Further results on a multiple-testing procedure for clinical trials.

O'Brien and Fleming (1979, Biometrics 35, 549-556) proposed a procedure for multiple testing in clinical trials for acute disease. We evaluated their procedure for chronic-disease models and found that it can be used without modification. Type I errors are unaltered and no loss of power is seen in comparison with single-sample procedures. However, average savings in sample size are less for the chronic-disease models than for the acute model. Finally, it is shown that for any specific disease model and fixed maximum number of inspections of the data, the percentage reduction in sample size can be readily predicted from the power.