Causal Effect Estimation on Restricted Mean Survival Time in Case-Cohort Studies via a Matching Design

2026-05-06

AI Summary

This study addresses the challenge of efficiently and accurately estimating the marginal causal effect on restricted mean survival time (RMST) while controlling for confounding bias in stratified case-cohort studies. The authors propose a novel template matching–based approach that overcomes the conventional requirement of equal sample sizes between exposed and unexposed groups. By flexibly adjusting for measured confounders through either propensity score or direct covariate matching, and integrating asymptotic theory with bootstrap-based variance estimation, the method substantially improves estimation efficiency. Simulation results demonstrate that the proposed estimator outperforms traditional matching methods in finite samples. The approach is successfully applied to the Atherosclerosis Risk in Communities (ARIC) study to estimate the causal effect of serum high-sensitivity C-reactive protein (hs-CRP) levels on event-free survival from coronary heart disease.

Abstract

In large observational studies, the case-cohort design is commonly used to reduce the cost associated with covariate measurement. For survival outcomes, literature has suggested that the restricted mean survival time (RMST) be a more appropriate marginal causal effect measure than the hazard ratio. In this paper, we develop a marginal causal effect estimation method for RMST difference under the stratified case-cohort design. We adjust for measured confounders using an innovative template matching design. Compared with conventional matching designs, template matching allows greater flexibility in the sample sizes of the exposed and unexposed groups. We establish the asymptotic properties of the proposed causal effect estimators and develop a bootstrap procedure to estimate their variances. By conducting comprehensive simulation studies, we evaluate the finite sample performance of the proposed estimators, demonstrate the advantage of template matching over conventional matching, and compare between matching on propensity score and matching on covariates. Finally, we apply the proposed methods to the Atherosclerosis Risk in Communities (ARIC) Study to estimate the marginal causal effect of serum hs-CRP level on the coronary heart disease-free survival.

Problem

• causal effect estimation
• restricted mean survival time
• case-cohort study
• matching design
• confounder adjustment

Innovation

• template matching
• restricted mean survival time
• case-cohort design
• causal effect estimation
• bootstrap variance estimation