←Back

259 – Clinical Trial Design II

Reference-Scaled Margin for Lot Consistency Study

at Merck

Before a test vaccine can be approved by regulatory agencies for public use, a lot consistency study is usually required for the manufacturer to demonstrate its ability to produce the vaccine consistently. Three (3) vaccine lots are typically required for the study of consistency in terms of clinical endpoints, e.g. geometric mean titers (GMTs). And lot consistency is demonstrated only if 95% confidence intervals for the 3 pairwise lot-to-lot differences are all contained within a pre-specified consistency margin. Thus, the success of a lot consistency study is heavily hinged upon the consistency margin. Currently, a fixed consistency margin of 1.5, i.e., an interval of [1/1.5, 1.5], is commonly required by regulatory agencies, although a larger margin of 2.0 has also been allowed on a case by case basis. One concern with fixed margin is that it does not take consideration of within lot variance for the vaccine. Because the width of confidence interval is a function of both lot-to-lot difference and within lot group variance, a larger lot-to-lot difference may pass the consistency test if variances within lot groups are small but a smaller lot-to-lot difference may fail if variances within lot groups are large. For vaccines with high within lot variability a fixed margin of 1.5 becomes too stringent and requires a substantially larger sample size for it to pass the consistency test. This paper proposes a reference-scaled lot consistency margin based on variability of an active reference included in the lot consistency study. This active reference is a vaccine with antigens similar to those in the test vaccine, ideally one that has been approved by regulatory agencies for public use. Thus, the reference, being a vaccine of similar antigens, can provide an estimate of variance within lot group for the test vaccine antigens. The proposed consistency margin gradually widens based on intrinsic property of the vaccine, namely, the variability of the antibody titers it elicits in its target population. This approach is scientifically more appealing than a subjective decision to allow the use of a consistency margin of 2.0. Additionally, the proposed margin controls Type I error rate because the margin based on reference variance is independent of the lot consistent hypothesis testing.