A Simulation Study Comparing the Performance of Time-Varying Inverse Probability Weighting and G-Computation in Survival Analysis

Abstract

Inverse probability weighting (IPW) and g-computation are commonly used in time-varying analyses. To inform decisions on which to use, we compared these methods using a plasmode simulation, based on the Effects of Aspirin in Gestation and Reproduction trial. In our main analysis, we simulated 1226 individuals, followed for up to 10 weeks. The exposure was weekly exercise, and the outcome was time to pregnancy. We controlled for 6 confounders: 4 baseline (race, ever smoker, age, and BMI) and 2 time-varying (compliance to assigned treatment and nausea). We sought to estimate the average causal risk difference by 10 weeks, using IPW and g-computation implemented using a Monte Carlo (MC) estimator and iterated conditional expectations (ICE). Across 500 simulations, we compared the bias, empirical standard error (ESE), average standard error, standard error ratio, and 95% confidence interval coverage of each approach. IPW (bias: 0.017; ESE: 0.039; coverage: 92.6%) and MC g-computation (bias: -0.012; ESE: 0.031; coverage: 94.2%) performed similarly. ICE g-computation was the least biased but least precise estimator (bias: 0.010; ESE: 0.058; coverage: 93.4%). When choosing an estimator, one should consider factors like the research question, prevalence of the exposure and outcome, and the number of time points being analyzed.