Identification of Treatment Responders with Multiple Longitudinal Outcomes

• Activity Number: 193
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Yumi Kondo* and Yinshan Zhao and John Petkau
• Companies: The University of British Columbia and The University of British Columbia and The University of British Columbia
• Keywords: longitudinal data ; latent class clustering ; clinical trials ; generalized linear mixed effect model ; multiple sclerosis ; Monte Carlo expectation-maximization algorithm

Abstract

Assessments of treatment effects on specific individuals in clinical trials are essential as a treatment may work well in some patients but it could be harmful to others with apparently the same disease. Identifying subset of responsive patients is a challenge in trials where a treatment efficacy is often measured by various continuous and count longitudinally-collected outcomes. Therefore, existing procedures often identify the subset based only on a single outcome. We propose a latent class clustering approach to identify the responsive patients based on a multiple longitudinal outcome mixture model. Our novel model assumes that, conditioning on a cluster label, a single outcome is modelled with the generalized linear mixed effect model, a rich class of models for longitudinal analysis, and the maximum likelihood estimates of our high dimensional model is attained by utilizing the Monte Carlo expectation-maximization algorithm. We will demonstrate our novel procedure based on a MS clinical trial dataset.