GSK PowerPoint templateBasic principle of stability study design
and analysis for product licensure
T I M S C H O F I E L D C M C S C I E N C ES , L LC
WO R KS H O P O N STAT I ST I C A L A N A LY S I S O F STA B I L I
T Y T EST I N G
2 1 ST A P R I L 2 0 2 1
Workshop on Statistical Analysis of Stability TestingWorkshop on
Statistical Analysis of Stability Testing
Some notes on vaccine stability • Stability study goals differ
throughout the product lifecycle
• The goal of a vaccine stability study is to estimate an important
stability parameter (e.g., slope) rather than demonstrate
conformance to specifications
• Stability study design and analysis should be effective at
meeting the study goal (e.g., reducing uncertainty)
• Vaccine properties such as potency should remain within
appropriate minimum and maximum requirements throughout shelf
life
• The goal of lot release is to demonstrate that a manufactured lot
will be effective throughout shelf life
2
Outline • Some basis concepts
• Studies supporting product licensure
• Determination of release potency
• Managing the release limit through design
• Summary
3
Workshop on Statistical Analysis of Stability TestingWorkshop on
Statistical Analysis of Stability Testing
Some basic concepts • Most vaccines are unstable Require storage
from 2-8°C (protein and conjugated polysaccharide vaccines)
to
-70 °C (mRNA vaccines)
• Kinetics are often 1st order (esp. potency) = −
Revealed in early accelerated stability studies
• Most meaningful attributes of vaccines are measured using highly
variable assays e.g., potency – in vivo for older vaccines (>50%
GCV)
or ln = ln −
4
Workshop on Statistical Analysis of Stability TestingWorkshop on
Statistical Analysis of Stability Testing
Some basic concepts (cont.) • Stability studies should be designed
to provide a suitable fit to
statistical (e.g., kinetics) models
• Statistical design is aimed at controlling (minimizing) the
uncertainty associated with the statistical model Uncertainty = z ⋅
Variability (e.g., 2-sigma) Lower uncertainty results in lower risk
Risk (e.g., probability of an OOS) can be visualized
as the area under a normal curve
Stability design elements include replication, time points, and
ranges between time points
5
USLLSL
Studies supporting product licensure
Long term stability of drug substance (DS)
Long term stability of drug product (DP)
Accelerated stability at conditions of handling, excursion, and
use
6
Workshop on Statistical Analysis of Stability TestingWorkshop on
Statistical Analysis of Stability Testing
• Study goal: Establish that DS can be stored at a specified
storage
condition for a specified period of time without impact on final
product quality
• Quality criteria: DS is suitable for filling into final container
throughout DS
shelf life
Note: this is a business risk
Studies supporting product licensure Long term stability of drug
substance
7
• Adequate potency to ensure meeting target DP potency
8
Studies supporting product licensure Long term stability of drug
substance (cont.)
Lower DP Specification Limit
Drug Substance Release Limit
Drug Product Release Limit
Drug Substance Shelf life
Drug Product Shelf life
Workshop on Statistical Analysis of Stability TestingWorkshop on
Statistical Analysis of Stability Testing
• Cumulative age studies Difficult to simulate the total age (DS +
DP) of an antigen during development
Some DS’s are stored 5 to 10-years
DP undergoes multiple temperature transitions
“Modeling” cumulative age is preferred to “studying” cumulative
age
Accelerated stability studies may be performed to experimentally
demonstrate cumulative age can be modeled rather than studied
e.g., Compare stability of DP which has been filled after
accelerated aging of DS and not
Can be verified as part of ongoing monitoring of vaccine
stability
Assess relationship of DP stability to age of DS
Studies supporting product licensure Long term stability of drug
substance (cont.)
9
• Goal: Establish shelf-life of DP, or
Develop a release model for DP
• Quality criteria: Satisfactory potency through product
shelf-life
Maintenance of levels of factors which might impact stability
Moisture of a lyophilized vaccine
pH of an aluminum adjuvanted vaccine
Studies supporting product licensure Long term stability of final
container product
10
Workshop on Statistical Analysis of Stability TestingWorkshop on
Statistical Analysis of Stability Testing
• Typical design of 3 lots tested at ICH timepoints Lower
specification limit is 3.0 log10 TCID50/mL
Studies supporting product licensure Example of shelf life
determination for a measles vaccine
11
Potency (log10 TCID50/mL) Time (Months) Lot 1 Lot 2 Lot 3
0 3.94 3.39 3.63 3 3.81 3.24 3.76 6 3.87 3.33 3.56 9 3.62 3.33
3.56
12 3.77 3.49 3.56 18 3.61 2.95 3.53 24 3.67 3.11 3.47 30 3.59 2.81
3.30 36 3.49 2.96 3.33
Workshop on Statistical Analysis of Stability TestingWorkshop on
Statistical Analysis of Stability Testing
Is shelf-life 12-months due to a measured value for Lot 2 below the
lower specification limit at 18-months?
