A comment on so-called individual criteria of bioequivalence

The idea of defining bioequivalence in terms of the proportion of individuals exhibiting a drug formulation discrepancy that does not exceed some prespecified limit seems natural enough. Since the derivation of formal statistical procedures for testing the associated hypotheses is more or less straightforward, it is surprising that the approach was pursued in a systematic manner not earlier then in 1990 [by Anderson and Hauck and Wellek]. In discussing the relative merits of the approach we will stress the importance of a careful understanding of its conceptual basis. It will be argued that the most severe limitation on the usefulness of such probability-based criteria of individual bioequivalence arises from the fact that large intraindividual formulation discrepancies provide evidence against equivalence only in definite absence of period effects.