Assessment of selection bias in estimates of relative bioavailability and intrasubject variability from...

The outcome selection of bioequivalence evaluations for abbreviated new drug applications results in bias towards unity of test-reference ratio estimates, and underestimation of intrasubject variability for the test drug product. In this study, the selection bias in the estimates of testreference ratio and intrasubject variability as function of the true testreference ratio, intrasubject variability, and sample sizes was derived, and the relationship of the selection bias with the true test-reference ratio, intrasubject variability, and sample sizes was evaluated using the derived functions for the selection bias. It was shown in this study that the selection bias decreases with the true test-reference ratio approaching unity, with decreasing intrasubject variability, or with increasing sample sizes of bioequivalence trials. As demonstrated in this study, the selection bias can reach such high levels that it can result in misleading optimism about interchangeability between the bioinequivalent generic versions of a reference drug product, especially for highly variable drug products. As demonstrated in this study, trial repeating raises the chances for test products of low test-reference ratio to meet the bioequivalence requirements, and as such inflates value of the expected selec tion bias. The a priori knowledge of intrasubject variability for the reference drug product may be used to infer the intensity of outcome selection, and as such to predict selection bias in test-reference ratio estimates for the test drug product.