Panic and Phobias 2

Reprinted from:
Panic and Phobias 2
Treatments and Variables
Affecting Course and Outcome
Edited by
Iver Hand and Hans-Ulrich Wittchen
Förcword by G.L.Klerman
Epilogue by I.M.Marks
Springer-Verlag
Berlin Heidelberg New York
London Paris Tokyo
1988
10. Panic Attacks in Nonclinical Subjects
J.
MARGRAF
and A.
EHLERS
Introduction
Sudden episodes of intense anxiety accompanied by a number of predominantly
somatic symptoms (now usually ca!led
panic attacks) are the primary feature of
the psychological disturbance termed
panic disorder in DSM-IIIR (APA 1987).
This diagnosis was introduced based on
the idea that panic attacks are a distinct
type of anxicty in ncccl of thcir own
diagnostic entity. However, recenl
research has shown that panic attacks are
not specific to panic disorder since thcy
also occu r 'icross a wide range of other
psychological disorders and evcn in nonclinical popu lations. Barlow et al. ( 1985)
studied IOS patients with the DSM-111
diagnoses simple phobia. social phobia,
generalized anxiety disorder, panic disorder. agoraphobia with panic attacks. obsessive-compulsive disorder. and major
depressive episodes. The great majority of
patients in each of these categories (at
least 83%) reported having experienced
panic attJcks. A.lthough the frequency of
a!!acks var!ed across diagnoses. thcre
were only fe\\' differences in terms of the
symptom p<Htern associated with the
:lttacks. Funhermore. symptom severity
was similar for patients with situational
(prcdictabk. öpcctcd) and spontancous
(unpredictable.
unexpected)
attacks.
These results are in line with those of the
Munich Follow-Up Study (MFS. Wittchen 1986) in \\hich panic attacks were
obser\·ed in 9.3% of a represcntativc
community sample, a percentage higher
than the combined frequencies of panic
disorder and agoraphobia with panic attacks.
If it is established that panic attacks are
not specific to people suffering from a
specific disorder, il is importanl to study
the distribution of the phenornenon in
the general population. There have been a
first few attempts to approach this qucstion using qucsti onnai rc scrccn ing
rnethods. Norton et al. (1985, 1986) initiated this line of resea rc h. They fou nd
surprisingly high prevalenccs of panic
attacks in noncl in ical subjects. About
OllC-third of thcir lWü 5amp lcs of undergradt1atC students rc portcd having expcrienccd at least one panic attack in thc
past year. Thcy concludcd that panic
attacks orten occur in presumably normal
peoplc and thal thcse panic attacks sh:1re
many similaritics with thosc or paticnts
who have well-detincd panic disorders
(Norton et al. 1985).
Aside from the quest ion of the distribution of" panic attacks in thc population,
there are at least two other important
reasons to study panic attacks in nonclinical su bjects. The tirst reason concerns
sampling bias. With the exception of l<1rgc
epidemiological studies such as 1he Epidc111iological Catchmcnt t\rc;1 Progra 111
(ECA, Regier et al. 1984) or the MFS
(\Vittchen 1986) pre vious studies or panic
attacks have exclusively investigated clinical populations. There are a number of
rcasons to cxpcct that thcse sampks
10-1
J.
M~1rgraf'
and A. Ehlcrs
represen t a biased selection or the total
population or persons with panic attacks.
Highly symptornatic inclividuals are morc
likcly to scck treatrnent or tobe detected
in clinical scrcenings, an cfTcct Motulsky
( 1978) termed "asccrtainment bias." This
bias incrcascs thc probability or pcrsons
with two disorders to be part of clinical
samples and thus may lead to mistaken
assumptions about the relationship between such disorders. For example, almost all agoraphobic patients seen in
clinical settings liave panic attacks (cf.
Mendel and Klein 1969; Thycr and 1-iirnlc
1985). In nonclinical community samples,
howcvcr. thc picturc looks quitc clilfcrcnt.
Weissman et al. (1986) and Wittchen
(1986) founcl that only a small percentage
or all subjccts meeting criteria for agoraphobia also exhibitcd panic disorder (ranging from 6% to 16% in thc various ECA
and MFS sitcs) and only anothcr 17% to
50% showed limited panic symptorns in
addition to agoraphohia.
Anothcr cxamplc for sarnpling hias is thc
postulatcd rclationship betwecn panic
disordcr and mitral valve prolapse (MVPJ.
\Ve havc argucd elscwherc ( Margrnf et al..
1988) Lhat the highcr prevalcncc of MVP
obsen·ect in some studics of panic disorder patients rerrcsents a problem or comorbidity rathcr than a true functional
relationship. If thc sampling bias inhercnt
in studying clinical samplcs is eliminated,
thc association hctwccn p;1nic attacks ancl
M\'P vanishcs. This was shown by I-lartman et al. ( 1982) and Dcvcrcux et al.
(1986). 11ho studied MVP patients and
their family 111embers who had not sought
treatment themseh·es. Family me111bers
with and without P.IVP were not different
from each othcr 11ith respect to numbcr
of panic attacks ancl othcr symptoms:
both groups were for less symptomatic
than thl.! original samplc of MVP paticnts
who hucl been refcrrecl to the clinic. Thus.
sampling bias ma1· strongly intluence the
r.:sults or r.:se:m:h on clinical samplcs. lt
is therefore irnportant to complcment
such research by studies of the characteristics or nonclinical subjects with the
same disturb:mce.
A second important reason to study
nonclinical panickers is related to the fact
that many of these people are infrequent
panickers. T hey experience fewer attacks
than patients who seek treatment for the
full -blown syndrome. lnfrequent panickers may form the basic population out
of which some people will continue to
develop the full clinical syndrome. If this
is the case, infrequent panickers offer a
uniquc opportunity to study possibie
vulnerability factors for panic att:icks.
