SESSION B

TITLE: Using Restricted Mean Survival Time for Classifications and Clinical Trials for Survival Endpoints
SPEAKER: Professor Ying Lu, Stanford University
Moderator: Naitee Ting

Abstract:

Because of the censoring, analyses of survival data have been focused on the difference in hazards function. For example, in a prospective clinical study that compares the difference between two groups in the time to a specific event (for example, disease progression, death), a hazard ratio estimate is routinely used to empirically quantify the between-group difference. Similarly, for a classification algorithm, we want to have subgroups that have the most different hazards function. When the hazard ratio of the two hazard functions is approximately constant over time, Cox model is a very powerful tool for these problems. However, the clinical meaning of such a ratio estimate is difficult, if not impossible, to interpret when the underlying proportional hazards assumption is violated.  When the assumption is not plausible, the hazard ratio is not a good metric to evaluate the treatment efficacy or classification efficiency. In this tutorial lecture, we will discuss several critical concerns regarding this conventional practice and propose an attractive alternative for quantifying the underlying differences between groups based on restricted mean survival time (RMST).  I will discuss various issues in employing RMST in practical analysis including the benefits of RMST in interpretation, using it as a classification efficiency metrics, using it in clinical trials, including statistical inference, selecting the truncation point, study design, power comparison, regression adjustment and extensions to competing risk and recurrent events settings. We will discuss the pros and cons of the RMST-based analysis and demonstrate that it is competitive to its hazard ratio-based conventional counterparts in many real world applications. This is a joint work with Professor Lu Tian in the Department of Biomedical Data Science, Stanford University.

 

Instructor Bio

Dr. Ying Lu, Ph.D., is a Professor of Biomedical Data Science, and by courtesy, of the Health Research and Policy and of Radiology, co-director of the Center for Innovative Study Design and Biostatistics Core of the Stanford Cancer Institute, Stanford University School of Medicine. He received his Ph.D. in Biostatistics from the University of California, Berkeley.  Professor Lu has considerable experience in statistical methodological research and led the planning and conduct of large multicenter clinical trials through the VA Cooperative Studies Program (Director of the Palo Alto Coordinating Center 2009-2016). Professor Lu’s current research interests include the statistical design and analytic methods for clinical trials, validation of biomarkers/medical diagnoses, meta-analysis, and medical decision making. He is the author of 245 peer-reviewed publications and editor of several books. He is the 2019 President Elect of WNAR, the 2014 President of the International Chinese Statistical Association (ICSA) and served as a member of DSMBs for clinical trials and FDA PCNS Drug Advisory Committee. Professor Lu was a fellow of ASA and biostatistics editor of JCO Precision Oncology.

 

This entry was posted in . Bookmark the permalink.

Leave a Reply

Your email address will not be published. Required fields are marked *