A Method of Biased Coin Randomization, Its Implementation, and Its Validation

Abstract

Many randomized studies in small patient populations and studies in early research (such as Phase I and Phase II trials) have small to moderate numbers of patients. In such studies the use of simple randomization or blocking on only one or two factors can easily result in imbalance between treatment groups with respect to one or more potentially prognostic variables. Baseline adaptive randomization methods (such as biased coin methods) can be used to virtually guarantee balance between treatment groups with respect to several covariates. One such method, which has been implemented in Splus, is discussed in detail. The impact of the baseline adaptive randomization method on the nominal distribution of the analysis of covariance test statistic is also discussed. Rather than relying solely on the assumption that the distribution of the analysis of covariance test statistic has its nominal distribution when adaptive randomization is used, a mechanism in Splus has been developed to perform a randomization test taking into account all of the constraints imposed by the chosen adaptive randomization procedure.