Most of our currcnt etiological rcsearch
on panic attacks is correlational in the
sense that groups of panic patients and
controls are compared in cross-sectional
designs. True experimental designs would
for instance involve attempts to produce
panic disorder in previously '' normal"
subjects. For obvious reasons such
studics cannot bc conductcd. Thus, we
cannot makc fi rm statcments abo ut causal
antecedcnts of the disorder. If infrequent
panickers are the basic popula tion for
panic disorder thcy should show 1vhatever
diathesis for panic exists. In this case we
can assume that characteri stics of patients
that are not found in infrequent p:mickers
are consequences rather tha n causes of
the disorder. On the other hand. infrequent panickers may represent a different
basic populat ion than frequent panickers
(panic disorder pat ients). lt is possible
that even though both groups show identical manifest symptoms (panic attacksJ
thcy suffer from different undc rlying
disturbances. In this case, infrequent
panickers shou ld be studied to gain insight into such a heterogeneity of the
panic atl::ick phenomenon.
Questionnaire Studies
Thc lirst studics of panic attacks in prcs umably normal populations were reported
Panic Altads in Nonclinical Sub.i.:cts
by Norton et al. (1985, 1986). In their
initial sludy, 186 studcnts wcrc sc.:rccncd
using the Hopkins Symptom Checklist
(HSCL-90, Derogatis et al. 1973) and a
specially designed anxiely questionnaire
asking for current level s of anxiety as weil
as frequ ency and sym1:ito111s of panic
attacks. A striking 34.4% of lhe subjects
reported having had one or more panic
attacks in the past year, and 2.2% reported
having had at least threc atlacks in thc
past 3 weeks. The symptoms reported lo
occur during these attacks were similar to
the ones described by clinical sampl es of
panic attack patients (ßarl ow et al. 1985;
Margraf et al. 1987). The most severe
symptoms were heart pounding, trcmbling, sweating, tlushing, and dizziness.
Subjects describing at least one panic
attack in the anxiety questionnaire scored
significantly higher than those withoul
attacks on the HSCL-90 subscales anxiety,
phobic.: anxiety, dcprcssion. intcrpcrsonal
sensitivity, somatization. and angcr/hostility. There were no significant difTerences with respec.:t to obsessive-c.:ompulsiveness. psychoticism, paranoid ideation,
or sleep difliculties.
In a second study, Norton et al. (1986)
screened 256 stude nts with a refin ed
version ot· their questionnaire. now
termed th e Panic Attack Questionnairc
(PAQ). Suhjects also complctcd thc S1;11cTrait Anxiety Jnventory (STAL Spielberge r et al. 1970), the Beck Depression
Inventory (BOI, Beck et al. 1961). and the
Profile of i\lood Stares (POMS. McNair et
al. 1981). In :iddition , subjects were either
gi\·cn thc Fear Survey Schedule (FSS-111,
Arrindell 1980) or the Fear Quest ion naire
(fQ. Marks and Math ews 1979). Very
similarly lO Lhe first SlUdy, 35.9% oi' thc
sampk reported having experienced at
least one panic altack in the past year. and
3.lt'O reponed having had at least three
attacks in the past 3 weeks. Panickers
scored significantly hi gher on state ancl
traic :.1nxie1y (STA I, anxicty scale of thc
PO~IS). depression (BOI, depression
I05
scale of thc POMS). littiguc (POMS), ;1nd
angcr (POMS). In c.:ontrast. there were no
difTerences compared to nonpanickers on
any ofthe FSS-111 or FQ subscales (agoraphobia, social phob ia, blood/injury
phobiil, aggression, animal phobia) or the
· POMS scales activity and confusion. Panickers and nonpanickers were similar in
the frequency of reported previou s treacments for any mental or physical disorder.
Similarly, thc two groups wcrc cornparable with respect to age, sex, or socioeconomic stalus. Panickers reported significantly more first-degrce relatives who had
panic attacks.
As in the first study, the most severe
symptoms of panic attacks were palpitations, trembling, sweating, clizziness. and
hot/c.:olc.J Oashes. Other characceristics of
panic attacks included a sucldcn onset in
thc majority of cases (59% unclcr 10 min).
an average of eight DSM-111 symptoms.
anc.J a widc variety or situat ional contcxts
in which attacks occurred. especially
social Situations. Thc grcat majority or
panickers reported havi ng expericnccd at
least OllC life Stressor at thc on se t of thc ir
panic attacks. Most frequentl y mentioned
wcrc dini c.:ult ies at work, fomily c.:rises,
and loss of a significant other. Subjects
who experienced some unpredictable attacks were different from those who
cxpcricnc.:cd only prcdic.:tablc atlac.:ks on 9
out of 40 comparisons. Those subjects
with unpredic.:table attacks reportcd more
attacks in more difTerent situations. as
weil as more severe feelings of unrealit y
and tachycardia.
Together. these studies show that panic
attacks may occur in more pcrsons than
previously assumccl and that suhjccts \\'ho
havc panic attacks rcport 111orc psychop;1thology than do nonpanickcrs. In add ition. the panic attacks experienced by
noncli nical panickers an d patients wi th
anxiety disorders arc \ery similar. \V hile
thesc studies have yielded some l'ascinating clala and initiated an imponant line or
researc h, they also pose some ne\\ chal-
106
J. \largrar and A. Ehlcrs
!enges. A tlrst problem is to establ ish the Panic disorder subjects showed more
reliability and validity of the PAQ as EMG and SCL reactivity to the two
compared to standard structured inter- imagcry tasks, in frcquent pa nickers more
view diagnoses. Norton et al. (1986) EMG and SCL reactivity to paced arithrcport ed that 22 out of 24 cases, previous- metic. There were no ditTerences in heart
ly identified as nonclinical panickers by rate react ivity or on the thought-listing
the PAQ. also met DSM-JII criteria for measure. With respect to panic symppanic. attacks in a structured interview. toms, panic disorder subjects scored signiHowever, they did not give information as ficantly higher on four and infrequent
to what interview was used, whether panickers scored higher on six out of 52
interviewers were blind to the qucstion- comparisons.
