. A total of 345 ADNI subjects
including 81 AD patients, 163 amnestic LMCI (aLMCI)
patients and 101 elderly NC were assessed. Predictor
variables included: 1) CSF measures and ratios of total
tau, p-tau181, Aβ1-42; 2) MRI volumetric measures of the
left and right hippocampus and entorhinal cortex; 3)
scores for RAVLT and ADAS subtests on memory. After a mean follow-up of 1.5 years, 50/163 aLMCI patients
converted to AD. Neuropsychological predictors were ADAS delayed recall, and total-immediate and 30-min.delayed recall of the RAVLT (classification accuracy = 81%) for the discrimination between aLMCI-AD and NC.
Unilateral hippocampus volume (left or right) improved prediction of aLMCI-AD converters vs. NC reaching a
re-sampling validated sensitivity of 88.9% and specificity of 96.76%. The ratio of CSF concentration of total
tau/Aβ42 added independently to prediction accuracy. Similar to the CSF-ratio of total tau/Aβ42, a recent
autopsy-confirmed AD and NC set of cases established the AD-biomarker signature including CSF-tau, Aβ42,
and ApoE genotype and this contributed to the left hippocampus volume plus memory model, with the
extended model reaching an overall classification accuracy of 95%. We next examined the association
between these CSF biomarkers and body mass index (BMI) in the same ADNI subjects and, for crossvalidation, in 56 patients with AD and 43 NC from the memory clinic at the Ludwig Maximilian University,
Germany (Ewers et al., Submitted, 2009a). Briefly, we found a low BMI is indicative of typical AD CSF
biomarker alterations in LMCI and AD. Finally, we collaborated with Hopkins investigators to investigate the
effect of CSF abnormalities on rate of decline in everyday function in NC, LMCI and AD (Okonkwo et al,
Submitted, 2009). Briefly, CSF t-tau, p-tau181p, and Aβ42 data from 114 NC, 195 LMCI patients, and 100 mild
AD ADNI subjects and their Functional Activities Questionnaire (FAQ) and ADAS-Cog data were analyzed. All
CSF analytes were associated with functional decline in LMCI, and all but t-tau/Aβ42 were associated with
functional decline in controls. Across all diagnostic groups, persons with a combination of tau and Aβ42
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
abnormalities exhibited the steepest rate of functional decline. Hence, persons with AD-like CSF tau and Aβ42
abnormalities are at greatest risk of functional impairment. Finally, we now characterize longitudinal changes of
CSF Aβ42, t-tau/Aβ42, p-tau181p/Aβ42 concentrations in 112 subjects (17 AD, 58 LMCI, 37 NC) who provided at
least 3 longitudinal samples: at baseline, 12 mos and either 24 or 36 mos. Thus far, the following annual rates
of change for each study group were observed for Aβ42, t-tau, p-tau181p, t-tau/Aβ42, and p-tau181p/Aβ42. AD:
Aβ42 (-2.6%, p<0.005), t-tau (-0.49%, p=0.74), p-tau181p(-3.0%,p<0.05), t-tau/Aβ42(2.27%, p=0.35), ptau181p/Aβ42(-0.06%, p=0.86); LMCI:Aβ42(-1.4%,p<0.005), t-tau(3.6%, p=0.052), p-tau181p(0.84%,p=0.53), ttau/Aβ42(5.0%,p<0.005), p-tau181p/Aβ42(2.6%, p<0.01); NC: Aβ42 (-1.9%, p<0.005), t-tau (3.2%, p=0.11), ptau181p(2.7%,p<0.05), t-tau/Aβ42(5.8%, p<0.005), p-tau181p/Aβ42(5.5%, p<0.0001). These data indicate that in
the AD cohort there is no significant annual % change for t-tau or the t-tau/Aβ42 and p-tau181p/Aβ42 ratios, but
there were small negative changes in Aβ42 and p-tau181p. In the LMCI cohort the small negative change in
Aβ42 and the positive change in the t-tau/Aβ42, p-tau181p/Aβ42 ratios were consistent with disease progression
for the group while in the NC group the most significant changes (increases) were in the two ratios t-tau/Aβ42
and p-tau181p/Aβ42. These data analyses are limited by the sparse number of study subjects and limited
longitudinal time periods (there were only 23 subjects, 16 NC and 7 LMCI, who provided 36 month samples;
there were 92 subjects, 25 NC, 51 LMCI and 16 AD who provided 24 month samples; 4 subjects provided both
month 24 and 36 samples). Further analyses on additional subjects and over a longer study period will be
necessary to provide more definitive statements and analyses regarding longitudinal changes of these CSF
biomarkers.
5.3.5. Assessment of Sulphatides, Homocysteine and Isoprostanes as AD Biomarkers in ADNI1
Sulphatides: Plans to validate CSF sulphatides as AD biomarkers in ADNI1 were based on promising reports
in 2004, but shortly thereafter we learned from David Holtzman that his colleagues could not replicate their
initial findings. Thus, we did not pursue sulphatides and we re-budgeted the savings from this change to
support the 2.5-fold increase in the number of CSF samples we obtained for >50% of ADNI1 subjects.
Homocysteine: ADNI1 included plans to study homocysteine in plasma and CSF. Using a validated enzyme
immunoassay methodology, we measured homocysteine in 813 BASELINE ADNI plasma samples. The data
summarized in Table 2 reveal that there was no significant difference in mean plasma homocysteine
concentration in AD vs LMCI, but there was between NC and AD (p<0.01)and NC and LMCI (p<0.01). These
Table 2: Baseline plasma homocysteine data for 813 ADNI subjects (μmole/L).
Baseline Plasma Homocysteine Data
AD
LMCI
NC
data are consistent with previous studies
193
393
227
N
showing an association between elevated
10.75
10.61
9.95
Mean
baseline
plasma
homocysteine
±3.27
±2.83
±2.80
SD
concentration and risk for development of
10.29-11.22
10.33-10.89
9.58-10.31
95% CI
AD (*14). Inclusion of homocysteine in the
p value, AD vs LMCI
0.599
logistic regression model described above
p value, AD vs NC
0.0075
for Aβ1-42, t-tau and APOε4 allele number
p value, LMCI vs. NC
0.0025
showed it was non-significant as a variable
in this model so the value of measuring
plasma homocysteine levels is uncertain, but additional analyses are underway for plasma homocysteine
including correlations with other biomarkers and longitudinal changes. Measurement of homocysteine
concentrations in CSF was achieved in 410 CSF samples collected at baseline using a validated enzyme
immunoassay (developed and performed by Merck) designed specifically to measure the much lower
concentrations present in this biofluid as compared to plasma. As shown in Table 3 below, there was no
significant difference between NC and either AD or LMCI in mean homocysteine values.
Isoprostanes: Since earlier studies suggested that F(2)-Isoprostanes reflect oxidative stress and may be AD
biomarkers (reviewed in *13,*14,*16), we proposed studies of the most relevant isoprostanes in ADNI1. To do
this, we developed and validated a semiautomated high-throughput HPLC tandem mass spectrometry assay
for plasma and urine 8-iso-PGF2a (*5). Concentrations of 8-iso-PGF2a in urine of 16 NC ranged from 55-348
ng/g creatinine. In 16 NC plasma samples, free 8-iso-PGF2a concentrations were 3-25 ng/L. Our fully validated
method enable analysis of >80 samples/day, and has the sensitivity to quantify 8-iso-PGF2a levels in NC
plasma and urine. We validated a liquid chromatography method with tandem mass spectrometry detection for
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
simultaneous analysis of 8-iso-PGF2a and 8,12-iso-iPF2a-VI (Korecka et al, Submitted, 2009) since these are
the most frequently studied isoprostanes in human CSF and brain tissue (*13,*14).
Table 3: Baseline CSF homocysteine data for 410 ADNI subjects (μmol/L).
Baseline CSF Homocysteine Data
An API 5000 triple quadrupole instrument with an
AD
LMCI
NC
APCI ion source was used here. Aliquots of CSF
100
196
114
N
(0.25 mL) were treated with a methanol:zinc
0.11
0.11
0.11
Mean
sulfate mixture followed by on-line cleaning on an
±0.02
±0.06
±0.03
SD
extraction
column.
Brain
samples
were
0.10-0.11
0.10-0.12
0.106-0.11
95% CI
homogenized and lipids extracted using Folch
p value, AD vs LMCI
0.38
solution. Solid phase extraction columns (C18)
p value, AD vs NC
0.66
were used for purification of isoprostanes.
p value, LMCI vs. NC
0.81
Chromatographic separation was performed using
an analytical column (Synergi C18 HydroRP) with 0.1% formic acid in water (A) and a mixture of methanol:
acetonitrile (80:20,B) (A:B-40:60, pH 6.15). The mass-spectrometer was operated in the negative ion mode
with MRM scanning. The quadrupoles were set to detect the molecular ions and high mass fragments m/z
353→193 (8-iso-PGF2a) and m/z 353→115 (8,12-iso-iPF2a-VI). We used this method to quantify both
isoprostanes in CSF from AD patients and age NC as well as in postmortem brains from AD and non-AD
controls. Our results (see Fig. 8A and 8B) do not confirm some previous reports that the CSF isoprostanes
studied here are useful AD biomarkers. However, we currently conduct similar isoprostane assays of ADNI
plasma samples to determine if plasma isoprostanes are informative for risk for conversion from LMCI to AD
since the elderly population studied here will have increased risk for cardiovascular and cerebrovascular
disease which could affect risk for AD.
Figure 8A: CSF 8-iso-PGF2a
Figure 8B: CSF 8,12-iso-iPF2a-VI
15
8-iso-PGF2α (pg/mL)
10
5
5.3±1.8
5.3±1.9
0
AD
Control
8,12-iso-iPF2α -VI (pg/mL)
40
p = 0.933
p = 0.219
30
20
10
17.4±5.0
0
18.6±5.2
AD
Control
5.3.6. Cross-sectional and Longitudinal Measures in NC, LMCI and AD Subjects in ADNI: In collaboration
with Petersen et al. (*10), we characterized NC, LMCI and mild AD subjects to assess the utility of
neuroimaging and chemical biomarkers in 819 subjects (229 NC, 398 with LMCI and 192 AD) enrolled at
baseline in ADNI and followed for 12 months. These studies show that the 12-month progression rate of LMCI
was as predicted, and the CSF measures heralded progression of clinical measures over 12 months.
5.3.7. Ventricular Expansion and CSF Biomarkers in NC, LMCI and AD Subjects in ADNI: A collaborative
study with the Thompson lab (*2) showed that lower CSF Aβ42 protein levels were correlated with lateral
ventricular expansion, and these studies show that ventricular expansion maps correlate with pathological CSF
and cognitive measures in AD.
5.3.8. Combined Analysis of PIB, PET, CSF Biomarkers and Cognition in ADNI Subjects: Collaborations
with the Jagust lab (*6) showed that PIB-PET was significantly correlated with Aβ42, t-tau, and p-tau181, while
FDG-PET only correlated with Aβ42. Thus, PIB PET and CSF biomarkers of Aβ agree with one another while
FDG-PET is modestly related to other biomarkers, but is better related to cognition.
5.3.9 Tensor Based Imaging, CSF Biomarkers and Cognition in ADNI Subjects: In a collaborative study
with Leow et al. (*7), we showed that serial MRI scans relate ongoing neurodegeneration to CSF biomarkers,
cognitive changes, and conversion from LMCI to AD.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
5.3.10. Hippocampal Volume Loss, APOE Genotype and CSF Biomarkers in early AD: A collaborative
study with the Weiner lab (*12) supports the concept that increased hippocampal volume loss is an indicator of
AD pathology and a potential marker for efficacy of therapeutic interventions in AD.
5.3.11 Comparing MRI and CSF Biomarkers in NC, LMCI and AD Subjects In ADNI: Collaborative studies
with Cliff Jack’s lab (*18,*19) showed that MRI and CSF tau/Aβ42 provide complimentary predictive
information about time to conversion from aLMCI to AD and combination of the two provides better prediction
than either source alone.
5.3.12. RBM Studies of CSF and Plasma from a Penn Cohort of Patients with AD and Other Disorders:
ADNI RBM “add-on” studies are in progress using plasma and CSF samples from the entire ADNI cohort, but
we completed similar studies funded by other sources using CSF and plasma from non-ADNI Penn cohorts of
patients with AD, frontotemporal lobar degeneration (FTLD), amyotrophic lateral sclerosis (ALS) or Parkinson’s
disease (PD) as well as NC to identify novel potential biomarkers. Interrogation of 1500 CSF and plasma
samples from these Penn cohorts is complete, and data analysis is in progress. We present here our data on
CSF samples collected from demented patients with longitudinal clinical characterization to autopsy
confirmation of AD, FTLD withTDP-43 positive inclusions (FTLD-TDP), or tau positive inclusions (FTLD-Tau),
and dementia with Lewy bodies. We interrogated these CSF samples with the RBM DiscoveryMap 189 analyte
panel using Luminex and identified CSF biomarkers associated with definite AD, FTLD-TDP, and FTLD-Tau
(Hu et al, Submitted, 2009). Several analytical strategies were used including significance analysis of
microarrays (SAM) and random forest analysis. Many analytes differed between AD and NC subjects, but few
differed between AD and non-AD dementias. Analytes that were significant by SAM were entered into a logistic
regression model by forward likelihood method, and the model achieved a sensitivity of 84.3% and specificity
of 84.5% for autopsy confirmed AD. Combining the SAM analytes with tau, Aβ42 and APOEε4 allele number in
the logistic regression model improved sensitivity to 97.1% and specificity to 89.7%. The new analytes
determined to be potential CSF AD biomarkers include: ENRAGE, fatty acid binding protein, resistin, IL-1alpha,
IgA, calcitonin, PDGF, C3, prolactin, IL23, TRAIL-R3, 1309, CgA, BMP6, NrCAM and AgRP. We also
investigated plasma biomarkers that distinguish between AD and elderly NC. Age and ApoEε4 allele number
achieved 60.5% sensitivity and 82.1% specificity. Combination of age and ApoEε4 allele number with plasma
analytes significantly improved sensitivity to 92.1% and specificity to 84.6%, and also captured over 90% of the
aLMCI cases. Plasma analytes in this RBM study that are potential AD biomarkers include: alpha-1antichymotrypsin, BDNF, MMP-7, NGAL, and PDGF. Thus, we have identified potential CSF and plasma AD
biomarkers, and completion of the RBM interrogation of the ADNI1 CSF and plasma samples in the coming 4
months will enable independent confirmation or rejection of these analytes for further study in ADNI2.
5.3.13. Round Robin Studies to Determine the Role of Plasma Aβ40/42 as AD Biomarkers: We recently
completed our Aβ plasma round robin study that includes 12 academic and industrial laboratories, including
ISAB sites (Figure 9). Each site used Innogenetics kits shown to be reliable for the quantification of
Figure 9 A,B &C: The plots below show the narrow range of % difference from the mean for Aβ42, Aβ40 and
Aβ42/Aβ40 measured in 15 plasma samples (from left to right in each plot below) in the ADNI Biomarker Core
Lab.
A. Plasma Aβ42
B. Plasma Aβ40
C. Plasma Aβ42/Aβ40
plasma Aβ (see see ADNI website for details). These studies show we can reliably assess measuring plasma
Aβ to predict conversion from LMCI to AD and/or to monitor AD progression in studies of plasma Aβ proposed
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
for ADNI2. Detailed statistical analyses are underway of the full data set, but Figure 9 above summarizes the
reproducibility data achieved by the Biomarker Core using the Innogenetics/Luminex multiplex assay for
plasma Aβ42/40 in this interlab round robin study. Thus, the Biomarker Core has demonstrated total
reproducibility (within + between day %CVs) ranging between 0.9 -4.9% for plasma Aβ42, 1.2 – 6.6% for Aβ40
and up to 7.6% for the ratio of Aβ42/Aβ40.
5.3.14. Leveraging and Outreach Within and Beyond ADNI by the Biomarker Core: As a result of the rapid
evolution and advances in biomarker technologies from the time the ADNI grant was written to the funding and
launch of the Penn Biomarker Core as well due to the changes in scope of ADNI1 after it began, we modified
resource allocation and plans as described above. Further, as summarized below, the ADNI Biomarker Core
successfully leveraged its assets to extend its resources to accomplish our ADNI1 Aims.
Leveraged Support for the ADNI Biomarker Core:
1. Innogenetics immunoassay reagents, ~$100,000 for reagents so far for the CSF biomarker studies.
Additional reagents will be provided as an “in-kind” donation for the year 1 and year 2 CSF sample
analyses. Also, the Biomarker Core recruited Innogenetics to join ADNI ISAB.
Leveraged “in-kind” support: $100,000
2. Luminex equipment: 2 Luminex systems donated by Luminex, Inc. at an estimated cost of
$100,000.00. Both systems are used for the CSF and plasma bead-based immunoassays for biochemical
biomarkers.
3. API5000 HPLC mass spectrometer. $420,000 of the total cost for this system was through internal
Penn resources independent of ADNI. This equipment is being used for analysis of plasma and CSF
samples.
4. CSF homocysteine analyses. CSF homocysteine analyses were performed by Merck Research
Laboratories at no cost to ADNI1.
5. ADNI repository space at the University of Pennsylvania Medical Center. Penn provided $400,000 to
renovate space for the ADNI biomarker biobank.
TOTAL LEVERAGED SUPPORT FOR THE ADNI BIOMARKER CORE: $1,032,330
With respect to outreach by the Biomarker Core, many studies reviewed above document the success of our
outreach and collaborative activities. Further, we worked with Dr. Tom Montine, Chair of the RARC in making
samples available to 7 RARC approved investigators for ADNI “add-on” studies (see ADNI website), and Dr.
Montine will to continue to Chair the RARC if asked to do so in ADNI2 (see Letter of Support). Also, in addition
to bi-weekly ADNI Executive Committee conference calls and presentations at ADNI meetings held 1-2 times
per year, Drs. Shaw and Trojanowski have given 14 invited lectures at national and international meetings on
their work in the ADNI Biomarker Core since 2005 and 13 of these have been within the last 2 years). Further,
they and other ADNI Core Leaders have met with FDA representatives to confer on how to use biomarkers as
surrogate markers in AD clinical trials and these meetings will continue in ADNI2. Finally, the Biomarker Core
partners with the Penn Center for Neurodegenerative Disease Research (CNDR) for a biomarker symposium
on 10/16/09 that is expected to draw >300 attendees. In summary, the ADNI Biomarker Core at Penn has
been extremely successful in its leveraging and outreach activities.
5.3.15. Concluding Remarks: The work summarized here demonstrates the exceptional progress made by
the ADNI Biomarker Core since 2004. Moreover, we pursued extensive collaborations with other investigators
within and outside ADNI to analyze ADNI data. In addition to the 9 peer reviewed publications cited above and
10 submitted papers, Biomarker Core investigators co-authored 10 invited reviews on ADNI and AD
biomarkers (*1, *3, *4, *8, *9, *11, *13, *14, *16, *17) including 3 with WW-ADNI colleagues (*3, *4, *17).
Hence, this Core is exceptionally well positioned to implement the Aims of ADNI2 and contribute to the mission
of ADNI2.
Biomarker Core Publications 2004-2009
1. Clark, C.M., Davatzikos, C., Borthakur, A., Newberg, A., Leight, S., Lee, V.M.-Y., Trojanowski, J.Q.
Biomarkers for early detection of Alzheimer pathology. NeuroSignals, 16:11-18, 2008. Paper accepted
prior to 4/7/08
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
2. Chou, Y.-Y., Leporé, N., Avedissian, C., Madsen, S.K., Parikshak, N., Hua, X., Shaw, L.M., Trojanowski,
J.Q., Weiner, M.W., Toga, A.W., Thompson, P.M., and the Alzheimer’s Disease Neuroimaging Initiative.
Mapping correlations between ventricular expansion, and CSF amyloid and tau biomarkers in 240
subjects with Alzheimer’s disease, mild cognitive impairment and elderly controls. Neuroimage, 46:394410, 2009. PLMCID: PMC2696357
3. Hampel, H., Shen, Y., Walsh, D.M., Aisen, P., Shaw, L.M., Zetterberg, H., Trojanowski, J.Q., and Blennow,
K. Biological markers of β-amyloid related mechanisms in Alzheimer’s disease. Exper. Neurol., In press,
2009.
4. Hampel, H., Blennow, K., Shaw, L.M., Hoessler, Y.C., Zetterberg, H., and Trojanowski, J.Q. Total and
phosphorylated tau protein as biological markers of Alzheimer’s Disease. Exper. Gerontol., In press,
2009.
5. Haschke, M., Zhang, Y.L., Kahle, C., Klawitter, J., Korecka, M., Shaw, L.M., Christians, U. HPLCatmospheric pressure chemical ionization MS/MS for quantification of 15-F2t-isoprostane in human urine
and plasma. Clin. Chem., 53:489-497, 2007. Paper accepted prior to 4/7/08
6. Jagust, W.J., Landau, S.M., Shaw, L.M., Trojanowski, J.Q., Koeppe, R.A., Reiman, E.M., Foster, N.L.,
Petersen, R.C., Weiner M.W., Price, J.C., Mathis, C.A., and the Alzheimer’s Disease Neuroimaging
Initiative. Relationships between biomarkers in aging and dementia. Neurol., In press, 2009.
7. Leow, A.D., Yanovsky, I., Parikshak, N., Hua, X., Lee, S., Toga, A.W., Jack, C.R. Jr., Bernstein, M.A.,
Britson, P.J., Gunter, J.L., Ward, C.P., Borowski, B., Shaw, L.M., Trojanowski, J.Q., Fleisher, A.S.,
Harvey, D., Kornak, J., Schuff, N., Alexander, G.E., Weiner, M.W., Thompson, P.M., and the Alzheimer's
Disease Neuroimaging Initiative. Alzheimer’s Disease Neuroimaging Initiative: A one-year follow up
study using tensor-based morphometry correlating degenerative rates, biomarkers and cognition.
NeuroImage, 45:645-455, 2009. PLMCID: PMC2696624
8. Mueller, S.G., Weiner, M.W., Thal, L.J., Petersen, R.C., Jack, C.R., Jagust, W., Trojanowski, J.Q., Toga,
A.W., and Beckett, L. Ways toward an early diagnosis in Alzheimer's disease: The Alzheimer's Disease
Neuroimaging Initiative (ADNI). Alzheimer’s & Dementia, 1:55-66, 2005. Paper accepted prior to 4/7/08
9. Mueller, S.G., Weiner, M.W., Thal, L.J., Petersen, R.C., Jack, C.R., Jagust, W., Trojanowski, J.Q., Toga,
A.W., and Beckett, L. The Alzheimer's Disease Neuroimaging Initiative. Neuroimaging Clin. N. Amer.,
15:869-877, 2005. Paper accepted prior to 4/7/08
10. Petersen, R.C., Aisen, P.S., Beckett, L.A., Donahue, M.J., Gamst, A.C., Harvey, D.J., Jack Jr., C.R.,
Jagust, W,J., Shaw, L.M., Toga, A.W., Trojanowski, J.Q., Weiner, M.W., and the Alzheimer’s Disease
Neuroimaging Initiative. Alzheimer’s Disease Neuroimaging Initiative (ADNI): Clinical characterization,
Neurol., In press, 2009.
11. Petersen, R.C., and Trojanowski, J.Q. Time for Alzheimer’s Disease biomarkers? Potentially yes for
clinical trials, but not yet for clinical practice. JAMA, 302(4):436-7, 2009.
12. Schuff, N., Woerner, N., Boreta, L., Kornfeld, T., Shaw, L.M., Trojanowski, J.Q., Thompson, P.M., Jack Jr,
C.R., Weiner, M.W., and the Alzheimer's Disease Neuroimaging Initiative. Hippocampal volume loss in
early Alzheimer’s disease in relation to ApoE genotype and biomarkers. Brain, 132:1067-1077, 2009.
PLMCID: PMC2668943
13. Shaw, L,M. PENN biomarker core of the Alzheimer's disease Neuroimaging Initiative. Neurosignals, 16:1923, 2008. PLMCID: PMC2696349
14. Shaw, L.M., Korecka, M., Clark, C.M., Lee, V.M..-Y., and Trojanowski, J.Q. Biomarkers of
neurodegenertaion for diagnosis and monitoring therapeutics. Nat. Rev. Drug Discovery, 6:295-303,
2007. Paper accepted prior to 4/7/08
15. Shaw, L.M., Vanderstichele, H., Knapik-Czajka, M., Clark, C.M., Aisen, P., Petersen, R.C., Blennow, K.,
Soares, H., Simon, A., Lewczuk, P., Dean, R., Siemers, E., Potter, W., Lee, V.M.-Y., Trojanowski, J.Q.,
and the Alzheimer's Disease Neuroimaging Initiative. Cerebrospinal fluid biomarker signature in
Alzheimer’s Disease Neuroimaging Initiative subjects. Ann. Neurol., 65:403-413, 2009.
PLMCID: PMC2696350
16. Trojanowski, J.Q. Searching for the biomarkers of Alzheimer’s. Practical Neurol., 3:30-34, 2004. Paper
accepted prior to 4/7/08
17. Vanderstichele, H., De Meyer, G., Shapiro, F., Engelborghs, B., DeDeyn, P.P., Shaw, L.M., and
Trojanowski, J.Q. Alzheimer’s disease biomarkers: From concept to clinical utility. In: Biomarkers For
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Early Diagnosis Of Alzheimer’s Disease, D. Galimberti, E. Scarpini (Eds.), Nova Science Publishers, Inc.,
Hauppauge, NY, pp. 81-122, 2008. Paper accepted prior to 4/7/08
18. Vemuri, P., Wiste, H.J., Weigand, S.D., Shaw, L.M., Trojanowski, J.Q., Weiner, M., Knopman, D.S.,
Petersen, R.C., Jack Jr, C.R., and the Alzheimer’s Disease Neuroimaging Initiative. MRI and CSF
biomarkers in normal, LMCI, AD: Diagnostic discrimination and cognitive correlations. Neurol., 73:287293, 2009. PLMCID: PMC2715214
19. Vemuri, P., Wiste, H.J., Weigand, S.D., Shaw, L.M., Trojanowski, J.Q., Weiner, M., Knopman, D.S.,
Petersen, R.C., Jack Jr, C.R., and the Alzheimer’s Disease Neuroimaging Initiative. MRI and CSF
biomarkers in normal, LMCI, AD: Predicting future clinical change. Neurol., 73:294-301, 2009.
PLMCID: PMC2715210
5.4. Research Design & Methods 2
To test our Biomarker Core hypotheses, we will implement the six Specific Aims of the Penn Biomarker Core in
ADNI2. Below, we outline the rationale for each of our Aims, the methods we will use to conduct the studies
proposed in these Aims, how data generated in these Aims will be interpreted and the significance of these
studies for understanding mechanisms of AD, for developing and validating informative AD biomarker tests and
for the mission of ADNI2.
5.4.1. Specific Aim 1: Continue to receive, aliquot, store, curate and track all samples collected from
subjects in ADNI1, GO and ADNI2.
Rationale: A key function of the ADNI2 Biomarker Core at Penn is to continue to receive, aliquot, store, curate
and track all samples collected from subjects in ADNI1, GO and ADNI2. This is essential to enable studies of
ADNI biofluid samples by investigators in this Core, as described in the subsequent Aims here, and by other
investigators who are approved by the RARC for “add-on” studies.
Methods: The same methods and SOPs implemented in ADNI1 proven to be reliable and effective over the
past 5 years will be used in ADNI2. These methods and procedures were summarized above in the Progress
Report, and they have been disseminated to all ADNI investigators and site personnel who have used them
effectively. They also have been distributed to WW-ADNI investigators to assist them with their SOPs, and to
promote harmonization of biomarker SOPs. To maximize broadest dissemination of these SOPs, they also are
available to download on the ADNI website by any investigator around the globe. Thus, the SOPs required to
implement this Aim in ADNI2 are up and running in the Biomarker Core as well as across all ADNI sites and
they are being adopted by collaborators in WW-ADNI. Below is a brief summary of biofluids archived in ADNI2.
BIOFLUIDS TO BE COLLECTED AND ALIQUOTS GENERATED IN GO AND ADNI2
ADNI-GO totals for plasma (1,412) and serum (1,412) = 2,824
ADNI-2 CSF totals = 1,560
ADNI-2 totals for plasma (7,192) and serum (7,192) = 14,384
GO and ADNI2 total collected CSF, plasma and serum samples = 18,768
Total # of aliquots banked from all the above CSF, plasma and serum samples = ~227,000
Interpretation: The continuing efforts by the ADNI Biomarker Core to receive, aliquot, store, curate and track
all samples collected from subjects in ADNI1, GO and ADNI2 is essential for all biofluid studies in ADNI2.
Significance: The significance of Aim 1 is that it implements a key function of this Core that is essential to
ADNI2.
5.4.2. Specific Aim 2: Continue biomarker studies of CSF Aβ42, total tau and taup181 as well as plasma
Aβ40/42.
Rationale: The Biomarker Core made substantial progress in validating optimal methods to measure and
interpret CSF Aβ42, total tau and p-tau181p as well as plasma Aβ40/42 using the Luminex/Innogenetics
immunoassay. As noted in Figure 1, our studies test the hypothesis that AD begins with Aβ deposition in
cerebral cortex, which in turn leads to synaptic dysfunction, neurodegeneration, and cognitive and functional
decline. Thus, Aim 2 will elucidate steps in the onset of EMCI/LMCI and progression to AD, improve diagnostic
methods for the early detection of AD and inform clinical trial design with high statistical power.
2
Please see the Clinical Core on exclusion/inclusion criteria for subjects diagnosed as NC, EMCI, LMCI and AD.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Methods: All of the methods to be used here were described above in the Progress Report, in publications
(*15,*17) and in reports posted on the ADNI website. Thus, all the methods needed to successfully implement
Aim 2 are up and running in this Core.
Interpretation: The analytical and biostatistical methods required to interpret the data generated in this Aim
were described in the Progress Report. Moreover, additional analytical strategies are presented in Aim 3 below
as well as in the Biostatistics Core led by Laurel Beckett. Thus, all the analytical methods needed to
successfully implement Aim 2 are up and running in this Core. Notably, we will focus on using the data
generated here to elucidate pathophysiological steps in the onset of EMCI/ LMCI and progression to AD, to
improve methods for the early and reliable detection of AD and to inform clinical trial design with high statistical
power as illustrated in several ADNI1 publications (*2,*6,*7,*10,*12,*15,*18,*19).
Significance: The significance of implementing Aim 2 is that it will enable the Biomarker Core to elucidate the
pathophysiology of progression from NC to EMCI/LMCI and thence to AD, as well as improve methods for
reliably detecting NC, EMCI/LMCI and AD. This will improve the design of clinical trials of new AD therapies.
5.4.3. Specific Aim 3: Validate promising new biomarkers including BACE, and analytes identified by
the RBM DiscoveryMAP panel.
Rationale: BACE-1 is a 501-amino acid-long type I transmembrane aspartyl protease that is considered to be
the primary β-site cleaving enzyme in the brain (*14) [76]. The extracellular domain contains two active
aspartyl residues at amino acid positions 93 and 289. Knockdown of BACE-1 in mice abolishes brain Aβ
production with no or mild phenotypic effects. In sporadic AD, there is an increase in BACE protein and BACE
enzymatic activity in postmortem brain that appears to correlate with Aβ load. Brain BACE activity also appears
to increase in humans with sporadic AD. Preliminary studies indicate an increase in CSF BACE enzymatic
activity in AD versus NC subjects. A 70 kDa protein corresponding to full-length BACE-1 has been suggested
as the source of BACE activity in human CSF and there is a modest increase in BACE activity in AD versus
NC subjects. Recently, ISAB collaborators at Merck developed a sensitive and specific assay for measuring
BACE-1 activity in CSF [76]. This assay uses a novel optimized BACE substrate, which, after cleavage, the
product can be detected using a neo-epitope antibody and a sensitive ELISA. Thus, given the preliminary data
on BACE [76], we will work with ISAB collaborators at Merck to use this assay in ADNI2 (see Letter of
Support). As for the rationale for the proposed RBM studies, this is summarized in the Progress Report
wherein we report the identification of novel and potentially informative plasma and CSF biomarkers that merit
further assessment here in collaboration with RBM (see Letter of Support).
Methods: BACE will be measured using the 2-step assay developed by Merck [76]. Briefly, cleavage of a
biotinylated peptide substrate is accomplished using either CSF or baculo BACE (bBACE). The extent of
enzymatic cleavage is detected using an ELISA. In the cleavage step, 25 μL of either bBACE (at up to 100 pM)
or 25 μL human CSF is added into a 96 well assay plate. To each of these wells, 25 μL of reaction buffer
containing 50 mM NaOAc, 0.01% BSA, 15 mM EDTA, 0.2% CHAPS, 1 mM Deferoxamine Mesylate and 20uM
Pepstatin A at pH 4.5 is added. The plate is gently agitated, and 100 μL of 200 nM substrate (biotinKTEEISEVNFEVEFR) is added to the wells. The plate is sealed, incubated at 37 °C and agitated at 40 rpm for
2.5 h. The reaction is arrested by adding 50 μL of 1 M Tris–HCl (pH 8.0). Next, the product of BACE enzymatic
cleavage from the above reaction “biotin-KTEEISEVNF” is measured by ELISA. The reaction mixture is
transferred into a streptavidin coated detection plate and incubated overnight at 4 °C. The plate is washed
three times with phosphate buffered saline with 0.1% Tween-20 (PBST) at pH 7.4 followed by addition of 100
μL of anti-NF c-terminal neo-epitope rabbit polyclonal antibody at 1:30,000 dilution in Superblock PBSTween20 (0.1%, v/v) and incubated for 1 h at room temp. Next, the plate is washed three times with PBST, and 100
μL of goat anti-rabbit IgG-AP at 1:30,000 dilution in 0.1% Tween-Superblock is added and incubated for 1 h at
room temp. The plate is then washed five times with PBST. The reaction is finally developed using 100 μL/well
of CDP-Star ready-to-use with Sapphire-II Enhancer substrate for 30 min at room temp. Luminescence counts
are measured in a LJL-Analyst (Molecular Devices, CA). Counts from CSF samples are converted to BACE
concentrations using coefficients determined by quadratic fit to the bBACE standard curve.
The RBM studies will be done as described (Hu et al, Submitted, 2009). Briefly, Luminex technology is used
to perform up to 100 multiplexed microsphere based assays in a single reaction vessel or well by combining
optical classification schemes, biochemical assays, flow cytometry and digital signal processing
hardware/software (see Figure 10 on the next page). Multiplexing is accomplished by assigning each analytespecific assay a microsphere set labeled with a unique fluorescence signature. To attain 100 distinct
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
microsphere signatures, 2 fluorescent dyes (red and fare red) are mixed in various combinations using ten
intensity levels of each dye (10 x 10 = 100). Each set of microspheres is encoded with a fluorescent
Figure 10: The Luminex components.
signature and assay-specific capture antibodies are
conjugated covalently to each unique set of microspheres.
Covalent attachment of the capture reagent to the
microspheres is achieved using carboxyl functional groups
located on the surface of each 5.6 µm microsphere and
primary amines within the capture reagent. Coupling
chemistry is performed on large numbers of individual
microspheres (107-109 microspheres/mL) simultaneously
and after optimizing the parameters of each assay
separately, multi-analyte profiles are performed by mixing up to 100 different sets of the microspheres in a
single well of a 384-format microtiter plate. A small sample volume (10-20 µL) is added to the well and allowed
to react with the microspheres. The assay-specific capture antibody on each microsphere binds the analyte of
interest. A cocktail of assay-specific, biotinylated detecting reagents is reacted with the microsphere mixture,
followed by a streptavidin-labeled fluorescent “reporter” molecule and then the multiplex is washed to remove
unbound detecting reagents. Next, the mixture of microspheres is analyzed with the Luminex instrument which
uses hydrodynamic focusing to pass the microspheres in single file through two laser beams. As each
microsphere passes through the excitation beams, it is analyzed for size, encoded fluorescence signature and
the amount of fluorescence generated in proportion to the analyte. Microsphere size is determined to eliminate
microsphere aggregates. Since each microsphere is encoded with a unique signature, the classification
identifies the analyte measured on each microsphere and a reporter signal is generated in proportion to the
analyte concentration. Data acquisition, analysis and reporting are performed in real-time on all microsphere
sets and 100 individual microspheres from each unique set are analyzed and the median value of the analytespecific, or “reporter,” fluorescence is logged. Using internal controls of known quantity, sensitive and
quantitative results are achieved with precision and the quantitative data generated is then mined for distinctive
patterns of biomarkers that indicate NC, EMCI/LMCI or AD status.
Interpretation: Multiple analytical approaches will be used to predict cognitive decline and AD using the BACE
and RBM data as described (*1,*2,*6,*7,*10,*12,*15,*18,*19), and similar approaches can be used to integrate
these data sets with those obtained in Aim 2. The Biostatistics Core discusses these issues in greater detail for
ADNI2 overall and with respect to the chemical biomarker data (8.4.2), and we will collaborate with Laurel
Beckett and colleagues in the design and implementation of Aim 3. Here, we briefly summarize how we
accomplish classifying and predicting the diagnostic category of a sample on the basis of protein quantitative
profiles. The first statistical challenge is the identification of biomarkers that characterize the diagnostic groups
(e.g. NC, EMCI, LMCI and AD). As described in the studies summarized above in our Progress Report, a
combination of algorithms can be used including nearest shrunken centroid, linear discriminant analysis (LDA),
support vector machines (SVM), and partial least square classification. The resultant classifiers can be further
analyzed with ANOVA to confirm optimal markers. Machine learning also will be applied to identify CSF and
plasma biomarkers that discriminate between diagnostic groups as described (De Meyer et al, Submitted,
2009). To address interactions between predictor variables, Random Forests classification methods will be
applied as described (Hu et al, Submitted, 2009). We also will use systems biology approaches (e.g. Ingenuity
and Pathway studio) to better understand pathway relationships between identified proteins and to find
functional connections between CSF and plasma biomarkers. Formal statistical analyses will be conducted on
the most informative proteins and we anticipate that ~20 peptides/proteins will be most informative based on
our current RBM data (Hu et al, Submitted, 2009). For each of the identified CSF and plasma analytes, the
distributions across the diagnostic groups will be compared through the group means or adjusted means from
either the original scale or the transformed scale using analysis of covariance (ANOCOVA) models. These
ANOCOVA models will include the diagnosis and other covariates including age, education, gender, ApoEε4
genotype as well as possible interactions among these factors. Depending on the outcome of these tests, the
differences between the diagnostic groups will be tested either by the main effect of diagnosis or the effect of
diagnosis at a fixed level of other covariates (i.e., ApoEε4 status) or through the adjusted least square means.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
A major analytic concern in these tests is the control of overall type I error rate due to the relatively large
number of CSF and plasma analytes tested. We will address this by: 1) Benjamini-Hochberg corrections to P
values; 2) seeking composite CSF and plasma markers based on the informative analytes and conducting the
primary analyses using these composite markers (instead of 10 or 20 individual markers). For example, we
may apply the technique of Xiong et al. [77] to search for the linear combination of informative analytes that
optimally discriminates between the diagnostic groups. We will consult with the Biostatistics Core on use of
cross-validation (e.g. training/test sets) to avoid bias in data-generated composites. Markers identified as
distinguishing diagnostic groups will be further examined for associations with trajectories of cognitive change
(8.4.2). Sample size and power for Biomarker Core hypotheses are given in the Biostatistics Core (8.4.2.4).
Significance: The significance of Aim 3 is that it will identify and validate novel CSF and plasma biomarkers
that could distinguish AD from NC and EMCI/LMCI, predict who among the NC and EMCI/LMCI subjects will
progress to AD, reflect the progression of AD, and possibly indicate the response of AD and/or LMCI subjects
to disease modifying therapies. These and other integrative analyses of ADNI data will significantly impact
basic and clinical AD research as well as efforts to develop meaningful therapies for AD.
5.4.4. Specific Aim 4: Partner with the ADNI ISAB and investigators outside ADNI in RARC approved
“add-on” biomarker studies that may include proteomic, metabalomic and lipidomic methodologies.
Rationale: The rationale for Aim 4 is to make optimal use of ADNI biofluids, ADNI data and the biomarker
skills and expertise of our ISAB and outside collaborators. Outreach beyond ADNI disseminates information on
ADNI resources that are available to non-ADNI investigators for research and data analysis.
Methods: We will use strategies similar to those that have worked so well in ADNI1 to partner with ADNI ISAB
members and other non-ADNI investigators in RARC approved “add-on” studies in ADNI2. Briefly, this occurs
through presentations the Core Leaders give at national/international meetings, by conferencing with ISAB
members to discuss biomarker data and projects, events held at Penn like the 2009 CNDR Retreat on
biomarkers, as well as contacts generated through the ADNI website. Indeed, our Progress Report documents
robust collaborations with non-ADNI investigators, and the RARC has approved ISAB-led CSF and plasma
proteomic “add-on” studies that will launch shortly and result in novel ADNI biomarker data sets.
Interpretation: The interpretive analyses for the studies in this Aim will be addressed using methods similar to
those described in our Progress Report (*1,*2,*6,*7,*10,*12,*15,*18,*19), and in Aims 2 and 3.
Significance: The significance of the studies proposed in Aim 4 is that they offer the prospect of identifying
additional novel CSF and plasma biomarkers that will enable the reliable distinction of AD from NC and
EMCI/LMCI, predict which NC and EMCI/LMCI individuals will develop AD, serve as progression markers of
AD, and possibly indicate the response of AD and/or EMCI/LMCI subjects to disease modifying therapies.
Hence, Aim 4 will have a very significant impact on basic and clinical AD research.
5.4.5. Specific Aim 5: Collaborate with all ADNI Cores/Investigators in analyses of biomarker, clinical,
imaging and autopsy data.
Rationale: The rationale for Aim 5 is similar to Aim 4 while our active collaborations with ADNI colleagues are
described in our Progress Report (*2,*6,*7,*10,*12,*15,*18,*19). Hence, we will continue these collaborations
to integrate biomarker, clinical, imaging and autopsy data in ADNI2.
Methods: We will use methods and procedures in Aim 5 similar to those in Aims 2-4 and in our publications
with other ADNI investigators cited above.
Interpretation: We will use interpretive and analytical strategies in here like those described in publications
with other ADNI investigators (*2,*6,*7,*10,*12,*15,*18,*19).
Significance: The significance of Aim 5 is that it integrates clinical and biomarker data to reliably distinguish
AD from NC and EMCI/LMCI, predict which NC and EMCI/LMCI individuals will develop AD, identify
progression markers of AD as well as biomarkers that may indicate the response of AD and/or EMCI/LMCI
subjects to disease modifying therapies. Thus, Aim 5 will have a very significant impact on basic and clinical
AD research as well as efforts to develop meaningful therapies for AD.
5.4.6. Specific Aim 6: Collaborate with WW-ADNI Sites in Europe, Japan, Korea, China and Australia in
joint biomarker studies and comparative analyses of previously collected and new biomarker data.
Rationale: The rationale for Aim 6 is similar to that for Aims 4 and 5, and we already have begun to
implement this Aim as exemplified by papers we published with WW-ADNI investigators (*3,*4,*15,*17).
