PAT general presentation

Transcription

NIR-Control of a Pan-Coating process
Process Analytical Technology PAT: Wo stehen wir heute?
Hotel Arte, March 4th 2010
Amazing where you can go
Outline
• The Product
•CQA (Critical Quality Attributes)
• The Process
•Research and Development
•Full Scale Production
• Development of the NIR Model
•Feasibility
•In-line Model
•Off-line Model
• The Future
The Product
The Product
O
O
C
C
CH3
O
O
R
OH
+
+
O
C
O
Glycolide
DL-Lactide
H
O
CH C
CH3
O
O
O
CH C
CH3
catalyst
C
CH3
O
O
heat
O
O
CH2 C
x
O
O
CH2 C
O
R
y
Poly(DL-lactide-co-glycolide)
A DL-lactide-co-glycolide polymer (PLG copolymer) was chosen as the coating
barrier material
Critical Quality Attributes (CQA)
•
•
•
•
•
•
•
•
Controlled release product
Start release of API after 70 days
Dissolution profiles
Dissolution depending on coating thickness
Coating thickness – CQA
Coating Process – Critical Process
Coating Endpoint - CPP (Critical Process Parameters)
Monitor and control the CPP
Long Time Dissolution Results
100
ND um coating
80
17 um coating
60
27 um coating
40
42 um coating
20
Dissolution Time, day
98
84
70
56
42
28
14
0
1
Cumulative Release, %
uncoated pellets
The Process
Granulation
(includes Drying and Final Blending)
Tableting
Coating
Filling and Packaging
Experimental Scale Fluid Bed Coater
Small Scale Coater
Sampling Location Small Scale Coater
Sampling Location Production Coater
The Installation
Inside the Coater
Coater Pan Compressed
Ionizer
Air Lines
Spray Nozzles
Warm
Air Inlet
NIR Probe
Mixing Element
Air Out
Polymer
In/Out Lines
Applications with Reflection Probes
Coating Process Controls
Coater
▪ Drum speed
Dehumidifier
Inlet air
▪ Flow rate
▪ Humidity
▪ Temperature
Acetone
burner
Outlet air
Temperature
Pump
▪ Spray rate
▪ Flow rate (Recorded)
Pellets
▪ NIR modeling
▪ Weight gain
▪ Visual inspection
Spray suspension
▪ Weight prepared (Recorded
for commercial production)
▪ Weight applied (Recorded
for development batches)
Referenzmethode
SEM (Scanning Electronic Microscopy)
SEM Experimental Procedure
x 2000 magnification (to acquire
accurate measurements of the
coating thickness
The measurements are reported
from four SEM images per tablet
SEM Pictures from different coating times
after 2 HR coating
after 3 HR coating
after 6 HR coating
Off-Line Configuration
FT-NIR Spectrometer MPA
Scheme of Integrating
Sphere
Integrating Sphere
0.5
Absorbance Units
1.0
1.5
2.0
Spectra of the Uncoated Pellets (Red), Coated
Pellets (Blue) and Coating Polymer (Black)
10000
9000
8000
7000
6000
5000
Wavenumber cm-1
C:\Kunden\Intervet_NIR\Coating.3
Coating
C:\Kunden\Intervet_NIR\Wirkstoff mit Coating.3
C:\Kunden\Intervet_NIR\Wirkstoff.0
Wirkstoff
Matrix-FE, Granulat
Wirkstoff mit Coating
Matrix-FE, Granulate
24/10/2003
Matrix-FE, Granulat
24/10/2003
24/10/2003
0.1
0.2
0.3
Absorbance Units
0.4
0.5
0.6
0.7
0.8
Expanded Spectra of Uncoated Pellets (Red)
and Pellets after 1 to 8 Hours Coating
11000
10000
9000
8000
7000
6000
5000
Wavenumber cm-1
C:\Kunden\Intervet
C:\Kunden\Intervet
C:\Kunden\Intervet
C:\Kunden\Intervet
C:\Kunden\Intervet
C:\Kunden\Intervet
C:\Kunden\Intervet
C:\Kunden\Intervet
C:\Kunden\Intervet
NIR\Messungen0401\Intervet
Österreich\Revalor LA
Tbl
Tbl
Tbl
Tbl
Tbl
Tbl
Tbl
Tbl
Tbl
L006
L006
L006
L006
L006
L006
L006
L006
L006
1B
1B
1B
1B
1B
1B
1B
1B
1B
0h
1h
2h
3h
4h
5h
6h
7h
8h
20
20
20
20
20
20
20
20
20
11
11
11
11
11
11
11
11
11
2003.0
2003.0
2003.0
2003 0
2003.0
2003.0
2003.0
2003.0
2003 0
Intervet
Intervet
Intervet
Intervet
Intervet
Intervet
Intervet
Intervet
Intervet
Österreich;Emissionsmessung;Rev 14/01/2004
0.0
0.2
0.4
Absorbance Units
0.6
0.8
1.0
1.2
1.4
Triplicate Sample Spectra
(Integrating Sphere)
9000
8500
8000
7500
7000
6500
Wavenumber cm-1
6000
5500
5000
4500
\\IAMFl02.d50.intra\uh2$\sunx\Revalor\NIR Model\D002 5.0 h.0
Kugelmessungen;Intervet;D002 5.0 h ;
14/09/2004
14/09/2004
14/09/2004
Feasibility Study
•
•
•
•
•
•
•
•
•
20 samples of coated tablets
Uncoated tablets
Pure PLG Co-Polymer
Clear spectral features to differentiate uncoated and coated tablets
Clear changes in different regions of the spectra depending on the coating
time in the plot of the second derivate
Quantitative calibration was performed using PLS algorithm
Clear correlation between NIR spectra and coating time
8 of the 20 samples were correlated to the reference values of coating
thickness
It was shown that NIR spectroscopy is a feasible method to measure and
predict the coating thickness
Off-Line Method and NIR-Model
• Total of 35 samples pulled from various coating times (known reference
value by SEM)
• Calibration Set (22 samples from three different batches)
• Validation Set (13 samples from two different batches)
• NIR equipment: Bruker MPA with integrated sphere in diffuse reflection