naire results, and the reliability of their Sandler et al. (1987) reported a compariinterview and questionnaire methods. A son of nonclinical panickers and control
second challenge is to go beyond mere subjects withoL1t panic attacks. Eighty
questionnairc asscssmcnt of psychopa- subjccts were recruited through screening
thology and to comparc nondinical pan- or collegc studcnts with a panic at tack
ickers. clinical panickcrs. and normal questionnaire. Subsamples included frecontrols 011 psychophysiological variables quent panickcrs, infrequcnt panickcrs.
or the response to strcssors. More recent and panic-free controls. Cardiovascular
studies hc.m:~ atternpted to addrcss these reactivity to a psychological (challenge
reaction time task) and a physical (cyissues.
cling) Stress task was assessed by measuring heart rate and blood pressure at
Laboratory studics
intervals before. during, and after the
tasks. All of the measures showed pro13cck and Scolt ( 1987) comparcd ten gressive dcclincs during thc prctask bascsubjects who had panic disordcr with ten linc pcriods. increases during the stress
infrequent panickcrs. All subjccts were tests. and declines during the post-task
recruited from thc community using rccovery phases. There were no differmedia <Jnnounccments. Infrequent pan- ences bctween groups with respect to
ickers were defined as subjects reporting cardiovascular reactivity. This is similar to
four of the 12 DSM-111 symptorns during what is seen in most so-called panic
typical attacks. but never having ex- induction stud ies of clinical samples (cf.
perienced three attacks in 3 consecuti ve Ehlers et al. l986a. 1986b. 1988: Margraf et
weeks. The Scree ning instrument was thc al. 1986a). In contrast to these studies.
Anxiety Disorder Interview Schedule however. differences in baseline heart
(A DIS-R. DiNardo et al. 1983). Laboratory rate levels were lacking. The only apparacligms \\·cre four 2-min tasks (neutral parcnt ditference was some modest cviimagcry. hosp italization imagcry, signal dcncc f"or slowcr rccovery from exercisc
detection. paced arithmetic). Assessments in those subjects reporting the most
included continuous measurcmcnt or frequent occurrence of panic att<Jcks.
trapezious electromyogram (EMG). skin lt is interesting to note that neither Sanconductance level (SCU. and heart rate. dle r et al. (1987) nor Beck and Scott (1987)
as weil as thought-listing and ratings ol' found difTerences in cardiovascular reacthe DSM-111 panic symptoms immcdiate- tivity. There were few reactivity difTerly following each task. Overall. thcre wcre cnccs between frequent and infrequcnt
relativcly !°<.:\\' dilTcrcnccs bctwccn panic panickcrs in 13cck and Scott's study and
disordcr subjects' anti inl'rcqucnt pan- no rcactivity differenc.es between panickers· res;ionses to the test paradigms. ickcrs and nonpanickers in Sandler et al. ·s
Panic Attacks in Nonclinical Subjects
study. This is in contrast to the consislenl
and strong differences on several of the
questionnaire measures in the two Norton studies. We have recently conducted
two studi es rrying to combine questionnaire and laboratory measures in a comparison of nonclinical panickers and controls.
The Marburg and Tübingen Studies
·Du ring late 1986 and early 1987 we conducted two independent studies of nonclinical panickers. The tirst study (Ehlers
and Meisner, in preparation) involved a
sample of 170 undergraduate students at
Philipps Univcrsity in Marburg in the
central parr of West Germany. The
second study (Margraf, Wrobel, and Jakschik, in prcriaration) involvcd a samplc
of 136 undergraduate studen ts at the
university of Tübingen in southcrn Germany. These studies pursued three specific goals:
1. To rcrlicate thc Norton et al. (1985. 1986)
lindings using a Gcrman translation or their
PAQ
2. To determinc the reliability and validity thc
qucstionnairc S(rccning mcthod ;is comparcd
to a stand:ird structurcd interview approad1
3. To comparc nonclinical panickcrs anc.J rnntrols
on a 4ucstionnairc battery and a psyd10physiulogical laborJtory assessment
or
sc~ee~·
r;
t<>rt~
•: ol · 9861
Tahlc 1. Odinitiun or a pani1.: alla1.:k giv1:11 in the
vcrsion or the panic allack qucstionnairc used in
thc Gcrman studics
A p:111ic attack (anxiety attack) is a discrete period
of suddcn onsct of intcnsc apprl'ltcnsion, fcar, or
tcrror, oftcn ;1ssociated with fcc lings ofimpending
doom. Thc following symptoms may be expericnccd:
Racing. f11Jl111<li11g. ur irregular lican 1>,·111
Di::iness ur ligl11he1ulecl11e.u
Slior1111.>ss of brealh
s,.,"a1i11g
Cl1es1 pain or disco111/iJn
Tre111bli11g or shaking
f/01 (f/1(1 cold jlashes
Choki11g or s11101heri11g se11sa1io11s
N11111b11ess or 1i11gli11g i n par1s 11/ 1//e btJdr
Fcar of 1~1·i11g
Fai11111ess
Nausea or abdo111i11al dis1ress
Feelings t!F 1111realizr or hci11g dewched
fror t!F losi11g comrol ur g11i11g cra:y
The questions on the following pagcs rcf.;r to
p;inic attads in situations th;1L werc not li fcthrcall·nini:. Th1.: atlacks h;1v..: lo be acco111pa11i..:d
by al least four of thc Symptoms listcd abovc.
Figure 1 gives an overview of"the studies.
Each study consisted or three phascs.
First, a !arge group of undcrgraduate
students was screcncd us ing our German
translation of thc P/\Q. Thc dclinition of
a panic attad givcn to thc suhjccts is
shown in Table 1.
STUDY l
STUDY 2
<Narourg)
<Tübingen)
~ AO
( Pente Al~!C" ~es~1ortne1re.
