A two-stage design for randomized trials with binary outcomes.

A two-stage rule for a randomized clinical trial with a dichotomous outcome is presented. It is based on a rule discussed by Ellenberg and Eisenberger (Cancer Treat Rep 69:1147-1154, 1985). The rule is shown to lead to a reduction in expected sample size if the test treatment is ineffective, with almost no loss in power. All the mathematical formulas required to design such a rule are given, as well as some time-saving approximations. The accuracy of the approximations is shown.