Joint Analysis of Accelerated Failure Time and Longitudinal Data Under a Case-Cohort Design

• Activity Number: 330
• Type: Topic Contributed
• Date/Time: Tuesday, August 2, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Lan Kong*+ and Xinxin Dong
• Companies: University of Pittsburgh and University of Pittsburgh
• Keywords: Joint model ; biomarker ; mixed effects model ; case-cohort

Abstract

A panel of biological markers is often measured over time to better understand the mechanism of a disease and aid in the development of effective treatments. In large cohort studies, it is prohibitive to measure multiple candidate markers over time for each individual. Case-cohort design provides a cost effective solution when the covariate of interest is expensive to measure and the event rate is low. Under the case-cohort design, biomarkers are measured for a subcohort that is randomly selected from the entire cohort and any additional cases outside the subcohort. Joint modeling of longitudinal marker and clinical outcome is an appealing technique to reveal the relationship between biomarker trajectory and the evolution of the disease. We propose a joint analysis of accelerated failure time (AFT) and longitudinal data from case-cohort studies. We use the shared latent parameters to link the longitudinal model and AFT model, and obtain the maximum likelihood estimators by Gaussian quadrature method. We evaluate the performance of our case-cohort estimator and its relative efficiency to full cohort estimator through simulation studies and give a numerical example for illustration.