N• 170
unoergraouat e stuaenc s
107
N•; 36
Jnoergraouate
stuce~:s
1
+
·a11na· 1n terv •,;ws. SCID
{Stn:cti.;re-J Cllr ·\:)' l l"J~erv tew
for DSM . So1tze-.l. "' •!11011" 1986)
cc:n;ar1son
:attery,
osyc~oonys : : 1091ca 1
taooratory assessment
· e1:~0 ·
~uesuonna!r ~
re l1a:)l l1 ly
N•43
1
aoout
sor.
s :va'I
SO~
oan1ckers.
co~trol s oy PAO
re!1ao11 1ty
N•43
N•SI
SC ID IPAO Jan1ckers
vs contro ls
SC IO/ PAO par.1c~ers
vs controls
s:~:y
Fi::. 1. Q,·en i.:" of thc tvlarburg and T iibingcn studies uf pani' :lltads in nondink:il subj.:(ls
108
J. i\largraf und J\. Ehlcrs
In phase two, we selecled subsamples of
PAQ determined panickers and nonpanickers or "controls" (about 50% each)
using slrict critcria (panickers : reporting
at least one spontaneous attack, at least
four sympto111s, attacks not only in socia l
situations: controls: no attacks or anxiety
symptoms). We then conducted blind
diagnostic interviews to determine th e
agreement between the interview and
questionnaire methods. The interview
was a Gerrnan translatio n of the Structu red Clinical In terview for DSM (SCID)
hy Spitzer and Williams (1986). In the
third phasc. subjccts mecting bolh PAQ
and SCID criteria for panic attacks ("nonclinical panickers") and controls were
com pared using an extensive questionnaire battery and a psychophysiological
laboratory asscssment involving a baseline and a hyperventilation task. In addition. two substudies assessed the retcst
reliabil ity of the PAQ and th e interrater
reliabilit~· of the SCID in our hands. In
thc foll owi ng. wc will prcscnt thc rcsults
or thc prel i111inary analyses con ducted so
far.
The retest reliabi lity oft he PAQ provecl to
be genera lly good. Sim ilarly, thc intcrrater reliability of thc SCID in our hands was
good. Table 2 summarizcs the resulls of
the re test reliabi lity study.
Information about the occurren ce, number, and intensity of attacks as weil as
stress at the onset of panic, avoidance
behavior, ancl fami ly history was give n
rel iably. In contrast, subjects were not
able to give reliable information abo ut
whether they had ever experi_e_nced un expected ("spontaneous") attacks, panicked
only in social situations, or experienced
most of their sympto ms within 10 min.
Table 3 compares the number of subjects
reporting panic attacks on the PAQ in our
two studies with the numbers reported by
Norton et al. (1985, 1986). Th e mean ages
or our samplcs were 24 (study 1) ancl 25
years (study 2), while 69% (study l) and
65% (stud y 2) of all subjects were female.
While we found a somewhat higher
percentage of panickers for the past year,
resul ts for the past 3 weeks close ly rcsembl e those of Norton et al. (1985, 1986).
Thus, their finding of a high percen tage of
nonclinical panickers is replicated using
the questionnaire method. However.
whcn using thc structurcd interview approach, a difTerent picture emerge. We
found that only 12 out of 23 (study L
Marburg) and 15 out of 29 (study 2. Tübingen) PAQ-determ ined panickers also met
SCID criteria for pan ic attacks. \Vhile we
thus had a high rate of false positives.
there were only a few false negatives:
Tahlc 2. Rc1c•;t.rcliabili11· or thc Pa nie :\ttack Qucstionnai rc (samplc si zc:
days with :1 rncan or 20 days)
ltcm
(llf
gf<lups
or itcms)
E1·c r h:td p:1111c ;1ltack
E1·cr hau .3 :irtacks in 3 wccks
E"er worri<?d for .J wccks aboul attack
:\umbcr or .it tacks past yc;1r
II=
39. retest int.:rv:il l.J-28
Rcliabi lity cocmcicn t Statistic
0.80
Kappa
0.80
0.82
Kappa
0 .85
K;1ppa
Spe;irm:m
:'-lost sympwms wi th in 10 mi n
E1·cr had un~xpec tcd panic
Panics only in social s ituations
0.33
Kappa
0.53
O..J3
Kappa
Kappa
Strc·s~
0.67- 1.0
Kappa
.11 llll<,'I o r panic (8 ilcllls)
·\gc or on<c!. lrcalmcnl. sclr-111cdica1ion. fomiliy
his1ory ..?1 oidancc bcha1·ior ( l.J i1c111s)
\\ cr:ti,:c· duralion. ;111\ic11·. numbcr or -~· 111pl0111s
0.65 - 0. 7:!
0. 70-0. 76
Kappa. Spearman
· Spear111:111. Pcarson
l'anic Allacks in Nonclinical Subjects
109
Tablc 3. Frequency of panic auacks dcterminctl by lhc Panic Allack Questionnairc (pcrccnt or all subjects)
Studr 1:
Marburg
(11
= 170)
S1udy 2:
Tübingen
(11 = 136)
Norton el al.
1985
1986
(11 = 186)
(11 = 256)
Panickers
(last yearJ
46
59
34
36
Panickers
(last 3 weeksJ
21
29
24
23
Three atlacks
in 3 weeks (liferimeJ
12
15.5
*
*
Threc auack.s in
last 3 weeks
2
2
3
• These results were not rcportetl by Norton et al.
pooled across both studies on ly 4 out of
42 PAQ nonpanickers met SCID criteria
for panic. Overall rates of agreement were
as low as 74°/o and 65% (kappa: 0.50 and
0.32, studies 1 and 2, respectively). A post
hoc analysis of those subjccts who ind icated panic attacks on the PAQ, but dicJ not
meet SCID criteria, revealed that disagreement was not of a pure "chance"
nature. Rather. it secmcd that thcsc folse
positives reported milder variants of the
same rhenomenon (cf. thc conccpt of
limitecJ symptom attacks in IJSM-Il!R)
and that the interview had a morc con servati\·e cut-off bctwcen panic and nonpanic.
