Simulating Data From Marginal Structural Models for a Survival Time Outcome

Area of Science:

• Causal inference
• Survival analysis
• Statistical modeling

Background:

• Marginal structural models (MSMs) are crucial for estimating causal effects of treatments on survival outcomes, especially with time-dependent confounding.
• Inverse probability of treatment weighting (IPTW) is a common method for fitting MSMs.
• Simulation studies are essential for evaluating statistical methods, but simulating data for MSMs with potential outcomes has been challenging.

Purpose of the Study:

• To propose a novel simulation method for marginal structural models (MSMs) that overcomes existing restrictions on the data-generating mechanism.
• To facilitate accurate performance evaluation of statistical methods for causal effect estimation in survival analysis.
• To enable simulation studies for MSMs without imposing limitations on the underlying data-generating process.

Main Methods:

• Developed a new algorithm for simulating data under marginal structural models (MSMs) for survival outcomes.
• The proposed method accommodates various MSM types, including logistic, Cox, and additive hazards models.
• The simulation allows for discrete or continuous treatment variables and conditional hazards based on baseline covariates.

Main Results:

• A simulation study was conducted to illustrate the utility of the proposed algorithm.
• The study compared confidence interval coverage for causal effect estimates derived from MSMs fitted via IPTW.
• The new simulation method provides a flexible tool for assessing MSM performance in diverse scenarios.

Conclusions:

• The proposed simulation method offers a flexible and unrestricted approach for generating data under marginal structural models (MSMs).
• This advancement supports more robust evaluations of causal inference methods in survival analysis.
• The method is applicable to various survival models and treatment variable types, enhancing simulation study capabilities.