Title: Estimating time-varying treatment effects on restricted mean survival time in large patient databases
Authors: Chi Hyun Lee - University of Massachusetts Amherst (United States) [presenting]
Wen Li - The University of Texas (United States)
Yu Shen - UT MD Anderson Cancer Center (United States)
Jing Ning - The University of Texas MD Anderson Cancer Center (United States)
Abstract: The restricted mean survival time (RMST), which is defined as the life expectancy up to a specific time point, has recently attracted substantial attention as an alternative to the hazard ratio for quantifying the treatment effect in clinical studies. We propose a flexible model to estimate the effect of treatment based on RMST. The effect of treatment is expressed as a function of restriction time to better characterize the dynamic trend of its effect on survival. To account for possible heterogeneity across patients in large databases, we incorporate the propensity scores for receiving treatment into the model. We further introduce an ensemble approach to aggregate estimators constructed based on subsamples of the observed failure times. We evaluate the finite sample performance of the proposed single model and ensemble-based approaches through simulations, and apply the proposed methods to the study of primary inflammatory breast cancer for assessing the effect of trimodality therapy on survival.