ABSTRACT

The concept of individual bioequivalence was introduced in Chapter 11 to address the issue of switchability. Although its objective is very clear, no consensus has been reached for a precise definition of individual bioequivalence. As a result, under the two-stage mixed-effects model given in Equation 11.2.3, various similar or different criteria have been proposed for evaluation of individual bioequivalence. According to the distribution measures, they can be grouped as either moment-based or probability-based criteria. On the other hand, relative to the manner of assessment, criteria can be classified into aggregate or disaggregate criteria. All these criteria, however, are based on the same three characteristics for quality assurance of the generic drug product that include difference in average bioavailabilities, variance of the subject-by-formulation interaction, and difference in intra-subject variabilities. For the disaggregate criteria, individual components are assessed separately for

individual bioequivalence. Therefore, the estimates of averages, the variance of the subject by formulation, and intra-subject variabilities must be obtained from the pharmacokinetic responses generated under the higher-order design for inference of individual bioequivalence. On the other hand, because the aggregate criteria are a composite index of the three components, it is not necessary to estimate individual parameters for evaluation of individual bioequivalence. In addition, as already indicated in Chapter 11, most of aggregate criteria are rather complicated functions of difference in average bioavailabilities, variance of the subject-by-formulation interaction, and difference in intra-subject variability. To resolve the analytical difficulty, the 2001 FDA guidance recommends a linearized form of MI3 proposed by Hyslop et al. (2000) for inference of individual bioequivalence. Recently, McNally et al. (2003) applied the technique of generalized p-value based on generalized pivotal quantities (Tsui and Weerahandi, 1989; Weerahandi, 1995) directly to MI3. Because inference of bioequivalence involves hypothesis testing as well as estimation, similar to average bioequivalence, most currently available statistical procedures for assessment of individual bioequivalence adopt the confidence interval (CI) approach.