Asymmetric group sequential boundaries for monitoring clinical trials

A group sequential method is proposed for the periodic analysis of accumulating data in clinical trials which are essentially one-sided as in treatment-placebo studies. The proposed boundaries retain the early stopping properties of the procedure described previously by DeMets & Ware (1980) when treatment appears to be ineffective or harmful but incorporate the more stringent requirements of the model proposed by O'Brien & Fleming (1979) for early rejection of the null hypothesis in favour of treatment.