|
Abstract:
|
(note that due to company policy, the study name and the disease won’t be released in this presentation) Subgroup analysis always has its challenges for both sponsors and health authorities from the submission perspective. This is especially true for cases where the planned analysis does not meet the primary endpoint, but the post hoc analysis demonstrates a treatment effect based on the sub-group. In our case, SAP pre-specified that if there is a significant interaction between the treatment and the subgroup with a pre-specified significant alpha, the treatment comparisons within each subgroup stratum will be examined. The study results indicate that there is a highly significant interaction between the treatment and age groups. The younger population demonstrates a favorable benefit/risk ratio in the experimental treatment, while the older population shows the opposite direction. Further biomarker exploratory analyses were performed within the older sub-populations. Machine learning was applied to identify the positive biomarkers. This talk will illustrate the positive biomarkers selection that can be interpretable from both statistical and clinical perspectives to make a sound t
|
