Maximum entropy constrained balance randomization for clinical trials.

A specialization of the biased coin clinical trial randomization schemes of Pocock and Simon (1975) and Efron (1971) is proposed which gives guidelines for the coin biasing probabilities of assignment to treatment. The scheme parameterizes the often conflicting goals of random treatment assignment and allocation for stratum balance in terms of a single trade-off parameter specifying a desired degree of determinism. Treatment radomization probabilities are obtained iteratively in an exponential form which maximizes randomization entropy subject to a specified balance constraint.