Individual measurements do not effectively address the stability of
the lot
Long Term Stability at 2-8 C
2.0
2.5
3.0
3.5
4.0
Time (Months)
Po te
nc y
(lo g
10 T
C ID
50 /D
os e)
Lot 1
Lot 2
Lot 3
Studies supporting product licensure Example of shelf life
determination for a measles vaccine (cont.)
12
• ICH analysis (Q1E) Goal: establish shelf-life
3-lots (ICH Guidance)
• Pooled or worst case
Shelf life is the intersection of the lower specification limit
(LSL) and the lower 1-sided 95% confidence bound on the regression
line
Studies supporting product licensure Example of shelf life
determination for a measles vaccine (cont.)
13
4
4.3
4.6
4.9
5.2
Time (Months)
Po te
nc y
Workshop on Statistical Analysis of Stability TestingWorkshop on
Statistical Analysis of Stability Testing
Example – live attenuated measles vaccine (cont.) Evaluation of Lot
2 alone
Residual Plot
-0.25 -0.20 -0.15 -0.10 -0.05 0.00 0.05 0.10 0.15 0.20 0.25
0.30
• No pattern in the residuals (deviations of the observed points
from the line)
• Root MSE (assay variability) is typical for the potency assay
(~0.15 log)
• The slope is statistically significant (P=0.01)
Studies supporting product licensure Example of shelf life
determination for a measles vaccine (cont.)
14
y = -0.0148x + 3.4058
2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0
0 6 12 18 24 30 36
Po te
nc y
Lot 1
Lot 2
Lot 3
Evaluation of Lot 2 alone (cont.) Shelf-life determination
( )
2 avg
Po te
nc y
(lo g
10 TC
ID 50
/d os
Lower Bound
Min. Req.
Shelf-Life = 20-months
Studies supporting product licensure Example of shelf life
determination for a measles vaccine (cont.)
15
Month CB 18 3.04 19 3.03 20 3.01 ≥3.00 21 2.99 <3.00 22 2.97 23
2.95 24 2.93
Workshop on Statistical Analysis of Stability TestingWorkshop on
Statistical Analysis of Stability Testing
Evaluation using all lots Establishing poolability Equality of
slopes is established using a formal test of
parallelism –
Equality of levels is tested after fitting a common slope –
equal levels if P-value is sufficiently high (P<0.0001 is not
>0.25)
Equality of Slopes
Lot 1
Lot 2
Lot 3
Studies supporting product licensure Example of shelf life
determination for a measles vaccine (cont.)
16
• Evaluation using all lots (cont.) Shelf-life determination
Utilizing a model with common slope but different levels
Use “worst case” lot (lot 2) to determine shelf-life
Shelf-Life Determination (All Lots)
lo g 1
0 TC
ID 50
/m L
Lot 1 Lot 2 Lot 3 Regression Lower Bound Min. Req.
Shelf-Life = 24-months
Studies supporting product licensure Example of shelf life
determination for a measles vaccine (cont.)
17
Month LB 21 3.05 22 3.04 23 3.02 24 3.01 ≥3.00 25 3.00 <3.00 26
2.99 27 2.97
Workshop on Statistical Analysis of Stability TestingWorkshop on
Statistical Analysis of Stability Testing
Comparison of approaches
Approach Estimate Comments
Individual Measurements 12-months
Does not effectively use all the data to show potency trend; post
12-month potencies are within specification.
Studies supporting product licensure Example of shelf life
determination for a measles vaccine (cont.)
18
2.0
2.5
3.0
3.5
4.0
Time (Months)
Po te
nc y
(lo g
10 T
C ID
50 /D
os e)
Lot 1
Lot 2
Lot 3
Individual Lot 20-months
Effectively uses data for the worst case lot to estimate
shelf-life; does not use the full power of the long term stability
study. Recommended if the loss rates are not comparable among
lots.