Methods: We will use methods and procedures here that are similar to those summarized above in Aims 2-5
as well as in our Progress Report and in publications with WW-ADNI investigators summarized above.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Interpretation: We will use interpretive and analytical strategies here like those described above as well as in
our Progress Report and publications with ADNI and WW-ADNI investigators (*2-*4,*6-*7,*10,*12,*15,*18,*19).
Significance: The significance of the studies in Aim 6 is that they will integrate clinical and biomarker data
from other WW-ADNI sites to enable the reliable distinction of AD from NC and EMCI/LMCI, predict which NC
and EMCI/LMCI individuals will develop AD, identify progression markers of AD as well as biomarkers that may
be able to indicate the response of AD and/or EMCI/LMCI subjects to disease modifying therapies. This will
have a very significant impact on basic and clinical AD research as well as efforts to develop meaningful
therapies for AD and also help establish a world wide network of AD biomarker research sites to accelerate the
pace of implementing lessons learned from ADNI2.
5.5. Literature Cited:
1.
Schneider, J.A., Z. Arvanitakis, W. Bang, and D.A. Bennett, Mixed brain pathologies account for most
dementia cases in community-dwelling older persons. Neurology, 2007. 69(24): p. 2197-204.
2.
White, L., B.J. Small, H. Petrovitch, G.W. Ross, K. Masaki, R.D. Abbott, J. Hardman, D. Davis, J.
Nelson, and W. Markesbery, Recent clinical-pathologic research on the causes of dementia in late life:
update from the Honolulu-Asia Aging Study. J Geriatr Psychiatry Neurol, 2005. 18(4): p. 224-7.
3.
Gomez-Isla, T., R. Hollister, H. West, S. Mui, J.H. Growdon, R.C. Petersen, J.E. Parisi, and B.T.
Hyman, Neuronal loss correlates with but exceeds neurofibrillary tangles in Alzheimer's disease. Ann
Neurol, 1997. 41(1): p. 17-24.
4.
Savva, G.M., S.B. Wharton, P.G. Ince, G. Forster, F.E. Matthews, and C. Brayne, Age, neuropathology,
and dementia. N Engl J Med, 2009. 360(22): p. 2302-9.
5.
Terry, R.D., E. Masliah, D.P. Salmon, N. Butters, R. DeTeresa, R. Hill, L.A. Hansen, and R. Katzman,
Physical basis of cognitive alterations in Alzheimer's disease: synapse loss is the major correlate of
cognitive impairment. Ann Neurol, 1991. 30(4): p. 572-80.
6.
Knopman, D.S., J.E. Parisi, A. Salviati, M. Floriach-Robert, B.F. Boeve, R.J. Ivnik, G.E. Smith, D.W.
Dickson, K.A. Johnson, L.E. Petersen, W.C. McDonald, H. Braak, and R.C. Petersen, Neuropathology
of cognitively normal elderly. J Neuropathol Exp Neurol, 2003. 62(11): p. 1087-95.
7.
Price, J.L. and J.C. Morris, Tangles and plaques in nondemented aging and "preclinical" Alzheimer's
disease. Ann Neurol, 1999. 45(3): p. 358-68.
8.
Petersen, R.C., Mild cognitive impairment as a diagnostic entity. J Intern Med, 2004. 256(3): p. 183-94.
9.
Dubois, B., H.H. Feldman, C. Jacova, S.T. Dekosky, P. Barberger-Gateau, J. Cummings, A.
Delacourte, D. Galasko, S. Gauthier, G. Jicha, K. Meguro, J. O'Brien, F. Pasquier, P. Robert, M.
Rossor, S. Salloway, Y. Stern, P.J. Visser, and P. Scheltens, Research criteria for the diagnosis of
Alzheimer's disease: revising the NINCDS-ADRDA criteria. Lancet Neurol, 2007. 6(8): p. 734-46.
10.
Chetelat, G., B. Desgranges, V. de la Sayette, F. Viader, F. Eustache, and J.C. Baron, Mild cognitive
impairment: Can FDG-PET predict who is to rapidly convert to Alzheimer's disease? Neurology, 2003.
60(8): p. 1374-7.
11.
Dickerson, B.C., I. Goncharova, M.P. Sullivan, C. Forchetti, R.S. Wilson, D.A. Bennett, L.A. Beckett,
and L. deToledo-Morrell, MRI-derived entorhinal and hippocampal atrophy in incipient and very mild
Alzheimer's disease. Neurobiol Aging, 2001. 22(5): p. 747-54.
12.
Drzezga, A., N. Lautenschlager, H. Siebner, M. Riemenschneider, F. Willoch, S. Minoshima, M.
Schwaiger, and A. Kurz, Cerebral metabolic changes accompanying conversion of mild cognitive
impairment into Alzheimer's disease: a PET follow-up study. Eur J Nucl Med Mol Imaging, 2003. 30(8):
p. 1104-13.
13.
Jack, C.R., Jr., R.C. Petersen, Y.C. Xu, P.C. O'Brien, G.E. Smith, R.J. Ivnik, B.F. Boeve, S.C. Waring,
E.G. Tangalos, and E. Kokmen, Prediction of AD with MRI-based hippocampal volume in mild cognitive
impairment. Neurology, 1999. 52(7): p. 1397-403.
14.
Killiany, R.J., T. Gomez-Isla, M. Moss, R. Kikinis, T. Sandor, F. Jolesz, R. Tanzi, K. Jones, B.T. Hyman,
and M.S. Albert, Use of structural magnetic resonance imaging to predict who will get Alzheimer's
disease. Ann Neurol, 2000. 47(4): p. 430-9.
15.
Petersen, R.C., G.E. Smith, R.J. Ivnik, E.G. Tangalos, D.J. Schaid, S.N. Thibodeau, E. Kokmen, S.C.
Waring, and L.T. Kurland, Apolipoprotein E status as a predictor of the development of Alzheimer's
disease in memory-impaired individuals. JAMA, 1995. 273(16): p. 1274-8.
16.
Shaw, L.M., H. Vanderstichele, M. Knapik-Czajka, C.M. Clark, P.S. Aisen, R.C. Petersen, K. Blennow,
H. Soares, A. Simon, P. Lewczuk, R. Dean, E. Siemers, W. Potter, V.M. Lee, and J.Q. Trojanowski,
PHS 398/2590 (Rev. 11/07)
Page
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
Weiner, Michael
Cerebrospinal fluid biomarker signature in Alzheimer's disease neuroimaging initiative subjects. Ann
Neurol, 2009, 65: p. 403-13.
Visser, P.J., P. Scheltens, F.R. Verhey, B. Schmand, L.J. Launer, J. Jolles, and C. Jonker, Medial
temporal lobe atrophy and memory dysfunction as predictors for dementia in subjects with mild
cognitive impairment. J Neurol, 1999. 246(6): p. 477-85.
Mattsson, N., J. Zetterberg, and O. Hansson, CSF biomarkers incipient Alzheimer disease in patients
with mild cognitive inpairment. JAMA, In press 2009.
Petersen, and C.R. Jack, Jr., MRI and CSF biomarkers in normal, LMCI, and AD subjects: predicting
future clinical change. Neurology, 2009. 73(4): p. 294-301.
McKhann, G., D. Drachman, M. Folstein, R. Katzman, D. Price, and E.M. Stadlan, Clinical diagnosis of
Alzheimer's disease: report of the NINCDS-ADRDA Work Group under the auspices of Department of
Health and Human Services Task Force on Alzheimer's Disease. Neurology, 1984. 34(7): p. 939-44.
Ingelsson, M., H. Fukumoto, K.L. Newell, J.H. Growdon, E.T. Hedley-Whyte, M.P. Frosch, M.S. Albert,
B.T. Hyman, and M.C. Irizarry, Early Abeta accumulation and progressive synaptic loss, gliosis, and
tangle formation in AD brain. Neurology, 2004. 62(6): p. 925-31.
Hardy, J.A. and G.A. Higgins, Alzheimer's disease: the amyloid cascade hypothesis. Science, 1992.
256(5054): p. 184-5.
Klein, W.L., W.B. Stine, Jr., and D.B. Teplow, Small assemblies of unmodified amyloid beta-protein are
the proximate neurotoxin in Alzheimer's disease. Neurobiol Aging, 2004. 25(5): p. 569-80.
Bigio, E.H., L.S. Hynan, E. Sontag, S. Satumtira, and C.L. White, Synapse loss is greater in presenile
than senile onset Alzheimer disease: implications for the cognitive reserve hypothesis. Neuropathol
Appl Neurobiol, 2002. 28(3): p. 218-27.
Nelson, P.T., E.L. Abner, F.A. Schmitt, R.J. Kryscio, G.A. Jicha, C.D. Smith, D.G. Davis, J.W. Poduska,
E. Patel, M.S. Mendiondo, and W.R. Markesbery, Modeling the Association between 43 Different
Clinical and Pathological Variables and the Severity of Cognitive Impairment in a Large Autopsy Cohort
of Elderly Persons. Brain Pathol, In press 2009.
Klunk, W.E., H. Engler, A. Nordberg, Y. Wang, G. Blomqvist, D.P. Holt, M. Bergstrom, I. Savitcheva,
G.F. Huang, S. Estrada, B. Ausen, M.L. Debnath, J. Barletta, J.C. Price, J. Sandell, B.J. Lopresti, A.
Wall, P. Koivisto, G. Antoni, C.A. Mathis, and B. Langstrom, Imaging brain amyloid in Alzheimer's
disease with Pittsburgh Compound-B. Ann Neurol, 2004. 55(3): p. 306-19.
Rowe, C.C., S. Ng, U. Ackermann, S.J. Gong, K. Pike, G. Savage, T.F. Cowie, K.L. Dickinson, P.
Maruff, D. Darby, C. Smith, M. Woodward, J. Merory, H. Tochon-Danguy, G. O'Keefe, W.E. Klunk, C.A.
Mathis, J.C. Price, C.L. Masters, and V.L. Villemagne, Imaging beta-amyloid burden in aging and
dementia. Neurology, 2007. 68(20): p. 1718-25.
Ikonomovic, M.D., W.E. Klunk, E.E. Abrahamson, C.A. Mathis, J.C. Price, N.D. Tsopelas, B.J. Lopresti,
S. Ziolko, W. Bi, W.R. Paljug, M.L. Debnath, C.E. Hope, B.A. Isanski, R.L. Hamilton, and S.T. DeKosky,
Post-mortem correlates of in vivo PiB-PET amyloid imaging in a typical case of Alzheimer's disease.
Brain, 2008. 131(Pt 6): p. 1630-45.
Clark, C.M., S. Xie, J. Chittams, D. Ewbank, E. Peskind, D. Galasko, J.C. Morris, D.W. McKeel, Jr., M.
Farlow, S.L. Weitlauf, J. Quinn, J. Kaye, D. Knopman, H. Arai, R.S. Doody, C. DeCarli, S. Leight, V.M.
Lee, and J.Q. Trojanowski, Cerebrospinal fluid tau and beta-amyloid: how well do these biomarkers
reflect autopsy-confirmed dementia diagnoses? Arch Neurol, 2003. 60(12): p. 1696-702.
Strozyk, D., K. Blennow, L.R. White, and L.J. Launer, CSF Abeta 42 levels correlate with amyloidneuropathology in a population-based autopsy study. Neurology, 2003. 60(4): p. 652-6.
Fagan, A.M., M.A. Mintun, R.H. Mach, S.Y. Lee, C.S. Dence, A.R. Shah, G.N. LaRossa, M.L. Spinner,
W.E. Klunk, C.A. Mathis, S.T. DeKosky, J.C. Morris, and D.M. Holtzman, Inverse relation between in
vivo amyloid imaging load and cerebrospinal fluid Abeta42 in humans. Ann Neurol, 2006. 59(3): p. 5129.
Jagust, W.J., S.M. Landau, L.M. Shaw, J.Q. Trojanowski, R.A. Koeppe, E.M. Reiman, N.L. Foster, R.C.
Petersen, W. M.W., J.C. Price, and C.A. Mathis, and the Alzheimer’s Disease Neuroimaging Initiative.
Relationships between biomarkers in aging and dementia. Neurol, In press, 2009.
Petersen, R.C. and J.Q. Trojanowski, Use of Alzheimer disease biomarkers: potentially yes for clinical
trials but not yet for clinical practice. JAMA, 2009. 302(4): p. 436-7.
PHS 398/2590 (Rev. 11/07)
Page
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
Weiner, Michael
Buerger, K., M. Ewers, T. Pirttila, R. Zinkowski, I. Alafuzoff, S.J. Teipel, J. DeBernardis, D. Kerkman, C.
McCulloch, H. Soininen, and H. Hampel, CSF phosphorylated tau protein correlates with neocortical
neurofibrillary pathology in Alzheimer's disease. Brain, 2006. 129(Pt 11): p. 3035-41.
Hansson, O., H. Zetterberg, P. Buchhave, E. Londos, K. Blennow, and L. Minthon, Association between
CSF biomarkers and incipient Alzheimer's disease in patients with mild cognitive impairment: a followup study. Lancet Neurol, 2006. 5(3): p. 228-34.
Hesse, C., L. Rosengren, N. Andreasen, P. Davidsson, H. Vanderstichele, E. Vanmechelen, and K.
Blennow, Transient increase in total tau but not phospho-tau in human cerebrospinal fluid after acute
stroke. Neurosci Lett, 2001. 297(3): p. 187-90.
Ost, M., K. Nylen, L. Csajbok, A.O. Ohrfelt, M. Tullberg, C. Wikkelso, P. Nellgard, L. Rosengren, K.
Blennow, and B. Nellgard, Initial CSF total tau correlates with 1-year outcome in patients with traumatic
brain injury. Neurology, 2006. 67(9): p. 1600-4.
Arai, H., M. Terajima, M. Miura, S. Higuchi, T. Muramatsu, N. Machida, H. Seiki, S. Takase, C.M. Clark,
V.M. Lee, and et al., Tau in cerebrospinal fluid: a potential diagnostic marker in Alzheimer's disease.
Ann Neurol, 1995. 38(4): p. 649-52.
Blennow, K., A. Wallin, H. Agren, C. Spenger, J. Siegfried, and E. Vanmechelen, Tau protein in
cerebrospinal fluid: a biochemical marker for axonal degeneration in Alzheimer disease? Mol Chem
Neuropathol, 1995. 26(3): p. 231-45.
Tapiola, T., M. Overmyer, M. Lehtovirta, S. Helisalmi, J. Ramberg, I. Alafuzoff, P. Riekkinen, Sr., and H.
Soininen, The level of cerebrospinal fluid tau correlates with neurofibrillary tangles in Alzheimer's
disease. Neuroreport, 1997. 8(18): p. 3961-3.
Bian, H., J.C. Van Swieten, S. Leight, L. Massimo, E. Wood, M. Forman, P. Moore, I. de Koning, C.M.
Clark, S. Rosso, J. Trojanowski, V.M. Lee, and M. Grossman, CSF biomarkers in frontotemporal lobar
degeneration with known pathology. Neurology, 2008. 70(19 Pt 2): p. 1827-35.
Grossman, M., J. Farmer, S. Leight, M. Work, P. Moore, V. Van Deerlin, D. Pratico, C.M. Clark, H.B.
Coslett, A. Chatterjee, J. Gee, J.Q. Trojanowski, and V.M. Lee, Cerebrospinal fluid profile in
frontotemporal dementia and Alzheimer's disease. Ann Neurol, 2005. 57(5): p. 721-9.
Jagust, W., B. Reed, D. Mungas, W. Ellis, and C. Decarli, What does fluorodeoxyglucose PET imaging
add to a clinical diagnosis of dementia? Neurology, 2007. 69(9): p. 871-7.
Minoshima, S., B. Giordani, S. Berent, K.A. Frey, N.L. Foster, and D.E. Kuhl, Metabolic reduction in the
posterior cingulate cortex in very early Alzheimer's disease. Ann Neurol, 1997. 42(1): p. 85-94.
Hoffman, J.M., K.A. Welsh-Bohmer, M. Hanson, B. Crain, C. Hulette, N. Earl, and R.E. Coleman, FDG
PET imaging in patients with pathologically verified dementia. J Nucl Med, 2000. 41(11): p. 1920-8.
Bobinski, M., M.J. de Leon, J. Wegiel, S. Desanti, A. Convit, L.A. Saint Louis, H. Rusinek, and H.M.
Wisniewski, The histological validation of post mortem magnetic resonance imaging-determined
hippocampal volume in Alzheimer's disease. Neuroscience, 2000. 95(3): p. 721-5.
Fox, N.C., R.I. Scahill, W.R. Crum, and M.N. Rossor, Correlation between rates of brain atrophy and
cognitive decline in AD. Neurology, 1999. 52(8): p. 1687-9.
Jack, C.R., Jr., R.C. Petersen, Y. Xu, P.C. O'Brien, G.E. Smith, R.J. Ivnik, B.F. Boeve, E.G. Tangalos,
and E. Kokmen, Rates of hippocampal atrophy correlate with change in clinical status in aging and AD.
Neurology, 2000. 55(4): p. 484-89.
Jack, C.R., Jr., D.W. Dickson, J.E. Parisi, Y.C. Xu, R.H. Cha, P.C. O'Brien, S.D. Edland, G.E. Smith,
B.F. Boeve, E.G. Tangalos, E. Kokmen, and R.C. Petersen, Antemortem MRI findings correlate with
hippocampal neuropathology in typical aging and dementia. Neurology, 2002. 58(5): p. 750-7.
Jagust, W.J., L. Zheng, D.J. Harvey, W.J. Mack, H.V. Vinters, M.W. Weiner, W.G. Ellis, C. Zarow, D.
Mungas, B.R. Reed, J.H. Kramer, N. Schuff, C. DeCarli, and H.C. Chui, Neuropathological basis of
magnetic resonance images in aging and dementia. Ann Neurol, 2008. 63(1): p. 72-80.
Silbert, L.C., J.F. Quinn, M.M. Moore, E. Corbridge, M.J. Ball, G. Murdoch, G. Sexton, and J.A. Kaye,
Changes in premorbid brain volume predict Alzheimer's disease pathology. Neurology, 2003. 61(4): p.
487-92.
Whitwell, J.L., K.A. Josephs, M.E. Murray, K. Kantarci, S.A. Przybelski, S.D. Weigand, P. Vemuri, M.L.
Senjem, J.E. Parisi, D.S. Knopman, B.F. Boeve, R.C. Petersen, D.W. Dickson, and C.R. Jack, Jr., MRI
correlates of neurofibrillary tangle pathology at autopsy: a voxel-based morphometry study. Neurology,
2008. 71(10): p. 743-9.
PHS 398/2590 (Rev. 11/07)
Page
53.
54.
55.
56.
57.
58.
59.
60.
61.
62.
63.
64.
65.
66.
67.
68.
69.
Weiner, Michael
Aizenstein, H.J., R.D. Nebes, J.A. Saxton, J.C. Price, C.A. Mathis, N.D. Tsopelas, S.K. Ziolko, J.A.
James, B.E. Snitz, P.R. Houck, W. Bi, A.D. Cohen, B.J. Lopresti, S.T. DeKosky, E.M. Halligan, and
W.E. Klunk, Frequent amyloid deposition without significant cognitive impairment among the elderly.
Arch Neurol, 2008. 65(11): p. 1509-17.
Mintun, M.A., G.N. Larossa, Y.I. Sheline, C.S. Dence, S.Y. Lee, R.H. Mach, W.E. Klunk, C.A. Mathis,
S.T. DeKosky, and J.C. Morris, [11C]PIB in a nondemented population: potential antecedent marker of
Alzheimer disease. Neurology, 2006. 67(3): p. 446-52.
Peskind, E.R., G. Li, J. Shofer, J.F. Quinn, J.A. Kaye, C.M. Clark, M.R. Farlow, C. DeCarli, M.A.
Raskind, G.D. Schellenberg, V.M. Lee, and D.R. Galasko, Age and apolipoprotein E*4 allele effects on
cerebrospinal fluid beta-amyloid 42 in adults with normal cognition. Arch Neurol, 2006. 63(7): p. 936-9.
Fagan, A.M., C.M. Roe, C. Xiong, M.A. Mintun, J.C. Morris, and D.M. Holtzman, Cerebrospinal fluid
tau/beta-amyloid(42) ratio as a prediction of cognitive decline in nondemented older adults. Arch
Neurol, 2007. 64(3): p. 343-9.
Li, G., I. Sokal, J.F. Quinn, J.B. Leverenz, M. Brodey, G.D. Schellenberg, J.A. Kaye, M.A. Raskind, J.
Zhang, E.R. Peskind, and T.J. Montine, CSF tau/Abeta42 ratio for increased risk of mild cognitive
impairment: a follow-up study. Neurology, 2007. 69(7): p. 631-9.
Jack, C.R., Jr., V.J. Lowe, S.D. Weigand, H.J. Wiste, M.L. Senjem, D.S. Knopman, M.M. Shiung, J.L.
Gunter, B.F. Boeve, B.J. Kemp, M. Weiner, and R.C. Petersen, Serial PIB and MRI in normal, mild
cognitive impairment and Alzheimer's disease: implications for sequence of pathological events in
Alzheimer's disease. Brain, 2009. 132(Pt 5): p. 1355-65.
Chetelat, G., B. Desgranges, V. De La Sayette, F. Viader, F. Eustache, and J.C. Baron, Mapping gray
matter loss with voxel-based morphometry in mild cognitive impairment. Neuroreport, 2002. 13(15): p.
1939-43.
de Leon, M.J., A. Convit, O.T. Wolf, C.Y. Tarshish, S. DeSanti, H. Rusinek, W. Tsui, E. Kandil, A.J.
Scherer, A. Roche, A. Imossi, E. Thorn, M. Bobinski, C. Caraos, P. Lesbre, D. Schlyer, J. Poirier, B.
Reisberg, and J. Fowler, Prediction of cognitive decline in normal elderly subjects with 2-[(18)F]fluoro-2deoxy-D-glucose/poitron-emission tomography (FDG/PET). Proc Natl Acad Sci U S A, 2001. 98(19): p.
10966-71.
Jagust, W., A. Gitcho, F. Sun, B. Kuczynski, D. Mungas, and M. Haan, Brain imaging evidence of
preclinical Alzheimer's disease in normal aging. Ann Neurol, 2006. 59(4): p. 673-81.
Reiman, E.M., R.J. Caselli, L.S. Yun, K. Chen, D. Bandy, S. Minoshima, S.N. Thibodeau, and D.
Osborne, Preclinical evidence of Alzheimer's disease in persons homozygous for the epsilon 4 allele for
apolipoprotein E. N Engl J Med, 1996. 334(12): p. 752-8.
Small, G.W., J.C. Mazziotta, M.T. Collins, L.R. Baxter, M.E. Phelps, M.A. Mandelkern, A. Kaplan, A. La
Rue, C.F. Adamson, L. Chang, and et al., Apolipoprotein E type 4 allele and cerebral glucose
metabolism in relatives at risk for familial Alzheimer disease. JAMA, 1995. 273(12): p. 942-7.
Fox, N.C., E.K. Warrington, and M.N. Rossor, Serial magnetic resonance imaging of cerebral atrophy in
preclinical Alzheimer's disease. Lancet, 1999. 353(9170): p. 2125.
Jack, C.R., Jr., M.M. Shiung, J.L. Gunter, P.C. O'Brien, S.D. Weigand, D.S. Knopman, B.F. Boeve, R.J.
Ivnik, G.E. Smith, R.H. Cha, E.G. Tangalos, and R.C. Petersen, Comparison of different MRI brain
atrophy rate measures with clinical disease progression in AD. Neurology, 2004. 62(4): p. 591-600.
Kaye, J.A., T. Swihart, D. Howieson, A. Dame, M.M. Moore, T. Karnos, R. Camicioli, M. Ball, B. Oken,
and G. Sexton, Volume loss of the hippocampus and temporal lobe in healthy elderly persons destined
to develop dementia. Neurology, 1997. 48(5): p. 1297-304.
Engler, H., A. Forsberg, O. Almkvist, G. Blomquist, E. Larsson, I. Savitcheva, A. Wall, A. Ringheim, B.
Langstrom, and A. Nordberg, Two-year follow-up of amyloid deposition in patients with Alzheimer's
disease. Brain, 2006. 129(Pt 11): p. 2856-66.
Josephs, K.A., J.L. Whitwell, Z. Ahmed, M.M. Shiung, S.D. Weigand, D.S. Knopman, B.F. Boeve, J.E.
Parisi, R.C. Petersen, D.W. Dickson, and C.R. Jack, Jr., Beta-amyloid burden is not associated with
rates of brain atrophy. Ann Neurol, 2008. 63(2): p. 204-12.
Jack, C.R., Jr., V.J. Lowe, M.L. Senjem, S.D. Weigand, B.J. Kemp, M.M. Shiung, D.S. Knopman, B.F.
Boeve, W.E. Klunk, C.A. Mathis, and R.C. Petersen, 11C PiB and structural MRI provide
complementary information in imaging of Alzheimer's disease and amnestic mild cognitive impairment.
Brain, 2008. 131(Pt 3): p. 665-80.
PHS 398/2590 (Rev. 11/07)
Page
70.
71.
72.
73.
74.
75.
76.
77.
Weiner, Michael
Mormino, E.C., J.T. Kluth, C.M. Madison, G.D. Rabinovici, S.L. Baker, B.L. Miller, R.A. Koeppe, C.A.
Mathis, M.W. Weiner, and W.J. Jagust, Episodic memory loss is related to hippocampal-mediated betaamyloid deposition in elderly subjects. Brain, 2009. 132(Pt 5): p. 1310-23.
Bennett, D.A., J.A. Schneider, R.S. Wilson, J.L. Bienias, and S.E. Arnold, Neurofibrillary tangles
mediate the association of amyloid load with clinical Alzheimer disease and level of cognitive function.
Arch Neurol, 2004. 61(3): p. 378-84.
Petersen, and C.R. Jack, Jr., MRI and CSF biomarkers in normal, LMCI, and AD subjects: diagnostic
discrimination and cognitive correlations. Neurology, 2009. 73(4): p. 287-93.
Sunderland, T., B. Wolozin, D. Galasko, J. Levy, R. Dukoff, M. Bahro, R. Lasser, R. Motter, T.
Lehtimaki, and P. Seubert, Longitudinal stability of CSF tau levels in Alzheimer patients. Biol
Psychiatry, 1999. 46(6): p. 750-5.
Fox, N.C. and P.A. Freeborough, Brain atrophy progression measured from registered serial MRI:
validation and application to Alzheimer's disease. J Magn Reson Imaging, 1997. 7(6): p. 1069-75.
Jack, C.R., Jr., R.C. Petersen, Y. Xu, P.C. O'Brien, G.E. Smith, R.J. Ivnik, E.G. Tangalos, and E.
Kokmen, Rate of medial temporal lobe atrophy in typical aging and Alzheimer's disease. Neurology,
1998. 51(4): p. 993-9.
Wu, G., S. Sankaranarayanan, K. Tugusheva, J. Kahana, G. Seabrook, X.P. Shi, E. King, V.
Devanarayan, J.J. Cook, and A.J. Simon, Decrease in age-adjusted cerebrospinal fluid beta-secretase
activity in Alzheimer's subjects. Clin Biochem, 2008. 41(12): p. 986-96.
Xiong, C., D.W. McKeel, Jr., J.P. Miller, and J.C. Morris, Combining correlated diagnostic tests:
application to neuropathologic diagnosis of Alzheimer's disease. Med Decis Making, 2004. 24(6): p.
659-69.
5.6. Human Subjects Research
1. Risks to the Subjects
a. Human subjects involvement and characteristics: All participants will be individuals over the age of
50, based on the natural history of AD and EMCI/LMCI. The Clinical Core of ADNI-2 will recruit and
follow individuals with EMCI/LMCI, AD and NC including women and minorities. Children will not be
included in this study as AD and EMCI/LMCI affect adults exclusively nor will we include special
classes of subjects such as pregnant women, prisoners, institutionalized individuals or other
vulnerable populations. The studies pursued in the Biomarker Core focus on biofluids obtained from
the members of the ADNI-2 cohort with informed consent obtained by the Clinical Core. Hence,
once the fluids have been obtained by the Clinical Core, the studies conducted in the Biomarker
Core present no further risk to these individuals and biomarker data are not provided to individuals
so there is no risk to them of misunderstanding test results.
b. Sources of materials: All research material from human subjects will be obtained by informed
consent together with clinical information provided by the ADNI-2 Clinical Core investigators. The
acquisition of these samples and maintenance of confidentiality of the information related thereto
are addressed fully in the Administrative and Clinical Cores.
c. Potential risks: For the reasons mentioned above, the biomarker studies to be conducted in this
Core do not constitute any additional risk to these subjects.
2. Adequacy of Protection Against Risks
a. Recruitment and informed consent: A valid consent is obtained for all ADNI-2 subjects for all fluids
obtained from them and the studies conducted on these fluids by the ADNI-2 Clinical Core. The
consent authorizes the storage and study of these fluids in the Penn Biomarker Core of ADNI-2 and
by other investigators whose “add-on” studies are approved by the ADNI-2 RARC.
b. Protection Against Risk: All samples are collected, banked and studied under human subjects
approval granted by the University of California and the University of Pennsylvania IRB
Committees. All Biomarker Core investigators participate in educational programs for the protection
of human research subjects. Penn has established an educational program specifically for this
purpose to meet NIH requirements. All Core personnel have completed this required program.
Subjects will be provided with a unique identifier to protect confidentiality. No personal identifiers
are used. All bioflud samples and data are maintained in password-protected, secure databases
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
that are backed up on a secure server. Paper files are maintained in a locked file cabinet
accessible only to Biomarker Core investigators.
3. Potential Benefits of the Proposed Research to the Subjects and Others
Subjects are not expected to benefit directly from participation in this study. However, the biomarker
data obtained in the Penn Biomarker Core of ADNI-2 will facilitate the rational diagnosis of AD,
EMCI/LMCI and NC status as well as the application of therapeutics for AD and EMCI/LMCI when
these biomarker assays are fully validated and become available for use in clinical studies including
clinical trials.
4. Importance of the Knowledge to be Gained
There is much to be learned about AD biomarkers before they can be used widely in research and
clinical practice and this is the goal of the Penn Biomarker Core in ADNI-2. The studies described in the
Aims of this Core are essential for both correlation with the clinical phenotype of the subjects studied in
ADNI-2 as well as to provide well-characterized biofluids to other investigators whose ADNI-2 “add-on”
studies are approved by the RARC. The risks of this study are minimal and the likelihood of this study
further contributing to our understanding of the pathogenesis of EMCI/LMCI and AD, their differential
diagnosis and distinction from NC, and other applications of AD biomarkers is great. This would have
value to individuals and families suffering from AD, EMCI/LMCI or related dementias as well as to the
general community.
5. Inclusion of Women
Women will be encouraged to participate in the study and the Progress Report of the ADNI-2 Clinical
Core documents the plans for the gender and ethnic mix of the ADNI-2 cohort.
6. Inclusion of Minorities
Minorities will be encouraged to participate in the study. AD and EMCI/LMCI affect individuals of
multiple racial and ethnic backgrounds therefore the ADNI-2 Clinical Core seeks a diverse recruitment
for study in the Biomarker Core and in all ADNI-2 studies so all the studies conducted in ADNI-2
accurately reflect the diversity of patients affected by AD and EMCI/LMCI.
a. Human biofluids previously collected by the ADNI-2 Clinical Core with informed consent and
banked in the Biomarker Core will be studied in this Core. This does not entail any risk to living
individuals and the identity of all subjects whose biofluids are studied in this Core is confidential.
5.7. Vertebrate Animals
None
5.8. Letters of Support
See Section on Consortium/Contractual Arrangements.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Core: 6
Title of Core (not to exceed 81 spaces): Genetics Core
Core Leader: Saykin, Andrew, J.
Position/Title: Professor, Indiana University
Mailing Address:
950 West Walnut Street, E124
Indianapolis, IN 46202
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Genetic variation has an important influence on brain structure, function and risk for developing AD and MCI.
The Genetics Core, an outgrowth of the working group that recently completed a genome wide association
study (GWAS), was added as part of the ADNI RC2 grant. The overall goal is to facilitate the investigation of
genetic influences on early disease trajectory including structural, functional and molecular imaging changes,
fluid biomarkers, and cognitive performance. Specific Aims include (I) sample processing, genotyping and
dissemination, (II) GWAS of longitudinal phenotypic data, and (III) providing a central resource, point of
contact and planning group for genetics in ADNI. For Aim I, the Core will receive and bank blood samples,
extract DNA and RNA, develop and store immortalized cell lines, perform genotyping with quality control and
sample verification, coordinate with the Informatics Core for data release, and disseminate cell line derived
DNA and other samples for RARC-approved projects. For Aim II, the Core will work with other ADNI cores
and collaborators to ensure that systematic analyses of genotype-phenotype associations are completed on
the ADNI data set. For Aim III, the Core will provide an organizational and informational resource for
investigators, industry partners and other parties interested in analyses of the ADNI genetics data, including
hosting conference calls, serving as a liaison to organizations with synergistic research activities, facilitating
collaborations for replication and combined imaging genetics analyses of MCI and AD risk and rate of
change, and bringing together experts to identify important and promising future directions including followup proposals for appropriate post-genomic analyses. In sum, the Genetics Core will serve the consortium by
fostering the broad dissemination of results and data ensuring the maximal scientific yield from the genetic
data being collected. The ability to map the relationship between genetic variation and the rich longitudinal
phenotypic data generated by ADNI is an extraordinary scientific opportunity for creating new knowledge
regarding MCI and AD that will ultimately impact health outcomes.
ADNI is a landmark longitudinal study including structural and molecular imaging, fluid biomarkers and
cognition in early AD and prodromal states. By facilitating the study of genetic influences on longitudinal
changes in ADNI phenotypes, the Genetics Core will help to maximize the scientific yield with information
relevant for disease risk/mechanisms, treatment development, and selection criteria for future clinical trials.
Indiana University
DUNS:
60-300-7902
Street 1:
950 W. Walnut Street
City:
Street 2:
Indianapolis
County:
Province:
R2 E124
Marion
State:
USA
IN-007
Country:
Zip/Postal Code:
IN
46202
Indiana University
DUNS:
60-300-7902
Street 1:
410 W. 10th Street
City:
Street 2:
Indianapolis
Province:
PHS 398 (Rev. 11/07)
County:
Country:
Marion
USA
HS 4000
State:
Zip/Postal Code:
IN
46202-3002
IN-007
Page 2
Form Page 2
Weiner, Michael
Name
Organization
Role on Project
Saykin, Andrew
saykin
Indiana University
Core Leader
Foroud, Tatiana
tforoud
Indiana University
Co-Leader
Shen, Li
li_shen
Indiana University
Co-Leader
Name
Organization
Role on Project
Bertram, Lars
Max-Planck Institute
Collaborator
Farrer, Lindsay
Boston Univeristy School of Medicine Collaborator
Green, Robert
Boston University School of Medicine Collaborator
Moore, Jason
Dartmouth Medical School
Collaborator
Potkin, Steven
UCI
Co-Leader (UCI subcontract)
Thompson, Paul
UCLA
Collaborator
No
Yes
Cell Line
PHS 398 (Rev. 11/07)
Page 3
Weiner, Michael
6. GENETICS CORE:
6.1. Specific Aims:
Genetic variation plays a strong role in brain structure [1-3], function [4-6], and risk for developing AD [7, 8].
APOE genotype was explicitly included in the design of ADNI1 given extensive evidence supporting its role in
AD risk. Because it was recognized that numerous other genes are likely involved in AD risk and longitudinal
trajectory, additional sources of funding were obtained to support a high density genome-wide association
study (GWAS) of the ADNI1 cohort. The Genetics Core is an outgrowth of the Genetics Working Group that
recently completed and publicly released the genome-wide scan with over 600,000 markers genotyped in the
818 ADNI1 samples (4/16/09) and subsequently completed initial analyses of these data (see Preliminary
Studies).
The Core will provide collaborative genetic resources leading to a deeper understanding of the genetic
influences on clinical trajectory, biological pathways involved in cognition, cognitive aging and decline, as well
as structural, functional and molecular imaging changes associated with dementia. In addition, the imaging
phenotypes and other quantitative traits developed by ADNI will serve as a discovery tool for genes associated
with dementia, dementia risk and related variables such as rate of clinical progression. The Genetics Core will
serve the consortium by providing a central point of communication linking the many parties within and outside
of ADNI interested in the interface of genetics, brain imaging and dementia. This will ensure the broadest
dissemination of results and data and the maximal scientific yield from the extensive data being collected. The
ability to map the relationship between genetic variation and the extremely rich phenotypic data created by
ADNI is an extraordinary scientific opportunity. This investigation will be initiated in the Grand Opportunities
(GO) project and continue in ADNI2, when 550 additional samples will be collected and analyses will be
required. In ADNI2, the Genetics Core will accomplish the following Specific Aims:
Aim I: Sample processing, genotyping and dissemination:
a. Receive and bank blood samples, extract DNA and RNA, and perform APOE genotyping for 550 new early
and late MCI, AD and control participants. Develop and store immortalized cell lines on new samples.
b. Perform genotyping using an updated Illumina Human BeadChip compatible with the array used in ADNI1.
c. Perform quality control and sample verification for GWAS; coordinate with the Informatics Core to make the
GWAS data rapidly available to the scientific community.
d. Disseminate cell line derived DNA and other samples for RARC-approved projects.
Aim II: Genome-wide analysis of multidimensional phenotypic data collected on the ADNI cohort:
a. Assemble and organize the specialized bioinformatics, genetics and computational expertise needed for
analysis of genetic/genomic variation in relation to ADNI phenotypic data.
b. Collaborate with ADNI cores (clinical, biomarker, informatics, biostatistics, MRI, PET) and with ISAB,
NCRAD, NIA Program, the NIA-funded AD Genetics Consortium, outside experts and other relevant
entities, to facilitate integrative analyses for hypothesis testing and novel discovery.
c. Perform comprehensive analysis of genetic influences on baseline data including identifying novel risk
markers and validating established markers that predict conversion and progression. Emphasis will be on
continuous phenotypic measures from structural, functional, and molecular neuroimaging, other biomarkers
CSF, blood, urine) and clinical and neuropsychological variables.
d. Perform comprehensive analysis of genetic influences on longitudinal data with the goal of identifying
genetic features associated with rate and characteristics of progression.
Aim III: Serve as a central resource, point of contact and planning group for genetics in ADNI:
a. Provide an organizational and informational resource for investigators, industry partners and other parties
interested in analyses of the ADNI genetic data.
b. Host regular conference calls for working groups, advisory panels, and other interested parties.
c. Serve as a liaison to the recently funded NIA AD Genetics Consortium (ADGC) and World Wide ADNI
investigators collecting and analyzing GWAS data as well as other large scale cohort studies that could
serve as replication samples or collaborate in other synergistic research activities. A major focus of the
Core will be to facilitate international collaborations for replication and combined imaging genetics analyses
of MCI and AD risk and rate of change.
d. Bring together an extended advisory group of experts to identify important and promising future directions.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
6.2.1. Genetic Factors in AD and MCI: Genetic factors play a key role in AD. A twin study by Gatz et al.
(2006) showed a high heritability, i.e., proportion of phenotypic variation caused by genetic factors, in AD [12].
This high heritability has led to a number of studies to identify genes that may be associated with AD. AD can
be classified into two groups: early-onset AD (EOAD) (onset < 65 years) and late-onset AD (LOAD) (onset ≥
65 years). EOAD accounts for a small percentage of cases (< 5%) and is primarily caused by mutations in
three genes that affect the cerebral levels of amyloid-β peptide: APP (amyloid precursor protein), PSEN1
(presenilin 1) and PSEN2 (presenilin 2) [13]. LOAD accounts for the majority of AD cases, but the genes
associated with it have been difficult to find. A list of top candidate LOAD genes can be found at
http://www.alzgene.org/ [7]. The ε4 allele of APOE gene is the strongest genetic risk factor for AD. As
compared to individuals with no ε4 alleles, the increased risk for AD is 2- to 3-fold in people with one ε4 allele
and about 12-fold in those with two ε4 alleles. The APOE ε4 allele is also associated with an earlier age of AD
onset [14]. A number of other genes have also been identified, but until very recently only APOE had been
consistently replicated (see Section 6.2.2). Many approaches are now being attempted to identify genes that
may play a role in AD. Genetic studies have also been performed in MCI patients. In a recent study of
amnestic MCI patients, Barabash et al. (2009) found an increased risk of MCI for participants with an APOE ε4
allele, as well as a higher risk to evolve to AD before 20 months in MCI patients with an ACT polymorphism but
lower risk in participants with a CHRNA7 polymorphism [15]. Many approaches are now being attempted to
identify genes that may play a role in the development and progression of AD and MCI. These techniques are
briefly explored in the following sections.
6.2.2. Genome-Wide Association Studies: Genome-wide association studies (GWAS) employ tests of
association between markers, called single nucleotide polymorphisms (SNPs), distributed across the genome
and a phenotype of interest, which could be dichotomous (affected, unaffected) or quantitative (longitudinal
change, imaging phenotypes). This approach has identified susceptibility loci in a number of diseases
including type 1 and type 2 diabetes, breast cancer and psychiatric disorders. An updated list of published
GWAS can be found at the National Human Genome Research Institute (NHGRI)’s catalog of published
genome-wide association studies (http://www.genome.gov/26525384). In AD, GWAS have confirmed a strong
association with APOE, but have found less convincing evidence implicating other genes. This suggests that
susceptibility genes other than APOE may have modest effects and require large sample sizes to detect them
([16-22]). Recently, two large GWAS have identified three loci: CLU, CR1 and PICALM to be strongly
associated with AD ([23, 24]). These loci also showed strong association in replication studies, further
supporting a role in AD susceptibility.
6.2.3. Imaging Genetics: Recent advances in brain imaging and high throughput genotyping techniques
enable new approaches to study the influence of genetic variation on brain structure and function [22, 25-28].
As a result, imaging genetics has become an emergent trans-disciplinary research field where genetic variation
is evaluated using imaging measures as quantitative traits (QTs) or continuous phenotypes. Imaging genetics
studies have certain advantages over traditional case control studies. QT association studies have been shown
to have increased statistical power and thus decreased sample size requirement [29]. Additionally, imaging
phenotypes may be closer to the underlying biological etiology of the disease making it easier to identify
underlying genes (e.g., [22]).
SNPs and other types of polymorphisms in single genes such as APOE have been related to neuroimaging
measures in both healthy controls and participants with brain disorders such as MCI and AD (e.g., [30, 31]).
However, analytic tools that relate a single gene to a few imaging measures are insufficient to provide
comprehensive insight into the multiple mechanisms and imaging manifestations of these complex diseases.
Although GWAS analyses are increasingly performed [32-35], effectively relating high throughput SNP data to
large scale image data remains a challenging task. As pointed out by Glahn et al. [36], prior imaging genetics
studies typically make significant reductions in one or both data types in order to complete analyses. For
example, whole brain studies usually focus on a small number of genetic variables (e.g., [37-42]), while whole
genome studies typically examine a limited number of imaging variables (e.g., [22, 43, 44]). This restriction of
target genotypes and/or phenotypes greatly limits our capacity to identify important relationships.
