Abstract
Health benefits of resistance exercise (RE), particularly in lowering cardiovascular disease (CVD) risks, are less understood in comparison to aerobic exercise (AE). Motivated by big data from the Aerobics Center Longitudinal Study (ACLS), we study the direct and indirect effects of RE on CVD risks. The primary outcome in our study, total CVD events (CVD morbidity and mortality combined), is modeled as a survival outcome. To investigate the pathway from RE to CVD outcome through potential mediators, we first conduct causal mediation analysis based on marginal structural models (MSMs). To fully account the information from repeated measurements of the mediators, we also adopt a joint model of the CVD survival outcome and multiple longitudinal trajectories of the mediators. Results show statistically significant direct effects of RE and AE on lowering the risk of total CVD events under each pathway. The causal effect of RE and AE on CVD risk is also studied across different age and gender groups. Furthermore, we produce a ranking for the relative importance of the potential risk factors for CVD, with total cholesterol ranking the highest.