Event Summary:
A one-day workshop organized by the Center for Interdisciplinary Research and
Consulting at the University of Maryland, Baltimore County (UMBC), and the
Center for Innovative Study Design (CISD) at Stanford University in the field of Clinical Trials and Regulatory Science.
Registration is now closed
⇒View Archives here⇐
Organizing Committee:
Yi Huang (UMBC)
Nagaraj Neerchal (UMBC)
Bimal Sinha (UMBC)
Tze Lai (Stanford)
Phillip Lavori (Stanford)
Ying Lu (Stanford)
Jie Chen (Novartis)
Joseph Heyse (Merck)
For more information:
Please contact Dr. Yi Huang, Department of Mathematics and
Statistics, University of Maryland Baltimore County (UMBC),
1000 Hilltop Circle, Baltimore, MD 21250.
E-Mail: yihuang@umbc.edu.
For any Technical difficulties Please contact Zana Coulibaly at czana1@umbc.edu.
Participant Information
Jenny Huang from Genentech/Roche.
Paper: Which test is more powerful: HR-based vs RMST-based?
When the primary endpoint is time to event in a randomized clinical trial, the hazard ratio (HR) is often used to measure the treatment effect. However, HR may depend on the censoring distribution and could be difficult to interpret when the proportional hazards (PH) assumption is violated. On the other hand, the contrast in restricted mean survival time (RMST) as a model-free metric for the treatment effect is an appealing alternative. However, there is a concern about the power of the test procedure based on RMST in detecting the group difference when compared to the HR-based test, ie., the popular logrank test. We investigate the asymptotical relative efficiency (ARE) between the two tests analytically as well as numerically under different alternatives. The main conclusion is that the two tests are comparable in most of the realistic settings, and the RMST-based test is more desirable when the PH assumption does not hold and the treatment benefit occurs early.