Statistical tests for population bioequivalence

In its 2001 guidance, the U.S. Food and Drug Administration (FDA) recommends that population bioequivalence (PBE) and individual bioequivalence (IBE) be assessed to address respectively the prescribability and switchability between a brand-name drug product and its new formulation or generic copy. For IBE, the FDA recommends a 2 × 4 crossover design and a statistical test procedure proposed by Hyslop, Hsuan and Holder (2000). The same method is also recommended in FDA (2001) for assessment of PBE under the 2 × 4 crossover design. However, we note that, asymptotically, FDA's PBE test has a size smaller than the nominal level and thus has a low power to detect PBE. In addition, the 2001 FDA guidance does not provide any statistical procedure for PBE under commonly used 2 × 2 or 2 × 3 crossover designs. In this paper, an asymptotically valid statistical test is derived for PBE under the 2 × 2, 2 × 3 or 2 × 4 crossover design, using the method of moments and linearization. A method of determining the sample size required to achieve a desired power of the PBE test is also proposed. Simulation results are provided to examine the performance of the proposed PBE test and FDA's test. Finally, an example is presented for illustration.

Duke Scholars

Author Shein-Chung Chow Biostatistics & Bioinformatics, Division of Biostatistics