Should Hazard Ratio Be Used to Quantify Treatment Effect?

• Activity Number: 452 - Section on Statistics in Epidemiology P.M. Roundtable Discussion (Added Fee)
• Type: Roundtables
• Date/Time: Wednesday, August 10, 2022 : 12:30 PM to 1:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Ronghui Xu*
• Companies: University of California, San Diego
• Keywords: causal inference; randomzied studies; observational studies; hazard contrast; time-varying

Abstract

Hazard ratio (HR) has long been used to quantify treatment effects in randomized clinical trials in cancer, HIV, and many other disease areas. It is also well understood and accepted among the medical community. On the other hand, there has been argument in the causal literature against the use of hazard function, as well as caution against the use of HR or other hazard contrasts when they change over time. Yet most recently (as of the writing of this abstract), the New England Journal of Medicine published a statistical analysis paper on the effectiveness of Covid-19 vaccines using time-varying HR. We hope that these different approaches and viewpoints can generate live discussion on this very important topic, as it affects many statistical applications and research.