Control groups for psychosocial intervention outcome studies.

In psychosocial outcome research, as contrasted to pharmacologic research, control groups receiving inert treatment, designed to raise expectations but otherwise provide no service, are almost never indicated; this is true because of methodologic as well as ethical reasons. Four types of comparisons suffice as alternatives: treatment versus no treatment, treatment versus minimal treatment, treatment A versus treatment B, and dismantling. When choices are made among these types of comparisons with power analysis and eight other factors taken into account, the questions of outcome research should be answerable with maximum economic efficiency, with maximum benefit to subjects, and without deception.