A G-Formula Estimator for Performing Causal Mediation Analysis with Survival Outcomes: Investigating the Relationship Between Statins, Cholesterol and Cardiovascular Diseases

• Activity Number: 414 - Advances in Estimation Methods
• Type: Contributed
• Date/Time: Tuesday, July 31, 2018 : 2:00 PM to 3:50 PM
• Sponsor: SSC
• Author(s): Denis Talbot* and Joseph A Delaney and Veit Sandfort and David M Herrington and Robyn L McClelland
• Companies: Universite Laval and University of Washington School of Public Health and National Institutes of Health and Heart and Vascular Center of Excellence and University of Washington School of Public Health
• Keywords: Biostatistics; Causal Inference; Mediation Analysis; Survival Analysis

Abstract

Statin drugs were initially developed for reducing blood cholesterol as a mean of improving cardiovascular health. However, there is biological evidence that statin treatment yields cardiovascular benefits via other pathways. We developed a semi-parametric g-formula estimator for performing a causal mediation analysis between statin use, cholesterol and cardiovascular events. The analysis was performed on data from the Multi-Ethnic Study of Atherosclerosis, a multi-center population-based prospective cohort study. Our g-formula estimator directly accommodates censored time-to-event outcomes, provides estimates on risk difference and relative risk reduction scales that are easy to interpret and requires relatively few parametric assumptions. We performed a simulation study to investigate the empirical properties of this estimator under four scenarios and modest sample size: 1) ideal circumstances, 2) no indirect effect, 3) mildly misspecified models and 4) lack of adherence for some participants. Small bias and well-calibrated confidence intervals were obtained under all scenarios considered.