Abstract
Randomized optimum designs of biased‐coin type are compared with other strategies for the sequential allocation of two or more treatments in a clinical trial. The emphasis is on the variance of estimated treatment contrasts. This variance, which depends on the design strategy employed, may be interpreted as the number of patients on whom information is lost. Simulations provide clear plots of the evolution of this loss during the course of the clinical trial. Copyright © 1999 John Wiley & Sons, Ltd.