Date on Master's Thesis/Doctoral Dissertation

5-2016

Document Type

Doctoral Dissertation

Degree Name

Ph. D.

Department

Bioinformatics and Biostatistics

Degree Program

Biostatistics, PhD

Committee Chair

Kulasekera, K.B.

Committee Co-Chair (if applicable)

Datta, Somnath

Committee Member

Folz, Rodney

Committee Member

Gaskins, Jeremy

Committee Member

Mitra, Riten

Author's Keywords

Personalized Treatments; Single Index Models; Multi-state models; Right-censoring

Abstract

This dissertation contains three research projects on personalized medicine and a project on multi-state modelling. The idea behind personalized medicine is selecting the best treatment that maximizes interested clinical outcomes of an individual based on his or her genetic and genomic information. We propose a method for treatment assignment based on individual covariate information for a patient. Our method covers more than two treatments and it can be applied with a broad set of models and it has very desirable large sample properties. An empirical study using simulations and a real data analysis show the applicability of the proposed procedure. We then extend this idea for treatment section for survival outcomes under right-censoring by introducing re-weighted estimation to adjust the bias caused by censoring. Series of empirical studies using simulations show the desirable performance of re-weighted estimation concept in treatment selection in finite sample cases. We provide a real data application of the proposed procedure to illustrate the applicability for right-censored data. Next we propose a novel method for individualized treatment selection when the treatment response is multivariate. The proposed method uses a rank aggregation technique to estimate an ordering of treatments based on ranked lists of treatment performance measures such as smooth conditional means and conditional probability of a response for one treatment dominating others. An empirical study demonstrates very desirable performances of the proposed method in finite sample cases. We also present a data analysis using a HIV clinical trial data to show the applicability of the proposed procedure for real data. Multi-state models are extensions of simple survival models that incorporate the progression of a subject in an interconnected system such as a disease network. An important measure arising from a mutistate model is the subjects’ state occupational probabilities given baseline covariates. In the final portion of this dissertation we introduce an inverse censoring probability re-weighted semi-parametric single index model based approach to estimate conditional state occupation probabilities of a given individual in an acyclic multistate model under right-censoring. Besides obtaining a temporal regression function, we also test the potentially time varying effect of a baseline covariate on future state occupations. We show that the proposed technique has desirable finite sample performances. Its performance is competitive when compared with two other existing approaches. We illustrate the proposed methodology using two different data sets. First we re-examine a well known data set on various event times tracking the progression of a sample of leukemia patients undergoing bone marrow transplant. Our second illustration is based on the functional status of a set of spinal cord injured patients undergoing a rehabilitation program.

Included in

Biostatistics Commons

Share

COinS