Interim decision-making strategies in adaptive designs for population selection considering post-progression survival magnitudes

The development of targeted therapies, which benefit only a subgroup of patients treated for a given type of cancer, has been extremely attractive to many investigators. Adaptive seamless phase II/III designs in oncology clinical trials with interim analyses for subpopulation selection could be used if pre-defined biomarker hypothesis exists. We consider the interim analysis using time-to-event endpoints, e.g., overall survival (OS) and progression-free survival (PFS), to identify whether the whole population or only the biomarker-positive population should be continued into the subsequent stage, whereas a final decision is based on OS data. In this paper, we propose the interim decision-making strategies in adaptive designs with correlated endpoints, considering post-progression survival (PPS) magnitudes. In our approach, the interim decision is made on the basis of predictive power, by incorporating information on OS as well as PFS. We consider PFS data only in making interim decision in order to supplement the immature OS data. Simulation studies assuming a targeted therapy show that our interim decision procedure performs well in terms of selecting the proper population, especially under a scenario in which PPS affects the translation of the benefit from PFS to OS.