A note on conditional and unconditional allocation concealment

Abstract

Allocation concealment is often misunderstood but, when understood properly, can offer a metric for evaluating the quality of nonadaptive randomization procedures. For example, appeal to allocation concealment tells us that the maximal procedure and the big stick procedure are both superior to permuted blocks. However, the standard definition of allocation concealment is not sufficient to properly evaluate adaptive randomization, be it covariate adaptive or response adaptive randomization. We therefore propose a new concept, namely conditional allocation concealment, to be distinguished from conventional or unconditional allocation concealment, to better understand the strengths and weaknesses of existing adaptive randomization methods. Better evaluation of randomization methods will predictably lead to better randomization, too, and this will in turn lead to more robust trials.