Subject-by-formulation interaction in bioequivalence: Conceptual and statistical issues

Purpose. The FDA has proposed replacing the current average bioequivalence criterion with population and individual bioequivalence criteria that consi der variances in addition to the difference of averages. One of these varia nces in the individual bioequivalence criterion measures subject-by-formula tion interaction, the extent to which the test-reference difference varies from person to person. This paper discusses conceptual acid statistical iss ues raised in various publications and presentations with respect to the pr esence and estimation of such an interaction. Methods. We focus on the importance of subject-by-formulation interaction, an understanding of what is a large interaction, and the assessment of the magnitude of this interaction in bioequivalence studies. Simulation studies , examples from the literature, and data from FDA files are used to demonst rate the magnitude of the interaction and its distribution under various co nditions. Results. The concept of a large interaction is tied to the concept of a lar ge mean difference. We suggest that an interaction greater than 0.15 is a c onservative criterion for a large interaction. Magnitudes of estimated inte raction are affected by variability, sample size, and the selection of data sets that pass average bioequivalence. Conclusions. Examples of substantial interactions are beginning to appear. More data is needed before reaching definitive conclusions regarding the fr equency and importance of observed interactions.