Non‐Randomized Trial: A Tutorial

Abstract To determine whether a new intervention has the postulated effect in a population of interest, one typically conducts a clinical trial comparing the intervention against a control. When assigning qualified subjects to the control or the intervention groups, a formal randomization process is preferred in order to remove the potential bias in subject allocation, to produce comparable groups in terms of both known and unknown prognostic factors at baseline, and to establish the validity of statistical tests of significance. This type of studies, so called randomized control study, is the standard by which all other studies are judged. Nonrandomized clinical trials arise from situations in which it is impossible or difficult to assign subjects to treatment by chance. Unlike randomized control studies, nonrandomized trials are vulnerable to bias because of the inability to control the balance of prognostic factors between the treatment groups. Because of this limitation, nonrandomized trials are primarily used for exploratory purpose. In this paper, a tutorial description of nonrandomized trials is provided. Basic features of study designs, statistical methods for analysis, and their limitations are discussed.