Uses and limitations of the restricted mean survival time: illustrative examples from cardiovascular outcomes and mortality trials in type 2 diabetes

The Restricted Mean Survival Time (RMST) has been advocated as an alternative, or supplement, to the hazard ratio (HR) for reporting the effect of an intervention in a randomized clinical trial. The RMST difference allows quantifying the postponement of an outcome over a specified (restricted) time interval and corresponds to the difference between the areas under the two survival curves for the intervention and the control arm.In this article, we present examples of its uses in a research and a clinical context. First, we show how the RMST difference can answer research questions about the efficacy of different treatments. We present estimates for the effects of pharmacological 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. We show how these measures can be used to mitigate uncertainty about the efficacy of intensive glucose control. Second, we demonstrate how the RMST difference can be used in the setting of a clinical consultation to guide the decision to start or discontinue a treatment.We then discuss the advantages of the RMST over the absolute risk difference, the number needed to treat, and the median survival difference. We argue that the RMST difference is both easily interpretable and flexible in its application to different settings. Lastly, we highlight its major limitations, 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.