Introduction to the Special Issue.

Abstract

Multi-armed trials are randomized experiments in which subjects are randomly assigned to more than two conditions. In contrast to standard twoarmed experiments that usually involve randomization to two conditions (usually treatment and control), multi-armed trials involve three or more arms (usually control and two or more active treatment conditions). This special issue of Evaluation Review includes five articles on multiarmed trials. The first three are retrospective works by pioneers of the method and its use in their respective fields. The last two articles describe contemporary applications. In this introduction, I describe an important recurrent theme, and then briefly comment on each article. Multi-armed experiments have benefits but also pose challenges. Consider first the benefits. Multiple treatment arms allow an evaluation of more causal contrasts compared to a standard two-armed study. One treatment may be compared to another, and each treatment may be compared to a control. Multiarmed trials expand the opportunity to test multiple hypotheses concerning which treatment works best under specific conditions, and for certain subgroups of interest. The additional information yielded through multi-armed experiments is valuable: in complex decision-making situations, it offers flexibility of solutions and helps to establish the external validity of causal inferences. However, we must temper our enthusiasm over these benefits with the reality that multi-armed trials engender complexities that are either irrelevant to, or more-simplified with, two-armed trials. They include the logistics of