ABSTRACT
Subgroup analyses are commonly used to assess the heterogeneity of treatment effects across different patient subpopulations defined by baseline characteristics. They are useful for guiding patient treatment selection, but they introduce analytic challenges and can lead to overstated and misleading results. We describe analysis methods for conducting subgroup, considerations for subgroup analysis in the design stage, and reporting and interpretation of subgroup analysis in randomized clinical trials. Specifically, we review the conventional interaction tests in regression settings, methods for detecting qualitative interactions, graphical methods, multivariate tests of interaction, as well as power consideration.
