Covariate information is usually available and often plays a critical role in a clinical study. The stratified permuted block design and the classical covariate-adaptive designs have been widely employed to balance important covariates in clinical trials. Both designs have some serious drawbacks. In addition, there is no theoretical justification of the covariate-adaptive designs in the literature. In this project, two new families of adaptive designs are proposed and their properties are studied. The first family of designs overcomes the drawbacks of the stratified permuted block design and the classical covariate-adaptive design, and hence provides better balance. The second family of designs is proposed to detect the interaction between treatment and covariate more efficiently. Also the investigator introduces a new technique (called "drift conditions") to study the asymptotic properties of covariate-adaptive designs. This project will produce new sequential tools for solving many practical problems. The proposed methods will be applied to some specific applications.

The objective of this project is to develop new methods for clinical trials based upon covariate information. With today's advanced technology, it becomes easier and easier to collect useful covariate information in sequential experiments. For example, scientists have identified many new biomarkers that may link to certain diseases over the past several decades. Since one is now able to collect information on important biomarkers (covariate information) of each patient, it becomes more and more important to incorporate information on covariates into the design of clinical trials. The investigator will propose two new families of adaptive designs and study their properties. The first family of designs overcomes the drawbacks of the classical covariate-adaptive designs. The second family of designs is proposed to detect the interaction between treatment and covariate more efficiently. Upon completion of this project, one will be able to apply new designs in clinical trials for personalized medicine. The research project will produce some advanced statistical tools, which may be applied in many fields including drug development, medical studies, industrial experiments, economics and finance.

Agency
National Science Foundation (NSF)
Institute
Division of Mathematical Sciences (DMS)
Type
Standard Grant (Standard)
Application #
1209164
Program Officer
Gabor J. Szekely
Project Start
Project End
Budget Start
2012-07-01
Budget End
2014-05-31
Support Year
Fiscal Year
2012
Total Cost
$110,000
Indirect Cost
Name
University of Virginia
Department
Type
DUNS #
City
Charlottesville
State
VA
Country
United States
Zip Code
22904