In health and social sciences, research questions often involve systematic assessment of the modification of treatment causal effect by patient characteristics, in longitudinal settings with time-varying or post-intervention effect modifiers of interest. In this work, we investigate the robust and efficient estimation of the so-called Counterfactual-History-Adjusted Marginal Structural Model (van der Laan and Petersen (2007)), which models the conditional intervention-specific mean outcome given modifier history in an ideal experiment where, possible contrary to fact, the subject was assigned the intervention of interest, including the treatment sequence in the conditioning history. We establish the semiparametric efficiency theory for these models, and present a substitution-based, semiparametric efficient and doubly robust estimator using the targeted maximum likelihood estimation methodology (TMLE, e.g. van der Laan and Rubin (2006), van der Laan and Rose (2011)). To facilitate implementation in applications where the effect modifier is high dimensional, our third contribution is a projected influence curve (and the corresponding TMLE estimator), which retains most of the robustness of its efficient peer and can be easily implemented in applications where the use of the efficient influence curve becomes taxing. In addition to these two robust estimators, we also present an Inverse-Probability-Weighted (IPW) estimator (e.g. Robins (1997a), Hernan, Brumback, and Robins (2000)), and a non-targeted G-computation estimator (Robins (1986)). The comparative performance of these estimators are assessed in a simulation study. The use of the TMLE estimator (based on the projected influence curve) is illustrated in a secondary data analysis for the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) trial.



Included in

Biostatistics Commons