6.2.4. Deep Resequencing: While GWAS have identified chromosomal regions and genes of interest,
association does not typically infer causation. As a result, targeted sequencing or deep resequencing of genes
or regions associated with a phenotype is often employed to catalog all sequence variation (SNPs,
insertion/deletions, etc), as a first step in determining which variation(s) directly contribute to disease risk. For
example, deep resequencing has helped to determine how sequence variations in genes such as nuclear
receptor 2E1 gene (NR2E1), dopamine- and cAMP-regulated phosphoprotein of molecular weight 32 kDa
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
(DARPP-32), synaptogyrin 1 (SYNGR1), brain-derived neurotrophic factor (BDNF) and glutamate receptor,
metabotropic 3 (GRM3) are associated with psychiatric disorders such as bipolar disorder, schizophrenia and
major depressive disorder ([45-49]). This technique has also helped identify variants associated with X-linked
mental retardation (XLMR) and epilepsy and mental retardation limited to females (EFMR) [50, 51]. Although
there has been only one study in AD thus far using this approach [52], deep resequencing techniques hold
great promise in uncovering new variants which may be involved in the pathogenesis of AD.
6.2.5. Copy Number Variation (CNV): CNVs are segments of DNA, ranging from 1 kilobase (kb) to several
megabases (Mb), for which copy-number differences have been revealed by comparison of two or more
genomes. These differences can be copy-number gains (duplications or insertional transpositions), losses
(deletions), gains or losses of the same locus, or multiallelic or complex rearrangements. CNVs have been
implicated in various disorders, such as autism, schizophrenia, bipolar disorder and cancer (reviewed in [5357]). Studies in EOAD have shown that APP gene duplications can cause EOAD with cerebral amyloid
angiopathy ([58, 59]). Theuns et al. (2006) identified three mutations in AD patients that showed a nearly two
fold neuron-specific increase in APP transcriptional activity [60]. A novel gene (ASAH2L) identified by
Avramopoulos et al. (2007), which is the result of partial duplication of the alkaline ceramidase gene (ASAH2),
showed reduced expression with increasing age, reduced expression in females across all ages, and further
reduction in LOAD patients [61]. In a study of two genetic CNVs, rs1937 and rs2306604 of mitochondrial
transcription factor A (TFAM) in an AD and Parkinson’s disease (PD) case-control study, Belin et al. (2007)
found significant genotypic and allelic association of rs2306604 to AD [62]. In a study examining gene loci in
Down syndrome (DS), Korbel et al. (2009) suggested the involvement of a CNV of a 1.95Mb interval including
APP [63]. Heinzen et al. (2009) performed a genome-wide scan of AD in 331 cases and 368 controls but did
not find any new SNP CNVs that were significant at genome-wide threshold. However, a duplication in the
CHRNA7 gene was found, which may warrant further investigation [64]. Future investigations into CNVs
leading to increased or decreased expression of genes involved in AD may result in the discovery of important
genetic variations associated with the development of AD or differential response to therapeutic strategies.
6.2.6. MicroRNAs: MicroRNA (miRNAs or uRNA) are a large family of short (21-25 nucleotides) non-coding
regulatory RNAs involved in post-transcriptional gene silencing. They hybridize to partially complementary
binding sites typically located in the 3’ untranslated regions (3’-UTR) of target messenger RNA (mRNA)
molecules and repress mRNA expression, either by interfering with translation or by targeting mRNA for
degradation [65]. The most recent version (13.0) of the miRNA database miRBase identifies 706 validated
miRNAs in the human genome [66], although there are estimated to be as many as thousand miRNA genes
[67]. miRNAs are abundant in the nervous system and have been shown to play a key role in synaptic
development and function [68]. miRNA-mediated inhibition may be an important part of synaptic plasticity in the
central nervous system and may play an important role in the formation of long-term memory [69]. Hence,
miRNAs have been studied in neurological diseases, such as Fragile-X syndrome, Rett syndrome, Parkinson’s
disease [PD], Down syndrome and AD (reviewed in [70-72]). A number of miRNAs have been found to be
differentially expressed and to alter the expression of other target genes in AD. Lukiw et al. (2006) found
elevated levels of miR-9, miR-125 and miR-128 and reduced levels of miR-124a in AD brains when compared
to age-matched controls [73]. Wang et al. (2008) found reduced levels of miR-107 early in the course of AD,
and a physiological miR-107 binding site on BACE1 mRNA that could regulate BACE1 levels [74]. Herbert et
al. (2008) found that miR-29a, miR-29b-1 and miR-9 could regulate BACE1 expression in vitro, as well as a
significant decrease of the miR-29a/b-1 cluster in AD patients with abnormally high BACE1 protein [75].
Boissonneault et al. (2009) found specific binding sites for miR-298 and miR-328 in the 3’-UTR of BACE1
mRNA and demonstrated that these miRNAs could exert regulatory effects on BACE1 expression in cultured
neuronal cells [76]. In a study to investigate the role of miRNAs in the regulation of APP gene expression,
Hebert et al. (2009) showed that miRNAs belonging to the miR-20a family (miR-20a, miR-17-5p, miR-106b)
could regulate APP expression in vitro and at the endogenous level in neuronal cell lines [77]. Altered
expression of miRNAs was also found in the CSF [78], as well as in blood mononuclear cells, of patients with
AD [79]. Thus, a number of studies have shown that miRNAs can regulate expression of various genes known
to be involved in AD, such as APP and BACE1, suggesting that miRNAs may play a role in AD pathology.
6.2.7. Gene Expression Data Analysis: Analysis of aberrant gene expression profiles in AD is difficult
because tissue samples for expression analysis cannot be acquired from living patients. Additionally, postPHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
mortem RNA degradation and protein modification may potentially confound the interpretation of data obtained
from neuropathological samples. Expression profiling of peripheral blood mononuclear cells (PBMC) may offer
advantages in deciphering gene regulation patterns because (a) circulating blood is easily accessible and cells
are readily procured by a simple venipuncture, (b) the central nervous system (CNS) communicates with the
immune system through multiple molecular, hormonal and neurotransmitter mechanisms which may be
expressed in the blood, (c) abnormal APP expression, altered levels of antioxidant enzymes, oxidative damage
to DNA, RNA and protein, deregulated cytokine secretion and augmented rates of apoptosis are features
shared by AD brain and blood lymphocytes, and (d) PBMC have been previously employed in the prognosis
and diagnosis of other neurological diseases [80, 81]. Liew et al. (2006) demonstrated that peripheral blood
cells share more than 80% of the transcriptome with each of the nine tissues: brain, colon, heart, kidney, liver,
lung, prostate, spleen and stomach, suggesting that peripheral blood cells can be used as an alternative to
biopsy tissue for molecular profiling in humans [82]. A number of studies have used PBMC for gene expression
studies in AD [80, 83-90]. Jasinska et al. (2009) analyzed expression profiles in brain tissues derived from
eight different brain regions and from blood in 12 vervet monkeys, a biomedically important non-human primate
model. They characterized brain regional differences in gene expression, focusing on transcripts for which
inter-individual variation of gene expression in the brain correlated well with variation in expression in the blood
from the same individuals. Using stringent criteria, they identified 29 genetic variants whose expression was
measurable, stable, replicable, variable between individuals, relevant to brain function, and heritable, and
identified these transcripts as candidate expression Quantitative Trait Loci (eQTL), differentially regulating
transcript levels in the brain [91]. Thus, genes altered in peripheral blood may also be altered in the brain.
6.2.8. Gaps to Be Addressed: The genetics of AD is a rapidly developing area. The meta-analytic AlzGene
database (www.alzgene.org), curated by Core collaborator Lars Bertram (Max Planck Inst.) and colleagues, is
dynamically changing and reflects the continuing evolution of leading candidate genes for AD. Despite many
biologically plausible and interesting candidates, the genetic signal in AD has largely been dwarfed by APOE,
the most widely replicated susceptibility marker for LOAD. As discussed in the Preliminary Studies sections
below (6.3), TOMM40, an interesting gene adjacent to APOE, has shown strong association with AD, as well
as target imaging phenotypes, as reported in the first GWAS publication from ADNI [22] and submitted
manuscripts from the IU and UCLA groups [10, 11, 92-94]. Presently, it is unclear to what extent TOMM40
plays a unique contributory role in AD or whether it serves solely as an additional marker of a known, important
locus (i.e., APOE). Further research including quantitative trait/phenotype studies will be extremely important to
clarify the role of TOMM40 in AD pathology. Furthermore, just prior to the submission of the present ADNI2
application, new replicated findings from two very large European GWAS studies implicated several additional
biologically plausible candidate genes (CLU, PICALM and CR1) [23, 24]. Our initial analyses indicate
relationships between structural MRI data and these target SNPs in the ADNI1 cohort. Other major candidate
genes in AlzGene predict variance in rate of change on MR scans as we reported at ICAD 2009 [9]. In sum, in
the combined ADNI1, GO and proposed ADNI2 data set there will be great potential for both discovery and
validation of imaging, clinical and biomarker associations with genetic variation. Replication of the ADNI1
GWAS during ADNI2 is a major goal, as is replication of the combined ADNI1/GO/ADNI2 sample of over 1550
participants by other similar cohort studies under development as part of worldwide ADNI efforts.
Identification of those at greatest risk for decline with an optimized panel of genetic markers, in combination
with imaging and other biomarkers, can help to reduce the sample size of clinical trials and may provide new
clues to fundamental disease mechanisms that could offer new therapeutic targets. However, there are myriad
challenges for GWAS in relation to quantitative phenotypes being obtained by ADNI. The emerging new transdisciplinary field of imaging genomics also has many important challenges such as optimal statistical modeling,
power and multiple comparison issues. Thus, a multidisciplinary team in needed in order to coordinate work
addressing these issues. We have assembled such a team, as described in Preliminary Studies, throughout
the research plan, and in the personnel and budget justification sections.
A number of promising molecular genetics and bioinformatics methodologies applied to the ADNI samples
could potentially provide exciting and novel information on an extremely well-characterized cohort of those at
high risk for AD. Longitudinal blood sample collection in ADNI will permit expression analyses, given that an
estimated 80% of the transcriptome is shared between PBMCs and brain [82] and promising results have been
reported in AD [80, 83-90], as described above (section 6.2.7). Although it is beyond the scope and budget of
ADNI2 to propose longitudinal gene expression from blood, expression studies in post-mortem brain tissue,
microRNA studies or targeted deep resequencing at this time, the Core will collect and bank the relevant data
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
and will provide a venue to foster new proposals and collaborations so that the most scientifically important
directions are followed. In summary, there has been exciting progress in genetics and genomics of AD and
ADNI is in a unique position to conduct and foster analyses of quantitative longitudinal imaging, biomarker and
clinical phenotypes.
6.3. Progress Report and Preliminary Data:
6.3.1. Creation of the Genetics Core: ADNI is a landmark study opening many new possibilities for clinically
meaningful and experimentally important research on structural and molecular imaging and biomarkers in AD
and prodromal states (EMCI, LMCI, asymptomatic older adults with high amyloid burden, abnormal CSF or
degenerating hippocampi). The great progress in MRI, FDG and amyloid PET imaging in AD is paralleled by
the availability of unprecedented technology for assaying the architecture of the human genome on cost
effective microarrays. Although genetic assessment beyond APOE genotyping was not included in ADNI1, in
August 2005 Dr. Weiner, cognizant of the potential importance and interest in academia and industry, tasked
an ADNI Genetics Working Group to develop plans and seek funding for genetic studies of the ADNI cohort.
Dr. Saykin, an ADNI site PI and neuroscientist working on strategies for integrating imaging and genomic data
in MCI and AD, was asked to chair this group and initiate planning for funding proposals. Multidisciplinary
members of the working group included Drs. Bryan DeChairo (Pfizer, ISAB representative, replaced by Elyse
Katz in 2009), Lindsay A. Farrer (BU), Tatiana M. Foroud (Indiana U, NCRAD), Robert Green (BU), Steven
Potkin (UCI), Eric Reiman (Banner, TGen), Andrew J. Saykin (Indiana U, Chair), Gerard D. Schellenberg (U
Penn), Rudolph E. Tanzi (MGH), John Q. Trojanowski (U Penn, Biomarker Core leader), Christopher van Dyck
(Yale), Michael W. Weiner (UCSF, ADNI PI), Kirk C. Wilhelmsen (UNC). The working group considered
potential data collection and analyses and recommended completing a GWAS using the then new Illumina
Human 1M BeadChip panel. Funds were raised via the FNIH from an anonymous foundation, Merck and Gene
Network Sciences and Pfizer contributed DNA extraction and other assistance with the project. Illumina
provided arrays at a strong discount. A supplement from the NIBIB provided initial support for data analysis.
Sufficient funds were raised by a challenge grant deadline to genotype the entire cohort of over 800 samples
on the Illumina 610 Quad array with over 620,000 markers. The arrays were performed by TGen (Phoenix), a
NIH Neuroscience Microarray Consortium site under the supervision of Drs. Matthew Huentelman and David
Craig. Sample processing, storage and distribution were provided by the NIA-sponsored National Cell
Repository for Alzheimer’s Disease (NCRAD). QC bioinformatics were performed at Indiana University (IU) by
Drs. Li Shen, Tatiana Foroud and staff in conjunction with Drs. Huentelman and Craig (TGen), Bryan Dechairo
(Pfizer) and Steven Potkin’s group at UC Irvine (UCI). The IU group worked closely with Dr. Arthur Toga and
the Informatics Core to quickly format, document and post the final genotyping results on the ADNI LONI web
site. The data was released to the scientific community on 4/16/09. Several initial GWAS analyses of ADNI
phenotypes have been completed and submitted for publication by the UCI, IU and UCLA groups and many
others are in various stages of completion. In 2009, the GO proposal was submitted and at that time it was
decided to include genetics as a new core to foster imaging genetics and GWAS research related to other
biomarkers in an integrated manner among and beyond the ADNI community.
6.3.2. Genotyping of the ADNI1 Cohort: The genotyping for 620,901 single nucleotide polymorphism (SNP)
and copy number variation (CNV) markers was completed on all ADNI participants using the following protocol.
7 ml of blood was taken in EDTA containing vacutainer tubes from all participants and genomic DNA was
extracted using the QIAamp DNA Blood Maxi Kit (Qiagen, Inc., Valencia, CA) following the manufacturer’s
protocol. Lymphoblastoid cell lines were established by transforming B lymphocytes with Epstein-Barr virus as
described by [95]. Genomic DNA samples were analyzed on the Human610-Quad BeadChip (Illumina, Inc.
San Diego, CA) according to the manufacturer’s protocols (Infinium HD Assay; Super Protocol Guide; Rev. A,
May 2008). Before initiation of the assay, 50ng of genomic DNA from each sample was examined qualitatively
on a 1% Tris-acetate-EDTA agarose gel to check for degradation. Degraded DNA samples were excluded from
further analysis. Samples were quantitated in triplicate with PicoGreen® reagent (Invitrogen, Carlsbad, CA) and
diluted to 50 ng/µl in Tris-EDTA buffer (10mM Tris, 1mM EDTA, pH 8.0). 200 ng of DNA was then denatured,
neutralized, and amplified for 22 hours at 37°C (this is termed the MSA1 plate). The MSA1 plate was
fragmented with FMS reagent (Illumina) at 37°C for one hour, precipitated with 2-propanol, and incubated at
4°C for 30 minutes. The resulting blue precipitate was resuspended in RA1 reagent (Illumina) at 48°C for one
hour. Samples were then denatured (95°C for 20 minutes) and immediately hybridized onto the BeadChips at
48°C for 20 hours. The BeadChips were washed and subjected to single base extension and staining. Finally,
the BeadChips were coated with XC4 reagent (Illumina), desiccated, and imaged on the BeadArray Reader
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
(Illumina). The Illumina BeadStudio 3.2 software was used to generate SNP genotypes from bead intensity
data.
After performing essential sample verification and quality control bioinformatics, the genotyping data of 818
ADNI participants were uploaded to the ADNI website (http://www.loni.ucla.edu/ADNI) and released to the
public on April 16, 2009. Based on the web access records on October 4, 2009, there have been 42,270
total downloads (counted as one per participant) by 94 different users in the past six months. Within
this short period of time, there were also several initial imaging/biomarker genetics studies completed on the
ADNI cohort using this data. These initial analyses are briefly summarized below.
6.3.3. Genetic Analyses of Baseline MRI Scans: Potkin et al. [22] completed two GWAS analyses on 381
participants from the ADNI study: a standard case-control analysis, and a novel approach using hippocampal
atrophy measured on MRI as an objectively defined, quantitative phenotype. A General Linear Model was
applied to identify SNPs for which there was an interaction between genotype and diagnosis on the
quantitative trait. The case-control analysis identified APOE and a new risk gene, TOMM40 (translocase of
outer mitochondrial membrane 40), at a genome-wide significance level of 10-6 (10-11 for a haplotype).
TOMM40 risk alleles were approximately twice as frequent in AD patients as controls. The quantitative trait
analysis identified 21 genes or chromosomal areas with at least one SNP with a p-value 10-6, which can be
considered potential ‘‘new’’ candidate loci to explore in the etiology of sporadic AD. These candidates included
EFNA5, CAND1, MAGI2, ARSB, and PRUNE2, genes involved in the regulation of protein degradation,
apoptosis, neuronal loss and neurodevelopment.
(For open access full text: http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0006501)
Nho et al. [92] completed a GWAS analysis on the ADNI cohort using a more stringent criterion for quality
control (136 AD and 181 HC involved). SNPs were tested for allelic associations using the chi-square test. A
permutation method was used to calculate the empirical P values for association significance to control for the
risk of false positives. The case-control GWAS analysis confirmed that SNP rs2075650 is strongly associated
with an increasing risk of AD using both asymptotic P values (p = 1.043 × 10-10) and corrected empirical P
values (p = 1× 10-4). This SNP is located in the TOMM40 gene on chromosome 19, roughly 13 kilobase pairs
distal to APOE, and exhibits strong linkage disequilibrium with APOE at D' = 0.93.
Shen et al. [10] developed a genome-wide, whole brain approach to investigate genetic effects on
neuroimaging phenotypes for identifying quantitative trait loci, and applied this technique to the ADNI 1.5T MRI
and genetic dataset. Using voxel-based morphometry (VBM) and FreeSurfer parcellation, 142 measures of
grey matter (GM) density,
volume, and cortical thickness
were extracted from baseline
MRI scans and investigated
as target phenotypes. GWAS
techniques, using PLINK,
were performed on each
phenotype
using
quality
controlled genotype and scan
data including 530,992 of
620,903 SNPs and 733 of
818 participants (175 AD, 354
MCI,
and
204
HC).
Hierarchical clustering and
heat maps [96] (Fig. 6.1.)
were used to analyze the
GWAS
results
and
associations were reported at
two significance thresholds Fig. 6.1. Heat maps of SNP associations with quantitative traits (QTs) at the significance level of p<10(p<10-7 and p<10-6).
As 7. GWAS results at a statistical threshold of p<10-7 using QTs derived from VBM/MarSBaR are shown. from each GWAS are color-mapped and displayed in the heat maps. Heat map blocks labeled
expected, SNPs in the APOE log10(p)
with "x" reach the significance level of p<10-7. Only top SNPs and QTs are included in the heat maps,
and TOMM40 genes were and so each row (SNP) and column (QT) have at least one "x" block. Dendrograms derived from
confirmed
as
markers hierarchical clustering are plotted for both SNPs and QTs. The color bar on the left side of the heat map
strongly
associated
with codes the chromosome IDs for the corresponding SNPs [10].
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
multiple brain regions. Other top SNPs showing significant associations with imaging phenotypes were
proximal to the EPHA4, TP63 and NXPH1 genes. Detailed image analyses of rs6463843 (flanking NXPH1)
revealed reduced global and regional GM density across diagnostic groups in TT relative to GG homozygotes.
Interaction analysis indicated that AD patients homozygous for the T allele showed increased vulnerability to
right hippocampal GM density loss. NXPH1 codes for a protein implicated in promotion of adhesion between
dendrites and axons, a key factor in synaptic integrity, the loss of which is a hallmark of AD. A genome-wide,
whole brain search strategy has the potential to reveal novel candidate genes and loci warranting further
investigation and replication.
Paul Thompson’s group (Core collaborator) performed several analyses of ADNI1 data. In a GWAS of
neurodegeneration, Stein et al. [11] mapped the 3D profile of temporal lobe degeneration in 742 brain MRI
scans of ADNI participants. After searching 546,314 genomic markers, 2 SNPs were associated with temporal
lobe volumes (P<5x10-7), and with increased atrophy in all 3 diagnostic categories. One significant SNP,
located in the GRIN2B gene which encodes the N-Methyl-D-Aspartate (NMDA) glutamate receptor NR2B
subunit, was significantly over-represented in AD and MCI patients versus controls (odds ratio=1.273; P =
0.039). This protein - involved in learning and memory, and excitotoxic cell death - has age-dependent
prevalence in the synapse, and is already a therapeutic target in Alzheimer's disease. Voxel-by-voxel, 3D
maps of genetic association with regional brain volumes, also revealed intense temporal lobe effects
(P=0.0257, corrected).
Stein et al. [94] completed an imaging genetics study on the ADNI cohort to find genes influencing brain
structure. They explored the relation between 545,871 single nucleotide polymorphisms in each of 252,407
voxels of the entire brain across 719 elderly subjects (mean age ± s.d. 75.44 ± 6.84 years; 421 male). Tensorbased morphometry was employed to measure individual differences in brain structure at the voxel level
relative to a study-specific template based on healthy elderly subjects and then a genome-wide association at
each voxel was conducted to identify genetic variants of interest. A novel method was developed to address
the multiple comparisons and computational burden associated with the unprecedented amount of data.
Several genes worthy of further exploration were identified using this method, including XKR4, PIP4K2A,
CSMD2, CADPS2, and PIP3-E. These genes have high relevance to brain and cytoskeletal structure; some
have been previously associated with psychiatric disease (Fig. 6.2.).
Ho et al. [97] used the ADNI MRI and genetic data to study how the obesity-associated FTO risk allele
affects human brain structure. They generated 3D maps of regional brain volume differences in 206 healthy
elderly participants scanned with MRI and genotyped as part of the ADNI. They found a characteristic pattern
of systematic brain volume deficits in carriers of
the obesity-associated risk allele versus non- carriers. FTO (fat mass and obesity-associated)
risk allele carriers had an average brain volume
deficit of around 10% in the frontal lobes and
15% in the occipital lobes. These regions also
showed significant volume deficits in participants
with higher BMI. These brain differences were
not attributable to differences in cholesterol
levels, hypertension, or the volume of white
matter hyperintensities; these were not
detectably higher in FTO risk allele carriers
versus non-carriers. The results of this study
reveal
that
a
very
commonly-carried
susceptibility allele for obesity is associated with
brain atrophy, a finding with significant health
Fig. 6.2. Map of most associated SNPs at each voxel. Each image
implications.
represents slices through the brain at 4 mm intervals from inferior to superior.
The top of the page represents anterior of the brain and the bottom of
represents posterior. The images are in radiological convention (left of the
image is the right side of the subject). There are 14,285 SNPs represented in
the map. Each SNP is colored differently, though there are only 127 color
gradations so some SNPs appear to be the same color. The gradient seen
from the inferior to superior is due to the numbering of SNPs, beginning at
the bottom of the brain.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
6.3.4. Genetic Analyses of Conversion Status and MRI Changes at 12 Month Follow-up:
Adjusted Annual Percent Change in
Hippocampal Volume
Adjusted Annual Percent Change in
Hippocampal Grey Matter (VBM)
APOE (Chr 19): rs429358 is the epsilon 4 allele marker
TOMM40 (Chr 19): translocase of outer mitochondrial membrane 40 homolog (LD with APOE)
CADH8 (Chr 16): cadherin 8, type 2; synaptic adhesion, axonal growth/guidance (no data in AD)
SLC6A13 (Chr 12): GABA transporter protein (no data in AD)
APOE (Chr 19): rs429358 is the epsilon 4 allele marker / TOMM40 in LD with APOE
MAD2L2 (Chr 1) mitotic arrest deficient-like 2 (mitotic spindle assembly)
LOC728574 (Chr 22): similar to retinitis pigmentosa GTPase regulator isoform C
QPCT (Chr 2): role in pyroglutamate formation of amyloid-related plaque-forming peptides?
GRB2 (Chr 17): growth factor receptor-bound protein 2; one isoform may trigger apoptosis
Fig. 6.3. Preliminary GWAS of annual percent change (APC) in hippocampal volume and GM density [9].
Annualized
Percent
Change in
Hippocampal
Volume:
Annualized
Percent
Change in
Hippocampal
Volume:
APOE
Epsilon 4
Dose
TOMM40
rs2075650
A/G
Fig. 6.4. Annual percent change in hippocampal volume assessed using the FreeSurfer software: Influence of number APOE epsilon 4
alleles (left) and TOMM40 genotype (right) [9].
Risacher et al. [98] analyzed baseline 1.5T MRI scans in 693 participants from the ADNI cohort divided into
four groups by baseline diagnosis and one year MCI to probable AD conversion status to identify neuroimaging
phenotypes associated with MCI and AD and potential predictive markers of imminent conversion. Nho et al.
[93] performed a follow-up GWAS analysis on the same sample using both the Illumina genotyping data and
APOE status data to identify genetic variants associated with the predisposition to AD and amnestic MCI. Four
SNPs were found to be strongly associated with AD: rs429358 on chromosome 19q13.32 (APOE, uncorrected
P value = 1.122 × 10-16, odds ratio (OR)=4.209, 95% CI=2.958-5.989), rs2075650 on chromosome 19q13.32
(TOMM40, uncorrected P value = 4.633 × 10-10, OR=3.050, 95% CI=2.128-4.370), rs1124544 on chromosome
22q13.32 (LOC728574, uncorrected P value= 9.604 × 10-8, OR=2.222, 95% CI=1.654-2.987), and rs6116375
on chromosome 20p13 (PRNP, uncorrected P value= 4.997 × 10-7, OR=0.467, 95% CI=0.347-0.630). Two
imputed SNPs were also found to be strongly associated with AD: rs6857 on chromosome 19q13.32 (PVRL2)
and rs4420638 on chromosome 19q13.32 (APOC1).
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Saykin et al. [9] examined genetic predictors of 12 month change in MRI hippocampal volume in the ADNI1
cohort using top candidates from the AlzGene meta-analytic database [7]. In this preliminary report at ICAD
2009, performed prior to completion of the full GWAS data set, we analyzed whether common variants in major
candidate genes could predict longitudinal changes on MRI. Baseline and 12 month T1-weighted MP-RAGE
scans acquired on 1.5T magnets from 627 ADNI participants (141 AD, 60 MCI-Converters (MCI-C), 241 MCIStable (MCI-S), 185 healthy controls (HC)) were analyzed using FreeSurfer software for automated
parcellation and SPM5 for voxel-based morphometry (VBM). Genetic data consisted of APOE alleles and the
Illumina 610 Quad array that includes over 620,000 features. For this analysis, common (minor allele
frequency, MAF > 0.2) SNPs from the top 30 candidate genes from the AlzGene database were examined.
Regression models were performed using SAS/Stat 9.3 to test the ability of SNPs to predict hippocampal
volume and GM density changes. Models including diagnosis group (AD, MCI-C, MCI-S, HC), SNP (0,1,2),
parental history of dementia (0,1,2), APOE epsilon 4 status (positive, negative) and all interactions were
computed. 732 SNPs were available for analysis with MAF > 0.2. A threshold of p < 0.00001 was employed to
reduce the likelihood of false discoveries. Using this model and criterion, 5 genes showed significant SNPs
associated with hippocampal volume changes (NEDD9, SORL1, DAPK1, IL1B, SORCS1). In addition, SNPs
from several other candidates genes showed less robust indications of possible association (at p<.0001:
MYH13, TNK1; at p<.001: ACE, PRNP, MAPT, PCK1, GAPDHS and APP). In addition to the candidate gene
analysis we also performed a preliminary GWAS to assess genome-wide associations between target SNPs
and the annualized rate of hippocampal volume decline. As expected, APOE and TOMM40 showed very
strong association with the rate of decline of hippocampal volume. Fig. 6.3. shows the genome-wide
associations and Fig. 6.4. the main effects of APOE and TOMM40 allelic status on hippocampal volume. In
sum, we found that variations in major candidate genes, as well as other novel loci, were significantly related to
12 month longitudinal change in hippocampal volume in the ADNI cohort. These initial findings suggest that
combining genetics and imaging with other clinical data could yield refined prediction models for those at
highest risk of progression. Replication and further investigation, as proposed here, is warranted. As putatively
disease modifying agents are developed, the ability to determine those at highest risk for progression will be
critical.
6.3.5. Preliminary Analysis of Imaging Associations with New AD Genes from the October 2009 French
and UK GWAS: As noted above, two large GWAS have identified three new genes (CLU, CR1 and PICALM)
as strongly associated with AD ([23, 24]). These loci also showed strong association in replication studies,
further supporting a role in AD susceptibility. These large case/control and replication studies however do not
have neuroimaging data, early or late MCI participants or longitudinal follow-up. Although these papers
appeared just as the present ADNI2 proposal was being completed, we were able to quickly examine selected
key regions of interest implicated in early AD to determine if these new gene loci were predictive of structural
changes. Fig. 6.5. shows the association of the PICALM SNP identified in the European GWAS and mean
bilateral entorhinal cortical (EC) thickness measured by FreeSurfer in the complete ADNI1 baseline sample
[98]. EC thickness was covaried for baseline age,
Figure 6.5
sex, education, handedness and total intracranial
volume. The number of G alleles was predictive
of the degree of cortical thickness reduction in
this region, one of the earliest known structural
changes in MCI and AD, at baseline across
groups (p<.002). Importantly, this SNP remained
significant after adding APOE status as a
covariate (p=.006), suggesting an independent
contribution to structural changes associated with
MCI and AD. Although this was a main effect of
SNP across groups, the data in Fig. 6.5. suggest
that the strongest driver for this effect is for the
combined AD and MCI-converter (probable AD
within 12 months) group. The Core will now
examine the other leading novel candidates and
other brain regions heavily implicated in AD as
well as associations with rate of longitudinal
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
change.
6.3.6. Preliminary GWAS of other ADNI
Figure 6.6
Phenotypes (FDG, PIB, CSF Biomarkers,
Cognition): At Indiana University Center for
Neuroimaging, work is in progress to perform
GWAS analysis on other available ADNI1
N = 377
biomarker data (e.g., FDG PET, PIB PET, CSF
AD+MCIc: 60 AA, 62 AG, 14 GG
MCI‐S: 75 AA, 55 AG, 12 GG
biomarkers), as well as neuropsychological
HC: 77 AA, 22 AG, 1 GG
assessments, available at the ADNI website. The
approach presented in [10] was employed to
perform these analyses. Selected initial findings
at significance level of p<10-7 are summarized
below.
For CSF biomarkers, both baseline (BL) and
Month 12 (M12) Aβ1-42 were strongly associated
with the following four SNPs on chromosome 19:
rs157580 (TOMM40, intron, p=3.73*10-9 at BL, p
= 1.71*10-8 at M12), rs2075650 (TOMM40,
intron, p=1.62*10-17 at BL, p = 3.72*10-14 at M12),
rs429358 (APOE, exon 4, p=1.36*10-31 at BL, p=3.5*10-24 at M12), and rs439401 (APOE, flanking_3UTR,
p=2.22*10-9 at BL, p=1.14*10-8 at M12). Fig. 6.6. shows the relationship between allelic variation in TOMM40
(rs2075650), already identified as a major AD candidate gene from several GWAS including our own, and
baseline Aβ1-42 levels adjusted for age, sex, education and handedness (as for imaging variables). Individuals
with a greater number of A alleles showed higher levels of CSF Aβ1-42. Although this effect appeared largely
due to linkage disequilibrium with APOE, further analyses are needed to determine if there are interactions with
APOE or other AD genes.
For PIB PET, the APOE SNP, rs429358, was strongly associated with SUVR of PIB in the Lateral Temporal
Cortex at M12 with p=1.68*10-8. For FDG PET, extensive FDG-SNP associations were identified between a
number of genome-wide SNPs and both the baseline values and longitudinal change in regional mean glucose
metabolism. For neuropsychological assessments, the APOE SNP, rs429358, was strongly associated with
about 30 BL/M12 neuropsychological assessments (10-17 < p < 10-7), the TOMM40 SNP, rs2075650, was
strongly associated with 6 psychometric variables (10-10 < p < 10-7), and other top SNPs included SNPs in the
CR1, CORO2B, E2F6, LOC642487, BARHL2, LOC728186, GLULL3, ZNF774, and SLC30A1 genes (10-8 < p
< 10-7). Further analyses are ongoing to address issues of multiple comparisons, as well as to perform
additional detailed brain mapping and/or refined statistical modeling.
6.3.7. Progress Report Publications:
Potkin, S.G., et al., Hippocampal atrophy as a quantitative trait in a genome-wide association study identifying
novel susceptibility genes for Alzheimer's disease. PLoS One, 2009. 4(8): p. e6501. (PMID: 19668339). (Open
access: http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0006501)
6.4. Research Plan and Methods:
Methods, organized by Specific Aim, if described in Preliminary Studies, are referenced but not repeated.
6.4.1. Sample Processing, Genotyping and Dissemination of Data and Biological Samples (Aim 1):
6.4.1.1. Blood Samples: These will be drawn by Clinical Core personnel at each ADNI site as defined in the
participant informed consent. Blood to be transformed into lymphoblastoid cell lines is received in yellow-top
blood collection tubes containing ACD solution (at least 8.5 ml). The PAXgene system from Qiagen
(http://www.preanalytix.com/) will be used for nucleic acid collection (DNA and RNA). The PAXgene method
integrates whole blood collection and nucleic acid stabilization and purification to enhance accuracy of
intracellular DNA and RNA analysis. The PAXgene Blood DNA system is an integrated system including a
blood collection tube and kit for DNA purification. The single standardized system reduces risk of sample mixup and cross-contamination. The PAXgene tube for whole blood is stable for 14 days at room temperature or
for 28 days at 2-8°C. Similarly, the PAXgene Blood RNA tube contains a reagent to immediately stabilize
intracellular RNA for 3 days at room temperature (18-25°C) and 5 days at 2-8°C. These tubes are stored
upright at room temperature for shipment.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
6.4.1.2. Sample Receipt and Processing: This will follow established procedures already implemented for
ADNI1 and GO. The NIA-sponsored National Cell Repository for AD (http://ncrad.iu.edu/), located in the
Department of Medical and Molecular Genetics of the Indiana University School of Medicine (IUSM) in
Indianapolis, will receive and process all DNA and RNA samples and develop immortalized cell lines. This
facility is directed by geneticist and core co-leader Dr. Tatiana Foroud. Kelley Faber, MS, CCRC, Clinical
Research Manager of NCRAD, will be responsible for ensuring all standard operating procedures (SOPs) and
standards established by the IU DNA and Cell Repository are met. All samples are bar coded and DNA
“fingerprinted” with a standard set of SNP markers that are also contained on the Illumina BeadChip array. The
bar code and fingerprint SNP set serve to verify the identity of samples to ensure matching of ADNI IDs and
results of genetic analyses. For long-term storage samples are frozen at -80°C and held in a state-of-the-art,
high security bio-repository opened at IUSM in 2009. It is located in close proximity to both Medical and
Molecular Genetics and the Imaging Genomics Laboratory in Radiology and Imaging Sciences.
6.4.1.3. Immortalized Cell Lines: Lymphoblastoid cell lines will be established as in ADNI1 by transforming B
lymphocytes with Epstein-Barr virus as described by [95] and stored in the secure NCRAD bio-repository.
6.4.1.4. APOE Genotyping: The Illumina 610 and 1M arrays, like most other current GWAS platforms, do not
include the two SNPs required for APOE genotyping, rs429358 (C112R) and rs7412 (R158C). The new
HumanOmni1 array does include these SNPs. Given their importance in AD genetics, these SNPs will be
independently analyzed using standard methods by NCRAD. This will permit early identification of the alleles
for the epsilon 2/3/4 isoforms.
6.4.1.5. Genome-Wide Panel: An updated Illumina BeadChip array that is SNP compatible with the 610 Quad
array used in ADNI1 will be employed. The 610 Quad will be discontinued during the next five years. The
newest BeadChip is the HumanOmni1-Quad which has 1,140,419 SNP loci per sample and 93% genomic
coverage as well as other highly desirable content related to copy number variation and other important
features. Of note, 359,530 markers are shared with the 610 Quad and Illumina has developed software to
impute the remaining SNPs. The basic methods for the Human BeadChip series (Illumina, Inc. San Diego, CA)
will follow the manufacturer’s protocols (Infinium HD Assay; Super Protocol Guide; Rev. A, May 2008, or latest
version). Details of the array processing for ADNI1 samples were described in Preliminary Studies (6.3.1) and
processing for the updated array will be similar. ADNI1 assays were performed at TGen. For GO and ADNI2,
we propose to have the Illumina process the arrays. Alternatively, another highly experienced molecular
genetics laboratory (e.g., Broad Institute, CHOP, CIDR, TGen) could perform the array processing. This
decision will be carefully considered when samples are ready with input from the Core’s Genetics Advisory
Committee, NIA program, ADNI leadership and Industry partners.
6.4.1.6. Initial Quality Control: As above, all samples are bar coded and DNA “fingerprinted” to verify sample
identity. Prior to analysis, quantity and quality of DNA (or RNA in the future) will be determined to meet
specifications for the planned microarray analyses. These methods were developed by the ADNI Genetics
Working group and NCRAD during ADNI1. After the GWA array is completed, extensive QC protocols will be
completed prior to releasing data and for specific imaging genetics or other biomarker analyses as previously
discussed under Preliminary Studies (6.3).
6.4.1.7. Dissemination of GWA Data Sets: All genetic data sets produced by the Core will be rapidly made
publicly available, in keeping with ADNI policy. APOE genotyping will be processed and made available at
regular intervals, at least quarterly, depending on rates of enrollment. Over 5 years the Core expects to
process and disseminate genome-wide data for 550 new ADNI samples. BeadChip assays will be processed
annually as a batch service and ~140 arrays per year are projected and budgeted. Randomly selected
samples (5-10%) from each enrollment year will be re-genotyped at the conclusion of ADNI2 to ensure there
are no systematic changes in genotyping methods. Rigorous initial QC procedures can be completed within 23 months or less after microarrays are processed. All of the necessary mechanisms are now in place for
posting the data on the ADNI LONI website (http://www.loni.ucla.edu/ADNI). This process represents a
collaboration between the Genetics and Informatics Cores and all procedures were developed and tested
during ADNI1. Within a week after the completion of final QC, data for the complete set of ADNI1 participants
was released to the scientific community on 4/16/09. As of 10/4/09, there have been 42,270 individual
genotype array records downloaded by 94 different users from the ADNI LONI website. The genetics
investigators and staff, prior to establishment of the Core, have responded to dozens of requests for
information or assistance from investigators, as well as inquiries from industry and media. Taken together, the
quantitative download data and requests for information or assistance reflect a remarkable degree of interest in
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
these data in a six month time frame. Following the NIH GWAS policy, ADNI data will also be made available
through the NCBI Database of Genotypes and Phenotypes (dbGaP;
http://www.ncbi.nlm.nih.gov/sites/entrez?db=gap).
6.4.1.8. Dissemination of Biological Samples: Requests for DNA, RNA or other specimens collected for
genetics will be evaluated by the ADNI Resource Allocation Review Committee (RARC), as is the case for all
limited resources such as serum, plasma, urine, CSF or brain tissue. The immortalized cell lines ensure a
renewable supply of DNA sets for near term and future research. The RARC Policy and Procedures are posed
on the ADNI Information website (http://www.adni-info.org). The RARC is chaired by Dr. Tom Montine. Once a
request is approved, a Material Transfer Agreement (MTA) must be completed. Then NCRAD will ship the
samples to the recipient. Some charges will apply to help defray the cost to NCRAD of providing this service.
6.4.2. Analysis of the Association between Genetic Variation and the Rich Multidimensional Phenotypic
Data Collected on the ADNI Cohort (Aim II).
6.4.2.1. General Considerations for Analysis of ADNI Genotypes and Phenotypes: The Genetics Core will
work with the ADNI cores and other groups of investigators, industry partners and other collaborators to help
ensure that systematic analyses of genotype-phenotype associations are completed on the ADNI data set.
The Core will provide or collaboratively organize the bioinformatics, genetics and computational expertise
required for such analyses. Given the relatively limited resources at present, it is anticipated the Core will work
with other groups to generate competitive grant applications for detailed genetic and genomic analyses. A
priority for initial analyses will be GWAS of longitudinal imaging phenotypes, as well as other important and
highly relevant phenotypes including clinical data (e.g., conversion/progression status and psychometric
performance), and CSF, blood and urine biomarker data. The ADNI genetic data can be examined at various
levels of granularity (e.g., SNP, gene, regional locus, biological pathway, or genome-wide basis) depending on
the scientific questions and purpose. Phenotypic data can similarly be analyzed on a global or granular level
(e.g., overall cortical atrophy, structure-based analysis, or voxel by voxel analysis). Dimension reduction can
be applied to both the neuroimaging and other ADNI data, as well as to the genomic data, as needed. It is
anticipated that a broad range of investigators will use the ADNI GWA data in various ways. Analyses will be
hypothesis-guided or data-driven, discovery-oriented or confirmatory. All of these approaches are desirable at
this stage of investigation and each will require somewhat different analytic strategies. The Core will ensure
that certain key relationships are assessed but the scientific community clearly will interrogate the combined
GWA data from ADNI1, GO and ADNI2 for many years to come.
6.4.2.2. Quantitative Cross-Sectional and Longitudinal Phenotypes: Although diagnostic class will be
assessed, a major strength of ADNI for genetic analysis is the continuous quantitative phenotypic data that can
yield greater power and precision for detecting associations. A non-exclusive list of important cross-sectional
phenotypes at baseline includes measures discussed in other Cores in the proposal such as structural MRI
(volume, cortical thickness, GM density, etc., particularly of medial temporal and neocortical lobe structures;
see recent ADNI publications in MRI Core), PET (metabolic measures and amyloid burden), advanced MRI
measures on subsets of patients (resting state BOLD fMRI connectivity, DTI, ASL perfusion), CSF biomarkers
(amyloid beta, tau, and ptau), neuropsychological test performance (verbal memory and learning, naming and
fluency, overall cognition on ADAS and MMSE), and clinical/functional status (CDR Sum of Boxes, ADL
scores). Longitudinal phenotypes will include change scores and eventually rate of change, slope and
trajectory scores for the same measures acquired at baseline. The Genetics Core will collaborate with the
Biostatistics Core on appropriate longitudinal models of ADNI variables for GWA analyses. See Section 8.4.2
of the Biostatistics Core for descriptions of the general approach for characterizing trajectories and examining
the role of potential predictors and for comparing phenotypic measures for performance.
6.4.2.3. Dimensionality Reduction: The quantitative phenotypic measures listed above are clearly not
independent across modalities, such as cognition and MRI, or even within modality, e.g. across structural MRI
variables. Multivariate dimensionality reduction using standard techniques such as principal components
analysis (PCA), factor analysis (FA) and cluster analysis as well as more advanced methods such as
independent components analysis (ICA), data mining and pattern recognition techniques will likely be able to
assist in optimizing analysis of genotype/phenotype relationships. Tools are available for all of these methods
that can be adapted to the ADNI data as needed.
