Abstract: Urn model is a most commonly used randomization method of adaptive design. In this article, with some true data from the clinical trial of a new drug treating oral ulcer, the generalized Friedman's urn model was used to calculate the allocation proportions between the two treatments groups, and simultaneously, the influence of the allocation proportions on the power of statistical test were explored trough computer simulation experiments. It was showed in the results that the effects of response adaptive randomization on power was not as serious as we assumed, because the allocation proportions between the two treatments groups is usually less than 3∶1 for most clinical trials. We hope this study can provide some references for the application of urn model in practical clinical trials.

Key words: Adaptive design, Urn model, Simulation examination