The comparison of nonclinical panickers
(P ..\Q and SCI D criteria) and controls on
the questionnaire battery yielded a numb.:r or pronounccd dilTerenc.:cs. Sincc
study 2 (Tübingen) used a more comprehensiYc battery. the pattern of'its rcsults is
shown in Fig. 2. Thc rcsults ol' stucJy 1
were generally similar. The questionnaires used in study 2 werc the Panic and
Agoraphobia Profile (PAP, cf. Margraf
and Ehlers 1987), Fear Survey Schedulc
(FSS. Arrindell 1980). Symptom Chccklist-90 (SCL. Derogatis 1977), Sclf'-rcport
ln\'entory of somatic symptoms (SISS.
King et al. 1986), state-trait anxiety inventory (STAl, Sp iel berger et al. 1970. trai t
form), 13eck Depression ln ventory (BOI.
Bcck et al. 1981), and the Mobility lnventory (MI, Chamblcss et al. 1985).
Thc different qucstionnaircs use very
different scales. For a standardized presentation, we computed thc difference
bctween the mcans of ranickers and
controls cJividcd by thc standarcJ c.Jcviation
of the control group. The bars in Fig. 2
thus indicate the di!Tcrencc hetween the
two groups in units ur lhe stantlarcJ cJeviation of the controls. The upper part of
Fig. 2 shows those scales 011 which the
two groups di!Tered signillcantly ( P <
0.05), the lower part scales without sign iticant dilferences. lt is important to note
that qucstionnaires mcasuring s i mil~1r
constructs also yiclckd similar rcsulls.
Thcrcf'urc. such scaks wc re groupec.J
together.
Nonclinic.:;tl panickcrs rcportcd cunsillcrably higher levels of phobophobia. agoraphobic kars (but not avoidance behavior). somatization. anxiousness. depressivencss. and injury phobia than nonpanickers of' comparabk agc, sex. and socioeconornic background. 01· the two depression scales. thc BOI that locusses more on
J. Margraf' and A. Ehlcrs
110
Slgn!rlcant Olfferences
2
stanoara
oevlattoos
or
control
grovP
-
PAP
F5S 5CL
5155 5CL
5155 5CL ST Al
pnobopnobla
agorapl\Oblc
rears
somatlzatlon
anxlousness
BOI SCL
SCL
FS5
depresston psycno- lnjury
tlclsm pllOt>la
No Slgnlflcant Olfferences
2
stanoarc
aevlattons
or
control
grau~
0
MI
PAP
avolaance
FSS SC1. PAP
soc1a1
anxlety
fS.S SCt.
PAP
host1 11 ty separaaggresslon
tlon
anxlely
SCl
FS.S
$ISS
PAP
otner varlat>les
Fii-!. 2. Cumparison of rianickcrs and con trols using a qucstionnaire battery (study 2. Tübingen). The
l•on rcrr.:<ent thc diffcrcncc hcl\\een thc scorcs ofthc two groups diviclcd hy lhc standard dcviatinn of
the c·11111rol !o!l"OllJl. Sc.::1h:s measuring sinular wnstrucls are groupcd logcthcr. The abbreviations
refcmn~ to thc qucsllon naircs containing the scales arc explaincd in the text. T he 11pper half of the
Jigurc sh ows scaks ) iclding signilicant dilf.::rcnccs bctween panickcrs and controls: phobophobia
a!;orarihob i~1 ( FSSL phobic anxicty (SCLl. total somatization disorder score (SISSl. somati z:it ion
!SCLl. ncurotic symp1oms (SISS). anxicty (SCU. trail anxiety (STAIJ. deprcssion (BDll. depression
!SCL>. ps\-choticism (SCU. and blood/injury phobia (FSS). The scales listcd in the lo11·er ha(fof the
ligurc ,·ieldc:d no signilican t di!Tcrcnccs bctwccn panickcrs and controls: mobility alone and mobility
accompanicd (MI>. gcncral avoida ncc {P/\P). social fcars (FSSl. intcrpersonal sensitivity (SCL). fe ar of
social cmb~1rrassmcn1 ( P:\ Pl. aggrcssion (FSS l. hosl ility (SCU. scparation nnxiety ( P.-\Pl. obsessi,·e·
rn111puJ,l\c'llC\.'> and paranoid idc:llion (S('I.). :111i111;1i phobia !l'SS). ~art!iuvascular. gastrointcstin:1l.
:111t! mus.:ubr :1w;m:ncss (SISS). l'car oi' loss or control :111d J'car uf somatic di$tress (PAP)
(P:\Pl.
Panic Attacks in Nonclinical Subjects
the cognitive concomitants of depression
yielded a strenger difference than the
depression scale of the SCL-90 that con·
tains more vegetative sym ptoms. The
ditTerence on the SCL-90 · psychoticism
scale is probably due to several ambiguous items that can be interpreted as
signs of psychotic ideation as weil as
indicating typical panic symptoms (e.g., a
fear of going crazy, losing control over
one's body, derealization). Somcwhat
surprisingly, there were no ditTerences in
terms of self-reported avoidance behavior,
social anxiety, hostility, or aggression.
Separation anxiety, which has been linked
causally to the development of panic
auacks (Klein 1980; cf. Margraf et al. 1986b
for a critique), was not heightened in
nonclinic:.il rianickers. lt should be noted
that the separation anxiety scale used
hcre has bccn shown to be highly sensitive to the separation anxicty found in
clinical panickcrs as weil as in agorapho·
bics (Margraf and Ehlcrs 1987).
The results fo r the baseline and hyperventilation tasks of the psychophysiological
laboratory assessment are summarized in
Fig. 3 (study 1) and 4 (study 2). We chose
hyperventilation (60 cyc les/mi n, 2 min) as
thc strcss task bccausc it has l'requcntly
been associated with panic attacks. Sep·
arate repcated measures ANOVA's (using
the Greenhouse-Geisser correction when
appropri:.ite J for the ditTerent dependent
variables showed signiticant baseline differences between panickers and controls
in self-reported anxiety and panic symptoms, but not control symptoms which
are not usu:illy associated with anxiety,
heart rate. systolic blood pressure, and
diastolic blood pressure. Ulood pressure
results ;1rc not included in thc ligurcs.