6.4.2.4. Overall Hypotheses and Strategies for Genetics Core:
(1) Candidate Genes: Genetic variation in leading candidate genes (see AlzGene meta-analytic database;
http://www.alzgene.org/), beyond APOE epsilon 4 status, contributes to the prediction of (a) risk of developing
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
AD (or MCI) and (b) rate of progression. The inclusion of the candidate genetic variation in models assessing
risk for and development of AD using the clinical, neuroimaging and CSF/plasma biomarkers collected by
ADNI will lead to more accurate prediction of disease. As of 10/06/09, the top 15 candidate genes are APOE
(ε2/3/4), CLU, PICALM, TNK1, ACE, TFAM, CST3, IL1B, CR1, hCG2039140, SORL1, CHRNB2, SORCS1,
DAPK1 and PRNP. Many other important genes are in the top 30. Based on these existing results, a candidate
gene approach will be taken for the ADNI data to cross-validate, confirm, and refine the prior findings (e.g.,
identifying sensitive quantitative traits or new biomarkers associated with these genes).
(2) Targeted Pathways: Targeted pathway analysis will help to identify specific relationships while providing
some protection from the extremely stringent statistical thresholds required for genome-wide significance
(discussed below). Examples: a) Amyloid burden on PIB or AV-45 PET (or amyloid beta CSF markers) is
associated with allelic variation in an ensemble of genes known to be highly involved with amyloid processing
including clearance, deposition and toxicity. b) Neurodegenerative changes, such as rate of hippocampal and
other volume loss on structural MRI or decreased synaptic integrity indicated by FDG PET, will be associated
with variation in a group of selected genes implicated in apoptotic, oxidative stress, neurotoxicity and
inflammatory mechanisms. c) Degree of impairment and rate of decline in verbal memory and learning is
associated with variation in a groups of genes implicated in long term potentiation, cognitive and language
development, as well as neurogenesis, including BDNF and other growth factors and adhesion molecules.
These are just a few selected examples of hypothesis-guided pathway analyses that will be performed. The
scientific community within and beyond ADNI can be expected to generate many additional hypotheses that
will evaluate relationships extending beyond single candidate genes or alleles that are more targeted in scope
than genome-wide statistical testing.
(3) GWAS and CNV Analyses: GWAS employing important continuous phenotypes (e.g., hippocampal
volume and GM density or CSF biomarkers), and appropriate statistical models controlling for APOE, will result
in identification of novel genes related to susceptibility to AD (and MCI) and rate of progression. Specific details
regarding GWAS techniques are discussed below (Section 6.4.2.5). CNV features related to diagnosis and
other phenotypes especially rate of conversion/progression will be important to detect in ADNI. This is an
emerging and challenging area with new tools being developed frequently. Illumina’s GenomeStudio has a
plug-in for CNV analysis (cnvPartition 2.3.4) which will be used for initial steps. Winchester et al. [99] in a paper
last month compared several software packages for further CNV analysis and emphasized the use of
association methods for CNV analysis. In that regard, Barnes et al. [100] have developed an R package,
CNVtools, which facilitates case/control association analyses of CNV data. PennCNV [101] is another useful
tool which implements a hidden Markov model for high-resolution CNV detection in GWA SNP data. We plan
for a close integration between GWAS and CNV analyses similar to those in previous studies [100, 102-104].
As previously discussed (section 6.2.5), CNVs represent potentially important target genetic variations
associated with AD. We plan to carefully assess the most effective tools to fully elucidate the relationship
between CNVs (rare, common, total burden) and target imaging, biomarker, and clinical phenotypes.
(4) Preparation for Future Experiments and Analyses: In the future, when samples are acquired and funding
is obtained for longitudinal RNA expression analyses, microRNA analysis, and targeted deep resequencing of
candidate regions, whether by ADNI or other teams of investigators, these methods are expected to yield
important new clues to fundamental mechanisms and pathways involved in early prodromal changes leading to
AD. At present, we only request resources to collect and bank samples for these future analyses each of which
represent a rapidly evolving technical area.
6.4.2.5. Methods for Genome-Wide Association Studies (GWAS)
(1) Genotype Calling: Illumina BeadStudio software is used to generate the genotyping call for each SNP or
CNV marker. The latest version appropriate for the specific array in ADNI2 will be employed.
(2) Software for GWAS: There has been an explosion of research, methods development and publications on
GWAS methodology and guidelines over the past few years (See Background and Significance for references).
A number of software packages have been released and in some cases extensively assessed and employed in
GWAS reports. We have adopted the widely employed PLINK package [32] (v1.06 at present), developed and
publicly released by Shaun Purcell and colleagues of MGH and the Broad Institute
(http://pngu.mgh.harvard.edu/purcell/plink/), as the main platform for GWAS modeling.
(3) Quality Control Procedures: There are a number of potential basic QC problems that can occur in
genotyping studies prior to actually performing any association analyses. These relate to plating errors, DNA
degradation, hybridization problems with the array, as well as errors made by calling algorithms [105].
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Missingness, heterozygosity and extremely low allele frequencies are important indicators of genotyping quality
and can be used as filters for samples and SNPs [106]. Another important SNP QC measure is detection of
departures from Hardy-Weinberg equilibrium (HWE) [107]. Inadvertent relatedness of samples can be readily
detected by shared genetic markers (we detected several sibling pairs in the ADNI1 data and randomly
excluded one sibling from each pair). Identity by descent and inbreeding coefficient calculations can identify
problem samples due to cryptic relatedness. After filtering out problem SNPs and samples, >99% SNP call
rates and <1% missing data indicate a high quality data set. PLINK and other packages easily perform all of
these steps and have been previously employed in our Preliminary Studies reported above (section 6.3).
(4) Minor Allele Frequency (MAF): Typically SNPs with MAF <0.05 will be excluded so that cell sizes are
sufficient for analysis. For some analyses, we may restrict to highly common variants with MAF’s greater than
10 or 15%.
(5) Population Stratification: Case-control GWAS can be confounded by undetected or uncontrolled
population structure [108]. Fortunately, well established procedures to detect and adjust for population
structure [109-112] are available, including genomic control (estimated inflation of test statistic). Structure [113]
and Eigenstrat [114] (Eigensoft 3.0; http://genepath.med.harvard.edu/~reich/Software.htm) are two widelyused software packages that correct for population stratification in GWAS. Eigenstrat employs PCA to model
ancestral differences, minimizing spurious associations and maximizing power to detect true associations.
When required, we will perform stratified analyses based on subpopulations identified by Structure or
Eigenstrat or by using all SNPs and including the dimensions from an eigen analysis as covariates in GLM
models. We have previously employed these approaches successfully with ADNI1 data [10, 22].
(6) Imputation: Estimation of the genotype for SNPs where the actual SNP was not genotyped on an array is
often desired, particularly to assess hypothesis-driven associations between target SNPs and outcome
variables [115]. Also, using densely imputed genotypes in a GWAS often has a slightly increased power
relative to assayed SNPs and is likely to bring detection closer to causal variants if present [116]. Another
application of imputation is to enable a combined or replicated analysis of the ADNI data and other similar data
sets genotyped on different platforms. Multiple methods and software packages have been developed for
imputation and a recent study compared four publicly available imputation programs (BEAGLE, IMPUTE,
MACH, and PLINK) [117]. MACH had the best overall performance [117] so we used MACH to impute the
complete ADNI1 dataset using the Hapmap CEU data in the analyses discussed in the Preliminary Studies
section (section 6.3). For GO and ADNI2 data, imputation will be performed as needed for analyses.
(7) Statistical Modeling: As discussed above, we will employ continuous quantitative phenotypes wherever
possible to increase power [118]. We will typically build models with categorical main effects of SNP, diagnosis
and APOE genotype. Age,
Figure 6.7: Greater Power of Quantitative Trait (QT) education and gender are
potential covariates, as are
Compared to Case/Control Analysis
baseline intracranial volume
(where appropriate) and the
principal components derived
from Eigenstrat for correction of
population
stratification.
Interactions of diagnosis x
SNP, SNP x APOE, diagnosis
x SNP x APOE will all be
examined. Diagnosis by SNP
interactions, controlling for
APOE, are of the greatest
interest
and
priority
[9].
Maternal and paternal family
history of dementia is also a
Graph shows the power distribution curves for QT analysis contrasted with a case/control design at
potential variable of interest.
p<.01 and 10^-7 (OR1.5). The x axis portrays the sample sizes and the y axis the power at each value
of the sample size for a 10% percentage variance explained for the QT, a QT MAF of 10% and a marker
Furthermore,
additive,
SNP MAF at 20%. The results for QT and case-control are displayed in black and grey respectively.
dominant,
and
recessive
Compared to QT, the case-control curves are shifted to the right indicating that much larger sample
sizes are required to reach the same power. From Potkin et al (2009) [118].
models of genetic contribution
to disease risk will all be
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
systematically evaluated.
(8) Gene-Gene Interactions (Epistasis): Dr. Jason Moore (Dartmouth), collaborator on the Core, is a
computational geneticist with special expertise in development of algorithms and software for modeling
epistatic interactions [119-125]. Dr. Moore will guide the quantitative examination of gene-gene interactions
and the use of the packages he has developed and disseminated (e.g., MDR, SyMOD, EVA; see
www.epistasis.org).
(9) Genome-Wide Significance and Multiple Comparisons: Setting the threshold for determination of whether
an obtained GWAS result is statistically significant has been somewhat controversial. Most investigators
employ a stringent threshold of 10-7 or 10-8 following guidelines from the Wellcome Trust Case Control
Consortium. Essentially a significance threshold of 10-7 is the equivalent of a Bonferroni adjustment, i.e. p =
0.05 / 500,000 SNPs, which is overly conservative given the extensive linkage disequilibrium or nonindependence of nearby genomic regions. A number of solutions have been proposed to determine the
effective number of independent statistical tests performed in a GWAS. This is an active area of research. We
plan to work closely with the Biostatistics Core, monitoring and testing the state-of-the-art methods and tools
available in the area. One promising approach is SLIDE (http://slide.cs.ucla.edu), an accurate and efficient
method, recently proposed by Han et al [126] to address the multiple testing correction in GWAS. This method
uses a sliding window to account for all correlation within a moving windowed genomic region. In simulations
using the WTCCC data [106], the error rate of SLIDE's corrected p-values was >20 times smaller than previous
methods, while SLIDE is orders of magnitude faster than the permutation test and other competing methods.
However, the SLIDE technique is primarily designed to investigate the number of effective multiple
comparisons between numerous genomic factors and limited target phenotypes. At present there is no
comprehensive theory to correct for multiple comparisons for GWAS of numerous inter-correlated outcome
variables (i.e. multi-modality neuroimaging data and other biomarkers). The Core and collaborators will
investigate the use of permutation and simulation techniques for multiple comparison corrections in both the
genotypic and phenotypic domains, while working on the problem of multi-dimensional multiple comparisons
across centers and particularly with the ADNI Biostatistics Core.
(10) Power Considerations: There is a substantial literature on power in GWAS designs with most authors
advocating for extensive sample sizes. However, most of this literature is directed toward a case/control rather
than Quantitative Trait (QT) or continuous phenotype approach which greatly enhances power. For example,
Potkin et al. [118] estimated power under different effect sizes (10-30%) and variations in allelic frequency for a
QT (10-20%), and compared them to a case-control design with an Odds Ratio (OR) of 1.5. The results shown
in Fig. 6.7. demonstrate the superiority of QT to a classic case/control design [118]. QT analysis, as proposed
here, is an excellent strategy for
identification of unanticipated genes that
modulate structural, functional or molecular
processes in those at risk for AD. For
estimation of power, we plan to use SLIP
(http://slide.cs.ucla.edu), a package like
SLIDE, which was developed by the same
team [126] and employs a sliding window
algorithm
to
estimate
power
after
accounting for linkage disequilibrium. We
are presently adapting SLIP and SLIDE for
imaging applications. For initial power
analyses for GWAS QT designs we
employed
Quanto
(http://hydra.usc.edu/gxe/) [127] (see Fig.
6.8.), Genetic Power Calculator and related Fig. 6.8. Power to detect QT association with a SNP was calculated using Quanto
1.2) assuming an expected final GWAS sample of ~1,500 individuals and
standard methods [128-131]. Similar to the (version
-7
a significance threshold of 1x10 , for a range of genetic effects (proportion of the
multiple
comparisons
considerations phenotypic variability assuming a linear model with standard normal residuals).
80% power to detect association when at least 2.2% of the
discussed above, there is also no Results indicated
2
variability (R ) is accounted for by the genetic effect. The required sample for a
comprehensive theory for power for a QT continuous phenotype analysis is about 10% of that needed for a case/control
GWAS of many inter-correlated outcome design.
QTs, along the scope proposed here. Here
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
again, we will use permutation and simulation techniques to investigate issues of power in these GWAS
analyses, while actively working on this issue with other centers, including the Biostatistics Core. Preliminary
results of GWAS for key MRI variables in ADNI1 indicated R2 for top associations in the 4-11% range.
Therefore, the combined ADNI1, GO and ADNI2 samples (total N > 1550) should be well-powered to detect
moderate to strong genetic effects based on the power estimation shown in Fig. 6.8.
(11) Reporting of GWAS Results: New standards for transparent and complete reporting of GWAS results
have recently been published [132]. We plan to follow these guidelines when reporting results, whenever
possible, and to encourage other groups using the ADNI data to adhere to these standards.
6.4.3. Serve as a Central Resource, Point of Contact and Planning Group for Genetics in ADNI (Aim III)
6.4.3.1. Resources and Coordination: The Genetics Core will provide an organizational and informational
resource for investigators, industry partners and other parties interested in analyses of the ADNI genetics data.
This will be accomplished through hosting regular conference calls (biweekly or as needed) for working groups
and interested parties and establishing an ADNI Genetics Wiki or related technology for dissemination and
coordination. Below are selected examples of important areas for effective coordination:
(1) Interaction with Other ADNI Cores: The Genetics Core will interact extensively with all other cores. For
example, collaboration with the Clinical Core will include logistics for sample collection and banking and joint
work to select optimal phenotypes for GWAS and targeted analyses. There are also interesting and important
bioethical considerations bridging the genetics and clinical areas. Core collaborator, Dr. Robert Green
(neurologist, ADC PI and medical geneticist), will serve as the Genetics Core point person with regard to these
issues. He has done ground breaking research into the psychosocial issues related to genetic testing (APOE)
and effects of disclosure [133]. Although no disclosure is planned in ADNI, we wish to be proactive in
considering the wide range of implications of genetic testing as the field rapidly evolves. It seems likely that
personalized neurotherapeutic strategies in the coming years will be based at least in part on genetic profiling
of well validated markers. With the MRI and PET Cores, the focus will be on imaging phenotype definition and
variable selection and collaboration on imaging genetics analyses and manuscripts. Plans are underway with
the Biomarker Core for collaborative analyses of genetic associations with CSF and plasma analytes. In the
future, there is great potential for collaboration with Neuropathology Core for combined GWAS and expression
studies in post-mortem tissue as it becomes available. Although the antemortem-postmortem sample is likely
to be small, the scientific yield could be very important relating genotypes to expression. The Genetics Core
and Biostatistical Core have also begun to examine areas for joint work including optimized modeling for
genotype-phenotype associations, relevant covariates, and power and multiple comparison issues. With the
Informatics Core, continued collaboration on data organization, annotation and dissemination are planned.
(2) AD Genetics Consortium: The Core will liaison with the recently NIA-funded AD Genetics Consortium
(U01 AG032984-01, PI: Gerald Schellenberg) to facilitate the ADGC’s use of ADNI data and in turn, ADNI
researchers having access to large scale genetics databases from other consortia and centers. Drs. Saykin,
Foroud, and Farrer are all co-PI’s and members of the Executive Committee of the ADGC, and participate in
biweekly ADGC conference calls, ensuring close coordination of the complementary genetics research
activities between ADNI and ADGC. The main focus of ADGC is large case/control and autopsy studies.
Although it is hoped that imaging and other biomarkers will play a larger role over time in the ADGC, ADNI is
the currently the primary data source for imaging genetics studies. Nonetheless, close coordination will clearly
benefit both the ADNI and ADGC and foster additional research.
(3) World Wide ADNI: As multiple ADNI-affiliated consortia develop globally, the potential for joint genetics
studies is great. Expanded sample sizes and determination of differences in GWAS results as a function of
different ancestral populations and admixtures will be highly important. Furthermore, GWAS analyses and
imaging genetics studies of international ADNI cohorts may provide important replication of newly uncovered
loci and target SNPs. The Genetics Core has begun discussions with affiliated international AD imaging
projects, all of whom plan to collect DNA, and collaborative imaging genetics studies will be a priority.
(4) AddNeuroMed: The Genetics Core has been collaborating with AddNeuroMed, a six site European
consortium (http://www.innomed-addneuromed.com/) that is collecting MRI and GWAS data harmonized with
ADNI methods. Approximately 300 sets of MPRAGE MRI scans and genotyping using the same Illumina array
as ADNI have now been processed by the Imaging Genomics Lab at the IUSM using the same analysis
pipeline as for ADNI data [98]. Regular contact and working exchanges have been taking place with Drs.
Simon Lovestone (PI), Andrew Simmons (Imaging leader) and Simon Furney (Bioinformatics leader) and the
ADNI Genetic Core team and replication and joint follow-up analyses are being planned.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
(5) Multi-Institutional Research in Alzheimer's Genetic Epidemiology (MIRAGE): Funded by the NIA since
1990, MIRAGE is directed by Genetics Core collaborators Drs. Lindsay Farrer (PI) and Robert Green (co-PI)
and includes MRI and GWAS data (http://www.bu.edu/alzresearch/research/genetics/mirage/index.html) [134,
135]. Core plans include processing MIRAGE scan and GWA data using the same processing protocols as
used for ADNI data (e.g., [98]), permitting replication of ADNI findings as well as new combined analyses.
(6) Other Replication/Extension Studies for Imaging Genetics: In addition to World-wide ADNI, AddNeuroMed
and MIRAGE, and ADGC sites with imaging data, the ADNI Genetics Core will actively pursue additional
replication and extension opportunities so that over time there will be ample statistical power and the ability to
replicate potentially important findings in multiple independent data sets. The Core is actively working on the
infrastructure needed to facilitate this process. A valuable resource to quickly and effectively investigate
additional large scale imaging genetic cohorts is priority access to Indiana University’s current supercomputing
grid and the new FutureGrid to open within a year.
6.4.3.2. Future Directions and Planning: As mentioned earlier, the Genetics Core will continue to provide a
framework to bring together an extended advisory group of experts to identify important and promising future
directions for genetics research on AD and MCI with emphasis on imaging phenotypes and biomarkers. As
ADNI and the field progresses, careful consideration as to the next steps and how to gain support for additional
efforts will be important. The Core will bring ADNI, ADGC, industry and other external experts together
regularly to consider future directions and potential proposals. Major themes will include formulating potential
post genomic screening analyses, functional annotation of SNPs/genes, further analysis of CNV regions, and
the identification of functional pathways to place emerging results within a systems biology perspective in order
to guide further assays and analyses. One examples of a potentially important approach under discussion in
the ADNI and AD genetics community is targeted deep resequencing to identify rare variants in DNA sequence
related to AD susceptibility. For ADNI, particularly interesting targets would be deep resequencing of genomic
regions associated with key neuroimaging/biomarker phenotypes. These target regions would be identified
based on the emerging imaging GWA results from ADNI, as well as future large-scale case/control studies
pointing to new candidate susceptibility genes. The Core will also explore potential studies of longitudinal RNA
expression from peripheral blood derived samples and eventually post-mortem samples as they become
available. MicroRNA analysis is another promising area that could be examined in ADNI samples collected
and banked by the core. Additional analyses such as epigenetic assays and modeling of gene-environment
interactions will also be carefully considered. Funds are not requested for these specialized studies at this
time. The Genetics Core will coordinate with experts in each area to assess the potential scientific
yield and feasibility for proposed follow-up studies which will also be peer reviewed when funding is
sought.
6.4.4. Potential Limitations, Review Considerations and Proposed Solutions: Although not included in
ADNI1, the Genetics Core was included in the GO proposal. Reviewers of the GO proposal, while very
favorable about the overall application and plans to collect additional genetics data for a larger GWAS,
expressed concern about the genetic component in terms of power and sample size issues and the range of
expertise needed for the techniques proposed. We wish to briefly address some of these concerns. Unlike the
present ADNI2 application, there was very minimal space in the GO application (15 pages total) leaving < 1
page for the Genetics Core including specific aims, rationale and methods. The Genetics Core was proposed
to support new sample collection and genotyping for the additional early MCI participants and likely future
directions for genetic studies were outlined. In fact, no funds were actually requested for most of these
anticipatory future plans (e.g. for targeted deep resequencing of regions expected to emerge from the ADNI1
GWAS). Most of the budget was for sample banking and performing the genotyping assays on the additional
samples to be recruited, along with some support requested for analysis of existing GWAS data that became
available in April 2009. There was also no space to describe the strong multidisciplinary team that had been
assembled. Here we briefly address these concerns.
Power and Sample Size. These fundamental issues are discussed at length in section 6.4.2.5.(10). Briefly,
the combined ADNI1/GO/ADNI2 data set would include over 1550 participants and longitudinal scans, fluid
biomarkers, and psychometrics in which subjects in part serve as their own control. The focus of the genetics
analyses will be on quantitative traits/continuous phenotypes, which have much greater power than case/
control designs (see Fig. 6.7.). We provide power calculations based on new ADNI1 data (see Fig. 6.8.)
indicating adequate power (.80 or better) to detect moderate to large genetic influences on baseline values and
rate of change measures. The early results summarized in Preliminary Studies further support our ability to
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
detect meaningful genetic influences on imaging, CSF and other phenotypic variables. In addition, we have
already obtained and begun to analyze 300 samples from AddNeuroMed using the same MRI and genetics
methods as ADNI with 100 or more additional samples expected in the next year and the likelihood of more
samples during ADNI2 (see attached letter from Dr. Simon Lovestone). Furthermore, we are in the process of
obtaining over 1000 volumetric MRI scans and GWAS data from the MIRAGE study. Dr. Lindsay Farrer, core
collaborator, reported that his team has just completed a preliminary GWAS of nearly 1000 Caucasian and 150
African MIRAGE subjects with MRI data and he expects to increase that sample by 200 after genotyping is
completed (see attached letter). In addition, Dr. Farrer has GWAS and MRI trait data on ~2600 Framingham
Study subjects, most (~ 2300) of whom are cognitively normal, and we are discussing combined and
replication analyses with ADNI data. Finally, we are actively pursuing other studies with GWAS and imaging
data sets. At a minimum we expect to have GWAS and MRI data processed within the Core on several
thousand subjects. This will be the largest imaging genetics GWAS ever performed in MCI/AD or other
disorders. Given the power of the QT approach there is a strong likelihood of novel discovery and replication.
Expertise. This is a highly transdisciplinary core with investigators, collaborators and consultants bringing
together expertise from the fields of imaging genetics (Potkin, Saykin, Shen, Thompson), molecular genetics
and biorepository operation (Foroud), computational genetics (Bertram, Farrer, Foroud, Moore), genetics
databases and software development (Bertram, Moore), clinical genetics (Green), computer science (Shen,
Thompson), and the neuropsychology (Saykin), psychiatry (Potkin) and neurology (Green) of MCI and AD. In
addition, the ADNI Genetics Working Group (see section 6.3.1) and GWAS team (TGen, IU, UCI et al) and the
ADGC (PI: G. Schellenberg), of which ADNI is a member, include many renowned experts in each of these
areas and beyond. All of this expertise is available to ADNI. The specific laboratories performing the molecular
assays and the bioinformatics will be determined in consultation with advisors, ADNI leadership, NIA program
officials and ADNI’s industry partners. All data will be rapidly made publicly available regardless of lab sites.
6.5. References:
1.
Panizzon, M.S., C. Fennema-Notestine, L.T. Eyler, T.L. Jernigan, E. Prom-Wormley, M. Neale, K.
Jacobson, M.J. Lyons, M.D. Grant, C.E. Franz, H. Xian, M. Tsuang, B. Fischl, L. Seidman, A. Dale, and
W.S. Kremen, Distinct Genetic Influences on Cortical Surface Area and Cortical Thickness. Cereb
Cortex, 2009.
2.
Chou, Y.Y., N. Lepore, M.C. Chiang, C. Avedissian, M. Barysheva, K.L. McMahon, G.I. de Zubicaray,
M. Meredith, M.J. Wright, A.W. Toga, and P.M. Thompson, Mapping genetic influences on ventricular
structure in twins. Neuroimage, 2009. 44(4): p. 1312-23.
3.
Peper, J.S., R.M. Brouwer, D.I. Boomsma, R.S. Kahn, and H.E. Hulshoff Pol, Genetic influences on
human brain structure: a review of brain imaging studies in twins. Hum Brain Mapp, 2007. 28(6): p.
464-73.
4.
Liu, J., G. Pearlson, A. Windemuth, G. Ruano, N.I. Perrone-Bizzozero, and V. Calhoun, Combining
fMRI and SNP data to investigate connections between brain function and genetics using parallel ICA.
Hum Brain Mapp, 2009. 30(1): p. 241-55.
5.
Koten, J.W., Jr., G. Wood, P. Hagoort, R. Goebel, P. Propping, K. Willmes, and D.I. Boomsma, Genetic
contribution to variation in cognitive function: an FMRI study in twins. Science, 2009. 323(5922): p.
1737-40.
6.
Wishart, H.A., A.J. Saykin, L.A. Rabin, R.B. Santulli, L.A. Flashman, S.J. Guerin, A.C. Mamourian, D.R.
Belloni, C.H. Rhodes, and T.W. McAllister, Increased brain activation during working memory in
cognitively intact adults with the APOE epsilon4 allele. Am J Psychiatry, 2006. 163(9): p. 1603-10.
7.
Bertram, L., M.B. McQueen, K. Mullin, D. Blacker, and R.E. Tanzi, Systematic meta-analyses of
Alzheimer disease genetic association studies: the AlzGene database. Nat Genet, 2007. 39(1): p. 1723.
8.
Bertram, L. and R.E. Tanzi, Thirty years of Alzheimer's disease genetics: the implications of systematic
meta-analyses. Nat Rev Neurosci, 2008. 9(10): p. 768-78.
9.
Saykin, A.J., L. Shen, S.L. Risacher, S. Kim, K. Nho, J.D. West, T.M. Foroud, and and the Alzheimer's
Disease Neuroimaging Initiative. Genetic predictors of 12 month change in MRI hippocampal volume in
the Alzheimer’s Disease Neuroimaging Initiative cohort: Analysis of leading candidates from the
AlzGene database. in Imaging Consortium, ICAD 2009, Vienna, Austria. 2009.
10.
Shen, L., S. Kim, S.L. Risacher, K. Nho, S. Swaminathan, J.D. West, T. Foroud, N. Pankratz, M.J.
Huentelman, D.W. Craig, B.M. DeChairo, S.G. Potkin, C. Jack, M. Weiner, A.J. Saykin, and ADNI,
PHS 398/2590 (Rev. 11/07)
Page
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
Weiner, Michael
Whole Genome Association Study of Brain-Wide Imaging Phenotypes for Identifying Quantitative Trait
Loci in MCI and AD: A Study of the ADNI Cohort submitted 2009.
Stein, J.L., X. Hua, J.H. Morra, S. Lee, A.J. Ho, A.D. Leow, A. Toga, J. Sul, H.M. Kang, E. Eskin, A.J.
Saykin, L. Shen, T. Foroud, N. Pankratz, M.J. Huentelman, D.W. Craig, J.D. Gerber, A. Allen, J.
Corneveaux, D.A. Stephan, J. Webster, B.M. DeChairo, S.G. Potkin, C. Jack, M. Weiner, and P.
Thompson, Genome-Wide Analysis Reveals Novel Genes Influencing Temporal Lobe Structure with
Relevance to Neurodegeneration in Alzheimer's Disease. submitted 2009.
Gatz, M., C.A. Reynolds, L. Fratiglioni, B. Johansson, J.A. Mortimer, S. Berg, A. Fiske, and N.L.
Pedersen, Role of genes and environments for explaining Alzheimer disease. Arch Gen Psychiatry,
2006. 63(2): p. 168-74.
Bertram, L. and M. Alireza, Chapter 9 Alzheimer's Disease Genetics: Current Status and Future
Perspectives, in International Review of Neurobiology. 2009, Academic Press. p. 167-184.
Kim, J., J.M. Basak, and D.M. Holtzman, The role of apolipoprotein E in Alzheimer's disease. Neuron,
2009. 63(3): p. 287-303.
Barabash, A., A. Marcos, I. Ancin, B. Vazquez-Alvarez, C. de Ugarte, P. Gil, C. Fernandez, M. Encinas,
J.J. Lopez-Ibor, and J.A. Cabranes, APOE, ACT and CHRNA7 genes in the conversion from amnestic
mild cognitive impairment to Alzheimer's disease. Neurobiol Aging, 2009. 30(8): p. 1254-64.
Abraham, R., V. Moskvina, R. Sims, P. Hollingworth, A. Morgan, L. Georgieva, K. Dowzell, S. Cichon,
A.M. Hillmer, M.C. O'Donovan, J. Williams, M.J. Owen, and G. Kirov, A genome-wide association study
for late-onset Alzheimer's disease using DNA pooling. BMC Med Genomics, 2008. 1: p. 44.
Beecham, G.W., E.R. Martin, Y.J. Li, M.A. Slifer, J.R. Gilbert, J.L. Haines, and M.A. Pericak-Vance,
Genome-wide association study implicates a chromosome 12 risk locus for late-onset Alzheimer
disease. Am J Hum Genet, 2009. 84(1): p. 35-43.
Bertram, L., C. Lange, K. Mullin, M. Parkinson, M. Hsiao, M.F. Hogan, B.M. Schjeide, B. Hooli, J. Divito,
I. Ionita, H. Jiang, N. Laird, T. Moscarillo, K.L. Ohlsen, K. Elliott, X. Wang, D. Hu-Lince, M. Ryder, A.
Murphy, S.L. Wagner, D. Blacker, K.D. Becker, and R.E. Tanzi, Genome-wide association analysis
reveals putative Alzheimer's disease susceptibility loci in addition to APOE. Am J Hum Genet, 2008.
83(5): p. 623-32.
Carrasquillo, M.M., F. Zou, V.S. Pankratz, S.L. Wilcox, L. Ma, L.P. Walker, S.G. Younkin, C.S. Younkin,
L.H. Younkin, G.D. Bisceglio, N. Ertekin-Taner, J.E. Crook, D.W. Dickson, R.C. Petersen, and N.R.
Graff-Radford, Genetic variation in PCDH11X is associated with susceptibility to late-onset Alzheimer's
disease. Nat Genet, 2009. 41(2): p. 192-8.
Grupe, A., R. Abraham, Y. Li, C. Rowland, P. Hollingworth, A. Morgan, L. Jehu, R. Segurado, D. Stone,
E. Schadt, M. Karnoub, P. Nowotny, K. Tacey, J. Catanese, J. Sninsky, C. Brayne, D. Rubinsztein, M.
Gill, B. Lawlor, S. Lovestone, P. Holmans, M. O'Donovan, J.C. Morris, L. Thal, A. Goate, M.J. Owen,
and J. Williams, Evidence for novel susceptibility genes for late-onset Alzheimer's disease from a
genome-wide association study of putative functional variants. Hum Mol Genet, 2007. 16(8): p. 865-73.
Li, H., S. Wetten, L. Li, P.L. St Jean, R. Upmanyu, L. Surh, D. Hosford, M.R. Barnes, J.D. Briley, M.
Borrie, N. Coletta, R. Delisle, D. Dhalla, M.G. Ehm, H.H. Feldman, L. Fornazzari, S. Gauthier, N.
Goodgame, D. Guzman, S. Hammond, P. Hollingworth, G.Y. Hsiung, J. Johnson, D.D. Kelly, R. Keren,
A. Kertesz, K.S. King, S. Lovestone, I. Loy-English, P.M. Matthews, M.J. Owen, M. Plumpton, W.
Pryse-Phillips, R.K. Prinjha, J.C. Richardson, A. Saunders, A.J. Slater, P.H. St George-Hyslop, S.W.
Stinnett, J.E. Swartz, R.L. Taylor, J. Wherrett, J. Williams, D.P. Yarnall, R.A. Gibson, M.C. Irizarry, L.T.
Middleton, and A.D. Roses, Candidate single-nucleotide polymorphisms from a genomewide
association study of Alzheimer disease. Arch Neurol, 2008. 65(1): p. 45-53.
Potkin, S.G., G. Guffanti, A. Lakatos, J.A. Turner, F. Kruggel, J.H. Fallon, A.J. Saykin, A. Orro, S.
Lupoli, E. Salvi, M. Weiner, and F. Macciardi, Hippocampal atrophy as a quantitative trait in a genomewide association study identifying novel susceptibility genes for Alzheimer's disease. PLoS One, 2009.
4(8): p. e6501.
Harold, D., R. Abraham, P. Hollingworth, R. Sims, A. Gerrish, M.L. Hamshere, J.S. Pahwa, V.
Moskvina, K. Dowzell, A. Williams, N. Jones, C. Thomas, A. Stretton, A.R. Morgan, S. Lovestone, J.
Powell, P. Proitsi, M.K. Lupton, C. Brayne, D.C. Rubinsztein, M. Gill, B. Lawlor, A. Lynch, K. Morgan,
K.S. Brown, P.A. Passmore, D. Craig, B. McGuinness, S. Todd, C. Holmes, D. Mann, A.D. Smith, S.
Love, P.G. Kehoe, J. Hardy, S. Mead, N. Fox, M. Rossor, J. Collinge, W. Maier, F. Jessen, B.
PHS 398/2590 (Rev. 11/07)
Page
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
Weiner, Michael
Schurmann, H. van den Bussche, I. Heuser, J. Kornhuber, J. Wiltfang, M. Dichgans, L. Frolich, H.
Hampel, M. Hull, D. Rujescu, A.M. Goate, J.S. Kauwe, C. Cruchaga, P. Nowotny, J.C. Morris, K. Mayo,
K. Sleegers, K. Bettens, S. Engelborghs, P.P. De Deyn, C. Van Broeckhoven, G. Livingston, N.J. Bass,
H. Gurling, A. McQuillin, R. Gwilliam, P. Deloukas, A. Al-Chalabi, C.E. Shaw, M. Tsolaki, A.B.
Singleton, R. Guerreiro, T.W. Muhleisen, M.M. Nothen, S. Moebus, K.H. Jockel, N. Klopp, H.E.
Wichmann, M.M. Carrasquillo, V.S. Pankratz, S.G. Younkin, P.A. Holmans, M. O'Donovan, M.J. Owen,
and J. Williams, Genome-wide association study identifies variants at CLU and PICALM associated
with Alzheimer's disease. Nat Genet, 2009.
Lambert, J.C., S. Heath, G. Even, D. Campion, K. Sleegers, M. Hiltunen, O. Combarros, D. Zelenika,
M.J. Bullido, B. Tavernier, L. Letenneur, K. Bettens, C. Berr, F. Pasquier, N. Fievet, P. BarbergerGateau, S. Engelborghs, P. De Deyn, I. Mateo, A. Franck, S. Helisalmi, E. Porcellini, O. Hanon, M.M.
de Pancorbo, C. Lendon, C. Dufouil, C. Jaillard, T. Leveillard, V. Alvarez, P. Bosco, M. Mancuso, F.
Panza, B. Nacmias, P. Bossu, P. Piccardi, G. Annoni, D. Seripa, D. Galimberti, D. Hannequin, F.
Licastro, H. Soininen, K. Ritchie, H. Blanche, J.F. Dartigues, C. Tzourio, I. Gut, C. Van Broeckhoven, A.
Alperovitch, M. Lathrop, and P. Amouyel, Genome-wide association study identifies variants at CLU
and CR1 associated with Alzheimer's disease. Nat Genet, 2009.
Glahn, D.C., T. Paus, and P.M. Thompson, Imaging Genomics: Mapping the Influence of Genetics on
Brain Structure and Function. Human Brain Mapping, 2007. 28: p. 461-463.
Meyer-Lindenberg, A. and D.R. Weinberger, Intermediate phenotypes and genetic mechanisms of
psychiatric disorders. Nat Rev Neurosci, 2006. 7(10): p. 818-27.
Bearden, C.E., T.G. van Erp, R.A. Dutton, H. Tran, L. Zimmermann, D. Sun, J.A. Geaga, T.J. Simon,
D.C. Glahn, T.D. Cannon, B.S. Emanuel, A.W. Toga, and P.M. Thompson, Mapping cortical thickness
in children with 22q11.2 deletions. Cereb Cortex, 2007. 17(8): p. 1889-98.
Cannon, T.D., P.M. Thompson, T.G. van Erp, M. Huttunen, J. Lonnqvist, J. Kaprio, and A.W. Toga,
Mapping heritability and molecular genetic associations with cortical features using probabilistic brain
atlases: methods and applications to schizophrenia. Neuroinformatics, 2006. 4(1): p. 5-19.
Potkin, S.G., J.A. Turner, G. Guffanti, A. Lakatos, F. Torri, D.B. Keator, and F. Macciardi, Genome-wide
strategies for discovering genetic influences on cognition and cognitive disorders: methodological
considerations. Cogn Neuropsychiatry, 2009. 14(4): p. 391-418.
Lind, J., A. Larsson, J. Persson, M. Ingvar, L.G. Nilsson, L. Bäckman, R. Adolfsson, M. Cruts, K.
Sleegers, C. Van Broeckhoven, and L. Nyberg, Reduced hippocampal volume in non-demented
carriers of the apolipoprotein E epsilon4: relation to chronological age and recognition memory.
Neurosci Lett., 2006. 396(1): p. 23-7.
Wishart, H.A., A.J. Saykin, L.A. Rabin, R.B. Santulli, L.A. Flashman, S. Guerin, A.C. Mamourian, D.
Belloni, C.H. Rhodes, and T.W. McAllister, Increased prefrontal activation during working memory in
cognitively intact APOE e4 carriers. American Journal of Psychiatry, 2006. 163: p. 1603-1610.
Purcell, S., B. Neale, K. Todd-Brown, L. Thomas, M.A. Ferreira, D. Bender, J. Maller, P. Sklar, P.I. de
Bakker, M.J. Daly, and P.C. Sham, PLINK: a tool set for whole-genome association and populationbased linkage analyses. Am J Hum Genet, 2007. 81(3): p. 559-75.
Balding, D.J., A tutorial on statistical methods for population association studies. Nat Rev Genet, 2006.
7: p. 781-91.
Hirschhorn, J.N. and M.J. Daly, Genome-wide association studies for common diseases and complex
traits. Nat Rev Genet, 2005. 6: p. 95-108.
Zondervan, K.T. and L.R. Cardon, Designing candidate gene and genome-wide case-control
association studies. Nat Protoc, 2007. 2: p. 2492-501.
Glahn, D.C., P.M. Thompson, and J. Blangero, Neuroimaging Endophenotypes: Strategies for Finding
Genes Influencing Brain Structure and Function. Human Brain Mapping, 2007. 28(6): p. 488-501.
Pezawas, L., B.A. Verchinski, V.S. Mattay, J.H. Callicott, B.S. Kolachana, R.E. Straub, M.F. Egan, A.
Meyer-Lindenberg, and D.R. Weinberger, The brain-derived neurotrophic factor val66met
polymorphism and variation in human cortical morphology. J Neurosci, 2004. 24(45): p. 10099-102.
Nichols, T.E. and B. Inkster, Comparison of Whole Brain Multiloci Association Methods, in OHBM’09:
15th Annual Meeting of Organization for Human Brain Mapping. 2009: San Francisco, CA.
Brun, C.C., N. Lepore, X. Pennec, A.D. Lee, M. Barysheva, S.K. Madsen, C. Avedissian, Y.Y. Chou,
G.I. de Zubicaray, K. McMahon, M.J. Wright, A.W. Toga, and P.M. Thompson, Mapping the Regional
PHS 398/2590 (Rev. 11/07)
Page
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
Weiner, Michael
Influence of Genetics on Brain Structure Variability - A Tensor-Based Morphometry Study. NeuroImage,
2009. in press.
Ahmad, R.H., M.D. Emily, and R.W. Daniel, Imaging Genetics: Perspectives from Studies of Genetically
Driven Variation in Serotonin Function and Corticolimbic Affective Processing. Biological psychiatry,
2006. 59(10): p. 888-897.
Shen, L., A.J. Saykin, M.K. Chung, and H. Huang, Morphometric Analysis of Hippocampal Shape in
Mild Cognitive Impairment: An Imaging Genetics Study. IEEE BIBE, 2007: p. 211-217.
Filippinia, N., A. Rao, S. Wetten, R.A. Gibson, M. Borrie, D. Guzman, A. Kertesz, I. Loy-English, J.
Williams, T. Nichols, B. Whitcher, and P.M. Matthews, Anatomically-distinct genetic associations of
APOE ɛ4 allele load with regional cortical atrophy in Alzheimer's disease NeuroImage, 2009. 44(3): p.
724-728.
Seshadri, S., A. DeStefano, R. Au, J. Massaro, A. Beiser, M. Kelly-Hayes, C. Kase, R. D'Agostino, C.
DeCarli, L. Atwood, and P. Wolf, Genetic correlates of brain aging on MRI and cognitive test measures:
a genome-wide association and linkage analysis in the Framingham study. BMC Medical Genetics,
2007. 8(Suppl 1): p. S15.
Baranzini, S.E., J. Wang, R.A. Gibson, N. Galwey, Y. Naegelin, F. Barkhof, E.W. Radue, R.L. Lindberg,
B.M. Uitdehaag, M.R. Johnson, A. Angelakopoulou, L. Hall, J.C. Richardson, R.K. Prinjha, A. Gass, J.J.
Geurts, J. Kragt, M. Sombekke, H. Vrenken, P. Qualley, R.R. Lincoln, R. Gomez, S.J. Caillier, M.F.
George, H. Mousavi, R. Guerrero, D.T. Okuda, B.A. Cree, A.J. Green, E. Waubant, D.S. Goodin, D.
Pelletier, P.M. Matthews, S.L. Hauser, L. Kappos, C.H. Polman, and J.R. Oksenberg, Genome-wide
association analysis of susceptibility and clinical phenotype in multiple sclerosis. Hum Mol Genet, 2009.
18(4): p. 767-78.
Kumar, R.A., K.A. McGhee, S. Leach, R. Bonaguro, A. Maclean, R. Aguirre-Hernandez, B.S.
Abrahams, E.F. Coccaro, S. Hodgins, G. Turecki, A. Condon, W.J. Muir, A.R. Brooks-Wilson, D.H.
Blackwood, and E.M. Simpson, Initial association of NR2E1 with bipolar disorder and identification of
candidate mutations in bipolar disorder, schizophrenia, and aggression through resequencing. Am J
Med Genet B Neuropsychiatr Genet, 2008. 147B(6): p. 880-9.
