Phase III human oncology trails should be stopped as soon a possible it there is clear evidence that one treatment is superior to the others or that no treatment will emerge as superior. We propose to study statistical methods for the early stopping of such trails and to implement these methods within a microcomputer package for interim clinical train monitoring. Besides a selection of group sequential methods for early stopping, the package will display ancillary data useful to statisticians, clinicians and ethicists concerned with moniriting an ongoing study. In addition, by using commercial software for a distributed data system, remote sites will swiftly be able to enter and display critical outcome and toxicity data relevant to an early stopping decision. Our Phase I SBIR will begin with a comparison, by computer simulation, of some proposals for group sequential methods. Criteria to be compared include power; sample size at stopping, and the allocation of type I error at each possible stopping point. Then, we will apply the proposed methods to real data from oncology trails to see if different methods might have led to different decisions. During the SBIR Phase I. we will design as SBIR Phase II computer system containing a variety of statistical features for interim monitoring of trails and implement some of them in a SBIR Phase I prototype system.
