Semiparametric Joint Modeling of Longitudinal Measurements and Time-to-event Data

• Activity Number: 314
• Type: Contributed
• Date/Time: Wednesday, August 11, 2004 : 9:00 AM to 10:50 AM
• Sponsor: General Methodology
• Author(s): Wen Ye*+ and Xihong Lin and Jeremy M.G. Taylor
• Companies: University of Michigan and University of Michigan and University of Michigan
• Keywords: semiparametric ; joint modeling ; longitudinal ; survival

Abstract

Longitudinal studies in medical research often generate both censored time-to-event data and repeated measurements on biomarkers. Recently, joint models using both types of data have been developed. Commonly, the longitudinal covariate is modeled by a linear mixed model. However, in some cases, the biomarker's time trajectory is not linear, such as the prostate specific antigen PSA profile after radio-therapy in prostate cancer study. We propose a two-stage regression calibration approach which models the longitudinal biomarker using a semiparametric mixed model, where covariate effects are modeled parametrically and the individual time trajectories are modeled nonparametrically using a population smoothing spline and subject-specific random stochastic processes. Estimates of the biomarker level and change rate at each time-event are then used as time dependent covariates in the second stage survival model. We also propose an approach to jointly estimate parameters in the two submodels in the two-stage method. The performance of the approaches is illustrated by application to a prostate cancer study.