Developing and comparing treatment strategies: an annotated portfolio of designs.

Useful clinical strategies are adaptive, specifying the sequence of treatments that are alternatives, what it means for the treatment to "work," the rules for abandoning a treatment, and the subsequent treatments. Because combinatorial complexity precludes comparison of every possible whole strategy, current experiment-based methods rely on comparisons among a few options at particularly crucial decision points, and strategies are pieced together from scraps of information. Nonexperimental methods for strategy development offer a seductive alternative, but their advantages may be illusory. Clinical investigators deploy a wide range of study designs to compare treatment strategies in mental health. This article organizes the types of designs by their purpose and annotates this list with comments on the strengths and weaknesses of each type. We conclude with some general comments on the overall process of development of treatment strategies.