Subgroup Identification in Longitudinal Studies

Wei, Yishu

doi:https://doi.org/10.21985/n2-qrkn-fh11

Work

Subgroup Identification in Longitudinal Studies

Public

Download PDF

Download All Files (.zip)

This dissertation focuses on subgroup identification in longitudinal studies. There are two different but related topics. In chapter two and chapter three, several longitudinal based methods for subgroup identification with enhanced treatment effect are proposed to correct the deficiency in measuring treatment effect by simply using a summary statistic. In chapter four, deficiency in the trajectory analysis caused by ignoring random effects in each latent class is discussed. In clinical studies, treatment effect may be heterogeneous among patients. It is of interest to identify subpopulations which benefit most from the treatment, regardless of the treatment's overall performance. In chapter two and chapter three we are interested in subgroup identification methods in longitudinal studies when nonlinear trajectory patterns are present. Under such a situation, evaluation of the treatment effect entails comparing longitudinal trajectories. We propose a tree-structured subgroup identification framework, termed interaction tree for longitudinal trajectories or IT-LT in short, which combine marginal longitudinal models with regression splines to model the nonlinear progression patterns among repeated measures. The general framework of IT-LT has the great flexibility to incorporate missing data, irregular observation times, and other complexities that are commonly encountered in longitudinal studies. Extensive simulation studies are conducted to evaluate its performance and an application to an alcohol addiction pharmacogenetic trial demonstrates its advantage. In chapter three, the multiple test based IT-LT and the multivariate multiple regression based IT-LT are proposed for studies in which every subject is followed with relatively large amount of repeated measurements. Extensive simulation studies are performed to evaluate their performance. Upon real application on the weekly PHDD (percent heavy drinking days) dataset, we find that longitudinal based methods may give more insightful subgroup estimates. Group based trajectory model (GBTM for short) is a widely used modeling procedure in social and biomedical sciences to classify subjects into latent groups and describe the latent trajectory within each group. However, two important assumptions in GBTM, equal error variances and no subject specific random effects, are often violated in real data analysis. In chapter four, we compare latent class mixed models (LCMM for short) to GBTM in modeling trajectory heterogeneity. Simulation studies are done to demonstrate the impact of such violation on the performance of GBTM. An application to the Coronary Artery Risk Development in Young Adults (CARDIA) Study is presented to show that GBTM might give misleading conclusions in real data analysis, while the latent class mixed model can yield unbiased estimates.

Creator