Reduction of Variability of Response-Adaptive Designs for Continuous Treatment Responses in Phase 3 Clinical Trials

Several response-adaptive designs are available to allocate sequentially entering patients between two competing treatments, so that a larger number of patients are treated by the better treatment. Most of the designs are for binary treatment responses. Among the few available designs for continuous responses, the continuous drop-the-loser (CDL) rule of Ivanovo et al. (2006) has lower variability of the proportion of allocation to any treatment. The present article provides some modifications to the existing CDL design and extends the CDL rule to provide two new designs that have more or less the same allocation to the better treatment as the CDL, but the variability is shown to be nearly 40% less than that of CDL. The applicability of the proposed designs is illustrated by redesigning some real clinical trials.