Combined Lots 24-months Effectively uses all the data from the long
term stability study. Recommended if the loss rates are comparable
among lots.
Workshop on Statistical Analysis of Stability TestingWorkshop on
Statistical Analysis of Stability Testing
Determination of release potency
• ICH stability evaluation does not ensure quality of manufactured
lots throughout shelf life
• ICH 1E: Determine shelf life from pooled (n = 3) or worst-case
lot
• However, stability lots do not represent the total distribution
of release potencies of future manufactured lots
• Manufactured lots which are released below the distribution of
stability lots are predicted to fall below the lower specification
limit (LSL) prior to expiry
Re le
as e
Di st
rib ut
io n
0m0m 24m
Manuf
19
• A vision for the vaccine analytical control strategy
Release Limits
Lower Specification Limit (LSL)
Upper Specification Limit (USL)
– Scientifically/clinically justified upper and/or lower
specification limits
– Release limits calculated to ensure quality at release and
throughout shelf-life
– Control limits formulated to help manage manufacturing
consistency
Time (mos.)
Control Limits
= / =
20
Accelerated Stability • Uses of accelerated stability
Development
Formulation development Mechanism of degradation
Method validation Handling and use protocols
Release modeling
Post licensure
Stability monitoring
• Guidelines WHO Guidelines on Stability Evaluation of Vaccines
(2006) WHO Guidelines on the stability evaluation of vaccines for
use under
extended controlled temperature conditions (ECTC, 2015)
= + + 2 2 + 2
LPI (22°C)
Shipping (15°C) Labeled Storage (2-8°C) Use – R/S 2-8°C
Note: ECTC includes 40°C for 3-days for WHO certification
Use of accelerated stability in the release model
22
5.30907.0645.13728.00.3leaseRe =⋅++=
Condition Temperature Loss Rate Std.Err. Time at Condition Loss
Error
Labeling, Packaging, & Inspection
20-25o C 0.0025/hr. 0.0005 24-hrs. 0.06 0.012
Self-Life 2-8o C 0.012/mo. 0.0017 24-mos. 0.29 0.041 Reconstitute
& Store 2-8o C 0.0031/hr. 0.0005 8-hrs. 0.025 0.004
Release Assay Variability 0.08 0.08 Total Loss = 0.3728 Total Error
= 0.0907
Clinical Minimum = 3.0 Minimum Release = 3.5
∑ ∑ +++= 2 Assay
2 b
Managing the release limit through design
• The “uncertainty” in the release limit calculation contributes:
1.645⋅0.0907 = 0.15 log (~30%) to the total loss +
uncertainty
• Results in an increase in the manufacturing target
Forces production towards the upper specification limit (i.e.,
greater risk of OOS)
Decreases capacity, and thus availability to the customer
24
5.30907.0645.13728.00.3leaseRe =⋅++=
Target Target
Workshop on Statistical Analysis of Stability TestingWorkshop on
Statistical Analysis of Stability Testing
n 1 6 9 12 1 44% 16% 13% 11% 2 41% 15% 12% 10% 3 39% 15% 12%
10%
Runs (r)
Managing the release limit through design Reduce potency assay
variability through strategic replication
• Variance components (2 and
2 ) are from assay validation • Combinations of number of
runs
(r) and number of reps within runs (n) can be explored to reduce
release assay variability
• The “Reportable Value” is the average
• The stability testing format needn’t be the same as release
testing • The goal of stability testing
is different than the goal of release testing
= + + 2 2 + 2
4413
Std
1.4
3.89
1 4413 Clin 500783 1 1.4 0 4 41.5000 3.72569 0.33647
TEST DATE
12/8/99
Std
1.7
3.78
2 4413 Clin 500783 1 2.0 0 3 23.0000 3.13549 0.69315
Std
2.1
3.47
3 4413 Clin 500783 1 2.8 0 3 21.3333 3.06027 1.02962
COMMENTS
Std
2.4
3.18
4 4413 Liq 500996 1 190.0 0 3 40.3333 3.69718 5.24702
Std
2.7
2.81
5 4413 Liq 500996 1 266.0 0 3 43.6667 3.77659 5.58350
0
2.9957322736
4.3820266347
Std
3.1
2.46
6 4413 Liq 500996 1 380.0 0 3 28.6667 3.35574 5.94017
8
2.9957322736
4.3820266347
Std
