Comparing Different Methods for Assessing When to Start Treatment

• Activity Number: 620
• Type: Invited
• Date/Time: Thursday, August 8, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Bryan E Shepherd*+ and Nate Mercaldo
• Companies: Vanderbilt University and Vanderbilt University
• Keywords: causal inference ; hiv ; dynamic treatment regimes

Abstract

Optimal timing of initiating antiretroviral therapy (ART) has been a controversial topic in HIV. Two highly publicized studies (Kitahata et al. 2009, N Engl J Med; and Sterne et al. 2009, Lancet) applied different analytical approaches to different observational databases and got different conclusions. Differences between the two cohorts' results could be due to differences between the patient populations, fundamental differences between the methods, or differences between technical details. For example, the two studies adjusted for different covariates, compared different thresholds, and had different criteria for qualifying measurements. If the methods were applied to the same cohort holding technical details constant, would their results be similar? In this study, we apply both approaches with simulations and using observational data from >20,000 HIV-infected persons in the CNICS cohort. We first hold technical details constant between the two methods and then repeat analyses varying technical details to understand their impact. Our study provides insights into assessing optimal dynamic treatment regimes.