Understanding Marginal Structural Models for Time-Varying Exposures: Pitfalls and Tips

Abstract

Epidemiologists are increasingly encountering complex longitudinal data, in which exposures and their confounders vary during follow-up. When a prior exposure affects the confounders of the subsequent exposures, estimating the effects of the time-varying exposures requires special statistical techniques, possibly with structural (ie, counterfactual) models for targeted effects, even if all confounders are accurately measured. Among the methods used to estimate such effects, which can be cast as a marginal structural model in a straightforward way, one popular approach is inverse probability weighting. Despite the seemingly intuitive theory and easy-to-implement software, misunderstandings (or "pitfalls") remain. For example, one may mistakenly equate marginal structural models with inverse probability weighting, failing to distinguish a marginal structural model encoding the causal parameters of interest from a nuisance model for exposure probability, and thereby failing to separate the problems of variable selection and model specification for these distinct models. Assuming the causal parameters of interest are identified given the study design and measurements, we provide a step-by-step illustration of generalized computation of standardization (called the gformula) and inverse probability weighting, as well as the specification of marginal structural models, particularly for time-varying exposures. We use a novel hypothetical example, which allows us access to typically hidden potential outcomes. This illustration provides steppingstones (or "tips") to understand more concretely the estimation of the effects of complex time-varying exposures.

Mortality Among Patients Undergoing Blood Transfusion in Relation to Donor Sex and Parity

“…This approach has been used extensively within pharmacoepidemiology and is conceptually identical to a marginal structural model. 15 , 22 , 31 , 32 , 33 , 34 For a nontechnical summary, as well as a detailed, formal discussion of the method and a directed acyclic graph, see eFigure 1 in the Supplement .…”

Using Propensity Scores for Causal Inference: Pitfalls and Tips

“…presence of time-varying confounding. 21,45 Conventional analytic approaches, including other PS methods, fail to estimate the effects of a time-varying exposure when prior exposure affects confounders of subsequent exposure. 46 Comparison of PSM and IPW…”

“…Lastly, IPW can be expanded to causal inference for a time-varying exposure in the presence of time-varying confounding. 21 , 45 Conventional analytic approaches, including other PS methods, fail to estimate the effects of a time-varying exposure when prior exposure affects confounders of subsequent exposure. 46 …”

Using marginal structural models to analyze the impact of subsequent therapy on the treatment effect in survival data: Simulations and clinical trial examples

“…The MSM was originally developed as a methodology for post‐hoc adjusting of the confounding covariates in epidemiological studies where randomization may not be available. However, recent studies demonstrated that the method can help to interpret the data from controlled, randomized clinical trials 19 . For well controlled clinical trials with subsequent therapies, the implementation of the MSM can be significantly simplified under certain situations, for example, analysis with a point‐treatment subsequent therapy, because most of the fixed nuisance covariates can be controlled through randomization and the subsequent therapy may be the only confounder that cannot be balanced at initial randomization.…”