New approaches to population bioequivalence (PBE) and individual bioequivalence (IBE) have been motivated by the limitations of average bioequivalence (ABE) to handle unequal variances and subject-by-formulation interaction. The criteria for PBE and IBE described in the Food and Drug Administration draft guidance employ aggregate criteria that combine information on differences in bioavailability between formulation means, and differences in bioavailability variation of formulations between and within subjects. Examples from replicate design studies have demonstrated that the trade-off in means offered by the scaled aggregate criterion may result in clinically unacceptable decisions in favor of IBE, although ABE does not hold. Concerning the statistical methods, there are at least three conceptual issues that are still unresolved. First, aggregate hypotheses on the logarithmic scale have no obvious translation into the original scale. Second, scaling corresponds to a modification of the bioequivalence acceptance limits, and is an issue independent of IBE, PBE, and ABE, but is handled differently for IBE and PBE than for ABE. Third, the proposed criteria do not mandate hierarchical testing (first means, then variances, lastly subject-by-formulation interaction). If, despite major clinical reservations, IBE is further pursued, statistical research should focus on disaggregate criteria that allow exact stepwise procedures for evaluating untransformed parameters.
