Methodological Considerations Affecting How to Treat the Baseline Measurement of an Outcome

• Activity Number: 257
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Jennifer Faerber*+ and Russell Localio
• Companies: University of Pennsylvania and University of Pennsylvania
• Keywords: ANCOVA ; Longitudinal data analysis ; baseline measure of outcome

Abstract

In analyzing clinical trials data in which the baseline measurement of the outcome is observed, some people may wish to use analysis of covariance (ANCOVA), where the observed baseline measurement is included as a covariate in the model. We simulated data for a longitudinal observational study to examine how well the fitted ANCOVA model performed compared to the standard outcome model for assessing the treatment effect when a baseline trend was incorporated into the model. Whether the baseline measure of the outcome was treated as a covariate or part of the outcome vector affected both the coverage of the true value and power. We discuss the analytical models as well as other methodological issues affecting the simulation results, including the sensitivity of the results relative to the size of the baseline trend and the correlation between the treatment effect and baseline measure.