Keywords: cancer immunotherapy, complex survival patterns, design and analysis, immune-oncology, nonproportional hazards patterns, sample size and power calculation
The presence of various complex survival patterns in immune-oncology (IO) studies violates the Proportional Hazards (PH) assumption required by conventional design and analysis strategies, making most conventionally designed IO studies underpowered or even falsely negative. An efficient study design and analysis strategy should not only salvage the power loss incurred as a result of various non-proportional hazards (NPH) patterns but also provide a remedy to reduce the occurrence of such patterns. In this paper, we first examine the potential primary root cause behind the NPH patterns then present a novel strategy to incorporate the root cause into study design and data analysis, aiming to salvage power loss when the PH assumption is violated. Finite sample efficiency of the proposed methods is evaluated in simulations. Compared with the standard practice based on the regular log-rank test, the proposed design and analysis strategy can achieve an adequate study power when various NPH patterns are present. In addition, implicit in the proposed strategy is the solution to dramatically enhance the study efficiency of IO trials.