Uses and Limitations of the Restricted Mean Survival Time: Illustrative Examples From Cardiovascular Outcomes and Mortality Trials in Type 2 Diabetes

Abstract

The restricted mean survival time (RMST) has been advocated as an alternative or a supplement to the hazard ratio for reporting the effect of an intervention in a randomized clinical trial. The RMST difference allows quantification of the postponement of an outcome during a specified (restricted) interval and corresponds to the difference between the areas under the 2 survival curves for the intervention and control groups. This article presents examples of the use of the RMST in a research and a clinical context. First, the authors demonstrate how the RMST difference can answer research questions about the efficacy of different treatments. Estimates are presented for the effects of pharmacologic or strategy-driven glucose-lowering interventions for adults with type 2 diabetes from 36 trials and 9 follow-up studies reporting cardiovascular outcomes and mortality. The authors show how these measures may be used to mitigate uncertainty about the efficacy of intensive glucose control. Second, the authors demonstrate how the RMST difference may be used in the setting of a clinical consultation to guide the decision to start or discontinue a treatment. They then discuss the advantages of the RMST over the absolute risk difference, the number needed to treat, and the median survival time difference. They argue that the RMST difference is both easy to interpret and flexible in its application to different settings. Finally, they highlight the major limitations of the RMST, including difficulties in comparing studies of heterogeneous designs and in inferring the long-term effects of treatments using trials of short duration, and summarize the available statistical software for calculating the RMST.