AbstractConventional phase 2 clinical trials are typically single-arm experiments, with outcome characterized by one binary “response” variable. Clinical investigators are poorly served by such conventional methodology. We contend that phase 2 trials are inherently comparative, with the results of the comparison determining whether to conduct a subsequent phase 3 trial. When different treatments are studied in separate single-arm trials, actual differences between response rates associated with the treatments, “treatment effects,” are confounded with differences between the trials, “trial effects.” Thus, it is impossible to estimate either effect separately. Consequently, when the results of separate single-arm trials of different treatments are compared, an apparent treatment difference may be due to a trial effect. Conversely, the apparent absence of a treatment effect may be due to an actual treatment effect being cancelled out by a trial effect. Because selection involves comparison, single-arm phase 2 trials thus fail to provide a reliable means for selecting which therapies to investigate in phase 3. Moreover, reducing complex clinical phenomena, including both adverse and desirable events, to a single outcome wastes important information. Consequently, conventional phase 2 designs are inefficient and unreliable. Given the limited number of patients available for phase 2 trials and the increasing number of new therapies that must be evaluated, it is critically important to conduct these trials efficiently. These concerns motivated the development of a general paradigm for randomized selection trials evaluating several therapies based on multiple outcomes. Three illustrative applications of trials using this approach are presented.
A matched crossover design for clinical trials