• NIR parameters: Resolution 8 cm-1; 64 scans, 3 measurements per sample
• Division of NIR spectrum into 9 regions based on polymer signals
• Regions 3 (7562 - 6657 cm-1) and 7 (5380 – 5056 cm-1) excluded
• Second derivate used (better variations)
• PLS (Partial-Least-Squares) Method used to predict coating thickness
• Model had RMSECV = 2.79 and R2 = 96.54
Division of NIR Spectrum into 9 Regions
1
2
3
4
5
6
7
8
9
3 finally chosen NIR optical Regions
-0.00006
-0.00004
Absorbance Units
-0.00002 0.00000
0.00002
0.00004
Plot of Second Derivative
6400
C:\Kunden\Intervet
C:\Kunden\Intervet
C:\Kunden\Intervet
C:\Kunden\Intervet
C:\Kunden\Intervet
C:\Kunden\Intervet
C:\Kunden\Intervet
C:\Kunden\Intervet
C:\Kunden\Intervet
6200
6000
5800
Tbl
Tbl
Tbl
Tbl
Tbl
Tbl
Tbl
Tbl
Tbl
5600
L006
L006
L006
L006
L006
L006
L006
L006
L006
1B
1B
1B
1B
1B
1B
1B
1B
1B
0h
1h
2h
3h
4h
5h
6h
7h
8h
20
20
20
20
20
20
20
20
20
5400
11
11
11
11
11
11
11
11
11
2003.0
2003.0
2003.0
2003 0
2003.0
2003.0
2003.0
2003.0
2003 0
5200
Intervet
Intervet
Intervet
Intervet
Intervet
Intervet
Intervet
Intervet
Intervet
5000
Österreich;Emissionsmessung;Rev
4800
Correlation Tablet Wt. Gain vs. NIR
Correlation of Pellet Wt Gain vs NIR
Prediction of Coating Thickness
Heifer Lot D002
20
35
y = 1.6475x + 1.116
R2 = 0.9871
Coating thickness, um
Coating thickness, um
25
15
10
5
0
0.00
5.00
10.00
Correlation of Pellet Wt Gain vs NIR
Prediction of Caoting Thickness
Heifer Lot D003A
15.00
Pellet net w eight gain per sam ple, m g
30
y = 1.7546x + 1.2241
R2 = 0.9967
25
20
15
10
5
0
0.00
5.00
10.00
15.00
20.00
Pellet net w eight gain per sam ple, m g
In-Line Method and NIR-Model
• A Bruker Matrix-F equipped with a Solvias Reflector probe was installed
• During the production of the coater qualification batches, both in-line and offline measurements were taken
• Same shaped NIR spectra with slightly different characteristics
•Smaller measuring spot size
•Lower throughput of the fiber probe
• Best results by using 128 scans per sample
• Scans were executed every four minutes
• The final off-line model was transferred to the Matrix-F and adapted for the
in-line fiber probe
• The off-line results were used as reference values for the calibration
• One of the qualification batches was used for the validation of the in-line
model
Overlay in-line and off-line spectra
NIR-Model update (off-line)
• 16 samples from each of five new batches (80) taken every 30 minutes
during the coating process
• Samples were measured off-line using a Bruker MPA with integrated sphere
• The coating thickness of the samples was analyzed by SEM to provide the
reference values
• Slight deviations in some spectra
• No perfect fit of prediction of high coating thickness
• To make the model more robust it was expanded with part of the data of the
new batches
• The updated model was validated with the second part of the samples
NIR-Model update (in-line)
• The off-line NIR results were used as reference values for the corresponding
in-line spectra
• An updated in-line NIR model was created
• Half of the data was used for calibration
• The other half was used as an independent test set for validation
• Led to prediction errors of 0.6 µm coating thickness and R2 = 99.62
Cross Validation of the new MATRIX-F Model
Predicted coating thickness by NIR in
comparison to the SEM reference values
35
30
25
Thic k ne s s (µ)
20
15
10
5
0
00:00
01:12
02:24
03:36
04:48
-5
Time
06:00
07:12
08:24
09:36
Incomplete Letter Comments from FDA
•
•
•
•
•
•
•
•
•
•
Define sampling technique from coater
Define how to run the NIR
Describe how the sample is taken and run in the NIR
Explain data pre-processing methods, and information about the SW
Identify the characteristic bands related to coating thickness in the NIR
spectra
Does the SW normalize data? Describe regions, ranks, factors, frequencies
Need SOP on upgrade of library procedures and the model
Provide robustness information as part of the NIR validation
Submit in the Annual Report the documentation for the update of the NIR
library
Describe the critical parameters of the Coating Process
Benefit for Customer
•
•
•
•
•
Strong support from FDA during product development (reduced time)
PAI (Pre Approval Inspection) science focused and PAT driven (no 483)
Only one circle review process
Product approval more than 6 months ahead of schedule
Product launch more than 6 months ahead of schedule
The Future Process
Granulation
(includes Drying and Final Blending)
Starting Materials
Tableting
Coating
Filling and Packaging
Real Time Release
Thank you for your attention.
Peter Bauer
Head of PAT Business Unit
Phone: +41 61 686 60 50
[email protected]
www.solvias.com

PAT general presentation

Transcription

Similar documents