A modified two one-sided test for reference-scaled average bioequivalence

C. Chiang${}^a$ and C.F. Hisao${}^b$

${}^a$Tamkang University, New Taipei, Taiwan, ${}^b$National Health Research Institutes, Taiwan

The US FDA has released the first revision of the draft guidance, “Statistical Approaches to Establishing Bioequivalence,” in late 2022. This revision recommends reference-scaled average bioequivalence (RSABE) for in-vitro permeation tests and assessments of narrow therapeutic index drugs. The FDA suggests that the corresponding test be based on an upper confidence bound using the Howe method for a linear combination of squared mean difference and variance(s). Since the statistical characteristics for this confidence bound are difficult to derive, the sample size determination often relies on simulations. In this study, we proposed a modified two one-sided test (MTSOT) as an alternative, accompanied by a formal method for sample size determination. Simulation results demonstrate that the proposed MTOST effectively controls the Type I error rate and provides sufficient power. Furthermore, it is more powerful than the Howe method under various scenarios.

Keywords: Reference-scaled average bioequivalence; Sample size determination; Two one-sided test