Qualitative versus Quantitative Treatment-by-Subgroup Interaction in Equivalence Studies with Multiple Subgroups.

Subgroup analysis has been discussed extensively for efficacy studies. When there are heterogeneous treatment effects across subgroups, an important question is whether the treatment-by-subgroup interaction is qualitative or quantitative, where the former is usually the concern. For equivalence, however, we found no formal definition of qualitative versus quantitative interaction. In this article, we define quantitative versus concordant and discordant qualitative treatment-by-subgroup interactions for equivalence studies for the first time. We identified the most concerning discordant quantitative interaction unique to equivalence studies, where a seemingly equivalent overall treatment effect can come from the offset of the opposite directions of inequivalence in subgroups where none of the subgroups are equivalent. We also evaluated an appropriate easy-to-apply criterion: a significance threshold (i.e., alpha) for the treatment-by-subgroup interaction that warrants more investigation of heterogeneity across subgroups for equivalence studies. Based on simulation, we recommend that for a typical two-way crossover PK bioequivalence study, if the interest is in quantitative interaction, then α= 0.1 (or even higher) is recommended as a threshold; if the interest is in qualitative interaction, which is usually the case in practice, then α= 0.05 is a sufficient threshold. Use of criterion can help regulators and sponsors identify important heterogeneity in equivalence assessment. [ABSTRACT FROM AUTHOR]