Assessment of randomization procedures in the presence of selection and chronological bias

Abstract

Every randomized clinical study needs a method of allocating treatments to patients to ensure the basis for a valid statistical inference. In the literature, the susceptibility of randomization procedures to both selection and chronological bias is already well studied. However, no optimality statement is given for the selection of a randomization procedure, when both types of bias are present simultaneously. Therefore, a linked assessment criterion for the selection of an appropriate randomization procedure is presented. The linked assessment criterion is based on desirability functions. For the calculation of the optimality score of a randomization procedure the importance of the investigated types of bias is considered. Several randomization procedures are compared for the sample sizes 12, 50 and 200. Finally, a recommendation for the choice of a randomization procedure is given in the presence of both selection and chronological bias.