A retrospective assessment of the 75/75 rule in bioequivalence.

The 75/75 rule was originally proposed by the U.S. Food and Drug Administration (FDA) as an alternative means of testing the bioequivalence of two formulations of a pharmaceutical agent. The rule specified that the ratio of test-to-reference formulation of a bioavailability measure arising in a bioequivalence study must be between 75 and 125 per cent of unity in at least 75 per cent of subjects to declare two formulations bioequivalent. The rule has garnered criticism in the literature and the FDA no longer uses the rule formally in assessing bioequivalence. The basis, however, for all criticism of the rule has been simulation arguments. In this paper, we derive the sampling model implied by the rule and place the rule in the framework of a statistical hypothesis test. We show how the significance level of the test depends upon variability of the formulations, and thus why the rule performs in the way that has received criticism.