A three-step procedure for assessing bioequivalence in the general mixed model framework.

Bioavailability data arising from a standard two-period cross-over study are routinely analysed to establish bioequivalence between test and reference formulations. Current regulatory guidelines only require evidence of equivalence in average bioavailability for the assessment of bioequivalence. Under normality assumptions, this is achieved by demonstrating equivalence between the formulation means (step 1). However, the equivalence of formulation variances should also be assessed to get evidence of population bioequivalence (step 2), since a difference in variability of bioavailability may also pose significant problems in drug safety and efficacy. On the other hand, even population bioequivalence does not ensure that an individual subject could be expected to respond similarly to the two formulations. Therefore, whenever individual bioequivalence is the ultimate goal, the magnitude of intra-subject correlation should always be examined as the final stage (step 3). In this paper, these three successive concepts of bioequivalence are cast into the general mixed model framework and a stepwise testing procedure for the global assessment of bioequivalence is proposed. In addition to this, important issues addressed in the regulatory guidelines, such as verification of the model assumptions and application of the log-transformation, are discussed. Lastly, an example is presented to illustrate the proposed three-step procedure on the original and log-transformed scale of measurement.