Meyer-Lindenberg, A., R.E. Straub, B.K. Lipska, B.A. Verchinski, T. Goldberg, J.H. Callicott, M.F. Egan,
S.S. Huffaker, V.S. Mattay, B. Kolachana, J.E. Kleinman, and D.R. Weinberger, Genetic evidence
implicating DARPP-32 in human frontostriatal structure, function, and cognition. J Clin Invest, 2007.
117(3): p. 672-82.
Iatropoulos, P., R. Gardella, P. Valsecchi, C. Magri, C. Ratti, D. Podavini, G. Rossi, M. Gennarelli, E.
Sacchetti, and S. Barlati, Association study and mutational screening of SYNGR1 as a candidate
susceptibility gene for schizophrenia. Psychiatr Genet, 2009.
Licinio, J., C. Dong, and M.L. Wong, Novel sequence variations in the brain-derived neurotrophic factor
gene and association with major depression and antidepressant treatment response. Arch Gen
Psychiatry, 2009. 66(5): p. 488-97.
Egan, M.F., R.E. Straub, T.E. Goldberg, I. Yakub, J.H. Callicott, A.R. Hariri, V.S. Mattay, A. Bertolino,
T.M. Hyde, C. Shannon-Weickert, M. Akil, J. Crook, R.K. Vakkalanka, R. Balkissoon, R.A. Gibbs, J.E.
Kleinman, and D.R. Weinberger, Variation in GRM3 affects cognition, prefrontal glutamate, and risk for
schizophrenia. Proc Natl Acad Sci U S A, 2004. 101(34): p. 12604-9.
Tarpey, P.S., R. Smith, E. Pleasance, A. Whibley, S. Edkins, C. Hardy, S. O'Meara, C. Latimer, E.
Dicks, A. Menzies, P. Stephens, M. Blow, C. Greenman, Y. Xue, C. Tyler-Smith, D. Thompson, K.
Gray, J. Andrews, S. Barthorpe, G. Buck, J. Cole, R. Dunmore, D. Jones, M. Maddison, T. Mironenko,
R. Turner, K. Turrell, J. Varian, S. West, S. Widaa, P. Wray, J. Teague, A. Butler, A. Jenkinson, M. Jia,
D. Richardson, R. Shepherd, R. Wooster, M.I. Tejada, F. Martinez, G. Carvill, R. Goliath, A.P. de
Brouwer, H. van Bokhoven, H. Van Esch, J. Chelly, M. Raynaud, H.H. Ropers, F.E. Abidi, A.K.
Srivastava, J. Cox, Y. Luo, U. Mallya, J. Moon, J. Parnau, S. Mohammed, J.L. Tolmie, C. Shoubridge,
M. Corbett, A. Gardner, E. Haan, S. Rujirabanjerd, M. Shaw, L. Vandeleur, T. Fullston, D.F. Easton, J.
Boyle, M. Partington, A. Hackett, M. Field, C. Skinner, R.E. Stevenson, M. Bobrow, G. Turner, C.E.
Schwartz, J. Gecz, F.L. Raymond, P.A. Futreal, and M.R. Stratton, A systematic, large-scale
resequencing screen of X-chromosome coding exons in mental retardation. Nat Genet, 2009. 41(5): p.
535-43.
PHS 398/2590 (Rev. 11/07)
Page
51.
52.
53.
54.
55.
56.
57.
58.
59.
60.
61.
62.
63.
64.
65.
66.
Weiner, Michael
Dibbens, L.M., P.S. Tarpey, K. Hynes, M.A. Bayly, I.E. Scheffer, R. Smith, J. Bomar, E. Sutton, L.
Vandeleur, C. Shoubridge, S. Edkins, S.J. Turner, C. Stevens, S. O'Meara, C. Tofts, S. Barthorpe, G.
Buck, J. Cole, K. Halliday, D. Jones, R. Lee, M. Madison, T. Mironenko, J. Varian, S. West, S. Widaa,
P. Wray, J. Teague, E. Dicks, A. Butler, A. Menzies, A. Jenkinson, R. Shepherd, J.F. Gusella, Z. Afawi,
A. Mazarib, M.Y. Neufeld, S. Kivity, D. Lev, T. Lerman-Sagie, A.D. Korczyn, C.P. Derry, G.R.
Sutherland, K. Friend, M. Shaw, M. Corbett, H.G. Kim, D.H. Geschwind, P. Thomas, E. Haan, S. Ryan,
S. McKee, S.F. Berkovic, P.A. Futreal, M.R. Stratton, J.C. Mulley, and J. Gecz, X-linked protocadherin
19 mutations cause female-limited epilepsy and cognitive impairment. Nat Genet, 2008. 40(6): p. 77681.
Coon, K.D., J. Valla, S. Szelinger, L.E. Schneider, T.L. Niedzielko, K.M. Brown, J.V. Pearson, R.
Halperin, T. Dunckley, A. Papassotiropoulos, R.J. Caselli, E.M. Reiman, and D.A. Stephan,
Quantitation of heteroplasmy of mtDNA sequence variants identified in a population of AD patients and
controls by array-based resequencing. Mitochondrion, 2006. 6(4): p. 194-210.
Cook, E.H., Jr. and S.W. Scherer, Copy-number variations associated with neuropsychiatric conditions.
Nature, 2008. 455(7215): p. 919-23.
Hastings, P.J., J.R. Lupski, S.M. Rosenberg, and G. Ira, Mechanisms of change in gene copy number.
Nat Rev Genet, 2009. 10(8): p. 551-64.
Zhang, F., W. Gu, M.E. Hurles, and J.R. Lupski, Copy number variation in human health, disease, and
evolution. Annu Rev Genomics Hum Genet, 2009. 10: p. 451-81.
Wain, L.V., J.A. Armour, and M.D. Tobin, Genomic copy number variation, human health, and disease.
Lancet, 2009. 374(9686): p. 340-50.
Alaerts, M. and J. Del-Favero, Searching genetic risk factors for schizophrenia and bipolar disorder:
learn from the past and back to the future. Hum Mutat, 2009. 30(8): p. 1139-52.
Rovelet-Lecrux, A., D. Hannequin, G. Raux, N. Le Meur, A. Laquerriere, A. Vital, C. Dumanchin, S.
Feuillette, A. Brice, M. Vercelletto, F. Dubas, T. Frebourg, and D. Campion, APP locus duplication
causes autosomal dominant early-onset Alzheimer disease with cerebral amyloid angiopathy. Nat
Genet, 2006. 38(1): p. 24-6.
Sleegers, K., N. Brouwers, I. Gijselinck, J. Theuns, D. Goossens, J. Wauters, J. Del-Favero, M. Cruts,
C.M. van Duijn, and C. Van Broeckhoven, APP duplication is sufficient to cause early onset Alzheimer's
dementia with cerebral amyloid angiopathy. Brain, 2006. 129(Pt 11): p. 2977-83.
Theuns, J., N. Brouwers, S. Engelborghs, K. Sleegers, V. Bogaerts, E. Corsmit, T. De Pooter, C.M. van
Duijn, P.P. De Deyn, and C. Van Broeckhoven, Promoter mutations that increase amyloid precursorprotein expression are associated with Alzheimer disease. Am J Hum Genet, 2006. 78(6): p. 936-46.
Avramopoulos, D., R. Wang, D. Valle, M.D. Fallin, and S.S. Bassett, A novel gene derived from a
segmental duplication shows perturbed expression in Alzheimer's disease. Neurogenetics, 2007. 8(2):
p. 111-20.
Belin, A.C., B.F. Bjork, M. Westerlund, D. Galter, O. Sydow, C. Lind, K. Pernold, L. Rosvall, A.
Hakansson, B. Winblad, H. Nissbrandt, C. Graff, and L. Olson, Association study of two genetic
variants in mitochondrial transcription factor A (TFAM) in Alzheimer's and Parkinson's disease.
Neurosci Lett, 2007. 420(3): p. 257-62.
Korbel, J.O., T. Tirosh-Wagner, A.E. Urban, X.N. Chen, M. Kasowski, L. Dai, F. Grubert, C. Erdman,
M.C. Gao, K. Lange, E.M. Sobel, G.M. Barlow, A.S. Aylsworth, N.J. Carpenter, R.D. Clark, M.Y. Cohen,
E. Doran, T. Falik-Zaccai, S.O. Lewin, I.T. Lott, B.C. McGillivray, J.B. Moeschler, M.J. Pettenati, S.M.
Pueschel, K.W. Rao, L.G. Shaffer, M. Shohat, A.J. Van Riper, D. Warburton, S. Weissman, M.B.
Gerstein, M. Snyder, and J.R. Korenberg, The genetic architecture of Down syndrome phenotypes
revealed by high-resolution analysis of human segmental trisomies. Proc Natl Acad Sci U S A, 2009.
106(29): p. 12031-6.
Heinzen, E.L., A.C. Need, K.M. Hayden, O. Chiba-Falek, A.D. Roses, W.J. Strittmatter, J.R. Burke,
C.M. Hulette, K.A. Welsh-Bohmer, and D.B. Goldstein, Genome-Wide Scan of Copy Number Variation
in Late-Onset Alzheimer's Disease. J Alzheimers Dis, 2009.
Liu, Z., A. Sall, and D. Yang, MicroRNA: an Emerging Therapeutic Target and Intervention Tool. Int J
Mol Sci, 2008. 9(6): p. 978-99.
Griffiths-Jones, S., H.K. Saini, S. van Dongen, and A.J. Enright, miRBase: tools for microRNA
genomics. Nucleic Acids Res, 2008. 36(Database issue): p. D154-8.
PHS 398/2590 (Rev. 11/07)
Page
67.
68.
69.
70.
71.
72.
73.
74.
75.
76.
77.
78.
79.
80.
81.
82.
83.
84.
85.
86.
Weiner, Michael
Griffiths-Jones, S., Annotating noncoding RNA genes. Annu Rev Genomics Hum Genet, 2007. 8: p.
279-98.
Kosik, K.S., The neuronal microRNA system. Nat Rev Neurosci, 2006. 7(12): p. 911-20.
Corbin, R., K. Olsson-Carter, and F. Slack, The role of microRNAs in synaptic development and
function. BMB Rep, 2009. 42(3): p. 131-5.
Chang, S., S. Wen, D. Chen, and P. Jin, Small regulatory RNAs in neurodevelopmental disorders. Hum
Mol Genet, 2009. 18(R1): p. R18-26.
Maes, O.C., H.M. Chertkow, E. Wang, and H.M. Schipper, MicroRNA: Implications for Alzheimer
Disease and other Human CNS Disorders. Current Genomics 2009. 10(3): p. 154-168.
Weinberg, M.S. and M.J. Wood, Short non-coding RNA biology and neurodegenerative disorders: novel
disease targets and therapeutics. Hum Mol Genet, 2009. 18(R1): p. R27-39.
Lukiw, W.J., Micro-RNA speciation in fetal, adult and Alzheimer's disease hippocampus. Neuroreport,
2007. 18(3): p. 297-300.
Wang, W.X., B.W. Rajeev, A.J. Stromberg, N. Ren, G. Tang, Q. Huang, I. Rigoutsos, and P.T. Nelson,
The expression of microRNA miR-107 decreases early in Alzheimer's disease and may accelerate
disease progression through regulation of beta-site amyloid precursor protein-cleaving enzyme 1. J
Neurosci, 2008. 28(5): p. 1213-23.
Hebert, S.S., K. Horre, L. Nicolai, A.S. Papadopoulou, W. Mandemakers, A.N. Silahtaroglu, S.
Kauppinen, A. Delacourte, and B. De Strooper, Loss of microRNA cluster miR-29a/b-1 in sporadic
Alzheimer's disease correlates with increased BACE1/beta-secretase expression. Proc Natl Acad Sci U
S A, 2008. 105(17): p. 6415-20.
Boissonneault, V., I. Plante, S. Rivest, and P. Provost, MicroRNA-298 and microRNA-328 regulate
expression of mouse beta-amyloid precursor protein-converting enzyme 1. J Biol Chem, 2009. 284(4):
p. 1971-81.
Hebert, S.S., K. Horre, L. Nicolai, B. Bergmans, A.S. Papadopoulou, A. Delacourte, and B. De
Strooper, MicroRNA regulation of Alzheimer's Amyloid precursor protein expression. Neurobiol Dis,
2009. 33(3): p. 422-8.
Cogswell, J.P., J. Ward, I.A. Taylor, M. Waters, Y. Shi, B. Cannon, K. Kelnar, J. Kemppainen, D.
Brown, C. Chen, R.K. Prinjha, J.C. Richardson, A.M. Saunders, A.D. Roses, and C.A. Richards,
Identification of miRNA changes in Alzheimer's disease brain and CSF yields putative biomarkers and
insights into disease pathways. J Alzheimers Dis, 2008. 14(1): p. 27-41.
Schipper, H.M., O.C. Maes, H.M. Chertkow, and E. Wang, MicroRNA Expression in Alzheimer Blood
Mononuclear Cells. Gene Regulation and Systems Biology, 2007. 2007(GRSB-1-Maes-et-al-2): p. 263274.
Maes, O.C., S. Xu, B. Yu, H.M. Chertkow, E. Wang, and H.M. Schipper, Transcriptional profiling of
Alzheimer blood mononuclear cells by microarray. Neurobiol Aging, 2007. 28(12): p. 1795-809.
Mohr, S. and C.C. Liew, The peripheral-blood transcriptome: new insights into disease and risk
assessment. Trends Mol Med, 2007. 13(10): p. 422-32.
Liew, C.C., J. Ma, H.C. Tang, R. Zheng, and A.A. Dempsey, The peripheral blood transcriptome
dynamically reflects system wide biology: a potential diagnostic tool. J Lab Clin Med, 2006. 147(3): p.
126-32.
Kalman, J., K. Kitajka, M. Pakaski, A. Zvara, A. Juhasz, G. Vincze, Z. Janka, and L.G. Puskas, Gene
expression profile analysis of lymphocytes from Alzheimer's patients. Psychiatr Genet, 2005. 15(1): p.
1-6.
Cosentino, M., C. Colombo, M. Mauri, M. Ferrari, S. Corbetta, F. Marino, G. Bono, and S. Lecchini,
Expression of apoptosis-related proteins and of mRNA for dopaminergic receptors in peripheral blood
mononuclear cells from patients with Alzheimer disease. Alzheimer Dis Assoc Disord, 2009. 23(1): p.
88-90.
Grunblatt, E., J. Bartl, S. Zehetmayer, T.M. Ringel, P. Bauer, P. Riederer, and C.P. Jacob, Gene
expression as peripheral biomarkers for sporadic Alzheimer's disease. J Alzheimers Dis, 2009. 16(3):
p. 627-34.
Kamagata, E., T. Kudo, R. Kimura, H. Tanimukai, T. Morihara, M.G. Sadik, K. Kamino, and M. Takeda,
Decrease of dynamin 2 levels in late-onset Alzheimer's disease alters Abeta metabolism. Biochem
Biophys Res Commun, 2009. 379(3): p. 691-5.
PHS 398/2590 (Rev. 11/07)
Page
87.
88.
89.
90.
91.
92.
93.
94.
95.
96.
97.
98.
99.
100.
101.
102.
103.
Weiner, Michael
Fenoglio, C., D. Galimberti, F. Cortini, J.S. Kauwe, C. Cruchaga, E. Venturelli, C. Villa, M. Serpente, D.
Scalabrini, K. Mayo, L.M. Piccio, F. Clerici, D. Albani, C. Mariani, G. Forloni, N. Bresolin, A.M. Goate,
and E. Scarpini, Rs5848 Variant Influences GRN mRNA Levels in Brain and Peripheral Mononuclear
Cells in Patients with Alzheimer's Disease. J Alzheimers Dis, 2009.
Gatta, L., A. Cardinale, F. Wannenes, C. Consoli, A. Armani, F. Molinari, C. Mammi, F. Stocchi, M.
Torti, G.M. Rosano, and M. Fini, Peripheral blood mononuclear cells from mild cognitive impairment
patients show deregulation of Bax and Sod1 mRNAs. Neurosci Lett, 2009. 453(1): p. 36-40.
Cortini, F., C. Fenoglio, I. Guidi, E. Venturelli, S. Pomati, A. Marcone, D. Scalabrini, C. Villa, F. Clerici,
E. Dalla Valle, C. Mariani, S. Cappa, N. Bresolin, E. Scarpini, and D. Galimberti, Novel exon 1
progranulin gene variant in Alzheimer's disease. Eur J Neurol, 2008. 15(10): p. 1111-7.
Coppola, G., A. Karydas, R. Rademakers, Q. Wang, M. Baker, M. Hutton, B.L. Miller, and D.H.
Geschwind, Gene expression study on peripheral blood identifies progranulin mutations. Ann Neurol,
2008. 64(1): p. 92-6.
Jasinska, A.J., S. Service, O.W. Choi, J. Deyoung, O. Grujic, S.Y. Kong, M.J. Jorgensen, J. Bailey, S.
Breidenthal, L.A. Fairbanks, R.P. Woods, J.D. Jentsch, and N.B. Freimer, Identification of Brain
Transcriptional Variation Reproduced in Peripheral Blood: an Approach for Mapping Brain Expression
Traits. Hum Mol Genet, 2009.
Nho, K., L. Shen, S. Kim, S.L. Risacher, T. Foroud, N. Pankratz, D.W. Craig, M.J. Huentelman, B.M.
DeChairo, S.G. Potkin, M. Weiner, A.J. Saykin, and ADNI, Genome-wide Association Analysis of the
ADNI Cohort Identifies a Putative Susceptibility Locus for Alzheimer's Disease. submitted 2009.
Nho, K., L. Shen, S. Kim, S.L. Risacher, T. Foroud, N. Pankratz, D.W. Craig, M.J. Huentelman, B.M.
DeChairo, S.G. Potkin, M. Weiner, A.J. Saykin, and ADNI, Genome-wide Association Study for
Alzheimer's Disease using a Longitudinal Diagnosis. submitted 2009.
Stein, J.L., X. Hua, S. Lee, A.J. Ho, A.D. Leow, A. Toga, A.J. Saykin, L. Shen, T. Foroud, N. Pankratz,
M.J. Huentelman, D.W. Craig, J.D. Gerber, A. Allen, J. Corneveaux, B.M. DeChairo, S.G. Potkin, C.
Jack, M. Weiner, and P. Thompson, Voxelwise Genome-Wide Association Study (vGWAS). submitted
2009.
Neitzel, H., A routine method for the establishment of permanent growing lymphoblastoid cell lines.
Hum Genet, 1986. 73(4): p. 320-6.
Sloan, C., L. Shen, J. West, H. Wishart, L. Flashman, L. Rabin, R. Santulli, S. Guerin, C. Rhodes, G.
Tsongalis, T. McAllister, T. Ahles, S. Lee, J. Moore, and A. Saykin, Genetic Pathway-Based
Hierarchical Clustering Analysis of Older Adults with Cognitive Complaints and Amnestic Mild Cognitive
Impairment Using Clinical and Neuroimaging Phenotypes. submitted 2009.
Ho, A.J., J.L. Stein, X. Hua, S. Lee, D.P. Hibar, A.D. Leow, I.D. Dinov, A. Toga, A.J. Saykin, L. Shen, T.
Foroud, N. Pankratz, M.J. Huentelman, D.W. Craig, J.D. Gerber, A. Allen, J. Corneveaux, D.A.
Stephan, J. Webster, B.M. DeChairo, S.G. Potkin, C. Jack, M. Weiner, C.A. Raji, O.L. Lopez, J.T.
Becker, and P.T. Thompson, Commonly carried allele within FTO, an obesity-associated gene, relates
to accelerated brain degeneration in the elderly. submitted 2009.
Risacher, S.L., A.J. Saykin, J.D. West, L. Shen, H.A. Firpi, and B.C. McDonald, Baseline MRI
predictors of conversion from MCI to probable AD in the ADNI cohort. Curr Alzheimer Res, 2009. 6(4):
p. 347-61.
Winchester, L., C. Yau, and J. Ragoussis, Comparing CNV detection methods for SNP arrays. Brief
Funct Genomic Proteomic, 2009: p. elp017.
Barnes, C., V. Plagnol, T. Fitzgerald, R. Redon, J. Marchini, D. Clayton, and M.E. Hurles, A robust
statistical method for case-control association testing with copy number variation. Nat Genet, 2008.
40(10): p. 1245-1252.
Wang, K., M. Li, D. Hadley, R. Liu, J. Glessner, S.F. Grant, H. Hakonarson, and M. Bucan, PennCNV:
an integrated hidden Markov model designed for high-resolution copy number variation detection in
whole-genome SNP genotyping data. Genome Res, 2007. 17(11): p. 1665-74.
Carter, N.P., Methods and strategies for analyzing copy number variation using DNA microarrays. Nat
Genet, 2007.
Korn, J.M., F.G. Kuruvilla, S.A. McCarroll, A. Wysoker, J. Nemesh, S. Cawley, E. Hubbell, J. Veitch,
P.J. Collins, K. Darvishi, C. Lee, M.M. Nizzari, S.B. Gabriel, S. Purcell, M.J. Daly, and D. Altshuler,
PHS 398/2590 (Rev. 11/07)
Page
104.
105.
106.
107.
108.
109.
110.
111.
112.
113.
114.
115.
116.
117.
118.
119.
120.
121.
122.
123.
124.
125.
Weiner, Michael
Integrated genotype calling and association analysis of SNPs, common copy number polymorphisms
and rare CNVs. Nat Genet, 2008. 40(10): p. 1253-1260.
Sun, W., F.A. Wright, Z. Tang, S.H. Nordgard, P.V. Loo, T. Yu, V.N. Kristensen, and C.M. Perou,
Integrated study of copy number states and genotype calls using high-density SNP arrays. Nucl. Acids
Res., 2009: p. gkp493.
Teo, Y.Y.a.b., Common statistical issues in genome-wide association studies: a review on power, data
quality control, genotype calling and population structure. Current Opinion in Lipidology, 2008. 19(2): p.
133-143.
Wellcome Trust Case Control Consortium, Genome-wide association study of 14,000 cases of seven
common diseases and 3,000 shared controls. Nature, 2007. 447(7145): p. 661-78.
Teo, Y.Y., A.E. Fry, T.G. Clark, E.S. Tai, and M. Seielstad, On the usage of HWE for identifying
genotyping errors. Ann Hum Genet, 2007. 71(Pt 5): p. 701-3; author reply 704.
Cardon, L.R. and L.J. Palmer, Population stratification and spurious allelic association. Lancet, 2003.
361(9357): p. 598-604.
Wawro, N., K. Bammann, and I. Pigeot, Testing for association in the presence of population
stratification: a simulation study comparing the S-TDT, STRAT and the GC. Biom J, 2006. 48(3): p.
420-34.
Yu, K., Z. Wang, Q. Li, S. Wacholder, D.J. Hunter, R.N. Hoover, S. Chanock, and G. Thomas,
Population substructure and control selection in genome-wide association studies. PLoS One, 2008.
3(7): p. e2551.
Ziegler, A., I.R. Konig, and J.R. Thompson, Biostatistical aspects of genome-wide association studies.
Biom J, 2008. 50(1): p. 8-28.
Miclaus, K., R. Wolfinger, and W. Czika, SNP selection and multidimensional scaling to quantify
population structure. Genet Epidemiol, 2009. 33(6): p. 488-96.
Pritchard, J.K., M. Stephens, and P. Donnelly, Inference of population structure using multilocus
genotype data. Genetics, 2000. 155(2): p. 945-59.
Price, A.L., N.J. Patterson, R.M. Plenge, M.E. Weinblatt, N.A. Shadick, and D. Reich, Principal
components analysis corrects for stratification in genome-wide association studies. Nat Genet, 2006.
38(8): p. 904-9.
Dai, J.Y., I. Ruczinski, M. LeBlanc, and C. Kooperberg, Imputation methods to improve inference in
SNP association studies. Genet Epidemiol, 2006. 30(8): p. 690-702.
Hao, K., E. Chudin, J. McElwee, and E.E. Schadt, Accuracy of genome-wide imputation of untyped
markers and impacts on statistical power for association studies. BMC Genet, 2009. 10: p. 27.
Nothnagel, M., D. Ellinghaus, S. Schreiber, M. Krawczak, and A. Franke, A comprehensive evaluation
of SNP genotype imputation. Hum Genet, 2009. 125(2): p. 163-71.
Potkin, S.G., J.A. Turner, G. Guffanti, A. Lakatos, F. Torri, D.B. Keator, and F. Macciardi, Genome-wide
strategies for discovering genetic influences on cognition and cognitive disorders: methodological
considerations. Cogn Neuropsychiatry, 2009. 14(4-5): p. 391-418.
Tyler, A.L., F.W. Asselbergs, S.M. Williams, and J.H. Moore, Shadows of complexity: what biological
networks reveal about epistasis and pleiotropy. Bioessays, 2009. 31(2): p. 220-7.
Sinnott-Armstrong, N.A., C.S. Greene, F. Cancare, and J.H. Moore, Accelerating epistasis analysis in
human genetics with consumer graphics hardware. BMC Res Notes, 2009. 2: p. 149.
Pattin, K.A., B.C. White, N. Barney, J. Gui, H.H. Nelson, K.T. Kelsey, A.S. Andrew, M.R. Karagas, and
J.H. Moore, A computationally efficient hypothesis testing method for epistasis analysis using
multifactor dimensionality reduction. Genet Epidemiol, 2009. 33(1): p. 87-94.
Moore, J.H. and S.M. Williams, Epistasis and its implications for personal genetics. Am J Hum Genet,
2009. 85(3): p. 309-20.
Thornton-Wells, T.A., J.H. Moore, E.R. Martin, M.A. Pericak-Vance, and J.L. Haines, Confronting
complexity in late-onset Alzheimer disease: application of two-stage analysis approach addressing
heterogeneity and epistasis. Genet Epidemiol, 2008. 32(3): p. 187-203.
Moore, J.H., Analysis of gene-gene interactions. Curr Protoc Hum Genet, 2008. Chapter 1: p. Unit 1
14.
Moore, J.H., N. Barney, C.T. Tsai, F.T. Chiang, J. Gui, and B.C. White, Symbolic modeling of epistasis.
Hum Hered, 2007. 63(2): p. 120-33.
PHS 398/2590 (Rev. 11/07)
Page
126.
127.
128.
129.
130.
131.
132.
133.
134.
135.
Weiner, Michael
Han, B., H.M. Kang, and E. Eskin, Rapid and accurate multiple testing correction and power estimation
for millions of correlated markers. PLoS Genet, 2009. 5(4): p. e1000456.
Gauderman, W.J. and J.M. Morrison, QUANTO 1.1: A computer program for power and sample size
calculations for genetic-epidemiology studies. 2006, http://hydra.usc.edu/gxe.
Stromberg, U., J. Bjork, P. Vineis, K. Broberg, and E. Zeggini, Ranking of genome-wide association
scan signals by different measures. Int J Epidemiol, 2009. 38(5): p. 1364-73.
Menashe, I., P.S. Rosenberg, and B.E. Chen, PGA: power calculator for case-control genetic
association analyses. BMC Genet, 2008. 9: p. 36.
Purcell, S., S.S. Cherny, and P.C. Sham, Genetic Power Calculator: design of linkage and association
genetic mapping studies of complex traits. Bioinformatics, 2003. 19(1): p. 149-50.
Krawczak, M., ASP--a simulation-based power calculator for genetic linkage studies of qualitative traits,
using sib-pairs. Hum Genet, 2001. 109(6): p. 675-7.
Little, J., J.P. Higgins, J.P. Ioannidis, D. Moher, F. Gagnon, E. von Elm, M.J. Khoury, B. Cohen, G.
Davey-Smith, J. Grimshaw, P. Scheet, M. Gwinn, R.E. Williamson, G.Y. Zou, K. Hutchings, C.Y.
Johnson, V. Tait, M. Wiens, J. Golding, C. van Duijn, J. McLaughlin, A. Paterson, G. Wells, I. Fortier, M.
Freedman, M. Zecevic, R. King, C. Infante-Rivard, A. Stewart, and N. Birkett, Strengthening the
reporting of genetic association studies (STREGA): an extension of the STROBE Statement. Hum
Genet, 2009. 125(2): p. 131-51.
Green, R.C., J.S. Roberts, L.A. Cupples, N.R. Relkin, P.J. Whitehouse, T. Brown, S.L. Eckert, M.
Butson, A.D. Sadovnick, K.A. Quaid, C. Chen, R. Cook-Deegan, and L.A. Farrer, Disclosure of APOE
genotype for risk of Alzheimer's disease. N Engl J Med, 2009. 361(3): p. 245-54.
Cuenco, K.T., R. Friedland, C.T. Baldwin, J. Guo, B. Vardarajan, K.L. Lunetta, L.A. Cupples, R.C.
Green, C. Decarli, and L.A. Farrer, Association of TTR polymorphisms with hippocampal atrophy in
Alzheimer disease families. Neurobiol Aging, 2009.
Cuenco, T.K., K.L. Lunetta, C.T. Baldwin, A.C. McKee, J. Guo, L.A. Cupples, R.C. Green, P.H. St
George-Hyslop, H. Chui, C. DeCarli, and L.A. Farrer, Association of distinct variants in SORL1 with
cerebrovascular and neurodegenerative changes related to Alzheimer disease. Arch Neurol, 2008.
65(12): p. 1640-8.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Core: 7
Title of Core (not to exceed 81 spaces): Neuropathology Core
Core Leader: Morris, John, C.
Position/Title: Professor, Washington University in St. Louis
Department, service, laboratory, or equivalent: Neurology
Mailing Address:
One Brookings Drive, Box 8111
St. Louis, MO 63130
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Building on the recent experience (2007-9) of the ADNI1-Neuropathology Core (ADNI1-NPC), the aims of
the ADNI2-NPC are an extension of the ADNI2 specific aims in that it will provide the "gold standard"
validation of the clinical diagnoses and imaging surrogates through neuropathological examination of ADNI1
and ADNI2 participants who come to autopsy. Specific Aim 1: Provide training materials and protocols to
assist clinicians at ADNI sites in obtaining voluntary consent for brain autopsy in ADNI participants. We
project that over the 5 year period of funding we expect pathological material on the following subjects who
come to autopsy (from surviving ADNI1 and newly enrolled ADNI2 participants): 10 non-demented controls,
20 MCI, and 50 AD deaths. Assuming an autopsy rate of 75% over the proposed funding cycle for ADNI2
(2009-14), we anticipate that at the minimum there will be autopsies in at least 7 non-demented controls, 15
MCI individuals, and 38 AD individuals during the five year period of this project. Specific Aim 2: Maintain a
central laboratory to provide uniform neuropathological assessments in all autopsied ADNI1 and ADNI2
participants in accordance with standard criteria and to promote clinical-neuroimaging-neuropathological
correlations. Neuropathologic assessment is necessary to validate clinical, imaging, and biomarker data. In
3 (33%) of 9 ADNI participants there was a combined neuropathologic diagnosis of dementia with Lewy
bodies (DLB) and Alzheimer’s disease (AD). These preliminary data indicate that the ADNI population is
neuropathologically heterogeneous and that co-morbidity may confound or explain variance in the data
generated by the different Cores. Specific Aim 3: Maintain a state-of-the-art resource for fixed and frozen
brain tissue obtained from autopsied ADNI participants to support ADNI's biomarker studies and make tissue
available to ADNI-approved investigators for research purposes. These specimens will facilitate the
validation of clinical, imaging, and biomarker data obtained during the course of the disease. Specific Aim 4:
Interact with ADNI's Data Coordinating Center to ensure entry of the Core's data into ADNI's database,
promote data sharing and collaborative research, and integrate the ADNI2-NPC with all ADNI2 components.
The ADNI2-NPC will provide the "gold standard" validation of the clinical diagnoses and imaging surrogates
through neuropathological examination of ADNI participants who come to autopsy, and maintain a state-ofthe-art resource for fixed and frozen brain tissue obtained from autopsied ADNI participants to support
ADNI's biomarker studies and provide tissue to ADNI-approved investigators for research purposes.
Washington University
DUNS:
06-855-2207
Street 1:
660 S. Euclid Avenue
City:
Street 2:
Saint Louis
Province:
County:
Country:
Campus Box 8111
Saint Louis City
US
State:
Zip/Postal Code:
MO
63110
1
DUNS:
Street 1:
Street 2:
City:
Province:
County:
Country:
State:
Zip/Postal Code:
PHS 398 (Rev. 11/07)
Page 2
Form Page 2
Weiner, Michael
Name
Organization
Role on Project
Morris, John C.
J_MORRIS
ADNI-NPC Director
Cairns, Nigel J.
Nigel_Cairns
ADNI-NPC Co-director
Name
Organization
Role on Project
No
Yes
Cell Line
PHS 398 (Rev. 11/07)
Page 3
Weiner, Michael
7. NEUROPATHOLOGY CORE:
7.1. Specific Aims:
Building on the recent experience (2007-9) of the ADNI1-Neuropathology Core (ADNI1-NPC), the aims of the
ADNI2-NPC are a continuation of the original specific aims and focus on providing the "gold standard"
validation of the clinical diagnoses and imaging surrogates through neuropathological examination of ADNI1,
GO and ADNI2 participants who come to autopsy. Neuropathologic diagnosis remains essential to validate
clinical diagnoses; otherwise, the data generated by the different clinical assessments, imaging modalities, and
biomarkers obtained from ADNI participants believed to have Alzheimer’s disease (AD) may be contaminated
by individuals who in fact do not have AD. As an example, in the Aβ vaccine trial AN1792 in persons with a
clinical diagnosis of AD, 1 of 9 participants who came to autopsy had progressive supranuclear palsy rather
than AD [23]. Variance in neuropathologic data arises because different neuropathologists use different
methods and interpret criteria differently so that a SINGLE NEUROPATHOLOGY CORE is ESSENTIAL to
maintain staining standards and uniform neuropathologic diagnoses [21, 22]
7.1.1. Specific Aim 1 [SA1]: Provide training materials and protocols to assist clinicians at ADNI sites in
obtaining voluntary consent for brain autopsy in ADNI participants. We project that over the 5 year period
of ADNI2 funding pathological material will be obtained from surviving ADNI1, GO and newly enrolled ADNI2
participants as follows, assuming an autopsy rate of 75%: at the minimum there will be autopsies in at least 7
non-demented controls, 15 MCI individuals, and 38 AD individuals during the five year period of this project.
During the next phase of ADNI we expect to improve the autopsy rate from 0% at the start of ADN1 to 75%
during ADNI2. We believe that this is realistic as the autopsy rate in the last year of funding was 71.5%.
7.1.2. Specific Aim 2 [SA2]: Maintain a central laboratory to provide uniform neuropathological
assessments in all autopsied ADNI1 and ADNI2 participants in accordance with standard criteria and to
promote clinical-neuroimaging-neuropathological correlations. Neuropathologic assessment at a single
NPC site is essential to maintain staining standards and uniform neuropathologic diagnoses. In addition, the
ability to have neuropathology helps determine the sequence of biomarker changes [20]. Although relatively
few ADNI1cases to date have come to autopsy, the number of cases will increase as participants continue to
age. Moreover, 3 of the current 9 autopsied cases have a combined neuropathologic diagnosis of dementia
with Lewy bodies (DLB) and Alzheimer’s disease (AD). These preliminary data indicate that the ADNI
population is neuropathologically heterogeneous and that comorbidity may confound or explain variance in the
data generated by the different ADNI Cores.
7.1.3. Specific Aim 3 [SA3]: Maintain a state-of-the-art resource for fixed and frozen brain tissue
obtained from autopsied ADNI participants to support ADNI's biomarker studies and make available to
ADNI-approved investigators access to the tissue and data for research purposes. Fixed brain
tissue/paraffin sections are available from cases that have come to autopsy. These and frozen brain samples
will facilitate the validation of clinical, imaging, and biomarker data obtained during the course of the disease.
7.1.4. Specific Aim 4 [SA4]: Interact with ADNI's Data Coordinating Center to ensure appropriate entry
of the Core's data into ADNI's database, promote data sharing and collaborative research, and
integrate the ADNI2-NPC with all ADNI2 components to support its administration, operations, and
progress toward goals. Data is currently being uploaded to the Laboratory of Neuroimaging (LONI),
University of California, Los Angeles, and the format of the neuropathologic data is based on the
neuropathology form of the National Alzheimer Coordinating Center and is used by all NIH-funded Alzheimer
Disease Centers.
The Alzheimer’s Disease Neuroimaging Initiative (ADNI) was established to determine the relationships among
the clinical, cognitive, imaging, genetic and biochemical biomarker characteristics of the entire spectrum of
Alzheimer’s disease (AD) as the pathology evolves from normal aging to dementia. ADNI will inform the
neuroscience of AD, identify diagnostic and prognostic markers, identify outcome measures which can be used
in clinical trials and will help develop the most effective clinical trial scenarios. The proposed project continues
the currently funded ADNI (ADNI1), a public/private collaboration between academia and industry to study
biomarkers of AD.
The goals of ADNI2 will be accomplished by: 1) continuing annual clinical/cognitive/MRI follow-up of the 503
normal controls and late MCI (lMCI) subjects previously enrolled in ADNI1; 2) enrollment of 200 early MCI
(eMCI) to bridge the gap between healthy controls and late MCI (lMCI). Additional enrollment of new healthy
controls (n=100), lMCI (n=100), and AD (n=100) subjects; 3) performance of F18 amyloid PET on all subjects,
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
together with FDG-PET, lumbar puncture for CSF, clinical/cognitive measurements and MRI, and
neuropathology on all cases that come to autopsy. All collected data will be processed, analyzed by ADNI
investigators including the Biostat Core, and made available to all qualified scientists in the world who request
a password, without embargo.
To achieve the goals of ADNI, the Neuropathology Core is essential to validate the clinical classifications and
diagnoses. If there is no neuropathologic validation, ADNI data are likely to be contaminated by individuals who
do not have AD, or, more commonly, comorbidities such as vascular disease and non-AD neurodegenerative
disorders [23]. Also, a single Neuropathology Core site is necessary because different neuropathologists use
different processing and staining methods, as well as different antibodies and interpret diagnostic criteria
differently. Even for the neuropathologic diagnosis of AD, not all sites use the same sets of criteria. The
literature has extensive data showing variability between different neuropathologists, sites, and countries [21,
22]. A single Neuropathology Core ensures uniformity and fidelity of staining and application of diagnostic
criteria to all ADNI participants who come to autopsy.
7.2.1. ADNI2 Neuropathology Core (ADNI2-NPC) : The ADNI-NPC capitalizes on the existing infrastructure
of the Washington University Alzheimer Disease Research Center (WU ADRC; P50AG05681, JC Morris, PI),
funded continuously by the National Institute on Aging since 1985. The ADRC’s Administrative (Dr. Morris) and
Neuropathology (Dr. Cairns) Cores provide the framework for the ADNI1-NPC and will continue to do so during
the period of ADNI2. Fidelity of data between ADNI1, ADNI2, and the National Alzheimer Coordinating Center
(NACC; U01AG016976,W. Kukull, PI) is maintained by using the same NACC Neuropathology Data Form as is
used by all Alzheimer Disease Centers (ADCs) to report neuropathological findings from autopsied cases, and
will remain the primary data collection instrument. In this way, the ADNI2-NPC will use standard criteria for
neuropathological diagnoses of dementing illness and existing protocols and procedures to achieve these
diagnoses. ADNI1-NPC does not interfere with or supersede neuropathological activities at any ADNI site. The
ADNI-NPC uses brain tissue obtained at the participating ADNI sites to provide a uniform neuropathological
assessment to support the clinical classifications and research aims of ADNI1 and the proposed ADNI2.
Funding of the Neuropathology Core started on 09-01-2007 and since that time the ADNI-NPC has become
fully operational and serves all ADNI sites. During the initial period of funding (09-01-2007 to 08-31-2009), the
ADNI-NPC has achieved its stated goals.. It has: (1) provided and implemented training materials and
protocols to assist clinicians at ADNI sites in obtaining voluntary consent for brain autopsy in ADNI participants;
(2) established a central laboratory to provide uniform neuropathological assessments in all autopsied ADNI
participants in accordance with standard criteria [1-15] and promotes clinical-neuroimaging-neuropathological
correlations; (3) established and maintains a state-of-the-art resource for fixed (9 of 9 cases) and frozen brain
tissue (8 of nine cases) obtained from autopsied ADNI participants to support ADNI's biomarker studies) and
developed a process wherein investigators may have access to the tissue and data for research purposes; and
(4) interacts with ADNI's Data Coordinating Center to ensure appropriate entry of the Core's data into ADNI's
database, promotes data sharing and collaborative research, and integrate the ADNI-NPC with all ADNI
components to support its administration, operations, and progress toward goals.
7.3. Preliminary Results:
7.3.1. ADNI Neuropathology Core (ADNI1-NPC): Progress since 09-01-2007: A highly motivated ADNI-NPC
Research Coordinator, Mrs Lisa Taylor-Reinwald, has contacted all participating ADNI sites to implement the
protocols established for obtaining autopsy consent and performing neuropathology services. Mrs TaylorReinwald continuously monitors the sites to encourage and facilitate autopsy consent in ADNI participants. In
addition, all ADNI-NPC documentation is available at the ADNI website. Where autopsy procedures do not
exist locally, arrangements have been put in place with the site PI and local hospital to harvest brain tissue and
forward to the ADNI-NPC in St Louis. To promote the goals of the ADNI1-NPC and to inform participating ADNI
sites, meetings were held concurrently in April, 2008, at the American Association of Neuropathologists Annual
Meeting, Washington, D.C. and at the American Academy of Neurology Meeting, Boston, MA. Interestingly, the
first ADNI participant to come to autopsy had neuropathologic diagnoses of dementia with Lewy bodies (DLB)
and coexisting Alzheimer's disease (AD). Of the 9 autopsies, 3 have combined DLB and AD (see Table 2 for
all neuropathologic diagnoses encountered in ADNI cases).
During the period of funding of ADNI1, there have been 22 participant deaths (Table 1). In the initial phase of
ADNI1 (09-01-2005 to 08-31-2007), when no resources were available for neuropathology, there were 6
participant deaths and no autopsies (autopsy rate = 0%). During the initial year of funding of the ADNI1
Neuropathology Core (09-01-2007 to 08-31-2008), there was notable improvement on the autopsy rate to
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
44.4%. In the most recent year of funding (09-01-2008 to 08-31-2009), our autopsy rate improved to 71.5%
(Table 1). Although the overall numbers to date are small, these data demonstrate that the Neuropathology
Core has established the administrative organization with the participating sites to harvest brains from ADNI
participants who come to autopsy. As expected, the numbers of ADNI participants who come to autopsy is
increasing as the period of the study lengthens and participants age.
Table 1. ADNI1 Autopsy Rates 09-01-2005 to 08-31-2009
ADNI1
ADNI1-NPC
Deaths Autopsies Autopsy Rate (%)
Funding Period
09-01-2005 to 08-31-2007
N0
6
0
0
09-01-2007 to 08-31-2008
YES
9
4
44.4
09-01-2008 to 08-31-2009
YES
7
5
71.5
Total (2005-2009)
22
9
40.9
Note: During the initial stage of ADNI1 the NPC had not been established and no autopsies
were performed on the 6 ADNI participants who expired during 2007 and the first half of 2008.
Autopsy rate = number of brain autopsies/total number of ADNI participants who died.