The responses to the hyperventilation
task were similar in both groups with the
exception of a greater increasc in selfr:ited Jn\iety in panickers. In study 2
(Tübingen! the EKG was monitored continuousl y throughout the .different paradigms and a rath er strong heart rate
III
in crease in response to hyperventilation
was observed. This was not the case in
study 1 (Marburg) because heart rate
could not be measured during but only
before and after paradigms.
Overall, the results of the Marburg and
Tübingen studies replicate earlier findings: There is a high numbcr of persons
with panic auacks in nonclinical samples.
These persons also show more seff.
rcported psychopathology, but not thc
cardiovascular differences typical for clinical cases of panic disorder. These replications are complemented by data on the
reliability and validity of the questionnaire screening method and results from a
more comprehensive battery of questionnaires. In addition, hyperventilation was
again shown to proclucc incrcases in
anxiety, panic symptoms, and heart rate.
Nonclinical panickcrs showcd high t:!r
baseline anxicty anti a greater rcspo nse to
hypervcntilat ion on the anx iety rating
scalc than nonpanickers.
Conclusions
Takcn together, published stud ies of
nonclinical or infrcqucnt panickcrs ancl
our own prc liminary results suggcst tha1
Norton et al. (1985. 1986) idcntificd a val id
phenomenon. Panic attacks occur relativcly frequently in nondinical subjects.
As in clinical studies, the exact proportion
depends in part on thc rncasures or
critcria we use to determine panic attacks.
In our studies, at least 501Vo ot· questionnaire-determined panickers did not meet
SC ID criteria ror panic attad;s. This
occurred in spite or the fact that the
suhjccts invitcd for the interview h;1cl not
only indicatcd a panic attack but also
reported on the PAQ at least one spontaneous anack, at least fou r symptorns
during att;icks. and panic attacks not only
in social situ:llions. Thus. the proport ions
or panickcrs givcn in Tablc J arc probably
uppcr limits or thc prcvalcncc or panic
112
J. Margraf and /\. Ehlcrs
Anxlety Rating
(0- 10)
3
-II-
Noncllnical Panlcl<ers
-0-
Controls
HV: hyper.ienttlatlon
baseline
HV
posl
18 - , - - - - - - - - - - - - - - ,
Panlc Symptoms
16
(0-18)
14
54
52
50
12
48
46
10
44
8
42
40
38
36
6
4
2
0
~--.-----...,.....----....--'
oasel 1ne
11
10
9
HV
State Anx l ety
( 20- 80]
34
32
oase11ne
posl
HV
post
110
Control Symptom5
(0-11)
100
8
7
Heart Rate
(bpm)
90
6
5
80
4
3
2
70
1
o.L.-~O----~=-=--=-==-=-=-~~~====~o~__J
basel 1ne
HV
post
baseline
HV
post
Fi~.
3. Sc!ected rcsults of thc 11S)'chophysiologic;il laboratory assessment in stud y J (~brburgl. Shown
:1rc sclf·r:ucd an:-;icty (on a 0-10 scalcl. numbcr of panic symrHoms (on a list of 18 symptomsl. state
;111 '\icty <ST ·\I. 20-1-iO ~ca!.:. Spiclbcrgcr ct al. 1970>. numbcr of control symptoms (on a list of 11 symr·
!Omsl. anJ hcan rate (in bcats per min) al hasclinc. during hype~cntilation (2 min. 60 cyclcs per minl.
and ;1t thc ~nd of thc lahoralory scssion. 1lcan rate was mcasurcd at baseline. immed iately before and
a/rcr h)·pcn cntilation. and al lhc end 01· lhc scssion. Black squarcs represenl panide rs. Of1!'1l squarcs
rcprcs.:111 't>ntrols
altads in nonclinical subjcc ts. I n spitc of
thc high number or "falsc positive"
rcsults. thc low proportion of "falsc ncgati\ c„ rc~ults ;tnd ils g.oocJ rctcst rcliabilit~·
makc thc PAQ a valid screening cJevicc.
Howcvcr. if one wants to assure compalibility \\'ith the diagnost ic standards in
clinical stucJies. a·structured inter\'iew has
lO complemcnt the questionnairc in its
prcsent t'orm. Nevertheless. it may be
l'anic Allacks in Nonclinirnl Subjc~ts
113
5...-------------,
Anxlety Rating
(0-10)
3
2
-----
Nonclinical Panicl<ers
-D-
Controls
HV: hyperventi lat i on
0
~-r---.---.---.---.---.--'
base l ine
HV
18...------------~
16
14
Panic Symptoms
CO- 18)
54~------------
52
Slate Anxiety
50
(20-80)
10
48
46
44
8
42
12
40
38
36
34
6
4
2
0
~--,-----.-----.--......J
basel ine
32...._-..-----..-----..-----'
baseline
hypervent l lat Ion
II . . . - - - - - - -- - - - - - - .
1o Contra 1 Symptoms
g (0-11)
8
7
6
hypervent1lat1on
1 10 T - - - - - - - - - - - - - - ,
Hearl Rate
100
(bpm)
go
80~
5
4
3
~~
70
o~--.------T-----r-~
baseline
hypervenli lalion
basel tne
HV
fig . .J. Sde..:tcd rcsults of the psychophysiological laborawry assessmcnt in stu dy 2 ( fübing..:11). Slww11
are the same 1·anables :is in Fig. 3 du ring a 12-min basclinc and a 2-min hypcr\'cntilation test (60 .:yclcs
per minl. For the hyperventilation pmadigm. self-report mcasurcs werc takcn immcdiatcly heforc anti
aftcr hypcr\'cn: i:~ tion. hcart rates wcrc c:!lculatcd l'rom thc EKG imm.:diatcly h..:l(>rc allll d11ri11.~ thc: last
20 s of hyper\'en:ilacion. Black squares represent panickcrs. 011e11 s1111arcs reprcscnt con tro ls
possible to de\·elop future forms of the
PAQ that agree better with instruments
such as the SCID. lt is also possible that
the low agreement was due to the lad; or
reliability of .:enain criteria for the diag-
nosis of panic in DSM-llIR. The fact that
information about the "spontant.:it(' and
the rapidity of onset of panic altacks was
not givcn reliably raises doubt as to the
usefulness of these critt!ria.
ll4
J. :vlargrar and A. Ehlcrs
What are nonclinical panickers like?
