Variable selection, response adaptive randomization, and covariate-adjusted response-adaptive randomization for personalized medicine
Abstract
As the increase of biomarker information, new methods are needed to seek the biomarkers which may be related with the disease outbreak. With the incorporation of the biomarker information, new clinical trial designs are demanded to detect important interactions among treatments and biomarkers, based on both efficiency and ethics. To find the genetic biomarkers related with targeted agents, it is necessary to rely on genetic association study. We proposed a two-steps-method by combining the principal component analysis and Tukey's 1-df method, to improve the problem of power loss in testing gene-gene and gene-environment interactions in association studies. We also apply the proposed methodology to a case-control study designed to investigate the association between Crohn's Disease (CD) and DNA variants in CD related genes. Both the simulated and real data examples demonstrated major power advantages for the proposed methodology over two alternative tests of association. In Response Adaptive Randomization (RAR) designs, patients are dynamically assigned to the treatment groups using a modified allocation probability based on the current observed data. We evaluated the bias of the treatment efficacy estimators in clinical trials with RAR designs. To gain a deeper understanding of the performance of RAR, we study the properties of the response rate estimators through both theoretical derivation and extensive simulation studies. Covariate-adjusted response-adaptive (CARA) designs are expected to become useful techniques for application in the field of personalized medicine. To introduce the mechanisms involved in CARA designs, we applied the logistic regression model to provide a clear exposition. We also conduct a simulation study to comprehend the operating characteristics of the proposed CARA designs for finite samples, including the measures of the degree of balance (in terms of allocation proportions), and ethics (in terms of treatment failure rate).
Subject Area
Biostatistics|Genetics|Pharmacy sciences
Recommended Citation
Wang, Yaping, "Variable selection, response adaptive randomization, and covariate-adjusted response-adaptive randomization for personalized medicine" (2014). Texas Medical Center Dissertations (via ProQuest). AAI3689796.
https://digitalcommons.library.tmc.edu/dissertations/AAI3689796