Placebo-controlled trials: when are they needed?

Abstract

Randomized controlled trials require that patients be assigned to either receive the medication under investigation or to a control condition. It is not necessary that the control arm of the trial involve a placebo; it may consist of a different drug or conventional treatment. However, this can lead to serious problems in interpretation if the trial concludes that there is no statistically significant difference between the groups. In studies where a placebo arm was not used, it is extremely difficult to determine if this lack of a difference was due to equal effectiveness of the treatments, or was the result of a Type II error; i.e., erroneously concluding that there was no difference when one, in fact, existed. This may have been due to poor design or execution of the trial, or to low power resulting from an insufficient number of subjects in the study. The need for placebo controls can be minimized if there is assurance that the trial is well-designed and executed, and this may require external review of studies, especially those which are not peer-reviewed. If a placebo arm is deemed necessary, then it may be necessary to ensure that certain condition are met. These would include (1) setting very high requirements for informed consent; (2) an external group that monitors the prevalence of adverse reactions; (3) requiring that attending physicians not enroll their own patients into a trial; (4) not paying physicians on the basis of the subjects completing the trial; (5) having a time limit for patients remaining in the placebo condition; and (6) using only drug-free or drug-resistant patients. Further, alternatives to strict randomization schemes, such as 'play the winner' strategies, should be explored further.