Thcrc arc a numhcr of variahlcs in thc
scll~rcport domain thal difl'crcntialc nonclinical panickcrs from cont rols. These
are primarily measures of phobophobia,
agoraphobic fears, somalization, anxiousness, and depression. The physio logical
variables assessed so far as weil as reactivity to laboratory stressors difTerentiate
· much less weil or not at all between
panickers and controls in nonclinical
samples. The most consistent ditTerence
found in our laboratory assessments were
lonicallr clcvatccl levels of sclf-rcportccl
anxiety and symptoms. Reactivity to
stress tasks ditTerentiated only poorly and
cardiovascular measures ditTerentiatcd
not at all in our studies and in that of
Sandler et al. (1987). Even the signiflcant
difTerences on laboratory parametcrs
reach a magnitude of only about one
Standard deviation (of the control group)
ancl arc thus much smallcr than so mc of
thc qucstionnairc diffcrenccs.
S..:veral or thc fcaturcs of nonclinical
panickcrs havc previously been founcl in
clinical panic disorder patients (e.g.,
phobophobia. somatization, gcncral anxiousness. dcprcssion). Howevcr. further
studies arc necded that directly comparc
nonclinical and clinical samples. lt is an
open question whclhcr thc infrequent
panickers stuclied by 13eck and Scott
(1987) rcprcscnl a clinical or a nonclinical
popu lation since all subjects were
recruited through media announcemcnts
IOr pcopk with panic attack s. In our
experience. infrequcnt panickers who
respond 10 such advertisements are more
similar to self-selected clinical cases than
to noncli nical subjects rccruited from
community screenings. This could be one
reason for the lack of ditTerences between
the two samples rerorted by 13eck and
Scott (1987).
The results of studics of nonclinical
panickers are consistent with the psychophysiologic:.11. cognitive. or psychological
models of ranic proposcd by several
researchers (e.g., Barlow 1986; Clark 1986;
Margraf et al. !986a, 1986b; van den Hout,
1988) in showing a number or postulatcd
causal factors for the development of
panic (c.g., fear of anxiety symptoms.
anxiety response to hyperve ntilation) to
be present in this population. They are
not consistent with views that assume
separation anxiety or active avoidance
behavior as necessary an tecedents of
panic attacks.
If we want to use these results to make
more causal statements about the development of panic attacks, we need prospcctive longitudinal studics. These
studies have to determine whether infrequent panickers are the basic population
out of which some subjects go on to
develop the full-blown clinical picture of
panic disorder or even agoraphobia with
panic attacks or whether the phe nomenon
of panic is heterogeneous, representing
different subgroups of underlying causes.
Eithcr possibility is of' high scientific and
clinical intcrest. In the first case, we havc
a fascinating opportunity to study possible
vulnerability factors in subjects at a high
risk to develop panic disorder. In the
second case. we may gain insights into
difTcrential etiologies of panic attacks
possibly connected to clinical outcome in
the long run. In addition, the longitudinal
study of noncl inical panickers may give us
information about possiblc factors protecting most of them from becoming
clinical "cases". We have recently started
such a prospcctive longitudinal follow-up
study of infrequent panickers at the
Clinical Research Unit ofthe Oepartment
of Psychology at Phil ipps University. On
the whole. the studies presented in this
chapter il lustrate the usefulness of supplementing the usual study of clinical
samples by investigating of panic attacks
in noncl inical subjects.
.-lck11011 /edge111e11rs: Preparation of this
chapter was supported in part by German
Research Foundation grant Eh 97/1- 1.
1
Panic J\ttacks in Nonclinical Subjects
Additional financial support by the
Department of Psychology of Philipps
University and the help of 1. Florin, G.
Jakschik, W. Lutzenberger, K. Meisner, ß.
Rockstroh, F. Schneider, and F. Wrobel is
gratefully acknowledged. ·
References
Amaican Ps:..:hiatric J\ssociation ( 1987) Diagnostic and statistical manual of mental disorders.
Third edition-revised . Ameri'can Psychiatrie
Press. Wash ington, DC
Arrindell \VA (1980) Dimensional structurc and
psychopathology correlales of the Fear Survey
Schedule (FSS-III) in a phobic population: a
lactorial ddinition of agoraphobia. lkhav Res
Ther 18 :229-242
Barlow OH (1986) A psychological model of
panic. In: Shaw BF. Cashman F, Segal ZV.
Vallis Tm (eds) Anxiety disord ers: th eory,
diagnosis. and treatment. Plenum. New York
ß;1rlow OH. Vermilvea J. ßlanchard Eß. Vcrmilyea BB. Di Nardo PA. Cerny 11\ (1985) The
phcnomcnon or panic. 1 J\bnurm l'sychol
9-1:320-328
Beck JG. Scott SK (1987) Frequen t and in frequent
panic: a compa ri son of cogniti ve and au1ono111ic reactivi ty. J Anx Disord 1:47-58
Beck AT. \\'ard CH, Mendelsun M. Mock 1.
Erbaugh J (196 1) An inven lory ror mcasuring
depressi on. :\rch Gen Psychiatry -1 :561-57 1
Chambless DL. Caputo GC. lasin SE. Gracely El.
Williams C ( 1985) The mubility invcntory for
agoraphob1;.1. lkhav Res Thcr 23 :35--1-1
C!Jrk DM (1986) A cognitive approach tu panic.
