Keywords: Bioequivalence, Binary Clinical Endpoint, ANDAs
According to current FDA guidance, a generic product (TEST) and the Reference Listed Drug (RLD) are considered to be bioequivalent if the 90% confidence interval for the difference in success or cure rates between the two products is within [-20%, 20%]. This method, however, is not equally sensitive for delimiting a difference between TEST and RLD across the response range. This is particularly problematic when the success rate for RLD is expected to be low. For example, if the success rate for RLD is expected to be lower than 20%, even if the success rate for the TEST is only half of the RLD, a sample size of approximately 150 per group can almost guarantee that the TEST would pass the equivalence test. In such a scenario, the TEST could theoretically be half as effective as the RLD, but the difference in success rates between the two products would still be less than the 20% difference limit allowed. Although the totality of the information in ANDA submissions provides other important context regarding the pharmaceutical and bioequivalence of TEST and RLD products that would likely prevent the approval of a truly inferior generic product, the current statistical method for binary clinical endpoints presents a potential vulnerability that could undermine public acceptance for generic drugs. This presentation will explore an alternative two-step statistical method for the bioequivalence study with binary clinical endpoint in ANDA submissions. Through simulation, we show that our proposed method can control the passing probability for products with relatively low success rate, and meanwhile maintain high power with reasonable sample size when TEST and RLD are actually equivalent.