Abstract:

To explore the comparative performance of the recently proposed bioequivalence (BE) approaches, FDA(s) and EMA(s), by the FDA working group on highly variable drugs and the EMA, respectively; to compare the impact of the GMR-constraint on the two approaches; and to provide representative plots of % BE acceptance as a function of geometric mean ratio, sample size and variability. Simulated BE studies and extreme GMR versus CV plots were used. Three sequence, three period crossover studies with two treatments were simulated using four levels of within-subject variability. The FDA(s) and EMA(s) approaches were identical when variability was < 30%. In all other cases, the FDA(s) method was more permissive than EMA(s). The major discrepancy was observed for variability values > 50%. The GMR-constraint was necessary for FDA(s), especially for drugs with high variabilities. For EMA(s), the GMR-constraint only became effective when sample size was large and variability was close to 50%. A significant discrepancy in the performances of FDA(s) and EMA(s) was observed for high variability values. The GMR-constraint was essential for FDA(s), but it was of minor importance in case of the EMA(s).