In seamless phase II/III adaptive trials, multiple doses are often evaluated in the phase II part of the study and one or two doses are selected to continue into the phase III part. This set up introduces complicated multiplicity issues on the final efficacy analyses, especially when different endpoints are used in the two phases, such as biomarkers used in phase II for dose selection, and efficacy endpoints used in phase III. In addition, subject in the dropped arms may not have efficacy follow-up in phase III. Potential type I error inflation on the final efficacy endpoint may arise due to various causes, such as the correlation between the biomarker and the efficacy endpoint.

In this presentation, we will evaluate the multiplicity issues related to such 2-stage designs by assessing the potential factors for type I error inflation in various scenarios and multiple testing methods that control the family-wise type I error rate for trials observing biomarkers in phase II and efficacy endpoints in phase III. An industry trial example will be used to illustrate the findings.