Date of Award

1-1-2012

Document Type

Campus Access Dissertation

Department

Epidemiology and Biostatistics

Sub-Department

Biostatistics

First Advisor

Bo Cai

Abstract

In clinical trials, time-to-event data (survival component) and longitudinal data (longitudinal component) are often collected. In order to model these two components simultaneously, the joint modeling approach which can reduce potential biases and improve the efficiency in estimating treatment effects becomes increasingly important. In this work, we proposed a joint model where a nonlinear mixed-effect model with both exponential shrinkage and a linear progression term with time is used to describe the change in tumor size and an accelerated failure time frailty model with either weibull or exponential distribution is used to describe the overall survival time. The survival and longitudinal components are linked through random effects with appropriate adjustments. The simulation study shows that the joint analysis is better than separate analysis in terms of the parameter estimate and goodness-of-fit. A real dataset is applied to our proposed model to see the applicability. The Weibull survival distribution is found to have a better fit than exponential survival distribution in term of the goodness-of-fit in this real dataset.

Share

COinS