Evaluation of log-transformation in assessing bioequivalence

The question of whether we should use log-transformed data in assessing bioequivalence between two formulations is studied through a simulation. Under a 2×2 cross-over design with different combinations of sample size and intrasubject variability, we examine the impacts of different scales of the data on the conclusion of bioequivalence as well as on the level of significance and power. We also investigate the distributions of AUC and Cmax when the joint distribution of plasma concentrations is either multivariate normal or lognormal. A comparison between population bioequivalence of average bioavailability and individual bioequivalence under a lognormal distribution is also given.