ABSTRACT
Chapter 5 focuses on the cornerstone of clinical trials, randomization. We emphasize the sanctity of randomization and the fact that where the randomization is placed can change the primary question being answered in a clinical trial. We then move to methods of randomization, beginning with simple randomization. Future treatment assignments are impossible to predict with simple randomization, but accidental imbalances in the numbers of patients assigned to treatment and control can result, which increases the probability of between-arm imbalances in important prognostic variables. Permuted block randomization guarantees equal numbers assigned to the arms after each block, but makes future assignments more predictable. One consequence of increased predictability is the possibility of introducing selection bias. Also, balance in numbers assigned to the two arms does not guarantee balance in prognostic variables. We discuss methods with that aim, including stratified randomization and covariate-adaptive randomization. There has been recent interest in changing randomization probabilities to try to increase the probability that patients will be assigned to superior arms. We discuss this so-called response-adaptive randomization (RAR) and offer concerns about its use.
