Recently, the European Medicines Agency (EMA) issued a new guideline on the investigation of bioequivalence (BE). In case of highly variable drugs, this guideline proposes that the acceptance limits for C(max) can gradually be expanded as a function of within-subject variability (CV(wR)). Actually, these BE limits exhibit leveling-off properties since they are not allowed to scale continuously, but only up to CV(wR)=50%. To avoid the risk of accepting two drug products which may differ significantly, this EMA guideline also proposes the use of a secondary constraint criterion on the geometric mean ratio (GMR) of the two products under comparison. Aim of this study was to explore the leveling-off properties of the new EMA limits in comparison to other approaches, as well as to assess the impact of the complementary GMR criterion on the ability to declare bioequivalence. Simulated bioequivalence studies and extreme GMR plots were used to assess the performance of the EMA limits. Three sequence, three period (3×3) crossover studies with two treatments (T and R) were simulated. The R product was considered to be administered twice, while the T only once (i.e., TRR/RTR/RRT). Among others, this study revealed the leveling-off properties of the new EMA limits. It was also shown that the complementary GMR-constraint is only effective when a large sample size is used and at regions of CV(wR) close to 50%. This GMR-criterion begins to be effective at sample sizes around 60 and becomes more prominent as the number of subjects participating in the BE study increases. For CV(wR) values lower than 50%, the GMR-constraint has no role. In case of within-subject variabilities greater than 50%, the impact of the GMR-constraint diminishes due to the leveling-off properties of the EMA limits. Compared to the classic 0.80-1.25 or the extended 0.75-1.33 criteria, the new EMA limits are more liberal at high CV(wR) values and allow greater differences between the two drug products to be declared bioequivalent. Finally, this study showed that the use of an approximate value (0.760) on the scaling factor proposed by EMA, has no impact on the performance of the new BE limits compared to other more accurate approaches.
Copyright Â© 2011 Elsevier B.V. All rights reserved.