Using the Restricted Mean Survival Time Difference as an Alternative to the Hazard Ratio for Analyzing Clinical Cardiovascular Studies

Abstract

ardiovascular trials often use the time to a clinical event as the primary end point when evaluating a new treatment, versus control, via clinical and statistical significance criteria. For the past 50 years, the hazard ratio (HR) has been routinely used for quantifying the treatment effect. However, it is difficult to interpret clinical significance using a ratio measure, such as HR, when there is no reference hazard available from the control arm. Moreover, valid HR analysis requires the proportional hazards (PH) assumption: that the ratio of hazard curves is constant over time. This assumption is hardly plausible in practice. When PH is not met, HR may lack statistical power to detect a true treatment effect. Furthermore, without PH, the estimated HR is not a simple average of HRs over time, and is even more difficult to interpret. In this article, we discuss the advantages of an alternative analytical procedure based on the restricted mean survival time (RMST) 1,2 via 3 examples.