Alternative approaches for assessing bioequivalence regarding normality assumptions
This paper concerns several alternative approaches for the assessment of bioequivalence between two formulations (a test formulation and a reference formulation) of a drug with respect to various situations depending upon whether the normality assumptions for the random subject and random error terms are met. When the two terms are both normal, a new approach is proposed by constructing an exact confidence region for (μ(R), μ(T)) where μ(T) and μ(R) are means for the test and reference formulations, respectively. The constructed confidence region is then compared with the acceptance region which is bounded by μ(T) = 0.8μ(R) and μ(T) = 1.2μ(R). This approach is shown to more rigorously meet the Food and Drug Administration (FDA) requirement compared with the customary method. When the random subject term follows a skewed but unknown distribution, an alternative approach is derived based on the individual subject ratios. A nonparametric bootstrap procedure is also proposed for a situation where the distributions of the two terms are both unknown. Two examples concerning two-way crossover experiments for the comparison of bioavailability between two formulations are presented. A brief discussion regarding tests for normality assumptions and future research in this area is given.