Keywords: adaptive method, hazard ratio, logrank test, prespecified procedure, proportional hazards, restricted mean survival time
Logrank (LR) /hazard ratio (HR) test/estimation approach has been routinely used in almost all cancer clinical trials with time-to-event outcomes. Although the LR test is a nonparametric test that is asymptotically valid, it is not the most powerful test when the pattern of the difference is non-proportional hazards (PH). It would be natural to seek alternative testing procedures that offer greater power than the LR test when the non-PH becomes the concern. On the other hand, estimating the magnitude of treatment effect is also an essential analytic task. Estimating the treatment effect is more challenging than testing. For example, some measures (e.g., HR) relies on a strong modeling assumption for their validity. Also, it would not be easy for clinicians/patients to balance risk and benefit of the treatment using the resulting HR due to the lack of a reference number from the control group. Restricted mean survival time (RMST)-based approach is getting more attention recently as an alternative to the conventional LR/HR approach. Especially for estimation, it does have advantages over the HR-based approach. Specifically, it does not require any modeling assumption. The RMST value from the control group helps clinicians/patients to interpret whether the resulting RMST difference (or ratio) is large enough for them to select that therapy. In this talk, we introduce adaptive, prespecified test/estimation approaches based on RMST that can potentially capture various patterns of difference between two survival curves. Operating characteristics of the proposed RMST-based methods are investigated via numerical studies, compared to HR-based approaches, such as LR test, weighted LR tests, and MaxCombo test.