Adaptive Designs in Substance Abuse Research, with Applications to VA and non-VA Research
A number of treatments for alcoholism and substance abuse have been shown to have efficacy, but challenges remain in terms of implementation and effectiveness. Two particular problems are that many patients who enter specialty care programs drop out within several weeks, before they have received an adequate course of treatment, and patients who do remain in treatment for more than a few weeks often exhibit considerable response heterogeneity, with some patients doing well but others either continuing to drink/use or relapse. In addition, moderately high rates of nonattendance and dropout are common in alcohol and substance abuse trials, even among patients who remain past the initial few weeks. Adaptive treatment models provide one approach to these problems. These models are designed to increase participation in treatment by providing flexible care that is adjusted over time on the basis of patient response to treatment, according to clearly operationalized, empirically derived decision rules. We describe the design, implementation, and analysis of a two-stage adaptive trial for alcoholism, in which patients were initially randomized to one of two possible definitions of ''initial response.'' Then, responders were randomized to one of two ''maintenance'' treatments, and non-responders to one of two ``rescue'' treatments. We also describe two ongoing adaptive treatment trials for alcoholism, being implemented at the Philadelphia VAMC.