Abstract:
|
It is common practice in randomized clinical trials with a survival primary endpoint to report the hazard ratio (HR) and difference in median survival. The understanding is that totality of trial results must be examined before it can be decided whether the new therapy is sufficient to modify current medical practice. However, using multiple measures to compare survival curves and to guide decision making can decrease the power of the study. This potential consequence is seldom considered at the study design stage; instead, trials are often powered to detect a clinically meaningful difference for one endpoint and on one scale (e.g., PFS HR). In this talk, I will demonstrate the statistical impact of changing the scientific question from "Is the HR comparing two treatment groups < x?" to "Are the HR and median survival comparing two treatment groups < x and > y, respectively?" Our discussion will be motivated by two recent trials in oncology, Merck's KEYNOTE-024 and Amgen's TRINOVA-1.
|