7.3.2. Neuropathologic Assessment of ADNI Participants at Autopsy: Brain tissue from 9 ADNI
participants has been received and all have been neuropathologically assessed by the ADNI1-NPC (Table 2).
Seven men and 2 women have come to autopsy. The mean age at expiration of the men was 82 (range: 65 to
89) years and the women were aged 79 and 85 years. One participant was an African American; the remainder
were white. The mean postmortem interval (time from death to snap freezing of brain tissue) was 6.7 hours
(range: 2.8 to 16.0). Of the 9 autopised cases, the clinical diagnoses at the time of expiration were DAT in 6
and MCI in 3. All 9 cases had AD according to the neuropathologic diagnostic criteria of Khachaturian,
CERAD, and NIA-Reagan Institute. In addition, 3 of the 9 cases (33%) had sufficient alpha-synuclein pathology
(Lewy bodies and Lewy neurites) to fulfill McKeith et al criteria for the neuropathologic diagnosis of DLB
(neocortical stage) [7,8]. Other comorbid pathologies were: argyrophilic grain disease (4R tauopathy) (n=1)
[14] and TDP-43 proteinopathy in the medial temporal lobe (n=1) [16]. The identification of cases with
comorbid molecular pathology is important for determining the potential contribution of other molecular
pathologies to the clinical phenotype. The presence of cases with an additional molecular pathology in this
sample, although representative of other larger series, indicates that the contribution of tauopathy, alphasynucleinopathy, and TDP-43 proteinopathy, and possibly other proteinopathies, will need to be assessed in
the ADNI series as more cases come to autopsy. If the neuropathologic sample is representative of the total
ADNI cohort of dementia patients, these preliminary data indicate widespread comorbidity which may
contribute to variance in the data obtained by the different Cores.
Table 2. Clinical and Neuropathologic Diagnoses at Expirartion
Neuropathologic diagnosis [N (%)]
Clinical
diagnosis
AD
AD + DLB
AD + AGD
AD + TDP-43
TOTAL (%)
DAT
2 (22)
3 (33)
1(11)
1 (11)
7 (78)
MCI
2 (22)
0 (0)
0(11)
0 (0)
2 (22)
Normal
0 (0)
0 (0)
0(11)
0 (0)
0 (0)
TOTAL (%)
4 (44)
3 (33)
1(11)
1 (11)
9 (100)
Note: N, number of ADNI cases. AD, Alzheimer disease; AGD, argyrophilic grain disease; DAT, dementia of
the Alzheimer type; DLB, dementia with Lewy bodies; MCI, mild cognitive impairment; TDP-43, TDP-43
proteinopathy in the medial temporal lobe. Mild small vessel disease (arteriolosclerosis and cerebral amyloid
angiopathy) was a feature of all cases but none had infarcts.
7.3.3. Neuropathologic Assessment of Pittsburgh Compound-B (PIB) Amyloid Imaged Imaged
Participant with Clinical and Biomarker Data. The longitudinal assessment of ADNI participants facilitates
the collection of multimodal data (clinical, neuropsychology, structural and funtional neuroimaging including
PIB-PET, CSF biomarkers, and neuropathology) which will help to determine the temporal sequence of the
pathophysiological changes in the AD brain. As an example of the explanatory power of this approach, we
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
have followed up one informative case that has come to autopsy [20]. The participant was longitudinally
assessed at the Washington University School of Medicine Memory and Aging Center who had serial clinical
and psychometric assessments over 6 years, CSF collection for biomarker assays, and PIB-PET amyloid
imaging prior to coming to autopsy. The participant was an 85-year old individual with a Clinical Dementia
Rating (CDR) of 0 (cognitively normal) at initial and his next 4 annual assessments (Fig. 1). Between ages 89
and 90 years, the clinical and psychometric findings and the CSF biomarkers indicated an AD phenotype but
PIB-PET amyloid imaging was negative. At autopsy there were foci of frequent neocortical diffuse Aβ plaques
(Fig. 2), sufficient to fulfill Khachaturian diagnostic criteria for AD, but few neuritic plaques or neurofibrillary
tangles. Postmortem biochemical analysis of the cerebral tissue confirmed that PIB-PET-binding was below
the level needed for in vivo detection (Fig. 3). This case study provides evidence that in some individuals, low
CSF Aβ42 may indicate the presence of Aβ aggregates, predominantly in diffuse plaques, in the absence of a
significant amount of Aβ deposits in fibrils. Clinical, cognitive, and CSF markers consistent with AD may
precede the presence of sufficient fibrillar Aβ plaques to allow detection by amyloid-imaging agents such as
PIB.
Fig. 1. Clinical and cognitive course of PIB-PETnegative case. T-1, first clinical assessment; CDR,
Clinical Dementia Rating; z score, the means of
four neuropsychological test composites: episodic
memory, semantic memory, working memory, and
visuospatial ability [20].
Fig. 2. Numerous diffuse Aβ plaques (arrowhead), but
only few ring-with-core plaques (arrow) and modest
cerebral amyloid angiopathy (double arrowhead) in the
frontal lobe [20].
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Fig. 3. Fluorescent Aβpleated sheet stains
label a spectrum of Aβ
structures in the frontal
lobe (upper panels) and
CA1
subfield
of
hippocampus
(lower
panels) of postmortem
brain
of
PIB-PETamyloid-negative
participant. Amyloid is
visible using 6-CN-PIB
and
X-34,
highly
fluorescent derivatives
of PIB and Congo red,
respectively;
the
monoclonal
antibody
6E10, targeting amino acids 1-16 (N-terminus) of Aβ identifies similar structures (asterisk) [20].
(Photomicrographs courtesy of Milos Ikonomovic and William Klunk, University of Pittsburgh)
This study highlights the utility of combining longitudinal clinical and psychometric evaluations and
independently obtained CSF biomarkers and amyloid imaging with neuropathological examination. Similar to
other amyloid-binding compounds, PIB binds strongly to fibrillar Aβ in compact/cored plaques and cerebral
amyloid angiopathy, and only weakly to amorphous cortical Aβ plaques, so it may not be sensitive enough to
detect amyloid lesions associated with prodromal and very early symptomatic AD which may be characterized
predominantly by diffuse Aβ plaques [24]. This case had an APOE ε3:ε3 genotype.
7.4. Methods:
In ADNI2, the Neuropathology Core will enhance the existing infrastructure put in place during ADNI1 to: (1)
improve the overall autopsy rate at ADNI sites; (2) improve the neuropathologic assessment of cases to
include site, size, and nature of vascular lesions, and to assess the presence of recently described
proteinopathies, including TDP-43 proteinopathy; and (3) facilitate multidisciplinary research on those cases
that have come to autopsy.
In this application, we use conservative death rate estimates because the demanding ADNI2 protocols may
result in healthier participants. We thus assume annual death rates of 1% for non-demented ADNI2
participants, 1% for MCI individuals, and 5% for AD individuals. From surviving ADNI1 and newly enrolled
participants, and assuming an autopsy rate of 75% over the proposed funding cycle for ADNI2 (2009-14), we
anticipate that at the minimum there will be autopsies in at least 7 non-demented controls, 15 MCI individuals,
and 38 AD individuals during the five year period of this project. .
7.4.1. Provide training materials and protocols to assist clinicians at ADNI sites in obtaining voluntary
consent for brain autopsy in ADNI participants [SA1]. As there may be personnel changes over time, there
is a continuing need to monitor each site to ensure that training and protocols for obtaining autopsies are in
place, so it is essential to maintain a dedicated Coordinator to ensure these functions are performed over the
period of the grant.
To obtain consent for an autopsy, the ADNI physician will lead a discussion about autopsy with all
participants (demented and non-demented) at their initial assessment (study partners and families are
welcomed in the discussion and required for AD participants). There are 3 objectives of the discussion: 1) to
convey information about the value of brain autopsy in confirming the clinical diagnosis and advancing
knowledge regarding MCI and AD; 2) to initiate consideration of the individual’s wishes concerning an autopsy;
and 3) to answer questions, misconceptions, or concerns about autopsy. The involvement of the physician in
these discussions emphasizes the importance of autopsy. The discussions are repeated on an annual basis if
the individual has not decided about autopsy, but are terminated once a decision is reached. There is no
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
pressure on an individual to decide; they are encouraged to involve family members, clergy, physicians, or
other appropriate persons in their decision-making. Participants are assured that a decision not to have
autopsy in no way jeopardizes their research participation or any other patient rights.
When voluntary consent is granted, more detailed information is provided about procedures to follow at time
of death, including telephone numbers to call and other guidelines (sample forms available in manual appendix
and on line). Participants are strongly encouraged to share this information with next-of-kin, Durable Power of
Attorney (DPOA) and private physicians. In many states, final legal authorization by the Legally Authorized
Representative (LAR) or next-of-kin must be obtained at time of death.
Each ADNI site is encouraged to establish an autopsy coordinator (typically a research nurse or coordinator)
who processes the autopsy consent, provides information as needed, and monitors the need to update any
information (e.g., change in residence) at the ADNI participant’s longitudinal assessments. The coordinator
also will develop procedures for that site to facilitate autopsies outside of usual hours (e.g., evenings and
weekends). The actual procedures are expected to vary in accordance with local needs and resources (one
model used by many ADCs is to provide 24-hour telephone access).
At the time of death, the autopsy coordinator (or a suitable representative) facilitates arrangements to ensure
the completion of the autopsy. The coordinator will notify the ADNI2-NPC, which in turn verifies that the site
neuropathologist has the dissection protocol and necessary materials to send the requisite tissue to the
ADNI2-NPC
The ADNI2-NPC, in addition to instructing site personnel at each ADNI Steering Committee Meeting in these
procedures, will be available at any time to answer questions. Contact information, including a 24-hour pager,
is available. At ADNI2 sites that already have ADRC/ADC neuropathology services; these will continue to
follow their own existing protocols. For ADNI2 sites that do not have established neuropathology services,
transportation costs from point of death to the autopsy suite, costs of the autopsy procedure, and shipment of
materials will be covered by ADNI2-NPC so that the decedent’s family and the individual ADNI site do not incur
extra expense.
Once the Participant has given consent (provisional or otherwise) the Acknowledgement of Autopsy
Authorization letter and supporting documentation will be sent to the following: Participant and/or family and/or
applicable other (e.g. Power of Attorney), nursing home, funeral home/transport service (as requested), and
the Participant’s private physician (as requested).
7.4.2. Maintain a central laboratory to provide uniform neuropathological assessments in all autopsied
ADNI1participants in accordance with standard criteria and to promote clinical-neuroimagingneuropathological correlations [SA2]. Where possible, each center will undertake its own brain assessment
and forward a standard set of fixed tissue blocks or sections and frozen tissue to ADNI2-NPC (see below). For
sites that do not routinely undertake neuropathologic studies, a separate brain removal protocol is available.
Financial Assistance with Block Sampling, Preservation, and Shipping Costs, ADNI2-NPC will fund all costs
in shipping frozen and fixed tissue samples to St. Louis. To assist participating centers and neuropathologists
with the costs of obtaining frozen tissue blocks and/or formalin-fixed paraffin wax-embedded tissue the
following costs will be reimbursed, if requested: (1) harvesting of frozen tissue and/or formalin-fixed paraffin
wax-embedded tissue blocks (*see list of brain regions below) $300; (2) harvesting formalin-fixed paraffin waxembedded tissue sections or frozen sections (*see list of brain regions below) $100.
Resources to defray the costs of sampling, tissue, processing, administration, and transport will be made
available to each center already undertaking neuropathology. These resources are to facilitate the provision of
the standard set of blocks for ADNI2-NPC. To minimize the burden on participating centers, formalin-fixed,
paraffin wax-embedded tissue blocks from the following 16 areas from the left cerebrum will be forwarded to
ADNI2-NPC: middle frontal gyrus, superior and middle temporal gyri, inferior parietal lobe (angular gyrus),
occipital lobe to include the calcarine sulcus and peristriate cortex, anterior cingulate gyrus at the level of the
genu of the corpus callosum, posterior cingulate gyrus and precuneus at the level of the splenium, amygdala
and entorhinal cortex, hippocampus and parahippocampal gyrus at the level of the lateral geniculate nucleus,
striatum (caudate nucleus and putamen) at the level of the anterior commissure, lentiform nuclei (globus
pallidus and putamen), thalamus and subthalamic nucleus, midbrain, pons, medulla oblongata, cerebellum with
dentate nucleus, and spinal cord.
In the unusual situation where it is impractical to forward a tissue block (e.g., if the block is used for
stereology), 10 paraffin wax sections (4-8 μm) from each block will be provided to ADNI2-NPC for systematic
neuropathology and diagnosis.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
To provide tissue for biochemical studies and to advance the aims of the Biomarkers Study, snap frozen
tissue will be dissected, frozen, and sent to ADNI2-NPC. The following coronal hemibrain slices (0.5 to 1cm
thick), where possible, will be taken: (1) Frontal lobe to include striatum; (2) Frontal and temporal lobe at the
level of the mamillary body; (3) Temporal and parietal lobes at the level of the lateral geniculate nucleus; and
(4) Occipital lobe to include the calcarine sulcus.
Histology: In all cases, the following stains will be performed at the ADNI2-NPC lab on the blocks indicated
above, and/or as requested by the neuropathologist: hematoxylin and eosin and modified Bielschowsky silver
impregnation. Routine immunohistochemistry will be performed using the following antibodies: ubiquitin (Dako),
tau (PHF1 and/or AT8), β-amyloid (4G8 and/or 10D5), and α-synuclein (LB509). In cases with ubiquitinpositive inclusions, the following additional IHC will be performed: TDP-43 and FUS [17-19].
Histology Review: Dr. Cairns reviews the histological slides in a systematic manner. The data are entered
into the NACC Neuropathology Data Form and transmitted to the Biostat and Informatics Cores (Cores 8 and
9) at the ADNI Co-ordinating Center. The NACC Neuropathology Protocol is included in the Appendix. The
final neuropathologic diagnosis and neuropathologic report will be forwarded to ADNI for entry into the central
database (LONI) and to the center that made available the tissue.
Neuropathologic Assessment and Diagnostic Criteria: The operational criteria for the classification of AD
and other pathologies defined by NACC will be applied to all ADNI2-NPC cases (and are currently applied to
all WU ADRC cases) 1-15]. The neuropathological diagnosis will be determined by Dr. Cairns and Dr Robert
Schmidt (Division of Neuropathology, WUSTL) using consensus neuropathologic criteria for AD, and for nonAD disorders. The NACC Neuropathology Form includes an entry for the diagnosis of AD by each of the 3 sets
of criteria: CERAD, NIA-Reagan, and Khachaturian. ADNI2-NPC cases thus will be diagnosed in accordance
with each of these criteria, as no consensus currently exists in favor of one set in relation to the others
(particularly for the incipient stages of AD addressed by the ADNI study). This will allow investigators maximal
utility in applying the neuropathological diagnoses most appropriate to their research aims [20].
7.4.3. Maintain a state-of-the-art resource for fixed and frozen brain tissue obtained from autopsied
ADNI participants to support ADNI's biomarker studies and make available to ADNI-approved
investigators access to the tissue and data for research purposes [SA3]. The ADNI1-NPC has already
purchased a -80°C freezer with 23 cubic feet capacity with CO2 back-up and telephone alarm and we envision
that this will be adequate for the projected number of harvested cases in ADNI2. ADNI2-NPC will maintain a
neuropathology computerized database in concert with Biostatistics and the Clinical Core of the Washington
University Neuroscience Blueprint Interdisciplinary Center Core (P30-NS057105). Information stored will
include macroscopic images of fresh and fixed brain, demographic data, diagnoses, semi-quantitative
morphometric data, neuropathology reports (in collaboration with Dr Schmidt, Chair, Division of
Neuropathology), bibliographic information, and data relevant to Core tissue banking activities. In addition,
neuropathology data will be transferred, after Biostatistics Core quality control and validation, to the National
Alzheimer Coordinating Center (NACC), University of Washington, Seattle, WA (U01-AG016976) and to the
ADNI2 Coordinating center for upload onto the LONI database.
Although no tissue requests have been obtained to date, this reflects the small number of brain autopsies
performed (n=9). As the autopsy rate increased to >70% in the preceding year, we envisage that the numbers
of autopsies will increase during the period of the GO grant (2 years) and the proposed ADNI2 (4 years) and
generate sufficient samples for multi-modal studies, similar to one which we have already undertaken in a
single case [20]. To ensure that all participating sites are aware of the archived ADNI tissue, each site will be
contacted individually and annually to alert each site of this resource and to solicit feedback on the
Neuropathology Core. In addition, a link to the Neuropathology Core will be made available on the ADNI
website which describes the procedure for qualified investigators to obtain tissue samples (see 7.4.4).
7.4.4. Interact with ADNI's Data Coordinating Center to ensure appropriate entry of the Core's data into
ADNI's database, promote data sharing and collaborative research, and integrate the ADNI2-NPC with
all ADNI components to support its administration, operations, and progress toward goals [SA4].
Data generated by the ADNI2-NPC will be transmitted securely to the ADNI2 Co-ordinating Center for storage,
management and distribution according to the ADNI2 procedures. These de-identified data will be available
with the relevant clinical, biological and imaging data on the Co-ordinating Center’s web site (LONI) and will be
made available to all qualified scientists in the world who request a password, without embargo.
The process by which investigators request access to ADNI1 and ADNI2-NPC tissue is based on the
established and successful procedures in place at the WU ADRC. Qualified investigators will initiate requests
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
for ADNI autopsy material by providing basic information (including a 3 page research summary and NIH
biosketch) about their research project to the ADNI2-NPC Tissue Committee (see below). The instructions and
forms are web-based (www.alzheimer.wustl.edu/adrc2/ResourcesDB/Intro.asp) for easy access. Prospective
investigators will be encouraged to consult with Drs. Morris and Cairns. Written reviews of the request from at
least 2 members of the ADNI2-NPC Tissue Committee, chaired by Dr. Cairns, or other experts recruited for a
particular protocol will be provided for discussion and approval by email vote of the Tissue Committee
conducted monthly or as requests dictate. The Tissue Committee will forward its recommendations to the ADNI
Executive Committee (see below) for final approval. The criteria used by reviewers will be: scientific merit,
feasibility, appropriateness of principal investigator qualifications, burden on ADNI samples, and
appropriateness to ADNI goals/themes.
ADNI2 Neuropathology Core Tissue Committee
ADNI2 Executive Committee
John Morris, WUSTL, MO
Michael Weiner, UC San Francisco/SFVAMC, CA
Nigel Cairns, WUSTL, MO
Laurel Beckett, UC Davis, CA
Eileen Bigio, Northwestern U, IL
Clifford Jack, Mayo Clinic, MN
Dennis Dickson, Mayo Clinic, FL
William Jagust, UC Berkeley, CA
John Trojanowski, UPenn, PA
John Morris, WUSTL, MO
Ron Petersen, Mayo Clinic, MN
Ron Thomas, UC San Diego, CA
Arthur Toga, UC Los Angeles, CA
John Trojanowski, UPenn, PA
7.5. BIBLIOGRAPHY.
1. Khachaturian ZS. Diagnosis of Alzheimer's disease. Arch Neurol 42: 1097-1105, 1985.
2. Mirra SS, Heyman A, McKeel D, Sumi SM, Crain BJ, Brownlee LM, Vogel FS, Hughes JP, van Belle G and
Berg L. The Consortium to Establish a Registry for Alzheimer's Disease (CERAD). Part II. Standardization
of the neuropathologic assessment of Alzheimer's disease. Neurology 41: 479-486, 1991.
3. Consensus recommendations for the postmortem diagnosis of Alzheimer's disease. The National Institute
on Aging, and Reagan Institute Working Group on Diagnostic Criteria for the Neuropathological
Assessment of Alzheimer's Disease. Neurobiol Aging 18: S1-S2, 1997.
4. Braak H and Braak E. Neuropathological stageing of Alzheimer-related changes. Acta Neuropathol (Berl)
82: 239-259, 1991.
5. Braak H, Alafuzoff I, Arzberger T, Kretzschmar H and Del Tredici K. Staging of Alzheimer diseaseassociated neurofibrillary pathology using paraffin sections and immunocytochemistry. Acta Neuropathol
(Berl) 112: 389-404, 2006.
6. Roman GC, Tatemichi TK, Erkinjuntti T, Cummings JL, Masdeu JC, Garcia JH, Amaducci L, Orgogozo JM,
Brun A, Hofman A, Moody DM, Obrien MD, Yamaguchi T, Grafman J, Drayer BP, Bennett DA, Fisher M,
Ogata J, Kokmen E, Bermejo F, Wolf PA, Gorelick PB, Bick KL, Pajeau AK, Bell MA, DeCarli C, Culebras
A, Korczyn AD, Bogousslavsky J, Hartmann A and Scheinberg P. Vascular Dementia - Diagnostic-Criteria
for Research Studies - Report of the Ninds-Airen International Workshop. Neurology 43: 250-260, 1993.
7. McKeith IG, Galasko D, Kosaka K, Perry EK, Dickson DW, Hansen LA, Salmon DP, Lowe J, Mirra SS,
Byrne EJ, Lennox G, Quinn NP, Edwardson JA, Ince PG, Bergeron C, Burns A, Miller BL, Lovestone S,
Collerton D, Jansen EN, Ballard C, De Vos RA, Wilcock GK, Jellinger KA and Perry RH. Consensus
guidelines for the clinical and pathologic diagnosis of dementia with Lewy bodies (DLB): report of the
consortium on DLB international workshop. Neurology 47: 1113-1124, 1996.
8. McKeith IG, Dickson DW, Lowe J, Emre M, O'Brien JT, Feldman H, Cummings J, Duda JE, Lippa C, Perry
EK, Aarsland D, Arai H, Ballard CG, Boeve B, Burn DJ, Costa D, Del Ser T, Dubois B, Galasko D, Gauthier
S, Goetz CG, Gomez-Tortosa E, Halliday G, Hansen LA, Hardy J, Iwatsubo T, Kalaria RN, Kaufer D,
Kenny RA, Korczyn A, Kosaka K, Lee VM, Lees A, Litvan I, Londos E, Lopez OL, Minoshima S, Mizuno Y,
Molina JA, Mukaetova-Ladinska EB, Pasquier F, Perry RH, Schulz JB, Trojanowski JQ and Yamada M.
Diagnosis and management of dementia with Lewy bodies: third report of the DLB Consortium. Neurology
65: 1863-1872, 2005.
9. Braak H, Ghebremedhin E, Rub U, Bratzke H and Del Tredici K. Stages in the development of Parkinson's
disease-related pathology. Cell Tissue Res 318: 121-134, 2004.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
10. Gelb DJ, Oliver E and Gilman S. Diagnostic criteria for Parkinson disease. Arch Neurol 56: 33-39, 1999.
11. Braak H and Braak E. Argyrophilic grains: characteristic pathology of cerebral cortex in cases of adult
onset dementia without Alzheimer changes. Neurosci Lett 76: 124-127, 1987.
12. Braak H and Braak E. Cortical and subcortical argyrophilic grains characterize a disease associated with
adult onset dementia. Neuropathol Appl Neurobiol 15: 13-26, 1989.
13. Togo T, Sahara N, Yen SH, Cookson N, Ishizawa T, Hutton M, de Silva R, Lees A and Dickson DW.
Argyrophilic grain disease is a sporadic 4-repeat tauopathy. J Neuropathol Exp Neurol 61: 547-556, 2002.
14. Cairns NJ, Bigio EH, Mackenzie IR, Neumann M, Lee VM, Hatanpaa KJ, White CL, III, Schneider JA,
Grinberg LT, Halliday G, Duyckaerts C, Lowe JS, Holm IE, Tolnay M, Okamoto K, Yokoo H, Murayama S,
Woulfe J, Munoz DG, Dickson DW, Ince PG, Trojanowski JQ and Mann DM. Neuropathologic diagnostic
and nosologic criteria for frontotemporal lobar degeneration: consensus of the Consortium for
Frontotemporal Lobar Degeneration. Acta Neuropathol 114: 5-22, 2007.
15. Mackenzie IR, Neumann M, Bigio EH, Cairns NJ, Alafuzoff I, Kril J, Kovacs GG, Ghetti B, Halliday G, Holm
IE, Ince PG, Kamphorst W, Revesz T, Rozemuller AJ, Kumar-Singh S, Akiyama H, Baborie A, Spina S,
Dickson DW, Trojanowski JQ and Mann DM. Nomenclature for neuropathologic subtypes of frontotemporal
lobar degeneration: consensus recommendations. Acta Neuropathol 117: 15-18, 2009.
16. Amador-Ortiz C, Lin WL, Ahmed Z, Personett D, Davies P, Duara R, Graff-Radford NR, Hutton ML and
Dickson DW. TDP-43 immunoreactivity in hippocampal sclerosis and Alzheimer's disease. Ann Neurol 61:
435-445, 2007.
17. Cairns NJ, Neumann M, Bigio EH, Holm IE, Troost D, Hatanpaa KJ, Foong C, White CL, III, Schneider JA,
Kretzschmar HA, Carter D, Taylor-Reinwald L, Paulsmeyer K, Strider J, Gitcho M, Goate AM, Morris JC,
Mishra M, Kwong LK, Stieber A, Xu Y, Forman MS, Trojanowski JQ, Lee VM and Mackenzie IR. TDP-43 in
familial and sporadic frontotemporal lobar degeneration with ubiquitin inclusions. Am J Pathol 171: 227240, 2007.
18. Neumann M, Rademakers R, Roeber S, Baker M, Kretzschmar HA and Mackenzie IR. Frontotemporal
lobar degeneration with FUS pathology. Brain 2009.
19. Neumann M, Roeber S, Kretzschmar HA, Rademakers R, Baker M and Mackenzie IR. Abundant FUSimmunoreactive pathology in neuronal intermediate filament inclusion disease. Acta Neuropathol 2009. In
press.
20. Cairns NJ, Ikonomovic MD, Benzinger T, Storandt M, Fagan AM, Shah A, Taylor-Reinwald L, Carter D,
Felton A, Holtzman DM, Mintun MA, Klunk WE and Morris JC. PIB-PET imaging of cerebral Aβ may lag
clinical, cognitive, and CSF markers of Alzheimer's disease: a case report. Arch Neurol 2009. In press.
21. Mirra SS, Gearing M, McKeel DW Jr, Crain BJ, Hughes JP, van Belle G, Heyman B and Alafuzoff I. Interlaboratory comparison of neuropathology assessments in Alzheimer's disease: a study of the Consortium
to Establish a Registry for Alzheimer's Disease (CERAD). J Neuropathol Exp Neurol 53: 303-315, 1994.
22. Alafuzoff I, Pikkarainen M, Al Sarraj S, Arzberger T, Bell J, Bodi I, Bogdanovic N, Budka H, Bugiani O,
Ferrer I, Gelpi E, Giaccone G, Graeber MB, Hauw JJ, Kamphorst W, King A, Kopp N, Korkolopoulou P,
Kovacs GG, Meyronet D, Parchi P, Patsouris E, Preusser M, Ravid R, Roggendorf W, Seilhean D,
Streichenberger N, Thal DR, and Kretzschmar H. Interlaboratory comparison of assessments of Alzheimer
disease-related lesions: a study of the BrainNet Europe Consortium. J Neuropathol Exp Neurol 65: 740757, 2006.
23. Holmes C, Boche D, Wilkinson D, Yadegarfar G, Hopkins V, Bayer A, Jones RW, Bullock R, Love S, Neal
JW, Zotova E and Nicoll JA. Long-term effects of Aβ42 immunisation in Alzheimer's disease: follow-up of a
randomised, placebo-controlled phase I trial. Lancet 372: 216-223, 2008.
24. Price JL McKeel DW Jr, Buckles VD, Roe CM, Xiong C, Grundman M, Hansen LA, Petersen RC, Parisi JE,
Dickson DW, Smith CD, Davis DG, Schmitt FA, Markesbery WR, Kaye J, Kurlan R, Hulette C, Kurland BF,
Higdon R, Kukull W and Morris JC. Neuropathology of non-demented aging: presumptive evidence for
preclinical Alzheimer disease. Neurobiol Aging 30: 1026-1036.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Core: 8
Title of Core (not to exceed 81 spaces): Biostatistics Core
Core Leader: Beckett, Laurel, A.
Position/Title: Professor, University of California, Davis
Department, service, laboratory, or equivalent: Public Health Sciences
Mailing Address:
University of California, Med Sci 1-C
One Shields Avenue
Davis, CA 95616
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
The Biostatistics Core will ensure sound designs and statistical analyses for the scientific aims of ADNI-2.
Aim 1. To provide analytic support for the planning and design of ADNI-2. Aim 2. To carry out interim and
final analyses of all ADNI data to address research questions on clinical change, imaging measures,
biomarkers, their relationship to each other and to clinical trials. Aim 3. To participate in the ongoing ADNI-2
operations. Aim 4. To provide intellectual leadership and foster communication among academic and
industry biostatisticians interested in ADNI. Aim 5. To develop new biostatistical methodology for the
analysis of high-dimensional imaging data, modeling of multi-process correlation, assessment of surrogate
marker potential for AD studies, and other analysis questions that arise in pursuing ADNI-2 research goals.
In ADNI-1 we demonstrated the relationship of imaging and fluid biomarker summaries to a pattern of
cognitive decline and brain atrophy across normal controls (NC), mild cognitive impairment (MCI) and
Alzheimer disease (AD). We also compared performance of biomarkers in detecting a shift in rate of change
and showed that the best PET and MRI summary measures have much better performance characteristics
than cognitive tests for detection of difference in rates of change. We showed that fluid and imaging
biomarker summaries predict conversion from MCI to AD. Even in NC with little if any clinical decline, some
people show PET and MRI change that is related to other imaging and fluid biomarker levels at baseline. In
ADNI-2 we will extend these analyses over longer follow-up, more people, and a more complete clinical
diagnostic spectrum to advance our understanding of fluid and imaging biomarkers as predictive, diagnostic
and progression markers from normal controls (NC) to mild cognitive impairment (MCI), including transition
through early MCI (EMCI), to AD. We will also examine performance of biomarkers as potential clinical trial
entry criteria, stratification tools, covariates, and supplemental or surrogate endpoints. We will develop new
statistical methods needed to address high-dimensional data on multiple trajectories over time.
Successful biostatistical analysis will lead to a more complete description and understanding of the complex
interplay of neurobiological processes leading to AD and cognitive decline including the role fluid and
imaging biomarkers can play in tracking the process. These analyses also have the potential to improve
design and analysis of clinical trials for the prevention and treatment of AD.
Regents of the University of California (UC Davis)
DUNS:
047120084
Street 1:
Dept. of Public Health Sciences
City:
Davis
Province:
Street 2:
County:
Yolo
USA
CA-001
Country:
One Shields Avenue, Med. Sci. Bldg. 1C
State:
Zip/Postal Code:
CA
95616
DUNS:
Street 1:
Street 2:
City:
Province:
County:
Country:
State:
Zip/Postal Code:
PHS 398 (Rev. 11/07)
Page 2
Form Page 2
Weiner, Michael
Name
Organization
Role on Project
Beckett, Laural
LABECKETT
UC-Davis
PI
Harvey, Danielle
DJHARVEY
UC-Davis
Co-Investigator
Name
Organization
Role on Project
No
Yes
Cell Line
PHS 398 (Rev. 11/07)
Page 3
Weiner, Michael
8. BIOSTATISTICS CORE (Co-leaders: Beckett and Harvey):
8.1. Specific Aims:
The overall goal for the Biostatistics Core of the Alzheimer’s Disease (AD) Neuroimaging Initiative renewal
(ADNI2) is to ensure that sound designs and statistical analyses are used to address the scientific aims of
ADNI2. We will build on our activities during the first ADNI funding period (ADNI1) and on our planned activities
for the recently funded Grand Opportunities (GO) grant. Thus, the Biostatistics Core for ADNI2 has the
following aims:
Aim 1. To provide analytic support for the planning and design of ADNI2, including preparatory phase and
pilot data analysis and the determination of appropriate sample size, and study design specifications (e.g.
stratification and enrichment of target populations), quality control methods, and analysis strategies to achieve
estimation and hypothesis testing objectives.
Aim 2. To carry out interim and final analyses of all ADNI data to address research questions on clinical
change, imaging measures, biomarkers, relationship of ADNI measurements to each other, predictive value,
implications for clinical trials, and optimization of prediction and design, and to participate in the presentation
and dissemination of study results at research meetings and conferences and in scientific publications.
Aim 3. To participate in the ongoing ADNI2 operations, including: close collaboration with the leadership of
the clinical, MRI, PET, biomarker, genetics and informatics cores; participation in Executive Committee
conference calls and meetings; and generation of regular reports for ADNI operations and coordination among
ADNI Cores.
Aim 4. To provide intellectual leadership and foster communication among academic and industry
biostatisticians interested in ADNI data analysis, via a monthly ADNI biostatistics conference call, web site,
email, web-based training sessions, and documentation of routine procedures for downloading, merging,
summarizing and reporting of ADNI data in SAS and R, including procedures for the automatic generation of
formatted summary reports.
Aim 5. To develop new biostatistical methodology for the analysis of high-dimensional imaging data,
modeling of multi-process correlation, assessment of surrogate marker potential for AD studies, and other
analysis questions that arise in pursuing ADNI2 research goals.
To accomplish these Aims, we have formed a Biostatistics Core led by two statisticians at the University of
California, Davis (UCD), funded directly through the Core: Laurel Beckett, Ph.D., Core Director (and Director of
the ADNI1 Biostatistics Core) and Danielle Harvey, Ph.D., Co-Director, and also a key participant in the ADNI1
Core. Drs. Beckett and Harvey bring expertise in longitudinal studies, Alzheimer’s disease research, multidimensional data reduction, and neuroimaging data. In addition, our Core will include Anthony Gamst, Ph.D.,
and Michael Donohue, Ph.D., from University of California, San Diego (UCSD), funded through the Clinical
Core. Drs. Gamst and Donohue will form a liaison to the Clinical Core’s data collection and quality control
functions; in addition, they bring expertise in clinical trials, cognitive outcomes, survival analysis and
conversion endpoints. These four statisticians have worked closely together for almost the entire funding
period of ADNI1 to develop statistical strategies and carry out analyses of ADNI data. We will also work closely
with other statisticians both in university sites (for example, Philip Insel, University of California, San Francisco
(UCSF)) and in industry, continuing our practice of collaboration to provide the best possible advice and
support for ADNI research.
Our experienced biostatistics group will thus enable the ADNI2 researchers to advance our understanding of
fluid and imaging biomarkers as predictive, diagnostic and progression markers from normal controls (NC) to
mild cognitive impairment (MCI), including transition through early MCI (EMCI), to AD.
8.2.1. Brief overview: Alzheimer’s disease is the most common cause of dementia in the elderly and a
substantial burden to patients, caregivers, and the health care system [1]. Approved treatments are few and of
limited efficacy, serving mostly to slow or delay progression, not to cure the disease, despite significant
research efforts by NIH and the pharmaceutical industry. A major barrier is that clinical disease assessment
yields measures of limited value for characterizing diagnosis and quantifying disease progression and drug
efficacy: clinical measures have substantial between- and within-person variation [2], and they likely lag far
behind the underlying pathology onset and progression [3]. As we present in greater detail in the Introduction
and other Cores, evidence to date suggests that the neurobiological development of AD affects multiple
systems and multiple parts of the brain. Researchers have proposed a number of approaches for in vivo
biomarkers based on either fluid samples or neuroimaging, to ascertain the current status and track the
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
progression of AD-related brain change. Many published papers have shown association with disease
progression for potential biomarkers; examples include serum and CSF markers for Aβ42, tau, and tau
phosphorylated at residue 181 (P tau) [4]; PET imaging for amyloid [5] and for glucose metabolism [6]; and
structural MRI [7].
The primary goal of ADNI1 was to validate and compare biomarkers for potential use as outcome measures
in clinical trials; thus three quarters of the ADNI1 participants were enrolled from narrowly defined amnestic
MCI and mild-to-moderate AD groups, and the remaining participants were clinically defined NCs [8]. All
participants had repeated clinical evaluation including cognitive and functional assessments, neurological
examination, and MRI, approximately every six months for two years. In addition, about half of the participants
had CSF samples at baseline and 12 months, and, independently selected, about half had FDG-PET.
Supplemental funding later supported PIB imaging for a subset of the cohort. Five MRI laboratories and three
PET laboratories then developed summary measures on the sequential images. This process yielded a rich,
multidimensional dataset with longitudinal data on many candidate biomarker summaries and on multiple
aspects of clinical outcome, finally allowing the testing of many existing hypotheses about markers for AD [9].
ADNI1 results to date (see other Cores and a brief overview below) suggest that a number of these candidate
biomarkers, considered individually, had good precision for characterizing change and correlated with clinical
change in the MCI and AD groups, thus supporting their potential for use in clinical trials. Based on the
literature and on findings from ADNI1, the ADNI team has developed a series of hypotheses about the
relationships among the imaging and fluid biomarkers, the sequence in which their trajectories unfold, and their
prognostic relationship to clinical outcomes (See Introduction and the Aims for each Core).
8.2.2. Significance of biostatistics to proposed research: The statistical challenges of ADNI2 in addressing
these scientific aims and hypotheses will be substantial, beginning with the sheer volume of data when ADNI1,
GO and ADNI2 are considered together. Careful distinction between a priori hypotheses and data-driven or
hypothesis-generating methods is essential, using modern methods for cross-validation and control of false
discovery rates in the latter setting [10]. Integration of biostatistical support into the other Cores is critical to
ensure that this distinction is clear and that researchers developing data-driven summaries or exploratory
prediction have access to best statistical practices.
A second challenge is accurate quantitative characterization of trajectories of biomarkers and clinical
measures, the foundation step for testing hypotheses and developing new measures about the neurobiology of
onset and early progression of AD. The measures we study pose difficulties for longitudinal analysis including
ceiling and floor effects in the pre-clinical and late stages; substantial within-person variation for some
measures; some decidedly non-normal distributions; missing data and unequal lengths of follow-up; and
potential non-linearity of trajectories as the follow-up period lengthens. The standard linear mixed models
approach [11] is likely to fail for at least a substantial number of the measures of interest, and even the usual
alternative, generalized linear models, may not provide adequate fits to the data.
Once we have accurately characterized the trajectories of individual biomarkers, we face a third statistical
challenge in addressing the core neuroscience questions: what is the sequence of the different changes, and
which measures then serve to predict future changes, at each stage? These questions will require models of
change that link multiple trajectories, with many potential predictors and multiple outcome measures of change.
ADNI biostatisticians have demonstrated that we can gain substantial power by modeling more than one
trajectory together [12]. But the gains in power come at a cost of complexity, and possibly higher dependence
on model assumptions such as normality and equality of variances. These questions need further study,
especially in the context of the challenges in modeling individual trajectories. Statisticians have considered
such questions in the context of growth and development, for example [13], but with smaller numbers of
predictors and outcome measures.
A fourth statistical challenge then arises: to use the results of our models to develop improved clinical trial
designs. Potential uses for markers of early onset and predictors of progression include more precise eligibility
criteria for better targeting; use of markers to stratify in the trial design; use of markers as covariates to reduce
unaccounted for variation that might obscure treatment effects; and use of markers as surrogate or
supplemental outcome measures in trials. Comparison of potential markers for these uses is complicated. A
marker that shows early change and is predictive of future decline might be very promising, but high
measurement variability and error could greatly diminish its usefulness in trials. Thus evaluation of markers’
potential for use in trials must reflect more than one aspect of its performance.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
We summarized four key challenges here, as examples of those we will face with ADNI2. The work of the
Biostatistics Core personnel on ADNI1 has addressed some of these challenges, and has prepared us well to
move to ADNI2 with longer follow-up, larger sample sizes, and a richer set of measures on participants.
The Biostatistics Core has achieved or made highly significant progress on all of the goals of the original
ADNI1 proposal.
8.3.1. Aim 1: To oversee the statistical quality control of the data for the study, including development
of quantitative standards for longitudinal imaging data. We worked with the MRI Core to design and
implement the analysis of the Preparatory Phase short-term stability data (see MRI Core; also Core
Publications 1 and 2, denoted [C1] and [C2]). We also advised the Biomarker Core on the design and analysis
of quality control studies. We developed rigorous strategies for evaluating and comparing the performance of
the candidate regions of interest and imaging summary measures identified and developed by individual labs,
many of whom use data-driven selection rules. To protect against overfitting bias, this strategy relies on
common training and test sets for use in comparisons across multiple labs and measures. A stratified
randomization scheme is used to randomize study participants in a 40/60 split to one of these two sets, with
strata defined by image type, age group and screening diagnostic category. Labs are allowed to use datadriven selection rules on the training set to identify candidate regions and summary measures, but
comparisons of various candidates are made on the test data set. We also developed a leave-k-out scheme
for those who wanted more power, but that would still be comparable across labs using a training set- test set
approach. We implemented these policies through the UCSD biostatistics group and the Clinical Core, so that
the assignment was available at the time of initial data upload to the LONI web site. We then worked closely
with all investigators using data-driven variable selection to ensure that these methods were implemented
correctly.
Dr. Beckett’s and Dr. Harvey’s doctoral student, Hao Zhang, M.S., carried out a detailed voxel-based
analysis of the short-term stability study data from the MRI Preparatory Phase, and developed a statistical
model that captures the key features of short-term variation. Under Zhang’s model, the voxel-level differences
in short-term studies of patients for which there is no reason to suspect clinical change have a marginal
distribution that is a mixture of two Gaussian distributions, both centered at zero but with different variances.
Spatial correlation is similar front-to-back and left-to-right, but somewhat smaller in the vertical direction, and
falls off quite rapidly with distances greater than adjacent voxels (close to but not quite autoregressive).
We worked closely with Dr. Ron Thomas, then Drs. Anthony Gamst and Michael Donohue, at UCSD, to
support their smooth process for clinical data and image and biomarker data upload, facilitate summary
descriptive analysis preparation for the labs and Cores, and transmit queries and potential errors for rapid
corrections in the primary data set. Detailed descriptive summaries allowed several groups to detect problems
in data or uploading.
8.3.2. Aim 2. To carry out statistical analyses to develop improved strategies for data processing for
longitudinal imaging data. Both the MRI and PET Cores conducted longitudinal analyses at the voxel level.
For MRI data, one lab conducted voxel-based morphometry (VBM) using statistical parametric mapping (SPM)
and two other labs tensor-based morphometry (TBM) [C6, C7, C8, and C9.] using in-house developed
software. PET data were analyzed using SPM and stereotactic surface projection (SSP) approaches. Most of
these voxel-based analyses were carried out directly by individual labs and results were passed to the
biostatistics core for additional analysis and comparison (see report for Aim 3, Section 8.3.3). All methods
included appropriate tests for multiple comparisons across voxels, e.g. permutation based tests or false
discovery rate adjustment. Results from individual labs are presented within Section 3.3 from the MRI Core
and Section 4.3 from the PET Core.
As part of the voxel-based analyses of ADNI1 we developed standards for pre-processing, longitudinal
registration, and normalization, in order to enable fair comparisons between imaging methods and processing
techniques. As part of this process we considered such issues as whether it is optimal to use single or groupspecific templates for normalization, and whether all methods should follow the same strict pre-processing
steps. The final standards were defined through regular monthly “voxel-based” conference call discussion and
the subsequent development of written plans jointly worked on by leaders of the voxel-based analysis labs.
