Introduction

Advancement of clinical medicine depends on accurate assessment of the safety and efficacy of new therapeutic interventions. Relevant data come from a variety of sources—theoretical biology, in vitro experimentation, animal studies, epidemiologic data—but the ultimate test of the effect of an intervention derives from randomized clinical trials. In the simplest case, a new treatment is compared to a control in an experiment designed so that some participants receive the new treatment and others receive the control. A random mechanism governs allocation to the two groups. Well-designed, carefully conducted randomized clinical trials are generally considered the most valid tests of the effect of medical interventions for reasons both related and unrelated to randomization. Randomization produces comparable treatment groups and eliminates selection bias that could occur if the investigator subjectively decided which patients received the experimental treatment. Clinical trials often use double blinding whereby neither the patient nor the investigator/physician knows which treatment the patient is receiving. Blinding the patient equalizes the placebo effect—feeling better because one thinks one is receiving a beneficial treatment—across arms. Blinding the investigator/physician protects against the possibility of differential background treatment across arms that might result from “feeling sorry” for the patient who received what was perceived, rightly or wrongly, as the inferior treatment. Determination of whether a patient had an event is based on unambiguous criteria prespecified in the trial’s protocol and applied blinded to the patient’s treatment assignment whenever possible. Because the experimental units are humans, and because randomization and blinding are used, these trials require a formal process of informed consent as well as assurance that the safety of the participants is monitored during the course of the study.