Analysis of dynamic restricted mean survival time based on pseudo-observations

TL;DR

This paper introduces a dynamic, conditional restricted mean survival time (cRMST) method based on pseudo-observations, enabling more accurate and time-sensitive predictions of patient survival and life expectancy during follow-up studies.

Contribution

The paper develops a novel estimation and testing framework for cRMST using pseudo-observations and landmark methods, enhancing dynamic survival analysis.

Findings

cRMST estimation is accurate in simulations

cRMSTd test performs well in hypothesis testing

Dynamic RMST model outperforms static models in prediction

Abstract

In clinical follow-up studies with a time-to-event end point, the difference in the restricted mean survival time (RMST) is a suitable substitute for the hazard ratio (HR). However, the RMST only measures the survival of patients over a period of time from the baseline and cannot reflect changes in life expectancy over time. Based on the RMST, we study the conditional restricted mean survival time (cRMST) by estimating life expectancy in the future according to the time that patients have survived, reflecting the dynamic survival status of patients during follow-up. In this paper, we introduce the estimation method of cRMST based on pseudo-observations, the construction of test statistics according to the difference in the cRMST (cRMSTd), and the establishment of the robust dynamic prediction model using the landmark method. Simulation studies are employed to evaluate the statistical properties of these methods, which are also applied to two real examples. The simulation results show that the estimation of the cRMST is accurate and the cRMSTd test performs well. In addition, the dynamic RMST model has high accuracy in coefficient estimation and better predictive performance than the static RMST model. The hypothesis test proposed in this paper has a wide range of applicability, and the dynamic RMST model can predict patients' life expectancy from any prediction time, considering the time-dependent covariates and time-varying effects of covariates.