An Introduction to Dynamic Treatment Regimes
Recent public investments for comparative effectiveness studies have drawn attention to appropriate methods for designing such studies and analyzing comparative effectiveness data. Of particular concern is how we can best understand heterogeneity in outcomes among patients and translate that understanding to clinical practice. Treatment of patients with chronic diseases or disorders involves a series of decisions made over time. Providers periodically adjust, change, modify, or discontinue interventions based on the patient's observed progress, side effects, compliance, and so on, with the goal of "individualizing" treatment to the patient. A "dynamic treatment regime," also referred to as an "adaptive treatment strategy," is a set of formal, sequential "decision rules" that operationalize this process, where each rule corresponds to a key decision point at which a treatment decision is to be made based on information on the patient up to that point. Dynamic treatment regimes leading to the most desirable outcomes can be estimated based on longitudinal data from suitably designed clinical trials using so-called "learning" methods. This presentation sets the stage for the following two talks by providing an introduction to dynamic treatment regimes, to so-called Sequential Multiple Assignment Randomized Trials (SMARTs), and to the basic premise of learning methods.