The cyclical and heterogeneous nature of many substance use disorders highlights the need to adapt the type or the dose of treatment to accommodate the specific and changing needs of individuals. This proposal is motivated by the Extending Treatment Effectiveness of Naltrexone and the Adaptive Treatment for Cocaine Dependence trials, sequential multiple assignment randomized trials (SMART) designed to find a (personalized) rescue treatment for alcohol or/and cocaine dependent patients. One of the main challenges in these trials is the high rate of noncompliance to the assigned treatments. This feature has made it virtually impossible for investigators to fully explore the possibility of building high quality treatment strategies using the data. Our overarching aim is to address this particular challenge through developing and subsequently applying new statistical methods to the data. A SMART trial is a multi-stage trial that can inform the design of an adaptive treatment strategy (ATS) which formalizes an individualized treatment plan and where current treatment strategy can depend on a patient's past medical and treatment history. An optimal ATS is one that maximizes a specified health outcome of interest. Existing methods in analyzing SMART data are limited to intention-to-treat (ITT) analyses. That is the treatment effect at each stage is estimated based on the treatment group to which an individual was randomized at that stage regardless of whether the individual complied with their assigned treatment. One major concern is that the relationship between the experimental manipulation and the outcome may be confounded by treatment noncompliance. We develop methodologies that can be used to adjust for noncompliance in analyzing data collected in SMARTs. Specifically, we extend the principal strata framework and Bayesian Copulas to multi-stage randomized trials setting and propose novel procedures that estimate the mean outcome under different ATSs. We also propose a novel Bayesian machine learning approach that can be used to construct deeply tailored (i.e., individualized) treatment strategies that take into account patients' demographic factors, measures of mental health and alcohol use, obsessive-compulsive drinking and alcohol craving scales, physical composite scores. Finally, we will develop easy-to-use, publicly available open-source software leveraging the R and Python languages that implements our methods. This will provide an expandable platform that will assist researchers in developing new optimal ATSs for patients suffering from alcoholism and other substance use disorders.

Public Health Relevance

This project aims to address the need for robust, rigorous and computationally efficient methods for analysis of sequential multiple assignment randomized trials (SMARTs) data in the presence of partial compliance. The methods developed in this project will improve the usefulness, interpretability and generalizability of the results obtained by SMARTs and help mental health and substance use disorder scientists to develop more potent adaptive treatment strategies to guide the individualization of mental health and substance use disorder treatments.

National Institute of Health (NIH)
National Institute on Drug Abuse (NIDA)
Research Project (R01)
Project #
Application #
Study Section
Biostatistical Methods and Research Design Study Section (BMRD)
Program Officer
Duffy, Sarah Q
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
University of Rochester
Biostatistics & Other Math Sci
School of Medicine & Dentistry
United States
Zip Code