Building Better Adaptive Interventions by Expanding SMART

June 27, 2019:

John DziakBehavioral interventions for prevention and treatment are an important part of the fight against drug abuse and HIV/AIDS. Among the challenges faced by scientists is how and when to alter the course of treatment for participants in the intervention. Adaptive interventions change based on evidence about what is best for the participant at a given time.

For over a decade, Methodology Center researchers have developed and applied sequential, multiple assignment, randomized trials (SMARTs), which are experimental designs that can be used to build adaptive interventions that address a variety of health and behavioral challenges, such as substance abuse abstinence, weight loss, ADHD management, and language acquisition. Recently, researchers have begun developing methods to evaluate SMARTs by using multiple measures of the outcome over time rather than only considering the outcome at the end of the study. For example, a researcher who is developing an adaptive intervention to promote abstinence from alcohol may want to consider alcohol usage rates every month for six months to decide how to construct the intervention. In a recent article in Multivariate Behavioral Research by Methodology Center Investigator John Dziak, Methodology Center Affiliates Daniel Almirall and Inbal “Billie” Nahum-Shani, and others, the authors develop and demonstrate a new method for evaluating a SMART using repeated measures of a binary outcome (such as substance use versus nonuse).

The authors apply their method to the ENGAGE SMART study, which was conducted to help develop an adaptive intervention for promoting treatment engagement among cocaine- and alcohol-dependent individuals. The authors found that certain designs correlated to increased abstinence rates during the first two months but abstinence rates that were equivalent to other designs by the end of the study. Had the investigators measured relapse solely at six months, they would not have observed the relapse differences during the early months, which may have practical or clinical significance. The authors go on to provide guidelines for using multiple binary measurements of the outcome while analyzing data from a SMART.

Lead author John Dziak discussed the importance of the study. “SMART is a valuable method because conditions such as addiction and many other health problems, are chronic and often need treatment over time. In many cases, the appropriate treatment could change depending on the individual’s experiences. SMART trials can help scientists decide which set of adaptive treatment rules will work the best. In a lot of the past SMART literature, ‘work the best’ just meant having the best expected outcome at the end of the study.  But considering short-term and long-term effects together might help clinicians make better decisions to fit an individual’s  goals.  Also, it allows scientists to study delayed effects, where an early treatment choice affects how well later treatments work, and that could render theoretical insight into the treatments.”


Dziak, J. J., Yap, J. R., Almirall, D., McKay, J. R., Lynch, K. G., & Nahum-Shani, I. (2019). A data analysis method for using longitudinal binary outcome data from a SMART to compare adaptive interventions. Multivariate Behavioral Research, 1-24.

Bookmark the permalink.