The Bootstrap Percentile in Food and Drug Administration Regulations for Bioequivalence Assessment

Abstract

Nowadays, the bootstrap is one of the most prominent representatives of resampling techniques, especially with respect to their daily use. It is typically applied in situations which are not analytically tractable due to the complexity of the underlying statistical estimation or testing problem, where it is common practice to run a Monte-Carlo algorithm instead of calculating the exact bootstrap. This makes the bootstrap an easy to handle and also a very flexible tool, which, of course, must always be critically looked at before being applied. This is also the case when using the bootstrap to calculate confidence intervals for statistically evaluating the bioequivalence of two formulations of a drug as recommended in the 1997 draft of the Food and Drug Administration (FDA) guidance (1). Here, the bootstrap seems to be a sensible way to cope with the analytical problems related to the recently proposed and rather complex measures for assessing population and individual bioequivalence. In this paper, we briefly review the different concepts of bioequivalence and two approaches for measuring the individual and population bioequivalence, namely the moment-based and the probability-based approaches, where the focus is on individual bioequivalence. We then concentrate on the bootstrap percentile interval, which has been considered by the FDA.