To perform group analyses and subsequently determine the relative power for detecting decline between the
VBM and TBM analysis, comparisons were performed at regional and global levels. For regional comparisons
a priori (anatomically) defined regions of interest (ROIs) were used to extract average volumetric
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
TBM/VBM/blood volume change scores. These ROI summaries were then analyzed as univariate summary
statistics by the core biostatistics group (see report for Aim 3, Section 8.3.3). For global level comparisons, a
training-test ROI generation approach was developed and employed. Optimal ROIs were generated from the
training dataset for a particular imaging technique/processing method by comparing controls/AD, controls/MCI
and MCI/AD or comparing baseline to follow-up scans and thresholding at a particular p-value. The suprathreshold regions were used as data-generated ROIs for determining the power to detect 25% longitudinal
changes in the test dataset again through the core biostatistics group. Although there are issues in using pvalue thresholds to generate optimal ROIs in this fashion we found that the optimal pattern was relatively
robust to the choice of threshold value. In general, we found that data-driven ROIs provided considerable
increased power over anatomic ROIs and could potentially be used to provide very powerful clinical markers in
Alzheimer’s drug trials.
8.3.3. Aim 3. To develop statistical models for longitudinal change in potential markers (both imaging
and biological analyte classes of AD biomarkers) and in cognition, to use these models to test the
statistical hypotheses about the relationship between markers and cognition, and to assess the
feasibility and added precision for using these models to design more efficient clinical trials based on
markers as well as clinical measures. We worked with the cores to identify key summary measures for PET,
MRI and clinical outcomes, and developed longitudinal models to capture change and variation in each of the
summary measures. These models were used to provide descriptive summaries of rates of change, to test a
series of hypotheses about predictors of change in the NC, MCI and AD groups, and to assess the potential for
improving the efficiency of clinical trials.
We summarize briefly some of our key findings for selected measures from each domain studied:
• CSF: Aβ42, Tau
• PIB: overall summary
• FDG PET: SumZ2PNS (a hypometabolic summary, Utah); ROI-avg (average glucose metabolism over
composite region of interest, UC Berkeley); CV-fROI (a cross-validated data-driven region, Arizona)
• MRI: hippocampal volume, ventricular volume (Freesurfer, UCSF)
• Cognitive testing: ADAS-COG total score (ADAS-COG); MMSE; CDR Sum of Boxes (CDR-SB); RAVLT
sum across five learning trials (RAVLT)
Table 1. Baseline means (standard deviations) for selected ADNI variables, estimated mean
difference between NC and AD, and percentage of difference “traveled” by mean MCI, (MCINC)/(AD-NC)x100%.
%
Variable Name
Baseline mean (SD) by diagnosis
NC to AD
difference
difference
NC
MCI
AD
NC to MCI
207 (53)
158 (51)
139 (36)
-68
72%
CSF Aβ42
CSF Tau
70 (28)
107 (55)
122 (60)
52
71%
PIB
1.53 (0.30)
1.89 (0.42)
2.02 (0.33)
0.49
73%
1944 (2432)
3273 (3969)
6711 (6178)
4717
27%
1.277 (0.13)
1.200 (0.13)
1.075 (0.13)
-0.202
38%
0.977
(0.064)
0.898
(0.082)
0.876
(0.052)
-0.101
78%
Hippocampus
(UCSF)
Ventricles (UCSF)
3320 (407)
2900 (513)
2586 (509)
-734
57%
17997
(10042)
22241
(11680)
24461
(11956)
6464
66%
ADAS-COG total
6.17 (2.94)
11.6 (4.48)
18.5 (6.36)
12.3
44%
Hypometabolic
FDG PET (Utah)
FDG PET ROI-avg
(UCB)
FDG PET CV-fROI
(Arizona)
PHS 398/2590 (Rev. 11/07)
Page
MMSE
CDR-Sum of boxes
RAVLT 5-trial total
29.1 (1.0)
0.03 (0.12)
43 (9.2)
Weiner, Michael
27.0 (1.8)
1.60 (0.88)
31 (9.1)
23.4 (2.0)
4.26 (1.65)
23 (7.5)
-5.7
4.23
-20.1
37%
37%
59%
Table 1 summarizes the baseline distribution of each of these key measures by mean and standard deviation
(SD). ADNI participants are similar to those in previous studies, with AD worse than NC and MCI intermediate
on all measures, but with substantial variation. For most measures, the SD is comparable to the difference in
means between NC and MCI or between MCI and AD, so that 15-20% of participants might actually look more
like the adjacent group. Table 1 also shows the estimated difference between an average NC and an average
AD participant; for example, an average AD participant scores about 20 points lower on the RAVLT. We then
estimated how far the MCI mean is from the normal mean, compared to how far the AD mean is, by calculating
the ratio of NC to MCI mean difference to the NC to AD mean difference and converting to percent. This ratio
reflects both the true underlying biological change, and the ability of a given scale to measure change. For
example, the MMSE is limited at upper end by ceiling effects. The hypometabolic PET score was skewed and
was log-transformed for later analyses. These cross-sectional estimates offer landmarks for us to consider in
our examination of observed rates of longitudinal change.
Table 2. Mean (standard deviation) of annualized change for selected ADNI variables, and the
NC to AD difference as a reference for impact of rate of change.
Variable Name
Annualized mean change by diagnosis
NC to AD
difference
NC
MCI
AD
-0.94
(18)
-1.4
(17)
-0.1
(14)
-68
CSF Aβ42
CSF Tau
3.45 (13)
2.34 (21)
1.24 (24)
52
PIB
0.098 (0.18)
-0.008 (0.18)
-0.004 (0.25)
0.49
Hypometabolic FDG
PET (Utah)
FDG PET ROI-avg
(UCB)
FDG PET x validated
ROI (Arizona)
-177 (1532)
752 (2950)
2993 (4040)
4717
-0.006 (0.06)
-0.015 (0.064)
-0.055 (0.067)
-0.20
-0.019 (0.037)
-0.047 (0.030)
-0.081 (0.047)
-0.101
-40 (84)
848 (973)
-80 (91)
1551 (1520)
-116 (93)
2540 (1861)
-734
6464
-0.54 (3.05)
0.0095 (1.14)
0.07 (0.33)
0.29 (7.8)
1.05 (4.40)
-0.64 (2.5)
0.63 (1.16)
-1.37 (6.6)
4.37 (6.60)
-2.4 (4.1)
1.62 (2.20)
-3.62 (5.6)
12.3
-5.7
4.23
-20.1
Hippocampus (UCSF)
Ventricles (UCSF)
ADAS-COG total
MMSE
CDR-Sum of boxes
RAVLT 5-trial total
Table 2 shows the mean and SD of annualized change for key summary measures. The rate of change for
measures hypothesized to show early change (CSF, PIB) is greater in NC than AD, with MCI intermediate. For
the measures hypothesized to change later in the course of AD development, however, the rate of change is
greatest in AD and less in NC than in MCI (FDG PET, MRI, cognitive measures.) The next-to-last column of
Table 2 shows the difference between NC and AD, from Table 1. Comparing this difference to the rate of
change for an average person, or a person changing one SD faster than average, gives an estimate of the
number of years it would take to go from NC to AD at a constant rate of change. For example, a NC declining
one SD faster than average in CSF Aβ42 would reach the mean level for AD in a little less than 4 years. These
estimates are helpful to us in study design and power calculations for ADNI2.
We examined predictors of decline two different ways: first, using regression models with difference or
change trajectory as the outcome, then, for MCI only, using survival models for time to conversion to AD. Too
few NC converted to MCI to allow statistical modeling via survival models.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Tables 3 and 4 present selected results from longitudinal models predicting trajectories for two key
outcomes: change in hippocampal volume (Freesurfer, longitudinal approach to all data available for individual,
UCSF) and change in ADAS-COG. We began with univariate random effects repeated-measures models in
which each predictor’s effects on baseline level and rate of change per year were assessed [11]. Next, we
examined the joint effects of selected CSF, PET and (for ADAS-COG) MRI measures on baseline level and
rate of change in models also including years of formal education (centered at 12) and an indicator for
presence of one or more ApoE4 alleles. In the multivariate models, all biomarkers were transformed to zscores centered at the mean in the NC and scaled to the SD for the NC. Thus a one-unit change corresponds
to an increase or decrease of one SD among the NC.
Table 3. Predictors of longitudinal change in hippocampal volume (Freesurfer, performed at
UCSF), based on repeated measures regression models, showing results for coefficient of effect
on annualized change. Univariate models were not adjusted for other predictors; multivariate
models included ApoE4 status (any E4 allele), education, baseline CSF Aβ42 and Tau, and FDG
PET ROI-avg unless otherwise specified. Predictors significant at 0.05 are highlighted in the
table.
Univariate model
Multivariate model
Predictor of change/yr
P value
Coefficient
P value
Normal controls
0.095
-27
0.082
Apoe4+
0.36
0.85
0.68
Yrs of education
0.001
6.1
0.38
CSF Aβ42
0.031
-6.6
0.40
CSF tau
0.72
5.4
0.34
FDG PET ROI-avg
Mild Cognitive Impairment
Apoe4+
Yrs of education
CSF Aβ42
CSF tau
FDG PET ROI-avg
Alzheimer’s disease
Apoe4+
Yrs of education
CSF Aβ42
CSF tau
FDG PET ROI-avg
<0.001
0.040
<0.001
<0.001
0.005
-36
-2.7
-0.66
-4.4
9.3
0.002
0.12
0.91
0.15
0.026
0.087
0.79
0.002
0.031
0.73
-29
-3.4
-1.3
-8.7
10.2
0.18
0.18
0.92
0.046
0.75
In univariate analyses, lower values of CSF Aβ42 and higher values of CSF Tau were associated with more
rapid hippocampal atrophy in all three participant groups. In addition, in the MCI group, presence of an E4
allele, higher years of education, and lower metabolism as measured by the ROI-avg region were also
associated with more rapid atrophy. Multivariate analyses, however, suggested that some variables might not
predict independently, although it should also be noted that the sample size was reduced considerably,
typically by three quarters, when individuals were required to have data on all predictors. For NC, no single
variable was a significant predictor of hippocampal decline when all variables were included in the same
model, although the coefficients were generally in the expected direction. MCI who were E4+ had hippocampal
atrophy approximately twice as fast as those who were E4-, other variables being equal. An MCI participant
with baseline PET ROI-average score one NC-SD better than the average MCI had atrophy about 30% less
rapid than an average MCI participant. Among the AD group, every one NC-SD higher baseline CSF tau level
was associated with nearly a 30% faster hippocampal atrophy rate. Taken together, these findings suggest that
abnormal values of the CSF biomarkers are indeed associated with more rapid atrophy, in all diagnostic
groups. Our findings are limited, however, by the fact that only half of the participants had PET and half had
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
CSF biomarkers, so only 25% had both. Further analysis with the larger samples and longer follow-up of
ADNI2 is needed to determine whether the lack of correlation in multivariate analyses reflects mediation or
partial mediation through other processes, or is an artifact of the small sample sizes available in ADNI1 to
study all markers simultaneously.
Table 4. Predictors of longitudinal change in ADAS-COG Total 11 score, based on repeated
measures regression models, showing results for coefficient of effect on annualized change.
Univariate models were not adjusted for other predictors; multivariate models included ApoE4
status (any E4 allele), education, baseline CSF Aβ42 and Tau, FDG PET ROI-avg, and
hippocampal and ventricular volume (Freesurfer, x sectional, performed at UCSF) unless
otherwise specified. Predictors significant at 0.05 are highlighted in the table.
Univariate model
Multivariate model
Predictor of change/yr
P value
Coefficient
P value
Normal controls
0.22
1.06
0.020
Apoe4+
0.19
-0.026
0.64
Yrs of education
0.82
0.10
0.63
CSF Aβ42
0.75
0.33
0.19
CSF tau
0.076
0.05
0.77
FDG PET ROI-avg
0.016
-0.25
0.27
Hippocampal volume
0.45
0.18
0.31
Ventricular volume
Mild Cognitive Impairment
Apoe4+
Yrs of education
CSF Aβ42
CSF tau
FDG PET ROI-avg
Hippocampal volume
Ventricular volume
0.005
0.82
<0.001
<0.001
<0.001
<0.001
<0.001
0.57
-0.004
0.058
0.20
-0.40
-0.014
0.38
0.24
0.96
0.83
0.16
0.040
0.94
0.070
Alzheimer’s disease
Apoe4+
Yrs of education
CSF Aβ42
CSF tau
FDG PET ROI-avg
Hippocampal volume
Ventricular volume
0.30
0.15
0.26
0.004
<0.001
0.79
0.95
-0.39
0.050
-1.39
0.43
-2.12
-0.08
0.43
0.82
0.79
0.12
0.17
0.005
0.90
0.47
In Table 4, we look at similar analyses for prediction of change in the ADAS-COG total score. ADAS-COG
increases, representing cognitive impairment, were associated in the NC’s with smaller baseline hippocampal
volume, in univariate models, and with presence of ApoE4 in the multivariate model. In the MCI group, lower
baseline CSF Aβ42, higher Tau, lower FDG-PET metabolism, smaller baseline hippocampal volume, and larger
ventricles were all associated with more rapid cognitive function worsening, in univariate models. In
multivariate models, only the FDG-PET measure remained significant, and each one NC-SD worse metabolism
was associated with a 0.40 point faster annualized rate of worsening on the ADAS-COG. Among AD patients,
higher CSF tau was associated in univariate models with faster ADAS-COG decline, but not after adjusting for
covariates. Lower baseline metabolism, however, remained significantly associated, with each one NC-SD
worse metabolism associated with a two-point worse annualized rate of cognitive performance decline.
Overall, the findings from univariate regression models confirm that cognitive decline in MCI is associated
with a number of biomarkers. Multivariate models suggest that they are likely not operating independently but
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
instead may either represent part of a sequence in which some mediate others at least in part, or may reflect
several aspects of a common progression of underlying neurobiological damage. The modest number with
data on all markers limits the power of multivariate models. Again, the larger sample sizes, especially of people
with all modalities of fluid and imaging biomarkers, and longer follow-up in ADNI2 are critical to allow us to
determine relative contributions at different disease stages. In particular, cognitive decline in the NC’s is subtle
and requires larger sample sizes and longer follow-up to separate out effects of different risk factors. Results
in AD suggest that the hypothesized later changes are likely to play more of a role as predictors than those
thought to take place earlier in the disease process. Results for other cognitive outcomes and FDG PET and
MRI summaries are in general agreement (not shown).
A second set of univariate and multivariate analyses examined predictors for conversion from MCI to AD.
Drs. Gamst and Donohue at UCSD developed survival models for time to conversion, and examined not only
fluid and imaging biomarkers, but also baseline cognitive function as a potential predictor, and adjusted for
whether participants were already taking cholinesterase inhibitors. Univariate models (not shown) suggested
that a number of baseline fluid and imaging biomarkers were associated with shorter time to conversion,
including hippocampal and ventricular volume and brain size; three complex summaries of FDG PET
hypometabolism from the University of Utah; and the P-tau/Aβ42 ratio. In addition, baseline cognitive function
measures were predictive. People who were on Acetylcholinesterase Inhibitors (ACHEI) were also more likely
to convert. Multivariate analyses (Table 5) showed that only ACHEI and baseline cognition achieved statistical
significance when all variables were included, suggesting that most of the impact of the biomarkers on
conversion is already occurring through the baseline cognitive scores.
Table 5. Results of survival models for time to conversion from MCI to AD: Table shows
predictors that had P values less than 0.10 in model. Ridge regression used to shrink coefficients
for smaller values.
Predictor variable
Coefficient
P value
Entire MCI Cohort with MRI
-0.100
0.002
Baseline FAQ
-0.112
0.002
Baseline ADAS COG
-0.064
0.049
Using ACHEI
0.056
0.091
Baseline MMSE
MCI Cohort with MRI, FDG PET
Baseline FAQ
Baseline ADAS COG
Baseline FDG PET ROI-avg
-0.090
-0.085
0.092
0.024
0.047
0.062
MCI Cohort with MRI, CSF
Baseline FAQ
Using ACHEI
Baseline ADAS
-0.124
-0.094
0.101
0.017
0.057
0.058
Finally, we examined the potential of the fluid and imaging biomarkers to improve clinical trials in several
different ways. We considered the possibility that they might be used as outcome measures, and calculated
the sample size that would be required in a two-arm, one-year clinical trial, with 80% power to detect a 25%
improvement in annual rate of decline. We also looked at the possibility of using the biomarkers as covariates
in clinical trials to reduce unexplained variation, and we looked at the possibility that they might be used to
improve screening or study design (See results in Clinical Core).
When comparing sample sizes across labs and measures, we noted that sample size calculations are a
function of σ/δ, a measure of precision, where we estimate δ by the sample mean change and σ by the sample
standard deviation. We computed the subject-specific contribution to this quantity, by calculating the squared
deviation from the mean change and dividing it by the square of the mean change. The square root of this
value was used in our analyses, which were restricted to individuals for whom complete data (across all
imaging biomarkers) were available and who were assigned to the independent test set for the cross-validated
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
measures. Friedman’s rank test was used to test the hypothesis that the precision to capture change was the
same across all measures. If this global test was found to be significant, post-hoc pairwise tests adjusted for
multiple comparisons were performed. Table 6 shows the results of the comparisons of sample sizes for a trial
in MCI subjects across the most promising MRI and PET biomarkers, based on data obtained from 69 MCI
subjects. Each different color in the table identifies a group of measures that were not significantly different
from one another. In particular, measures of brain change and hippocampal atrophy required the fewest
subjects. The cross-validated functional ROI (CV-fROI) from the Reiman lab required the fewest subjects out
of the PET measures and was comparable to many of the top MRI measures. Results for AD trials were
similar (data not shown).
Table 6: 1.5T MRI vs. PET sample size calculations and comparisons: MCI (69 test subjects)
Lab
Modality
Variable
Sample Size
Jagust
PET
Average ROI
4605
Foster
PET
logSUMZ2PNS1
2176
Foster
PET
logSUMZ2PR1
1629
2
Fox
MRI
VBSI
284
Schuff (FreeSurfer)
MRI
ventricles
277
Reiman
PET
CV – fROI3
249
Schuff (FreeSurfer)
MRI
hippocampus
202
Fox
MRI
BSI4
177
Thompson
MRI
CV - % change3
73
1
Measures of glucose hypometabolism, log transformed
2
Ventricular boundary shift integral as a percentage of baseline brain volume
3
Cross-validated summaries from data-driven approaches
4
Boundary shift integral as a percentage of baseline brain volume
Another component of this aim is statistical methodology development to better address the overall
hypotheses of ADNI1. Drs. Beckett and Harvey supervised a Statistics doctoral student whose work focused
on the development of statistical models for studying the progression of tissue abnormalities over time. These
models incorporated the local spatial correlation and anatomical location of the data, as well as repeated
measurements over time on the same individual. Her methods may be applied to voxel-level white matter
hyperintensity data from MRI and voxel-level glucose hypometabolism data from PET, for example, and allow
for the detection of spatial patterns of disease progression. She is in the process of writing up her papers for
publication. A second student’s ongoing dissertation research, supervised by Drs. Beckett and Harvey in
consultation with Dr. Kornak at UCSF and motivated by questions that arose from the voxel-based analyses
being conducted for ADNI1, focuses on the linkage through a latent variable between two high-dimensional
processes observed through MRI: a tissue shrinkage process (atrophy) and a tissue damage process
(changes in MRI intensity). Finally, Drs. Harvey and Beckett have defined a standardized framework for the
comparison of biomarkers based on a set of pre-determined criteria, such as strong signal-to-noise (precision)
and clinical validity. The sample size comparisons presented in Table 6 are one illustration of this strategy for
the criterion of signal-to-noise.
Other examples include comparisons across biomarkers of clinical
differentiation at baseline, comparisons of correlations with cognitive decline, and comparisons of associations
with clinical progression to dementia. A paper based on this approach is currently in preparation.
8.3.4. Aim 4. To serve as an advisory group for other researchers interested in using the database to
assess additional markers or to design clinical trials. The ADNI Biostatistics Core holds a monthly
conference call, with biostatistics faculty from UC Davis, UC San Diego, and UCSF on the call, and is open as
well to industry, NIH and FDA biostatisticians. Minutes of this call are posted on the LONI website. Statistical
operations, analytic plans, results, and new methodology questions are discussed. New results are shared
electronically.
ADNI Biostatisticians participate in the semi-monthly ADNI Executive Committee calls, the MRI and PET
calls, and all national meetings of ADNI researchers.
National interest in ADNI data has been strong, so we developed introductory materials and organized
training sessions to help with this process. The ADNI Biostatistics Core has offered two successful web-based
training sessions in the past two years; one in August 2007 focused on navigating the image database while
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
the other in September 2008 focused on the clinical and image summary database along with an introduction
to statistical methods for analysis of these data. Both sessions were well attended (50-60 participants) by
individuals from industry, government, and academia, and requests continue to be made for additional training
sessions.
In addition, Dr. Harvey at UC Davis and Drs. Gamst and Donohue at UC San Diego fielded 15 questions, on
average, each month from outside researchers interested in the database but needing help to understand or
access it; each of these contacts often involved additional follow-up questions, but eventually led to others
making use of the database.
Our expertise in the database has also led to outside researchers consulting us on grant proposals related
to ADNI and Dr. Kornak (Rosen, UCSF P.I.) and Dr. Harvey (Crane, UW P.I.) have each been named as coinvestigators on (now funded) ADNI related proposals.
Biostatistics Core Publications 2004-2009
C1. Leow AD, Klunder AD, Jack CR Jr, Toga AW, Dale AM, Bernstein MA, Britson PJ, Gunter JL, Ward CP,
Whitwell JL, Borowski BJ, Fleisher AS, Fox NC, Harvey D, Kornak J, Schuff N, Studholme C, Alexander GE,
Weiner MW, Thompson PM; ADNI Preparatory Phase Study. Longitudinal stability of MRI for mapping brain
change using tensor-based morphometry. Neuroimage, 31(2):627-40, 2006. PMID: 16480900
C2. Mueller SG, Weiner MW, Thal LJ, Petersen RC, Jack CR, Jagust W, Trojanowski JQ, Toga AW, Beckett
L.: Ways toward an early diagnosis in Alzheimer’s disease: The Alzheimer’s Disease Neuroimaging Initiative
(ADNI), Alzheimer’s & Dementia 1:55-66, 2005. PMID: 17476317
C3. Mueller SG, Weiner MW, Thal LJ, Petersen RC, Jack C, Jagust W, Trojanowski JQ, Toga AW, Beckett
L. The Alzheimer’s Disease Neuroimaging Initiative. Neuroimaging Clin N Am, 15(4):869-77, 2005. PMID:
16443497
C4. Mueller SG, Weiner MW, Thal LJ, Petersen RC, Jack C, Jagust W, Trojanowski JQ, Toga AW, Beckett
LA. Ways toward an early diagnosis in Alzheimer’s disease: The Alzheimer’s Disease Neuroimaging Initiative.
Cognition and Dementia, 5(4):56-62, 2006. [In Japanese; See Appendix.]
C5. Jack CR Jr, Bernstein MA, Fox NC, Thompson P, Alexander G, Harvey D, Borowski B, Britson PJ,
Whitwell J, Ward C, Dale AM, Felmlee JP, Gunter JL, Hill DL, Killiany R, Schuff N, Fox-Bosetti, S, Lin C,
Studholme C, Decarli CS, Gunnar Krueger, Ward HA, Metzger GJ, Scott KT, Mallozzi R, Blezek D, Levy J,
Debbins JP, Fleisher AS, Albert M, Green R, Bartzokis G, Glover G, Mugler J, Weiner MW, ADNI Study. The
Alzheimer’s disease neuroimaging initiative (ADNI): MRI methods. J Magn Reson Imaging, 27(4):685-91,
2008. PMID: 18302232
C6. Hua X, Leow AD, Lee S, Klunder AD, Toga A, Lepore N, Chou Y-Y, Brun C, Chiang M-C, Barysheva M,
Jack Jr. CR, Bernstein MA, Britson PJ, Ward CP, Whitwell JL, Borowski B, Fleisher AS, Fox NC, Boyes RG,
Barnes J, Harvey D, Kornak J, Schuff N, Boreta L, Alexander GE, Weiner MW, Thompson PM, the
Alzheimer’s Disease Neuroimaging Initiative. 3D characterization of brain atrophy in Alzheimer’s disease and
mild cognitive impairment using tensor-based morphometry. NeuroImage, 41(1):19-34, 2008. PMID:
18378167.
C7. Leow AD, Yanovsky I, Parikshak N, Hua X, Lee S, Toga AW, Jack CR Jr, Bernstein MA, Britson PJ,
Gunter JL, Ward CP, Borowski B, Shaw LM, Trojanowski JQ, Fleisher AS, Harvey D, Kornak J, Schuff N,
Alexander GE, Weiner MW, Thompson PM, Alzheimer’s Disease Neuroimaging Initiative. Alzheimer’s Disease
Neuroimaging Initiative: A One-Year Follow-up Study Using Tensor-Based Morphometry Correlating
Degenerative Rates, Biomarkers and Cognition. Neuroimage. 45(3):645-55, 2009. PMID: 19280686
C8. Hua X, Lee S, Yanovsky I, Leow AD, Chou Y-Y, Ho AJ, Gutman B, Toga AW, Jack CR Jr, Bernstein MA,
Reiman EM, Harvey DJ, Kornak J, Schuff N, Alexander GE, Weiner MW, Thompson PM and the Alzheimer’s
Disease Neuroimaging Initiative. Optimizing power to track brain degeneration in Alzheimer’s disease and mild
cognitive impairment with tensor-based morphometry: an ADNI study of 515 subjects. NeuroImage
(2009);48:668-681. PMID: 19615450
C9. Ho AJ, Hua X, Lee S, Leow AD, Yanovsky I, Gutman B, Dinov ID, Leporé N, Stein JL, Toga AW, Jack Jr,
CR, Bernstein MA, Reiman EM, Harvey DJ, Kornak J, Schuff N, Alexander GE, Weiner MW, Thompson PM
and the Alzheimer’s Disease Neuroimaging Initative. Comparing 3 Tesla and 1.5 Tesla MRI for tracking
Alzheimer’s disease progression with tensor-based morphometry. Human Brain Mapping, 2009, Sept 24, E
pub ahead of print. PMID: 19780044
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
C10. Landau SM, Harvey D, Madison CM, Koeppe RA, Reiman EM, Foster NL, Weiner MW, Jagust WJ, the
Alzheimer’s Disease Neuroimaging Initiative. Associations between cognitive, functional, and FDG-PET
measures of decline in AD and MCI. Neurobiology of Aging, In Press, 2009. NIHMS 132130.
C11. Petersen, R.C., Aisen, P.S., Beckett, L.A., Donohue, M.J., Gamst, A.C., Harvey, D.J., Jack Jr., C.R.,
Jagust, W,J., Shaw, L.M., Toga, A.W., Trojanowski, J.Q., Weiner, M.W., and the Alzheimer’s Disease
Neuroimaging Initiative. Alzheimer’s Disease Neuroimaging Initiative (ADNI): Clinical characterization, Neurol.,
In press, 2009.
8.4. Methods:
The ADNI Biostatistics Core has 5 aims for ADNI2. Methods will be discussed separately for each aim.
8.4.1. Aim 1.To provide analytic support for the planning and design of the third phase of ADNI,
including such issues as sample size, stratification and enrichment of target populations, preparatory
phase and pilot analysis, planning for quality control. We will follow the pattern laid out successfully with
ADNI1 to ensure careful study design and high quality data. We will continue to coordinate closely with the UC
SD group to ensure quality of the data, to randomize new participants to study arms, and to ensure that
training-set test-set assignments are made and shared with researchers developing data-driven measures. We
will also work closely with the Cores as experimental protocols are developed to ensure quality control of the
data and that sufficient sample sizes are acquired for testing hypotheses related to those methodologies (for
example, the experimental protocols for the MRI core).
8.4.2. Aim 2. To carry out interim and final analyses of all ADNI data to address research questions on
clinical change, imaging measures, biomarkers, relationship of ADNI measurements to each other,
predictive value, implications for clinical trials, and optimization of prediction and design, and to
participate in presentation of results in abstracts, meeting and conference presentations, and papers.
The Introduction and other Cores of ADNI2 lay out a series of interrelated aims and hypotheses, for which the
Biostatistics Core will carry out most statistical analyses. Many of these will build on the approaches used for
ADNI1. A notable difference from ADNI1, and strength of ADNI2, is that most participants will have data from
all modalities of fluid and imaging biomarkers, thus reducing the need to address missing data by examining
only subsets, and increasing our statistical power (see Section 8.4.2.4 below). For simplicity of presentation,
we group the analyses into broadly similar classes. We give examples and discuss the analytic principles for
each class, then present sample size calculations.
8.4.2.1. Analytic strategies: Hypotheses comparing groups: The first broad class of analyses will test
hypotheses about cross-group differences at baseline or in one-number change summaries like boundary shift
integrals. Example: EMCI is intermediate between NC and MCI clinically (Clinical Core) and in imaging
measures (MRI, PET Cores). We will address these comparisons using standard linear models if the outcome
is normal and homoscedastic (e.g. ANOVA to compare hippocampal volume) and generalized linear models
otherwise. Models for cross-group comparisons may also be adjusted for important covariates such as
ApoE4+, age, education, or other markers.
8.4.2.2. Analytic strategies: Trajectories of change: The second broad group of analyses will test
hypotheses about trajectories of change and their predictors and relationships, including trajectories both for
cognitive measures and for biomarkers. For example, the PET core hypothesizes that FDG measures of
glucose metabolism will predict cognitive decline, and the Biomarker Core hypothesizes that baseline CSF
markers will predict metabolic decline and cortical atrophy. The first analytic step will be to develop longitudinal
models that provide accurate descriptions of the overall patterns of change in outcome measure and
heterogeneity of trajectories, while still allowing for possible missing data, unequal spacing of observations,
and within-person correlation. For most outcomes, we will use random effects repeated measures models for
longitudinal data [11]. If the assumptions of these models are systematically violated for some outcomes we
will consider a generalized linear models approach instead (this was necessary in ADNI1 for CDR and MMSE.)
We will also test for non-linear trajectories, for example accelerating rates of change over the longer-term
follow-up in ADNI2. The longitudinal models will then form the basis for addition of predictors, adjustment for
covariates, and examination of predictor effects both on starting level and rates of change. The scientific
hypotheses and available data will determine our approach to handling predictors that may themselves change
over time. Time-varying covariates may be used to see whether current level of the predictor correlates directly
with current level of the outcome, beyond the value of knowing the baseline predictor level. More elaborate
models [12] can take variation into account while estimating the relationship between two trajectories for two
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
different processes. In some cases, we will only have two time points on an outcome of interest; in this case,
we can calculate change scores (difference, percent change) and fit ordinary regression models with the
change summary as the outcome.
8.4.2.3. Analytic strategies: Conversion: A third set of analyses will address hypotheses about conversion
(MCI to AD, NC to MCI) and other time-to-event data. For these analyses, the primary approach will be survival
analysis, as used in ADNI1 and described in section 8.3.3. We anticipate using Cox proportional hazards
models but will consider alternatives such as accelerated failure time models if the Cox assumptions are
violated.
The first three statistical challenges mentioned in Section 8.2 will pose potential problems for the modelbuilding efforts of the first three classes of analyses. High-dimensional data are a general problem in imaging,
which Dr. Harvey’s research efforts and those of our graduate students continue to address. Genetics data
also run the risk of large numbers of predictors. The statistical literature is very rich in efforts to address this,
and we are mindful that we need to exert care to protect against false discovery without sacrificing too much
power [10]. We plan to continue the training-set, test-set approach first implemented in ADNI1 as another way
to avoid bias, and to include genetics analyses in the same paradigm. We have mentioned issues in modeling
trajectories; all models will be carefully validated both analytically and graphically, and alternatives considered
as needed. If generalized linear models are insufficiently flexible, we will explore other model families both for
the fixed and the stochastic parts of the model. Finally, we believe the challenges of modeling complex
processes offer a rich field for continued biostatistical research, and we will pursue this research area for
methodological advances (Aim 5 below).
8.4.2.4. Power Calculations: We present power calculations for the three major classes of analyses: 1)
comparisons across diagnostic groups; 2) assessing predictors of change; and 3) assessing predictors of
clinical progression from EMCI or LMCI to dementia. All calculations assume α=0.05 and a two-sided test.
Calculations are also presented in parentheses for α=0.01 to give some indication of power after accounting
for multiple comparisons. The Clinical Core has projected that we will have a total of 1226 subjects at baseline
(352 Normals, 300 EMCI, 424 LMCI, and 150 AD) and a total of 1153 subjects available for longitudinal
analyses (331 Normals, 282 EMCI, 399 LMCI, and 141 AD). Power is generally provided for the smallest
subset (AD) and an intermediate size group (Normals) unless otherwise stated. All other subgroup analyses
will have power between or larger than that presented for those two groups. A difference from ADNI1, and
substantial argument for the impact of the ADNI2 proposal, is that by having data on all fluid and imaging
biomarkers from all participants, we will no longer lose power by having to restrict analyses to subgroups. Thus
we should have adequate data to provide more definitive answers to many questions raised in our Progress
summary but limited by the sample size in subgroup analysis restricted to those with complete data.
Diagnostic Group Comparisons: With a total of 1226 subjects, we will have 80% power to detect an effect size
(variance in the means divided by the within group variance) as small as 0.009 when comparing baseline
levels of measures across the four diagnostic groups (effect size =0.013 with α=0.01). When we consider
comparisons of measures of change across the diagnostic groups, we will have 80% power to detect an effect
size as small as 0.010 (0.014 with α=0.01).
Assessing Predictors of Change: In analyses restricted to AD subjects, we will have 80% power to detect an
association accounting for 5.4% of the variability in the outcome (7.9% for α=0.01). If other variables account
for 10-20% of the variability, we will still have 80% power to detect a contribution of 4.3-4.8% to the variability
(6.3-7.0% for α=0.01). In the Normal group, we will have 80% power to detect an association accounting for
2.4% of the variability in the outcome (3.5% for α=0.01). If other variables account for 10-20% of the
variability, we will still have 80% power to detect a contribution of 1.9-2.1% to the variability (2.8-3.4% for
α=0.01).
Assessing Predictors of Change (longitudinal outcomes): The rate of change in the outcome we can detect for
a one-standard deviation increase in the predictor is the product of the correlation coefficient (square root of R
square) and the between-person standard deviation. ADNI1 cognitive and some imaging data have betweenperson standard deviations under 10, while others have between-person standard deviations around 115 (for
both Normals and AD). Therefore, in analyses restricted to AD subjects, we will have 80% power to detect a
change of about 0.5 units for measures with smaller between-person standard deviations or 6.2 units for
measures with larger between-person standard deviations. In Normals, the corresponding change is 0.24 units
or 2.8 units.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Assessing predictors of clinical progression: For analyses restricted to EMCI subjects, we will have 80%
power to detect a hazard ratio as small as 2.6 when comparing the worst quartile of a marker to all others,
assuming 20% have progressed to dementia after 4 years of follow-up (3.5 if α=0.01). For analyses restricted
to LMCI subjects, we will have 80% power to detect a hazard ration as small as 1.7 when comparing the worst
quartile of a marker to all others, assuming 40% have progressed to dementia after 4 years of follow-up (1.92 if
α=0.01).
8.4.2.5. Clinical Trial Design: A fourth set of analyses will examine ways that we can improve clinical trials
design. Biomarkers may be used in a number of ways to improve clinical trial design (the fourth challenge in
8.2). Early diagnosis of AD would allow rapid initiation of therapy, before brain damage becomes irreversible.
Identification of a high-risk, very-early-disease subgroup would also improve the design and conduct of
prevention trials by targeting the group with the highest conversion rate. We will assess the potential impact of
candidate biomarkers and composite diagnostic strategies on early treatment, assuming various levels of
treatment efficacy and potential side effects. We will also quantify their impact on the design of prevention trials
using the ADNI data on conversion from MCI to AD and using current clinical trial design as a basis for
possible design of future trials. When sufficient conversion data become available for the NC group, we will
extend this approach.
A second possibility for improved design is to use biomarkers in regression modeling to reduce unexplained
variation in clinical outcomes and increase the efficiency of clinical trials. Zhang, et al [14] recently proposed a
general framework for using auxiliary (baseline) covariates to improve the efficiency of inference in clinical
trials. Their semi-parametric approach incorporates information from baseline covariates associated with the
outcome to improve inference regarding the particular parameter of interest, e.g. the rate of decline or the
treatment effect, and can be seen as an extension of the “pretest-posttest” analysis of Leon, et al [15] to more
general regression models, including those used in the analysis of multivariate longitudinal data. Covariate
adjustments of a more basic flavor have been covered extensively in the literature [16-21] and improve the
efficiency of parameter estimates by exploiting correlations between baseline covariates and outcomes to
reduce the variance of parameter estimates of interest. We have applied the basic technique to both survival
and rate of change data from ADNI and we plan to continue this work, supplementing the basic paradigm with
both composite measures from Aim 1 and extensions of the novel semi-parametric techniques of Zhang, et al
[14]. The results thus far have been promising. For instance, a randomized placebo controlled study in a
population with Mild Cognitive Impairment (MCI) using change in Alzheimer’s Disease Assessment Scale
Cognitive subscale (ADAS-Cog) over time might require 15-18% fewer subjects if the model includes effects
for a baseline brain volumetric scan [22].
Third, we can address the possible use of ADNI biomarkers as surrogate markers [23]. Not only will ADNI2
and GO add to available data, but also the increasing use of treatment in our cohorts will provide useful data.
Moreover, ongoing clinical trials will provide additional data that may add to our knowledge base about the
performance of fluid and imaging biomarkers in the randomized trial setting.
8.4.2.6. Comparison of performance of biomarkers: Our strategy for comparison of biomarker performance
will follow the model laid out by Dr. Harvey for the ADNI1 comparisons (Section 8.3.3.) and will make use of
the test sets if any markers were developed by data-driven approaches in the training set. We will compare
performance across a common set of participants. We begin by identifying the person-specific contribution to
the relevant summary measure of performance, such as residuals to quantify errors of prediction. We then
form an array with one row per person in the comparison set and one column per measure, and enter the
contribution, then compare the results using suitable approaches for blocked data.
8.4.3. Aim 3. To participate in the ongoing ADNI2 operations, including: close collaboration with the
leadership of the clinical, MRI, PET, biomarker, genetics and informatics cores; participation in
Executive Committee conference calls and meetings; and generation of regular reports for ADNI
operations and coordination among ADNI Cores. We will continue to participate in all ADNI2 operations, as
we have with the ADNI1. We will add participation in the Genetics Core calls and operations, and work to link
phenotypic models developed by the Biostatistics Core with specialized genetics and epigenetics variables and
models developed by the Genetics Core. Our successful interactions with other ADNI cores and researchers,
including those in industry, have been taken as a model with other ADNI international partners, and Dr. Beckett
has served as a consultant for Japan-ADNI.
8.4.4. Aim.4. To provide intellectual leadership and foster communication among academic and
industry biostatisticians interested in ADNI data analysis, via a monthly ADNI biostatistics conference
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
call, web site, email, web-based training sessions, and documentation of routine procedures for
download, merging, summary and reporting of ADNI data in SAS and R. We will continue to hold monthly
conference calls and to post results and statistical programming and documentation to the LONI website. The
complexity of the dataset has made it critical to share our experiences, and we will continue to do so.
8.4.5. Aim 5. To develop new biostatistical methodology for the analysis of high-dimensional imaging
data, modeling of multi-process correlation, assessment of surrogate marker potential for AD studies,
and other analysis questions that arise in pursuing ADNI2 research goals. Our primary goal in this study
period is to use the increasing richness of the ADNI data, with longer follow-up and larger sample sizes, to
facilitate more sophisticated approaches to the discovery and validation of consistent multivariate patterns of
change in imaging and fluid biomarkers. We will then develop new biostatistical methods for two uses of these
patterns; first, to get a better understanding of the multivariate change process, and second, to improve clinical
trial design. We anticipate having sufficient conversions to MCI in the NC group by late in ADNI2 to permit
using both change models and survival models. We will explore the development of new analytic methodology
for simultaneous modeling of multiple change processes, expanding on previous work [11]. We will also
examine the possible development of composite measures for better prediction [12].
A second area for research is the development of improved methods for clinical trials. There are not only
biomarker assessment challenges here, but also statistical modeling challenges. We have discussed some of
our planned work; we will expand on these methods to develop approaches specific to Alzheimer’s and to the
multidimensional markers of ADNI.
References
1. Schneider, J.A., Z. Arvanitakis, W. Bang, and D.A. Bennett, Mixed brain pathologies account for most
dementia cases in community-dwelling older persons. Neurology, 2007. 69(24): p. 2197-204.
2. Schneider LS. Assessing outcomes in Alzheimer Disease. Alzheimer Dis Assoc Disord. 2001 Aug;15
Suppl 1:S8-18.
3. Knopman, D.S., J.E. Parisi, A. Salviati, M. Floriach-Robert, B.F. Boeve, R.J. Ivnik, G.E. Smith, D.W.
Dickson, K.A. Johnson, L.E. Petersen, W.C. McDonald, H. Braak, and R.C. Petersen, Neuropathology of
cognitively normal elderly. J Neuropathol Exp Neurol, 2003. 62(11): p. 1087-95.
4. Shaw, L.M., H. Vanderstichele, M. Knapik-Czajka, C.M. Clark, P.S. Aisen, R.C. Petersen, K. Blennow, H.
Soares, A. Simon, P. Lewczuk, R. Dean, E. Siemers, W. Potter, V.M. Lee, and J.Q. Trojanowski,
Cerebrospinal fluid biomarker signature in Alzheimer's disease neuroimaging initiative subjects. Ann Neurol,
2009, 65: p. 403-13.
5. Chetelat, G., B. Desgranges, V. de la Sayette, F. Viader, F. Eustache, and J.C. Baron, Mild cognitive
impairment: Can FDG-PET predict who is to rapidly convert to Alzheimer's disease? Neurology, 2003. 60(8): p.
1374-7.
6. Drzezga, A., N. Lautenschlager, H. Siebner, M. Riemenschneider, F. Willoch, S. Minoshima, M.
Schwaiger, and A. Kurz, Cerebral metabolic changes accompanying conversion of mild cognitive impairment
into Alzheimer's disease: a PET follow-up study. Eur J Nucl Med Mol Imaging, 2003. 30(8): p. 1104-13.
7. Killiany, R.J., T. Gomez-Isla, M. Moss, R. Kikinis, T. Sandor, F. Jolesz, R. Tanzi, K. Jones, B.T. Hyman,
and M.S. Albert, Use of structural magnetic resonance imaging to predict who will get Alzheimer's disease.
Ann Neurol, 2000. 47(4): p. 430-9.
8. Mueller SG, Weiner MW, Thal LJ, et al. The Alzheimer's disease neuroimaging initiative. Neuroimaging
Clin N Am. 2005 Nov;15(4):869-77, xi-xii.
9. Hampel H, Bürger K, Teipel SJ, et al. Core candidate neurochemical and imaging biomarkers of
Alzheimer's disease. Alzheimers Dement. 2008 Jan;4(1):38-48. Epub 2007 Dec 21.