3.4
1.79
7 4413 Liq 500996 1 532.0 0 3 21.3333 3.06027 6.27664
8 4413 Lyo 500997 1 14.0 0 3 79.3333 4.37366 2.63906
CURRENT STANDARD
0610246
3.9
36
39
49
59
61
49.00
1911
2358
2367
9 4413 Lyo 500997 1 19.6 0 4 60.5000 4.10264 2.97553
5.5
38
42
44
46
48
44.00
2420
Lyo
2.6
4.37
10 4413 Lyo 500997 1 28.0 0 3 51.3333 3.93834 3.33220
7.8
18
27
34
35
39
32.00
2496
Lyo
3.0
4.10
11 4413 Lyo 500997 1 39.2 0 4 37.0000 3.61092 3.66868
11.0
18
21
24
27
36
24.00
2640
Lyo
3.3
3.94
12 4413 Lyo 500997 1 56.0 0 3 20.6667 3.02852 4.02535
2.5
4.7
15.6
10
16
16
18
27
16.67
2600
Lyo
3.7
3.61
13 4413 Lyo 500997 1 78.4 0 4 24.0000 3.17805 4.36182
4
3.1
22.0
6
10
11
14
16
11.67
2567
Lyo
4.0
3.10
14 4413 Standard 610246 1 3.9 0 3 49.0000 3.89182 1.36098
31.2
4
5
6
7
10
6.00
1872
Lyo
4.4
3.18
15 4413 Standard 610246 1 5.5 0 3 44.0000 3.78419 1.70475
Lyo
4.7
2.61
16 4413 Standard 610246 1 7.8 0 3 32.0000 3.46574 2.05412
PU LYO STANDARD
0500997
14.0
72
78
80
80
89
79.33
11107
13934
13705
17 4413 Standard 610246 1 11.0 0 3 24.0000 3.17805 2.39790
14.0
79.33
19.6
48
62
62
70
DRY
60.50
11858
Clinical
0.3
3.73
19.6
60.50
28.0
38
45
52
57
63
51.33
14373
Clinical
0.7
3.14
28.0
51.33
39.2
DRY
41
38
37
32
37.00
14504
Clinical
1.0
3.06
39.2
37.00
56.0
17
17
21
24
27
20.67
11573
Clinical
1.4
2.20
56.0
20.67
78.4
21
23
23
29
DRY
24.00
18816
Clinical
1.7
2.12
Dilution
-0.4620354596
-0.7133498879
Regression
1
12.8073264701
12.8073264701
67.5922150161
0
0.63
0.49
-0.4620354596
-0.7133498879
-1.0498221245
-0.8439700703
Residual
174
32.9694004741
0.1894793131
0.35
0.43
-1.0498221245
-0.8439700703
IVRP
ED50
Low
High
-0.8675005677
-0.0100503359
Total
175
45.7767269441
0.42
0.99
-0.8675005677
-0.0100503359
0.5
0.52
0.17
1.55
-0.4307829161
-0.4462871026
0.65
0.64
-0.4307829161
-0.4462871026
0.6
0.44
0.15
1.31
-0.5108256238
0.1823215568
Coefficients
CONCORD
&A
2
5.231518
4.89642
187.0765
133.8099
5.2315176241
4.896420136
0.3350974881
4
4.498066
5.426619
89.84317
227.3793
4.4980655948
5.4266195481
-0.9285539534
6
5.126741
4.708723
168.4672
110.9104
5.1267410721
4.7087226681
0.418018404
Theoretical Concordance
Measured Concordance
• Stability variability is managed through strategic selection of
time points
Test at beginning and end of study Significant reduction in
variability (25%)
Note: assumes linear kinetics
0 1 2 3 4 5
26
Managing the release limit through design Reduce slope variability
through strategic stability study design
Design 1 Design 2 1 1 2 1 3 4 4 4
=0.45 Reduction=25%
Time (Hours)
• Calibration to a standard
• Reference to an unincubated sample (-70º C) from the same
lot
• Staged testing Appropriate when the slope is the parameter of
interest, and there is
significant run-to-run variability
Withdraw samples at stability time points, store at -70° C, test
together with unincubated samples
Requires sophisticated statistical design and analysis is
required
Analysis of covariance
Managing the release limit through design Other opportunities to
reduce variability
Workshop on Statistical Analysis of Stability TestingWorkshop on
Statistical Analysis of Stability Testing
Some current barriers • Lack of separate release and
end-of-shelf-life limits
• Setting shelf life based on individual values Last time point
within specification Modeling individual values using a prediction
interval versus a confidence interval
• Incorrect kinetics modeling Zero versus first order
• Sequential stability studies A study holding a DS through shelf
life, then formulated and held through DP shelf
life End-to-end modeling versus end-to-end studies
28
Summary
• Drug substance (bulk) evaluation addresses whether the
intermediate with be suitable for filling into drug product (DP)
throughout its shelf life
• Cumulative age may be addressed post licensure, or using
accelerated stability to simulate suitability of aged DS