Bt!hav Res Ther 24 :-161--170
Oerogat is LR (1977) SCL-90. Adm inistration.
scoring and procedures. Manual-! for the
r(evised 1 \'<.!fSion and otha instrumcnts of lhe
psychopath ology rating scale scrics. Johns
Hopkins L"ni\·ersity School of Mcdicine. llal timore
Derogatis LR. Lipman RS. Covi L (1973! HSCL90. An outpatien t psychiatric raling scalc:: preliminary rerort. Psycho pharm acol Bu ll 9: 13-25
De\·aeux RB . Kramcr-f-ox R. ll rown WT. Shcar
\IK. l l:1rtm,1n N, Kliglield I'. Lutas EM, Spiucr
\IC. Litwin SD (1986) Relation bctwccn clinical
k~1lurcs ut' thc mitral v;tlvc prolapsc sy 11tlru1111.:
and c:chu.::irdiographicall y documcntcd mitral
\,il\'c proL1pse. 1 Am (oll Cardiol 8:763-772
Oi:"\ardo P.- \. O'ßricn GT. ßarlow 01 1. \Vaddcll
\IT, lfön.:h;ird Eß !19S3 l Reliabili1y or L)SM 111 an.\iety disordcr c:l!egorics using a ncw
115
structure interview. J\rch Gen Psychiatry
40: 1070-1075
Ehlers A. Margrafl, Roth WT (1986a) Experimental induction of panic attacks. In : Hand 1.
Wittchcn H-U (eds) Panic and phobias. Springer, ß erlin Heidelberg New York
Ehlers A, Margraf J, Roth WT, Taylor CB,
Maddock Rl, Sheikh 1, Gossard D, Blowers
GH , Agras WS, Kopefl BS (1986b) Lactate
infusions and panic attacks: do patients and
controls respond dilforen tly'! Psychiatry Res
17:295-308
Ehlcrs J\. Margraf 1, Roth WT (19SSJ lnteract iun
of expectancy and physiological stressors in a
laboratory model of panic. In: Heilhammer D.
Florin 1. Weiner H (eds) Neurobiological
approaches to human disease. Huber. Toronto
Hartman N. Kramer R. Urown WT, Devt! reux RB
(1982) Panic disordcr in patients with mitral
valvc prolapsc. /\111 J l'sychiatry 139 :669-670
King R, Margrar 1, Ehlers A, Madd oc k RJ (1986J
Panic disordcr - ovcrlap with symptoms of
somatization disordcr. In: Hand 1, Wittchen
H-U (eds) Panic anti phobias. Springer, Be rlin
J-lcidclberg New York
Klein DF (1980) /\nxicty reconccptualizcd.
Compr Psychiatry 21 :411-427
Margrar 1. Ehlers /\ ( 1987) Fcar or lear in pan ic
disordcr and agoraphobia: thc l'anic and i\goraphobia Profile (l'AP). 17th Annual meeting of
the European Association for ßehaviour Therapy, Amsterdam, August 1987
Margraf 1. Ehlcrs /\, Roth WT (1988) t-.iiml valve
prolapse and panic disordcr: a rev icw or their
relationship. Psychosom Mcd in press
Margraf 1. Ehlcrs A. Roth WT (]986a) Sodium
lactatc infusions and panic attacks: a rcview and
nitiquc. 1',ycl10so111 Mcd -18:~.1-:'I
Margrar J. [hlcrs /\. Roth WT ( 1986bl Hiologic'a l
modcls of panic disorder and agoraphobia: a
revicw. ßehav Res Th cr 24 :553-567
Margrar J, Taylor Cß. Ehlcrs A. Roth \\T, Agras
WS (1987) Panic attacks in thc natural .::nvironmcnt. J i\erv Ment Dis 175:558-565
Marks IM. \lathews AM (1979) Briefstandard
scll~rating for phubic patients. Uchav R<!s Ther
17: 263-267
McNair o;-.1. Lorr M. Drop plc man LF (1981)
Profile of' \food Statcs. Edits, San Di.:go
l\kndc l JGC. Klein DI' 11%9) ,\1p.; i<:ty attac~s
with suhsequent ;1goraphohia. Compr l'sychiatry IO : l'JO-l'J:i
i\lotulsky :\(i ( l'J78) B i;1~<.: d asc<.:rla i11111c11 l and
thc natural history ur discas~s. N Eni;I J M.:d
298: 11%-1190
Nonon UR. Harrison B, l lauch J. Rhodes L (1985)
Characteristics of pcupk with infrc(Jut:lll panic
attacks. J :\bnorm Psychol 94: 216-22 1
116
J. Margrar and A. Ehlcrs
Nort on G R. Dorward J. Cox BJ ( 1986) Factors
associ:1tcd with panic at t:1cks in nondinical
suhj,·ri-. lkh:n· Tlll'r 17 :2.1'1-2 52
l{,•gi.:1 IJ:\. My..:rs JK. Kra111er M. Rubins LN.
Blazer DG. l-lough RL. Eatu n WW. Lode 13Z
( 1 98 ~ ) Thc NIMI 1 epi dcmiologic catchmcnt
area program - historical context. major objectives. and study population charactcristics.
Arch Gen Psychiatry 41 :934-941
Sandler L. Wilson KG. Ramsum D. Asmundson
G. Ashton G. Larsen D. Schumacher B (1987)
Psychophysiological characteristics of individuals with frequent panic attacks (Abstract).
Psychop hysiology 24 :609
Spielherg.:r CD. Gorsuch RL. Lushcnc RE (1970)
Stalc-Trait Anxiely Jnvenlory. Cunsulting Psydwlni;ists Press. Palo Alto
Spitzer RL Williams JB (1986) Struct ured Clinical
Interview fo r DSM (SCIDJ. New York Statc
J>syd1iatric Inst itute. New York
Thyer 13A. ll irnlt.: J (1985) Temporal relationship
bctween panic attack onset and phobic avoidance in agora phobi a. Behav Res Ther
23:607-608
van den Hout MA (1988) The explanation of
experimental punic. In: Maser J. Rachman S
(eds) Panic and cognition. Plen um. New York
Weissman MM. Leaf PJ. Blazer DG. Boyd JH.
Florio L ( 1986) The relat ionshi p between panic
disorder and ago raphob ia: an ep idem iologic
pcrspective. Psyc hopharmacol 13ull 22:787- 791
Wittchen 11-U ( 1986) Epidcmiology of panic
attacks and pan ic disorders. In: Hand I. Wiuchcn 11-U (eds) Panic and phobi:is. Springer.
13crlin Heidelberg New York