Industry Perspective of Subgroup Analyses
View Presentation View Presentation
*Joe W Bero, Boston Scientific 


A clinical study is powered to detect a treatment effect based on the overall intended patient population with an expected mix of gender, diabetic status, low/high risk patients, age groups, etc. Specific subgroup analyses were generally pre-specified in the protocol to assess the treatment consistency and it is not intended to make claim treatment effect for each of the subgroups. The treatment-by-subgroup interaction should be investigated and the statistical testing for the treatment effect pooling over the subgroups of interest should be deployed instead of performing tests for each subgroup. Any post-hoc subgroup analyses are hypothesis generation and should be used for investigative purposes in preparation for future studies.