[Review and use of decision rules for bioequivalence trials]

To declare bioequivalent two different formulations of one active drug, bioavailability studies are conducted, usually based on area under the plasma concentration-time curves and peak concentrations. The decision follows a statistical basis with right statement of the hypotheses of bioequivalence that are described. This procedure allows to control the consumer risk of falsely accepting bioequivalence while minimizing the new formulation risk of erroneously rejecting bioequivalence. Six decision rules meeting these criteria are reviewed and compared with numerical data: classic confidence interval; symmetric confidence interval; Hauck-Anderson method; two one-sided tests procedure; bayesian method; non parametric confidence interval. The six rules all have very similar performance. However, the bayesian procedure which gives a probability of the location of the true relative bioavailability could likely complete the two one-sided tests procedure and/or the classic confidence interval method that are recommended in regulatory guidelines. Some other statistical points that have received different interpretation in the international regulations are finally evoked and discussed.