Marginal Structural Models for a Continuous Outcome When the Risk of Death Depends on Treatment

• Activity Number: 4 - Recent Advance of Causal Inference in Failure Time Settings
• Type: Invited
• Date/Time: Sunday, July 28, 2019 : 2:00 PM to 3:50 PM
• Sponsor: ENAR
• Author(s): Judith Lok*
• Companies: Boston University, Dept of Mathematics and Statistics
• Keywords: Causal inference; Marginal Structural Models; Risk/benefit; Combining survival and continuous outcomes; Quantile regression; HIV/AIDS

Abstract

We will present an extension of Marginal Structural Models that incorporates both a continuous outcome and death, by combining the continuous outcome with death into one single outcome. We aim to provide clinicians with a risk/benefit profile in the form of an assessment of the treatment effect on both overall survival and the "survival-adjusted” median continuous outcome: the threshold such that half of the patients are alive with a value of the continuous outcome that is above this threshold; the other half of the patients have either died or have a value of the continuous outcome below the threshold. Our method builds on Marginal Structural Models and on quantile regression. We will illustrate our method by estimating how statin treatment affects the survival-adjusted median neurocognitive score in HIV-infected US patients on antiretroviral treatment (ART). Most HIV-infected patients are currently on ART, and statins are the most commonly prescribed drug class worldwide and have been shown to help prevent cardiovascular disease. Our results will inform whether neurocognitive side effects need to be taken into account when deciding on statin treatment for HIV-infected patients.