STATISTICAL TOOLS FOR ASSIGNING POTENCY TO REFERENCE STANDARDS

Transcription

1 STATISTICAL TOOLS FOR ASSIGNING POTENCY TO REFERENCE STANDARDS Yi Tsong, Xiaoyu Dong & Meiyu Shen Division of Biometrics VI, Office of Biostatistics, CDER, FDA *This presentation represents the speaker s professional opinion. It does not represent the regulatory position of U.S. FDA

2 OUTLINES 1. Introduction 2. Four review experience 3. Concluding Remarks 2013 CMC Strategy Forum 2

3 INTRODUCTION Reference material Representing the characteristic of the product Often taken from the lots with values around the means of batches selected Stored at low temperature condition in order to have long stability Requalification at each time to be used Changes of reference material Needs to extend shelf life Out of specification Changes in process 2013 CMC Strategy Forum 3

4 Review Experience (1) Determine the qualification acceptance criterion and suitability for use of the secondary reference standard (SRS). The sponsor proposed; AC = Mean ± 2SD based on the primary RS If the mean of secondary RS results are within AC, then it is suitable for use CMC Strategy Forum 4

5 Review Experience (1) - Continued Statistical Review Issues: What Mean and SD in AC (Mean± 2SD) represents? - If Mean and SD represent the mean and SD of all lots used in RS, then AC means the interval that 95% of the individual lots included (theoretically). In this case, SRS lot is just as good as any lot of the population of lots used in determine the RS lot. - If Mean and SD represent the mean and SD of K test values of the RS lot, then AC means the interval that 95% of the individual test values of RS lot included (theoretically). In this case, Mean(SRS) is accepted if it is as good as any test value within the 95% of the possible test values of RS lot. It is a comparison between mean and the interval of individuals CMC Strategy Forum 5

6 Review Experience (1) - Continued In order to establish SRS as equivalence to RS Needs to show the 2ndry RS is as good as the primary RS lot. It means Mean(SRS) almost = Mean(RS) in expectation How to do it? One may propose the follow the equivalence test approach Assuming a small but reasonable difference δ between two RS lots. Want to assure that SRS is no more different from RS by δ but adjusted for the small number K of test values Considering taking K values of the SRS. Want to assure that the 95% CI of mean difference (between RS and SRS) lies within δ 3SD/ K and δ+3sd/ K 2013 CMC Strategy Forum 6

7 Review Experience (1) Statistical Review Issues: In order to establish a new RS because of changes in process Use Mean and SD of all lots used in RS, then AC means the interval that 95% of the individual lots included (theoretically). Compare individually prepared K assays of SRS with AC to make sure all individual assays are within AC. Otherwise needs to re-establish specifications 2013 CMC Strategy Forum 7

8 Review Experience (2) Modification of manufacturing process: Requests to qualify a primary reference standard obtained from a production scale drug substance lot To replace the current primary reference standard All values are within the historical range of the lots. The purity of the new reference materials is higher than the current standard reference but the difference is minimal. FDA approved the change of reference standard 2013 CMC Strategy Forum 8

9 Review Experience (3) Requalification of the primary RS The assay results of the primary RS was out of AC during requalification. The sponsor claimed The RS is suitable for continued use because the P-value from the linear regression is non-significant. The original AC established from the drug substance batches are not suitable. Proposed to revise the AC based on the data from the RS CMC Strategy Forum 9

10 Review Experience (3)- continued Statistical Issues: Test for linear regression Y(t) = α + βt +ε No linear trend No trend Assay results have large variability, so it tends to have nonsignificant P-value using linear regression. Needs to calculate the confidence interval at all time point to check for out of specification Plot against t indicates the variability increases with time RS Specification is determined by several lots that are representative of the the substance or drug product. The RS lot with out of bound values may net be suitable to be used CMC Strategy Forum 10

11 Review Experience (4) Extending the expiry of the reference material beyond 96 months The sponsor provides two sets of long-time stability data with two storage conditions 120 month data from 3 drug substance batches in storage condition #1; 96 month data for 2 drug substance batch stored in condition #1 96 month data for 1 drug substance batch stored in conditions #1 and #2. Batches stored in condition #1 can t pool, but support 108 month shelf life Comparison between slopes of one batch stored in two conditions is not statistically significant Support shelf life of 108 months for storage condition #2. Reviewer recommendation: 108 months for condition #1 and 96 months for condition # CMC Strategy Forum 11

12 CONCLUDING REMARKS There are various reasons for the need of change reference standard Showed and discussed four examples There are different approaches to address the needs Discussions between the sponsor, regulatory reviewer (chemist, biologist and statistician) are essential to address the needs 2013 CMC Strategy Forum 12

13 THANKS FOR PARTICIPATING 2013 CMC Strategy Forum 13