The sequential version of Mady�s decision theoretic model for comparing two medical treatments is modified by considering the patients who are in the experiment and also the patients who do not participate in the experiment, while the experiment is in progress. The consideration of the patients who do not participate in the experiment leads to major changes in the results. Both the minimax and Bayesian solutions are obtained and are compared with Mady�s equivalent results.

Bayesian, decision theory, minimax, sequential analysis, three phases clinical trial.