Joint Modeling of Survival Time and Longitudinal Outcomes with Flexible Random Effects

• Activity Number: 660
• Type: Invited
• Date/Time: Thursday, August 8, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Jianwen Cai*+ and Jaeun Choi and Donglin Zeng and Andy Olshan
• Companies: The University of North Carolina at Chapel Hill and Harvard Medical School and The University of North Carolina and The University of North Carolina at Chapel Hill
• Keywords: Gaussian mixtures ; Generalized linear mixed model ; Maximum likelihood estimator ; Random effect ; Simultaneous modeling ; Stratified Cox proportional hazards model

Abstract

In biomedical research, it is common for both survival time and longitudinal outcomes to be collected for a subject, along with the subject's characteristics or risk factors. Joint analysis of longitudinal outcomes and survival time is used to find important variables for predicting both longitudinal outcomes and survival time which are correlated within the same subject. Random effects are introduced to account for the dependence between survival time and longitudinal outcomes due to unobserved factors. A Gaussian distribution with mean zero is conventionally assumed for random effects. Misspecifying normality assumption can lead to serious bias in estimation. In this talk, we relax the normality assumption of random effects by assuming the underlying distribution to be unknown. We propose to use a mixture of Gaussian distributions as an approximation in the estimation. Weights of the mixture components are estimated with model parameters using the Expectation-Maximization (EM) algorithm. The proposed method performs well in finite samples. We illustrate our approach with data from the Carolina Head and Neck Cancer Study (CHANCE).