Determining the probability of an important difference in bioavailability.

Bioavailability studies usually compare formulations of a drug under the premise of equal bioavailabilities. Analyses of such studies may be questioned when large differences exist, but are not statistically significant, or when small differences are found which do reach statistical significance. In either case, the question of clinical significance is not addressed and remains problematic regardless of the outcome. The important question is, "Given the results of the study, what is the probability that a clinically important difference exists?" To obviate dependence on statistical significance, we propose a methodology which permits computation of the probability that a clinically important difference in bioavailability exists between two products, regardless of the results of conventional tests of hypothesis. The methodology permits a strong statement which defines the likelihood of a meaningful difference, rather than the "nonstatement" of current approaches. The approach is illustrated by a study in which a 2% (nonsignificant) difference in bioavailability was observed. Using the proposed methodology the analysis concluded that the odds are about 5 to 1 against there being a 20% difference in bioavailability between the formulations.