10. Hastie T, Tibshirani R, Friedman J. The elements of statistical learning. New York: Springer, 2001.
11. Laird NM, Ware JH. Random-effects models for longitudinal data. Biometrics 1982; 38:963-974.
12. Beckett LA, Tancredi DJ, Wilson RS. Multivariate longitudinal methods for complex change processes.
Statistics in Medicine 2004; 23:231-239.
13. Demerath EW, Schubert CM, Maynard LM, Sun SS, Chumlea WC, Pickoff A, Czerwinski SA, Towne B,
Siervogel RM. Do changes in body mass index percentile reflect changes in body composition in children?
Data from the Fels Longitudinal Study. Pediatrics. 2006 Mar;117(3):e487-95.
14. Zhang M, Tsiatis AA, Davidian M. Improving efficiency of inferences in randomized clinical trials using
auxiliary covariates. Biometrics 2008 64(3): 707-715.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
15. Leon S, Tsiatis A, Davidian M. Semiparametric efficient estimation of treatment effect in a pretest –
posttest study. Biometrics 2003 59, 1046-1055.
16. Senn, S. (1989). Covariate imbalance and random allocation in clinical trials. Statist in Med 8, 467–475.
17. Hauck, W. W., Anderson, S., and Marcus, S. M.. Should we adjust for covariates in nonlinear regression
analyses of randomized trials? Controlled Clinical Trials 1998 19, 249–256.
18. Koch, G. G., Tangen, C. M., Jung, J. W., and Amara, I. A. Issues for covariance analysis of dichotomous
and ordered categorical data from randomized clinical trials and non-parametric strategies for addressing
them. Statistics in Medicine 1998 17, 1863–1892
19. Tangen CM, Koch GG. Complementary nonparametric analysis of covariance for logistic regression in a
randomized clinical trial setting. J Biopharm Stat 1999 Mar; 9(1):45-66.
20. Lesffre, E. and Senn, S. (2003). A note on non-parametric ANCOVA for covariate adjustment in
randomized clinical trials. Statistics in Medicine 22, 3586–3596.
21. Grouin, J. M., Day, S., and Lewis, J. (2004). Adjustment for baseline covariates: An introductory note.
Statistics in Medicine 23, 697–699.
22. Donohue M, Aisen P, Gamst A, Weiner M. Using the Alzheimers disease neuroimaging initiative (ADNI)
data to improve power for clinical trials. International Conference on Alzheimer’s Disease (ICAD). Chicago, IL.
July 2008.
23. Fleming TR (2005) Surrogate Endpoints and the FDA's Accelerated Approval Process. Health Affairs 24
(1) 67-78.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Core: 9
Title of Core (not to exceed 81 spaces): Informatics Core
Core Leader: Toga, Arthur
Position/Title: Professor, University of California, Los Angeles
Department, service, laboratory, or equivalent: Neurology
Mailing Address:
635 Charles E. Young Drive South, Suite 225
Los Angeles, CA 90095-7334
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
The Informatics Core will enhance, extend and improve the infrastructure developed during phase one of this
project. Building upon, what arguably could be claimed as one of the more successful informatics efforts for
archival, database and delivery systems created for multisite trials, we will further this mission by 1)enriching
the content and infrastructure of the database. 2) enable the upload, search and retrieval of processed data,
3) create a more sophisticated and intelligent query engine, 4) couple tools for analysis to the database, and
5) develop a robust intuitive and compelling online training system. The goals are to enable users of all
levels of expertise to leverage the enormous potential of this project by providing immediate, efficient and
most importantly easy-to-use tools for finding, obtaining and using these data. We will continue to refine and
augment the infrastructure to insure reliable and secure data access. The upload, quarantine, q/c and
archival responsibilities will also be enhanced with additional online error checking, validation and reporting
mechanisms. Finally we have plans to provide additional management tools so that all activity can be
analyzed to identify problem areas and successes
All the clinical, cognitive, biomarker, and imaging data is centralized in the Informatics Core for distribution to
all investigators, industry partners, and all qualified scientists throughout the world who request ADNI data.
The Regents of the University of California - Los Angeles
DUNS:
09-253-0369
Street 1:
11000 Kinross Ave Ste 102
City:
Street 2:
Los Angeles
Province:
County:
Los Angeles
USA
CA-030
Country:
State:
Zip/Postal Code:
CA
90095
DUNS:
Street 1:
Street 2:
City:
Province:
County:
Country:
State:
Zip/Postal Code:
PHS 398 (Rev. 11/07)
Page 2
Form Page 2
Weiner, Michael
Name
Organization
Role on Project
Toga, Arthur
TOGA22
UCLA
PI
Name
Organization
Role on Project
No
Yes
Cell Line
PHS 398 (Rev. 11/07)
Page 3
Weiner, Michael
9. NEUROINFORMATICS CORE:
9.1. Specific Aims:
9.1.1. Specific Aim 1 [SA1]: Enrichment of the ADNI Database Content and Infrastructure - a) Secure
archival of new imaging modalities and pulse sequences. New de-identification, categorization and translation
software will be developed to accommodate new neuroimaging modalities, pulse sequences, and file formats
to 1) ensure subject-identifying information is removed or replaced (e.g. HIPAA); 2) evaluate and categorize
MR images correctly so searches on modality and weighting are supported, 3) detect new tracers in PET
scans, and; 4) continue providing image data in popular file formats. b) Integrate greater richness of clinical
and biological data into the archive supporting more comprehensive options for selecting data of interest.
Additional clinical and biological data, including genetic data, will be integrated into the archive to support
searching and grouping results across more data elements. The incorporation of these data will increase the
overall utility of the archive. c) Enhancement of project administrative management tools within the IDA. These
management tools will govern the evaluation, approval/disapproval, on-going review and reporting needs of the
ADNI core leaders and the ADNI Data Sharing and Publication Committee.
9.1.2. Specific Aim 2 [SA2]: Incorporation of Processed data and its full Provenance - a) Enable the
efficient archiving of processed data, including atlases, from the most commonly utilized analyses packages
and methods. b) Develop flexible mechanisms for streamlining the capture of analytic sequence descriptors
that can be expanded to accommodate new analytic approaches. c) Create a system to record the full
provenance of the analyses for linkage to the analyzed data. d) Create an interface in which investigators may
graphically explore the analysis workflows created by other investigators and see examples of analyzed data
generated by the same.
9.1.3. Specific Aim 3 [SA3]: Enhanced, adaptive and intelligent query - a) Develop a context sensitive,
intuitive and interactive query system within the ADNI database system that supports pre- and post-query
refinement and customizable display of results in both graphical and tabular formats. This new approach will
1) Enable searches across multiple, global Alzheimer's disease efforts. Initially we will add connections with
AIBL (Australian Imaging, Biomarkers and Lifestyle), a project that has already archived imaging data at LONI,
to be followed by European AD efforts. 2) Provide query templates that address generalized searches adapted
from usage logs and direct requests from users. 3) Support queries combining elements of unprocessed and
processed data. 4) Support queries of the stored analysis results. 5) Support increased control by users over
query design, reuse and refinement.
9.1.4. Specific Aim 4 [SA4]: Tools for Efficient Workflow Processing of Data - a) Create tools supporting
improved access to ADNI data stored in the IDA including: 1) an interface for obtaining ADNI data, and; 2) the
direct import of ADNI data to the LONI Pipeline. In the case of the Pipeline, where execution may take place on
the LONI cluster, ADNI users may take full advantage of the Pipeline’s capabilities without having to download
data to their local systems. Further, processing provenance details performed on the data in the LONI Pipeline
will also be gathered and usable for inserting processed data into the IDA.
9.1.5. Specific Aim 5 [SA56]: Database Training - We will develop a suite of online tools to train novice users
as well as those experienced but not well versed in newer capabilities. This suite of online how-to manuals,
videos, wizards and community resources will be intuitive and comprehensive, with examples and test cases.
This will lower the learning curve for those wishing to utilize ADNI data, allow them to quickly understand the
variety of data available and learn best practices for finding, obtaining and using ADNI clinical, biological and
image data.
The increasing ability to obtain digital information in medical and biological neuroimaging research has lead to
a vast increase of scientific data from across a variety of spatial and temporal scales [26]. With each new
technological advance, neuroscientific data may be collected with finer resolution per unit time and render
more detailed forms of biologically-relevant information [3]. However, very often the raw and pre-processed
versions of these data are not available to researchers outside of the team that collected them. Concerns over
the sharing of the primary data have, for some, prohibited their availability [13]. Study meta-data (the data that
describes how the data were obtained, the parameters, experimental design, etc.) may be incomplete and
restrict the scope of future use.
Considerable attention was given to neuroimaging databases by the Organization of Human Brain Mapping
(OHBM) [11], who expressed concern about the quality of brain imaging data being deposited into such
archives, how such data might be re-used, and the potential for their being represented in new publication. The
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
question of data ownership, in particular, was a primary concern in initial attempts to archive data [7]. A recent
data ownership controversy [1] has highlighted anew the still tenuous nature of data ownership, re-use,
research ethical standards, and the pivotal role that peer-reviewed journals play in this process [10].
Perhaps one of the most influential successes in sharing brain image data is the ADNI collection
(http://www.loni.ucla.edu/ADNI/). It posses each of the characteristics noted in Arzberger et al.[2]. These
include: 1) the openness of the data archive—that access to information contained in a database is generally
unrestricted with respect to its user-base; 2) the database is transparent and there is evidence of active data
dissemination where it is clear what the database contains and that its contents experience ongoing access
over a period of time; 3) that there is an assignment and assumption of formal responsibilities among the
stakeholders; 4) that technical and semantic interoperability exists between the database in question and other
online resources; 5) curation systems governing quality control, data validation, authentication, and
authorization are in place; 6) there is demonstrated operational efficiency and flexibility; 7) the database insists
upon respect for intellectual property and other ethical and legal requirements; 8) there exists management
accountability which includes approaches to funding; 9) the archive is built upon a solid technological
architecture; and 10) users of the archive receive reliable support in data deposition and access. Beaulieu [4]
has elaborated on many of these characteristics and how they relate to what constitutes a trusted
neuroscience digital resource. Additional issues involve HIPAA compliance [14], concern over incidental
findings [12], anonymization of facial features [6] and [19], and skull stripping [29].
The advantage of having such large collections of data in one place is that they can be used to construct
detailed population-level atlases of brain morphometry or function. By population-level, we mean any form of
brain atlas assembled from image voxel intensity, geometry, or other attributes across large, representative
samples of human subjects warped to fit a known spatial reference frame. These include probabilistic
anatomical atlases [17], [24], [25], white matter fiber atlases [27], [18], and cortical surface atlases [22]. These
can also refer to functional maps or to the relation between functional results and anatomical features. Brain
atlases can be constructed to incorporate data describing multiple aspects of brain structure or function at
different scales from different subjects, at different times, yielding a comprehensive description of the brain in
normal or disease populations [21], [22]. However, the complexity and variability of brain structure, especially
in the gyral patterns of a dynamically changing (due to disease progression among other things) human cortex,
can present challenges in creating standardized brain atlases that reflect the anatomy of a population [23].
Alzheimer's atlases can potentially contain thousands of structure models, composite maps, average
templates, and visualizations of structural variability, asymmetry and group-specific differences. Although some
steps necessary for building atlases may be fully or partially automated, significant effort is required in their
creation. An infrastructure that provides the ability to archive, annotate and share these atlases will facilitate
their use in collaborative analyses and studies.
The Alzheimer’s Disease Neuroimaging Initiative (ADNI) was established to increase knowledge of the
mechanisms of Alzheimer’s Disease (AD) through the use of neuroimaging, thereby informing the development
of treatment strategies aimed at slowing down or preventing neuronal death. Phase I of ADNI has been
instrumental in helping to identify clinical, neuroimaging, and biomarker outcome measures and longitudinal
changes and the prediction of disease transitions. Interest in utilizing data from the ADNI project available
through the LONI Image Data Archive (IDA) has been extraordinary. To date, more than 1,000 investigators
across the world from industry, research and government have been granted access to ADNI data with over
500,000 MR/PET scans shared and the clinical, and biomarker data disseminated widely. Phase I of the ADNI
project can be considered an overwhelmingly successful first step in large-scale neuroimaging and the sharing
of that information with a larger community studying the disease and its treatments. Continued support for and
improvements in administering these data combined with the needs implicit in handling new imaging protocols
and additional meta-data stores will require new and expanded methods for managing, processing and
interacting with these rich and valuable data sets.
9.2.1. The Laboratory of Neuro Imaging (LONI): Investigations into brain structure and function require a
diverse array of tools to archive data for the purpose of creating, analyzing, visualizing, and interacting with
models of the brain. The laboratory houses a large super computer, many individual workstations and a data
archival system having over four petabytes of capacity. LONI has developed a high-end, secure, and easy to
use database structure for the specific purpose of large-scale brain data archiving and dissemination. LONI
has played a central role in the ADNI project, serving as the Informatics Core, responsible for receiving data
from participating ADNI centers, databasing images, biomarker, cognitive, genetics and their meta-data, and
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
making it accessible to ADNI the larger scientific community. The database technology and infrastructure
supporting it will be greatly enhanced and brought to bear in the ADNI2 project proposed here.
9.3. Preliminary Results:
The overall architecture subserving the ADNI informatics systems is illustrated in Figure 1 and the specifics in
Figure 2. These graphics show the functional elements necessary to provide efficient and intuitive access to
image, biomarker, cognitive, demographic and meta-data collected as part of this study. During the early years
of ADNI1, we designed and deployed a robust infrastructure to provide these services.
9.3.1. Infrastructure: Service continuity, deterministic performance and security were fundamental objectives
that governed the design of the subsystems providing the ADNI data archiving web service. Figure 2
illustrates the flow of data, initiated by a user on the Internet, from request to fulfillment. Best practices in
redundant, modular systems design as well as role-specific optimizations were implemented to meet the
current service level agreement (SLA) with the ADNI consortium and to accommodate future growth. Over the
course of ADNI Phase I, the IDA system has experienced 99.9% uptime ensuring that users around the globe
have continual access to data. In the following section, we describe the configuration, performance
optimizations and systems policies in place to guarantee that invaluable ADNI data are appropriately and
securely archived and remain accessible to authorized scientists and clinicians.
Network Design. The underlying network infrastructure utilized by the subsystems described herein has no
Figure 1. This graphic illustrates the major elements and functions of the Image Data Archive (IDA), the acquisition site and
communications between the two.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Figure 2. The figure above illustrates the data flow occurring when a user accesses the ADNI services over the web. In both examples, the web
request initially goes through redundant hardware load balancers, where it is analyzed and re-routed to the appropriate server pool. In Example 1,
the user accesses the static site http://www.loni.ucla.edu/ADNI. The request is sent to a pool of Apache web servers where it is rendered and the
results sent back to the user. Under certain circumstances, the Apache web servers query SQL server pools for stored procedures before returning
the results to the user. In Example 2, queries are made to the ADNI image data archives which are re-routed to the Tomcat application server pool.
Here, dynamic content is rendered and sent back to the user. Depending on the transaction type (upload or download), a different application server
group handles the request. A query or write request may be sent to the appropriate database server groups.
single points of failure and uses a layer 3, routed design to maintain network connectivity in the case that a
specific path fails. Multiple physical and Internet backbone connections, distribution switches, core and edge
routers are used and a firewall appliance protects and segments the network into role-specific de-militarized
zones (DMZ), permitting only authorized ingress and egress network traffic through. For instance, only HTTP
and HTTPS packets are allowed into and out of the web service DMZ while only specific database traffic is
permitted into and out of the database DMZ. Fault-tolerance and security were the guiding principles behind
the network topology design.
Server Reliability and Redundancy. In accordance with the fault-tolerance design of the network
infrastructure, multiple, highly-available web, application and database servers were configured to ensure
service continuity in the event of a single system failure as well as provide improved performance through realtime load balancing across N machines. Highly-available load balancing appliances intercept and analyze all
requests and route the web traffic to the appropriate resource. The load balancers maintain and monitor a
heartbeat connection to the servers and, in the case of a server failure, dynamically route all new connections
to available systems. A highly-available, fault-tolerant storage system in a separate network DMZ ensures that
these servers are able to access data reliably and securely.
Web Security and Data Integrity Checking. To augment the network-based security practices above and
to ensure HIPAA compliance, the web and application servers utilize SSL encryption for all ADNI data
transfers. Client-side de-identification of files using a signed applet was developed internally and only
successfully de-identified is data transferred to the servers. Post-transfer, we perform cyclic redundancy
checking on the files to detect unintended changes to the data. As with the network design and server
configuration practices above, these optimizations guarantee the integrity of the data.
Performance Optimizations. Significant system- and process-level performance optimizations were
implemented to meet the I/O requirements of transferring medical imaging data. Primarily, our approach
leveraged a modular systems design that facilitates future server additions to scale with increasing
performance demands. The application server pool is bifurcated between the upload and download server
groups; the database server pool is subdivided between query-only and write-only server groups. The role
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
specificity of each server pool allows for particular optimizations that would otherwise be inappropriate for
general servers. Automated bandwidth throttling based on the amount of data downloaded prevents any one
user from monopolizing the service. Data downloaded through the data archive service are also always
compressed. In comparison, no bandwidth throttling is implemented on data uploads and we also allow the
user to select compression options for improved uploading performance over slow connections. If a
connection during data transfer is lost, retries are automated to further simplify the process for the user. As
illustrated in Figure 2, we have discreet web server, application and database server pools. Improving the
performance of any pool would only require adding new servers to that specific group with minimal software
modifications.
Data Protection. As an added layer of security, sophisticated backup mechanisms are in place to protect the
integrity of ADNI data. The storage system for the ADNI data archive, for instance, utilizes a block-based
point-in-time snapshot feature that automatically and securely creates an internal copy of all files at the time of
creation or modification. In the event of data loss or corruption in the archive, we can readily recover copies of
files stored in the snapshot library without resorting to external backups.
To augment the data snapshot functionality above, we perform nightly incremental and monthly full backups
of the entire ADNI data repository. This automated backup is stored on tape in our secondary data center
located in another campus building to protect against data loss in the case of a catastrophic event in our
primary data center where the storage subsystem is housed.
We provide a tertiary level of protection against data loss by performing completely independent weekly tape
backups of the entire collection. We deposit the tape volumes to an offsite vaulting service (Iron Mountain).
This multi-pronged approach to data protection minimizes the risk of loss and ensures that a pristine copy of
the ADNI data archive is always available.
9.3.2. The Image Data Archive (IDA): The LONI Image Data Archive (IDA) serves as the central relationaldatabase repository for ADNI (Figure 3). The IDA was designed to ensure: 1) protection of patient privacy
through integrated data de-identification components; 2) strict access controls to ensure data are only
accessible to authorized individuals; 3) tracking of all data accesses to provide an audit trail so that project
managers may understand who and in what way their data are being accessed; 4) ease of use through a
platform-independent, user-friendly interface; 5) automated capture of image acquisition metadata; 6) a file
format translation engine which supplies on-demand image file format conversions; 7) an embedded image
viewer for evaluating image quality. These qualities help to address the issue of trust in the archive by
satisfying depositors that the data are being securely maintained, protecting subject identification, and
providing users with easy to use tools for viewing and manipulating the image data.
The IDA currently stores image files acquired in ADNI1, and metadata that describes the imaging protocols,
subject demographics and a subset of clinical assessments. The raw image data (data acquired by the MRI
and PET image scanners) is encoded in the DICOM, ECAT and HRRT file formats, and is created by different
scanner manufacturers and models (e.g., SIEMENS Symphony, GE SIGNA Excite, PHILIPS Intera, etc.). The
pre-processed and post-processed data (e.g., raw data processed by intensity-correction programs and mask
data produced by skull-stripping programs) is encoded in the DICOM and NIFTI file formats and also archived
in the LONI IDA. The variability of modalities, scanners, and file formats requires customized de-identification
programs to remove and/or replace patient-identifying information in the image files before they are archived,
mapping programs to copy metadata from image files to the IDA database supporting queries, and translation
programs to convert the different image file formats into those requested by ADNI investigators. Because the
technicians who operate the scanners and upload the image data to the LONI IDA can be susceptible to
human errors (e.g., misidentifying phantom scans as patient scans), we integrated image inspection code into
the upload process to alert technicians to possible errors. The non-uniformity of scanner sequences requires
the incorporation of image recognition software to determine the weighting of MRI scans so that this
information is available for queries.
Early in ADNI Phase I, obtaining imaging data required individual users to first query the database directly
and then select/add images of interest into data collections from which they could be download. Over the
course of ADNI1, the concept of shared data collections was introduced and with it, the ability to access ADNI
collections (e.g. collections “created” by the ADNI project and automatically populated with new data as it is
archived) was developed. Using ADNI shared collections, users may easily determine which images they have
previously downloaded and obtain all newly archived images without querying the database. Shared
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
collections, which are grouped by patient diagnostic group and visit, are also useful in ensuring the multiple
analysts tasked with analyzing the same images at each time point are all obtaining the same source data.
Summary tools (Figure 3D) help facilitate project management by providing a mechanism by which summary
and cumulative information about image archiving and downloading activities may be obtained for a specific
acquisition site and/or time period or for all acquisitions sites and time periods.
Figure 3. An example of end-to-end data management: archiving, querying, inspecting and managing data: A. archiving data: contribution to the
LONI IDA involves two basic steps: 1) de-identification—an HIPAA compliant manner for removing patient-identifying information involving the LONI
Debabeler [20], capable of de-identifying common medical imaging file formats, and; 2) data transmission efficient transfer of many files at once,
bundling files together so that file I/O is minimized. All data are securely transferred using the HTTPS protocol. B. Meta-data search. Meta-data
searches include the research group (control, patient, etc), gender, age, scan dates, scanner modality and accompanying scan parameters, among
other criteria. C. The LONI image viewer: users can examine data in multiple planes using a convenient and intuitive web-based viewer. D. Data
Management: Core leaders need to see, at a glance archiving activities including uploads, downloads, etc. The management component provides
a graphical glimpse into archiving activities for a specified time period. E. Shared data collections, automatically populated with newly archived data
give investigators a jump start on obtaining preferred data sets and determining whether new data are available.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
9.4. Methods:
9.4.1. Project Goals Under ADNI Phase II: Under ADNI Phase II, we will enhance the IDA to better support
and further the utility of the data in ADNI Phase II through, 1) Secure archival and dissemination of new
imaging modalities and/or protocols; 2) Efficient archiving of processed data and related provenance
information; 3) Enhanced query with integration of additional clinical and other pertinent study metadata; 4)
Streamlined mechanism for accessing ADNI data from external tools and from within the LONI Pipeline; 5)
Creation of a family of training and support tools. We will accomplish these goals under the following Specific
Aims:
Enrichment of the ADNI Database Content and Infrastructure [SA1]: ADNI data is dynamic and growing.
Efficient data management tools are critical for serving the ADNI community as the amount and the complexity
of the data increase. ADNI Phase I exposed the need of many investigators for a quick and easy way to select
and download large amounts of data in bulk. With existing scans averaging 20MB per scan and new protocols
such as DTI as much as 20 times greater, it is critical that robust hardware and software solutions are in place
to continue supporting the archival and dissemination of these data. ADNI investigators span the globe leading
to a need for providing data around the clock. Ensuring reliable, fast and user-friendly access to the existing
and planned image data by enhancing the infrastructure and software will be a focus during ADNI2.
The continued ability to de-identify, map, translate and categorize new ADNI2 imaging protocols such as
diffusion tensor imaging (DTI) and functional MRI (fMRI) will require system customizations. Because DTI is a
relatively recent neuroimaging technology, the DICOM file format does not yet provide standardization for
many of the DTI acquisition parameters. As a result, scanner manufacturers encode these parameters in
proprietary DICOM fields (as well as entirely private encodings embedded inside DICOM fields). We will
develop programs for the IDA to read these acquisition parameters in order to ensure HIPAA compliance and
so that ADNI collaborators may query on them. New mappings will have to be developed to map DTI and fMRI
metadata into the IDA database in a manner that fits the needs of the ADNI community. Archiving DTI scans
(typically about ten times the size of structural scans) will require the addition of more storage space as well as
faster servers to handle the increased file I/O. Functional MRI image scans are accompanied by additional
task-related data files which will need to be decoded, de-identified, and stored for dissemination to the ADNI
community. DTI and fMRI image scans include dimensions larger than structural scans (DTI has multiple
magnetic field directions and fMRI has a time dimension), which will require modifications to the integrated
viewer so that the ability to perform image inspection is not interrupted. Further, new translation code to
convert the DTI and fMRI scans into multiple file formats is needed for continued support of users accustomed
to obtaining ADNI data in their preferred file formats. Archiving new PET sequences may necessitate building
methods for differentiating between different tracers so that 1) image metadata labels correctly label the image,
and; 2) the images may be queried appropriately.
Enriching the ADNI database content by incorporating additional clinical and biological data will require the
development of new programs for reading in, validating and associating the new data with the existing records
plus interfaces and methods for disseminating these data to the users. Specifically this will include:
• Expanding the data model for non-neuroimaging biological data (genotypes, SNP sequences,
pharmaceutical data related to the subject, etc.)
• Expanding “Sanity Checking” and QC methods for uploaded images
• Leverage LONI-developed file conversion software to allow users to seamlessly convert data among
common neuroimaging file formats (e.g. DICOM, Analyze, NIfTI, etc.) for new imaging modalities
• A redesigned image data viewer for inspecting 4D datasets
Enhancement of project administrative management tools will expand the administrative control aspects of
the site for:
• Provision of summary and detailed archive activities across new imaging modalities
• Geographic display of upload/download activities
• New management features for administering data user activities including changing access status,
communicating with data users and managing annual reporting requirements and manuscript submission
• Expanding the XML schemas and XML document creation programs to support/supply new metadata
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Incorporation of Processed data and its full Provenance [SA2]: A description of how data were obtained
is often crucial for assessing its quality and usefulness, as well as enabling analysis in an appropriate context.
Many commonly available software packages exist that provide either complete analyses or enable specific
steps in neuroimaging data analysis. The recording of the data generation and processing provenance [5] [15]
is, however, not often practiced. Many software packages, for instance, possess diverse input and output
requirements, utilize different file formats, run only under particular computer environments, or are appropriate
for only certain types of data. The accurate preservation of data integrity during study data transactions or to
document any database normalization operation also falls under the domain of provenance. Recording the
history of data, its processing, curation, alterations or addendums to it and including this information in
databases can aid in the fidelity of the independent reproduction of results or, if viewed as meta-data itself, can
be used as predictor variables in multi-center trials to examine how acquisition or processing parameters
influence experimental results. The main difficulty is the need of a system to capture provenance information
accurately, completely, and with minimal user intervention. Provenance can be divided into two subtypes, data
provenance and processing provenance. Data provenance is the meta-data that describes the subject, how an
image of that subject was collected, what instrument was used, what settings were used, and how the sample
was prepared. Some data provenance is typically captured at the site where the data is collected, in the
headers of image files or in databases that record image acquisition [9] [16]. Processing provenance is the
metadata that defines what processing an image has undergone and may include multiple levels of detail; for
example, how the image was skull-stripped, what form of inhomogeneity correction was used, how it was
aligned to a standard space, etc. A complete provenance model will capture all this information.
We will develop tools for capturing, processing and mapping provenance information from XML documents
into the database. This effort will include a) developing XML schemas (XSDs) for capturing provenance
information, b) programs for processing provenance XML documents to 1) validate the content; 2) associate
processed data with the archived image(s) from which derived; 3) map the provenance data into the database
to support querying, c) graphical interfaces for building provenance XML templates and documents. The LONI
Pipeline is an example of a workflow environment that automatically generates an XML document containing
processing provenance information (executable, parameters, inputs, etc). Other analyses methods either have
parameter files, scripts or other execution trails from which provenance can be extracted. By documenting the
executable provenance of each module and the workflow itself as workflow provenance, any workflow
application can become a mechanism for capturing processing provenance. One of the real strengths of this
system is the capacity to easily recreate the processing applied to a file by viewing its provenance file,
extracting the workflow, and rerunning it. We will develop mappings between XML files describing provenance
and the IDA XSD, enabling the provenance XSD to be used as input when archiving processed data in the
IDA.
We will create and integrate data models and infrastructure for archiving and annotating digital atlases in a
manner that preserves the relationships to original data from which they were derived and retains the data
provenance information. Both the approaches described above will support the archiving of atlases and the
ability to search for atlases using demographic and provenance metadata. Coupled with the new
developments on the ADNI database infrastructure, this aim will greatly enhance how ADNI researchers
understand and interact with the neuroimaging data contained in the LONI IDA.
Enhanced, adaptive and intelligent query [SA3]: With the abundance of clinical and biological data to
become available under ADNI2, novel and powerful mechanisms for interrogating these data are critical for
investigators’ ability to interact with, organize and obtain data of interest. Expanding the number and types of
ADNI data integrated into the IDA will greatly improve the granularity by which investigators may explore the
ADNI data, enabling more precise and robust queries to be posed. Integrating data obtained from different
external entities involves developing methods for validating, interpreting and storing these data such that links
to existing imaging and related data are maintained. Each data source requires a customized approach and
careful attention to detail if continued provision of reliable information is to be made.
We will create a highly flexible, powerful and interactive query system that both incorporates the richer
selection of data described in SA1 and supports intuitive interrogation of the same. Completely redesigned
advanced database search pages will add support for searching on:
o Additional clinical and biological data
o Multiple AD projects (e.g. ADNI & AIBL)
o Multiple time points (subjects who have scans for visit 1 & visit 2)
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Joint modality searches (subjects who have both MRI & PET scans)
Image quality rating
Attributes of unprocessed scans during pre/post-processed image searches (find post-processed masks
derived from images scanned on a Siemens 3 Tesla scanner)
o Numeric measures associated with post-processed data
o Processing provenance
The new search module will support interactive refinement of searches from within the search results, the
ability to save and reuse searches and graphical representation of search results and the processing
provenance. These new and enhanced features will enable experienced users to create more powerful and
fine-grained searches while supporting naïve users with little or no background in using the database. To
improve cooperation between investigators we will add the ability for user-to-user data collection sharing
(allows users to share their personal data collections with team members). Principles of faceted metadata
searches, which have been shown to be flexible, intuitive and easy to use on complex collections of data [28]
[8] will be leveraged for this aim.
Tools for Efficient Workflow Processing of Data [SA4]: Typically, image analysts develop an image
analysis workflow that is used to process many image files. The LONI Pipeline (http://pipeline.loni.ucla.edu),
for example, was developed to accommodate the many software packages that exist to provide specific
methods for neuroimaging studies. The LONI Pipeline provides a tool for the organization, dissemination, and
use of neuroimaging analysis procedures formed through the linking of these algorithms into a cohesive
methodology. The environment already has mechanisms in place for dealing with the diverse input and output
requirements of various software packages, for accommodating different file formats, and for allowing
procedures to run across hardware environments as diverse as desktop PCs and supercomputers. The ability
to directly import image data and metadata from the ADNI database (IDA) into the Pipeline will greatly
streamline workflow processing.
We will develop an application programming interface (API) to allow direct access to ADNI data archived in
the IDA. The proposed API will consist of two main interfaces. Both interfaces will use HTTPS and require the
user's IDA user name and password (Figure 4). The first interface retrieves a list of all the collections the user
has access to. This “collection retrieval API” will return references to each image series in each collection.
Once an image series is selected, the second interface provides two methods to obtain information about the
series. This “series retrieval API” consists of a method to retrieve descriptive meta-data in the human-readable
XML file format as well as a method to download the image files (e.g., DICOM) of the series. Our first
implementation will use the Java coding language because of its cross-platform usability.
Because neuroimaging analysis software packages differ in the types of input files they can process, we will
develop a file translator that can convert files downloaded from the IDA into files acceptable to these software
packages. From the most common file formats (ANALYZE, NIFTI, MINC), we propose to pick a primary
(preferred) file format for each of the most widely-used software packages and customize the conversions for
each package. Even if the downloaded file is in the preferred file format for a software package, we will
perform checks (check for byte-swapping requirements, correct negative voxel spacings, fill in missing
metadata values) and correct when needed. In cases where the downloaded file cannot be converted without
user intervention (e.g., lack of support for 32-bit data), we will identify the conversion issue and if possible
suggest work-arounds (e.g., downsample 32-bit data to 16-bit data).
Database Training [SA5]: We will develop a series of online training materials, and 'wizards' that help the
user gain familiarity with the more sophisticated aspects of the IDA. Training on usage of the various ADNI
Phase II-specific features of the IDA and related software will include on-line tutorials and a comprehensive
User Manual. In addition, IDA staff will provide support through FAQ’s, Blogs and RSS feeds on new
developments. This will enable researchers to quickly and efficiently utilize the ADNI database (IDA) and
software tools and ultimately improve their productivity. Online support, tutorials and discussion forums will be
established for remote 24/7.
As was established under ADNI Phase I, for ADNI Phase II we will have our web, database, neuroinformatics
and programming staff available via e-mail and phone during regular business hours (9am-5pm, Pacific Time,
Monday-Friday). An e-mail group that will include such personnel that will receive all help, support, and training
inquiries and will address these in a timely fashion according to the subject of the request will remain in place.
Co-investigators, pharmaceutical companies and outside researchers can contact this e-mail group with direct
questions. Requests will be logged and posted on an online neuroinformatics core web-forum page for future
o
o
o
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
reference. All requests and feedback related to the neuroimaging database can be posted online at the
neuroinformatics core web-forum pages first. Notably this will include enhancements to the ADNI Project web
site such as:
• IDA User Manual
• On-line glossary of terms
• Graphical overview of the IDA features
• Contextual help
Figure 4. Illustration of the proposed IDA API. (top) The collection API retrieves accessible collections of image
series for the user. (bottom) Information about an image series is available with the series API to (1) download
image meta-data in XML and (2) download the image files of the series.
9.4.2. Validation and Testing: We will continuously assess progress toward meeting each aim, evaluating
system performance and user experience in order to meet objectives on time, ensure continual access and
reliability and meet user needs. Our software development, database management, web development and
system support team will continue holding weekly meetings to plan and evaluate progress on ADNI Phase II
tasks. Specific validation and testing of each aim will include:
Image de-identification and metadata mapping: At the beginning of ADNI Phase I, we worked closely with
the MR and PET Cores to establish which elements of the image data files would be removed/replaced to
protect patient privacy in a HIPAA compliant manner. We will do the same for the new imaging modalities
proposed under ADNI Phase II. Likewise, the process of selecting elements to map from the images and into
the database to support meaningful queries will be performed with close consultation with the MR core and
domain experts. Validating the approach will involve testing the de-identification and mapping programs using
phantom data from each of the scanner models that will be used in ADNI Phase II.
File format translations: New file format translations will be tested internally by comparing original and
translated pixel data and metadata to ensure the both the image portion of the files and the metadata portion of
the files are valid and properly formatted. Further testing may be performed through loading translated image
data into common image viewing and analysis programs and assessing their utility.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
Incorporate additional clinical & biological data & enhanced queries: Early in ADNI Phase I, we
surveyed investigators at each of the funded analysis sites to determine which clinical metadata elements
would be most useful to integrate into the IDA for supporting richer queries. We will do the same during ADNI
Phase II, extending outreach to the greater ADNI community of over 1,000 investigators. As part of this
process, data values will be validated for accuracy (data type, range, etc) and transformed into user-friendly
values for dissemination purposes. We will conduct usability testing of new query interfaces to ensure they are
useable by novice, intermediate and power users. As new data elements are incorporated and the query
interfaces enhanced, announcements in the form of e-mail, Bulletin Board postings, tutorials and/or RSS feeds
will be provided.
Incorporation of processed data and provenance: We will develop controlled vocabularies to validate the
accuracy of metadata used to populate XML descriptor files. Controlled vocabularies will be published as XML
documents and continually updated to ensure the vocabularies cover emerging standards, technologies and
methods.
Tools for Efficient Workflow Processing of Data: The collection and series API's will be integrated into the
Pipeline framework and we will conduct performance testing to achieve the fastest data transfer rates possible.
If needed, we will dedicate and optimize new hardware so that large studies can be downloaded from the IDA
to the Pipeline without impacting IDA web users. We will also integrate the file converter into the Pipeline
conversion GUI and perform testing by downloading different data sets from the IDA into commonly-used
Pipelines.
Training. We will incorporate user feedback into the training materials to continually improve the content.
9.4.3. Timeline:
Aim
Task
Month 1-3 Month 4-12 Year 2
SA1: Enrichment of the ADNI Database Content and Infrastructure
De-Identification programs
Mapping programs
Archiving new image sequences
Clinical-Biological Data Survey
Data Modeling & Transform
SA2: Incorporation of Processed data and its full Provenance
Data modeling (atlas data)
Create controlled vocabs
Controlled vocab management
XML schemas
Provenance capture tools
SA3: Enhanced query
Development
Usability Testing
Implementation
SA4: Tools for Efficient Workflow Processing of Data
Develop, test, release API
SA5: Training
On-going support
Training survey
Training videos
PHS 398/2590 (Rev. 11/07)
Page
Year 3
Year 4
Year 5
Weiner, Michael
9.5. Bibliography:
1. Abbott A, Neuroscientist: my data published without authorization are ‘misleading’: Max Planck researchers
charged with misusing data, Nature, 454(7200):6–7, 2008.
2. Arzberger P, Schroeder P, Beaulieu A, Bowker G, Casey K, Laaksonen L, Moorman D, Uhlir P & Wouters
P, Science and government: An international framework to promote access to data, Science,
303(5665):1777–1778, 2004.
3. Bandettini P, Functional MRI today, Int. J. Psychophysiol, 63:138–145, 2007.
4. Beaulieu A, Voxels in the brain: neuroscience, informatics and changing notions of objectivity, Soc. Stud.
Sci., 31(5):635–680, 2001.
5. Bidgood Jr WD, Horii SC, Prior FW & Van Syckle DE, Understanding and using DICOM, the data
interchange standard for biomedical imaging, J. Am. Med. Inform. Assoc., 4(3):199–212, 1997.
6. Bischoff-Grethe A, Ozyurt IB, Busa E, Quinn BT, Fennema-Notestine C, Clark CP, Morris S, Bondi MW,
Jernigan TL, Dale AM, Brown GG & Fischl B, A technique for the deidentification of structural brain MR
images, Hum. Brain Mapp., 28(9):892–903, 2007.
7. Editorial, Whose scans are they anyway?, Nature, 406(6795):443, 2000.
8. English J, Hearst M, Sinha R, Swearingen K & Yee K, Hierarchical faceted metadata in site search
interfaces. In CHI '02 Extended Abstracts on Human Factors in Computing Systems (Minneapolis,
Minnesota, USA, April 20 - 25, 2002). CHI '02. ACM, New York, NY, 628-639, 2002.
9. Erberich SG, Silverstein JC, Chervenak A, Schuler R, Nelson MD & Kesselman C, Globus MEDICUS —
federation of DICOM medical imaging devices into healthcare grids, Stud. Health Technol. Inform.,
126:269–278, 2007.
10. Fox PT, Bullmore E, Bandettini PA & Lancaster JL, Protecting peer review: correspondence chronology
and ethical analysis regarding Logothetis vs. Shmuel and Leopold, Hum. Brain Mapp., 30(2):347–354,
2009.
11. The Governing Council of the Organization for Human Brain Mapping, Neuroimaging databases, Science,
292(5522)1673–1676, 2001.
12. Illes J, Kirschen MP, Edwards E, Standford LR, Bandettini P, Cho MK, Ford PJ, Glover GH, Kulynych J,
Macklin R, Michael DB, Wolf SM, Working Group on Incidental Findings in Brain Imaging Research, Ethics.
Incidental findings in brain imaging research, Science, 311(5762):783–784, 2006.
13. Koslow SH, Opinion: sharing primary data: a threat or asset to discovery?, Nat. Rev. Neurosci., 3(4):311–
313, 2002.
14. Kulynych J, Legal and ethical issues in neuroimaging research: human subjects protection, medical
privacy, and the public communication of research results, Brain Cogn., 50(3):345–357, 2002.
15. Mackenzie-Graham AJ, Van Horn JD, Woods RP, Crawford KL & Toga AW, Provenance in neuroimaging,
NeuroImage, 42(1):178–195, 2008.
16. Martone ME, Zhang S, Gupta A, Qian X, He H, Price DL, Wong M, Santini S & Ellisman MH, The cellcentered database: a database for multiscale structural and protein localization data from light and electron
microscopy, Neuroinformatics, 1(4):379–395, 2003.
17. Mazziotta JC, Toga AW, Evans A, Fox P & Lancaster J, A probabilistic atlas of the human brain: theory and
rationale for its development. The International Consortium for Brain Mapping (ICBM), NeuroImage,
2(2):89–101, 1995.
18. Mori S, Oishi K, Jiang H, Jiang L, Li X, Akhter K, Hua K, Faria AV, Mahmood A, Woods R, Toga AW, Pike
GB, Neto PR, Evans A, Zhang J, Huang H, Miller MI, van Zijl P, Mazziotta J, Stereotaxic white matter atlas
based on diffusion tensor imaging in an ICBM template, NeuroImage, 40(2):570–582, 2008.
19. Neu SC & Toga AW, Automatic localization of anatomical point landmarks for brain image processing
algorithms, Neuroinformatics, 6(2):135–148, 2008.
20. Neu SC, Valentino DJ & Toga AW, The LONI debabeler: a mediator for neuroimaging software,
NeuroImage, 24(4):1170–1179, 2005.
21. Roland PE & Zilles K, Brain atlases—a new research tool, Trends Neurosci., 17(11):458–467,1994.
22. Toga AW & Thompson PM, Maps of the brain, Anat. Rec., 265(2):37–53, 2001.
23. Toga AW & Thompson PM, New approaches in brain morphometry, Am. J. Geriatr. Psychiatry, 10(1):13–
23, 2002.
24. Toga AW, Thompson PM, Mega MS, Narr KL & Blanton RE, Probabilistic approaches for atlasing normal
and disease-specific brain variability, Anat. Embryol. (Berl), 204(4):267–282, 2001.
PHS 398/2590 (Rev. 11/07)
Page
Weiner, Michael
25. Toga AW, Thompson PM, Mori S, Amunts K & Zilles K, Towards multimodal atlases of the human brain,
Nat. Rev. Neurosci., 7(12):952–966, 2006.
26. Van Essen DC, Windows on the brain: the emerging role of atlases and databases in neuroscience, Curr.
Opin. Neurobiol. 12(5):574–579, 2002.
27. Wakana S, Jiang H, Nagae-Poetscher LM, van Zijl PC & Mori S, Fiber tract-based atlas of human white
matter anatomy, Radiology, 230(1):77–87, 2004.
28. Yee K, Swearingen K, Li K & Hearst M, Faceted metadata for image search and browsing. In Proceedings
of the SIGCHI Conference on Human Factors in Computing Systems (Ft. Lauderdale, Florida, USA, April
05 - 10, 2003), CHI '03. ACM, New York, NY, 401-408, 2003.
29. Zhuang AH, Valentino DJ & Toga AW, Skull-stripping magnetic resonance brain images using a modelbased level set, NeuroImage, 32(1):79–92, 2006.
PHS 398/2590 (Rev. 11/07)
Page