[Subgroup analysis and forest plots: Limitations and interests]

In the era of personalized medicine, treatment effect can no longer be considered homogeneous for all patients. Subgroups analyses are generally performed in addition to main analysis in order to assess treatment effect in subpopulations. When planned, analyzed, presented and interpreted correctly, results of subgroups analyses can provide important information to guide treatment strategy. Several literature reviews have highlighted a lack of consistency in the analysis and interpretation of results. Indeed, subgroups analyses face different methodological issues (alpha risk inflation, power decrease…). Methodological principles for planification and interpretation of subgroups analyses in a clinical trial are presented in this methodological paper. Recommendations are proposed to assist in the interpretation of results of subgroups analyses.