To the Editor.
—Dr Schulz and colleagues wrote, "In reports of trials that had apparently used unrestricted randomization, the differences in sample sizes between treatment and control groups were much smaller than would be expected due to chance."1 They speculated that "nonrandom manipulation of comparison groups... may have occurred." For example, according to Schulz et al, out of 96 trials, and if unrestricted randomization was used, about half of trials should have had a disparity between the numbers of subjects in the treatment and control groups that fall outside the 50% prediction interval (see Figure 1 in Schulz et al1). Only eight did. However, their prediction intervals are based on a cumulative binomial distribution. This model does not reflect randomized trial conduct in practice.Consider a realistic example of a drug trial in which an investigator prepares a random assignment list in advance— usually, but not necessarily, by
Riet GT, Kessels AGH. Restricted Randomization in Randomized Controlled Trials. JAMA. 1995;274(23):1835. doi:10.1001/jama.1995.03530230021014