Restricted understanding.

Statisticians these days will often point out the benefit of calculating a restricted mean survival time (RMST). So...what is RMST? Many trials use the Cox proportional hazards model to determine if there is a difference in time-to-event outcomes. This model assumes that there is a proportional benefit in the treatment arm over time—think two Kaplan Meier curves that separate nicely and stay separated over time with a clear difference between arms. If only all results were this pretty. When we instead see curves that cross or are closely intertwined, the assumption of proportional benefit has been violated, and interpreting the hazard ratio can be unclear or misleading. In the same vein, the difference between median survival times between two arms doesn't necessarily yield the benefit expected for the average patient. RMST takes a more holistic approach by calculating the areas under the survival curves over a time period to produce a result that's often more intuitive to interpret. For example, in the above trial, RMST was 9.1 years with docetaxel versus 8.8 years without. This means, over a 10 year period, roughly 3.5 months of life is gained, on average, when docetaxel is added to radiation and ADT. | QuadShot 2021