Comparison of Two Approaches to Adjust for Rescue Medication Use Arising Post Randomization in Clinical Trials

• Activity Number: 14
• Type: Topic Contributed
• Date/Time: Sunday, July 31, 2011 : 2:00 PM to 3:50 PM
• Sponsor: International Chinese Statistical Association
• Author(s): Lixia Jiao*+
• Companies: sanofi-aventis
• Keywords: Conditional inference ; Longitudinal data ; Time-dependent covariates ; Confounding

Abstract

In certain randomized clinical trials patients may take non-randomized rescue medication due to insufficient therapeutic effect or an exacerbation of symptoms. Typically this rescue medication use reduces the observed treatment effect in intention-to-treat analysis. In this paper we focus on two competitive approaches that take account of rescue medication effect to achieve greater precision on treatment comparisons in trials with repeated quantitative outcomes. One proposed approach is to develop a multivariate (bivariate) mixed model characterizing the evolving relationship of the primary outcome measure and the use of rescue medication as a confounding outcome measure, where both variables are restricted to the left-hand side of the linear model. The other approach is to enter the rescue medication use as a time-dependent covariate in a general linear model. We illustrate these method