Back to Annual Meeting Page		American Public Health Association 133rd Annual Meeting & Exposition December 10-14, 2005 Philadelphia, PA

Abstract #120592

Maya Petersen, Division of Biostatistics, University of California, Berkeley, Earl Warren Hall #7360, Berkeley, CA 94720, 510-642-3241, mayaliv@gmail.com

Marginal structural models are an established method of analysis that effectively addresses confounding in longitudinal data. However, these models do not allow estimation of how treatment effects may change as a result of time-varying covariates. We developed a generalization of marginal structural models, History-Adjusted Marginal Structural Models (HA-MSM), that incorporates time-dependent effect modification, and identifies an optimal dynamic treatment strategy.

We illustrate this method using an example drawn from the treatment of HIV-infected patients experiencing incomplete viral suppression on antiretroviral therapy. Lack of effective and well-tolerated alternative regimens and the desire to protect future treatment options may result in a decision to delay switching for some individuals. We apply HA-MSM to observational clinical data to investigate how long to wait and how to decide when to switch regimens. HA-MSM allow estimation of how the effect of non-suppressive therapy may differ depending on time-varying covariates, and identify a decision rule for switching that is expected to optimize patient outcome.

Learning Objectives:

Presenting author's disclosure statement:

I wish to disclose that I have NO financial interests or other relationship with the manufactures of commercial products, suppliers of commercial services or commercial supporters.