Methods and Timing of Randomization

Abstract

Randomization of study participants avoids potential bias in assigning treatment, balances assignments by treatment group on measured prognostic factors and tends to do the same for unmeasured factors, and assures that statistical tests of significance are valid. There are two classes of randomization, and each has several different methods: fixed allocation randomization (simple, blocked, and stratified) and adaptive randomization (baseline adaptive, minimization, response adaptive). The pros and cons of each method and recommendations for their use are summarized in Table 15.2. Regardless of the method used, randomization will not be successful unless it is done properly (e.g., blinding to assignment is maintained; the number of ineligible participants randomized in error is small), and it is important to randomize as closely as possible to the start of treatment to minimize the number of participants withdrawn from the study. Some potential issues in invasive studies that are not of concern or are less problematic in drug studies include differences across sites in the skill level and experience of the interventionalist/surgeon and how those affect the participant’s response, the invasive technique may continue to develop and change while the study is ongoing, and the degree to which the interventionalist/surgeon follows the protocol may vary.