Covariate Adjusted Designs for Combining Efficiency, Ethics and Randomness in Normal Response Trials

This paper deals with the problem of allocating patients to two competing treatments in the presence of covariates in order to achieve a good trade-off between efficiency, ethical concern and randomization. We propose a compound criterion that combines inferential precision and ethical gain by flexible weights depending on the unknown treatment effects. In the absence of treatment-covariate interactions, this criterion leads to a locally optimal allocation which does not depend on the covariates and can be targeted by a suitable implementation of the doubly-adaptive biased coin design aimed at balancing the roles of randomization, ethics and information. Some properties of the suggested procedure are described.