• Accelerated stability studies are used throughout development and
post licensure to address development questions and to forecast
impact on product which has been exposed to elevated
temperatures
• Statistical methods make use of all the data to obtain a reliable
estimate of product shelf-life
• A release model can be used to help assure a commercial lot will
maintain adequate potency after handling, shipment, long-term
storage, and use
29
Workshop on Statistical Analysis of Stability Testing
1. Schofield, TL (2009) Vaccine stability study design and analysis
to support product licensure; Biologicals 37 (2009) 387-396.
2. Schofield, TL (2009) Maintenance of vaccine stability through
annual stability and 3. comparability studies; Biologicals 37
(2009) 397-402 4. Who Guidelines for Stability Evaluation of
Vaccines (2006) 5. WHO Guidelines on the stability evaluation of
vaccines for use under extended controlled temperature
conditions
(ECTC, 2015) 6. Fairweather WR, Mogg R, Bennett PS, Zhong J,
Morrisey C, Schofield TL. (2003) Monitoring the stability of
human
vaccines. Journal of Biopharmaceutical Statistics; 13: 395-413. 7.
Schofield, TL, et.al. (2006) Monitoring the stability of human
vaccines, presented at WCBP, SF 8. Gorko, MA (2003) Identification
of Out-of-Trend Stability Results, Pharmaceutical Technology; 27(4)
9. Noël C, Charles S, Francon A, Flandrois JP (2001) A mathematical
model 10.describing the thermal virus inactivation.
Vaccine;19:3575-82. 11.Yu, B, Zeng, L (2015) Evaluating the
comparability of stability at long-term storage temperature using
accelerated
stability data, IABS Statistical Meeting, September 29-30 12.Sidor,
L, Burdick, R, Cowley, D, Kendrick, BS, (2011) Demonstrating
comparability of stability profiles using
statistical equivalence testing, BioPharm International; 24 ,36-42
13.Burdick, RK, Sidor, L, (2013) Establishment of an equivalence
acceptance criterion for accelerated stability studies,
Journal of Biopharmaceutical Statistics; 23, 730-743
References
30
Thank you
Questions?
[email protected]
31
Basic principle of stability study design and analysis for product
licensure
Some notes on vaccine stability
Outline
Studies supporting product licensureLong term stability of drug
substance
Studies supporting product licensureLong term stability of drug
substance (cont.)
Studies supporting product licensureLong term stability of drug
substance (cont.)
Studies supporting product licensureLong term stability of final
container product
Studies supporting product licensureExample of shelf life
determination for a measles vaccine
Studies supporting product licensureExample of shelf life
determination for a measles vaccine (cont.)
Studies supporting product licensureExample of shelf life
determination for a measles vaccine (cont.)
Studies supporting product licensureExample of shelf life
determination for a measles vaccine (cont.)
Studies supporting product licensureExample of shelf life
determination for a measles vaccine (cont.)
Studies supporting product licensureExample of shelf life
determination for a measles vaccine (cont.)
Studies supporting product licensureExample of shelf life
determination for a measles vaccine (cont.)
Studies supporting product licensureExample of shelf life
determination for a measles vaccine (cont.)
Determination of release potency
Accelerated Stability
Use of accelerated stability in the release model (cont.)
Managing the release limit through design
Managing the release limit through designReduce potency assay
variability through strategic replication
Managing the release limit through designReduce slope variability
through strategic stability study design
Managing the release limit through designOther opportunities to
reduce variability
Some current barriers