When a drug (biological) product is going off patent protection, a pharmaceutical/biotech company usually files a regulatory submission for generic (biosimilar) approval. For the approval of a generic drug (biosimilar product), the United States Food and Drug Administration (FDA) requires a bioequivalence study (biosimilar studies) to be conducted for bioequivalence (biosimilarity) assessment, demonstrating that the generic drug (biosimilar product) is bioequivalent (biosimilar) to the innovative drug product. For bioequivalence (biosimilarity) assessment, a two one-sided tests (TOST) procedure or a confidence interval (CI) approach is commonly used. However, although TOST is operationally equivalent to the CI approach, TOST is not generally equivalent to the CI approach. For example, when the study endpoint is a binary response, both the sample size estimation procedure and the result of bioequivalence test may be different. In this article, a comparison between the method of TOST and the CI approach in terms of true positive rate and sample size was made theoretically. The confusion matrix is also provided to show that TOST and CI approach may give different result when the outcome variable is binary.
Author(s): Peijin Wang, and Shein-Chung Chow