Bioequivalence for highly variable drugs: regulatory agreements, disagreements, and harmonization

Abstract

Regulatory authorities introduced procedures in the last decade for evaluating the bioequivalence (BE) for highly variable drugs. These approaches are similar in principle but differ in details. For example, the Food and Drug Administration (FDA) and the European Medicines Agency (EMA) recommend differing regulatory constants. The constant suggested by FDA results in discontinuity of the BE limits around the switching variation at 30% observed within-subject variation of the reference product. The regulatory constant of EMA does not have these problems. The Type I error reaches 6–7% around the switching variation with the EMA constant but 16–17% with the FDA constant. Various procedures were recently suggested, especially for the EMA approach, to eliminate the inflation of the Type I error. Notably, the so-called Exact algorithms try to amalgamate the positive features of both EMA and FDA procedures without their negative sides. The computational procedure for the EMA approach is simple and has a straightforward interpretation. The procedure for the FDA approach is based on an approximation, has a bias at small degrees of freedom, and requires a suitable computer program. All regulatory agencies impose a second requirement constraining the point estimate of the ratio of geometric means. In addition, EMA and Health Canada impose an upper limit for applying the recommended procedures. These expectations have psychological motivation and political rationale but no scientific foundations. Their inclusion results in incorrect and misleading interpretation of the principal criterion which involves confidence intervals. Different regulatory authorities expect to apply their approaches either to both AUC and C_max or only to AUC or only to C_max. Rational resolution of the disharmonization is needed.