DATA MONITORING IN COMPLEX CLINICAL-TRIALS - WHICH TREATMENT IS BETTER

This paper addresses the problem of defining which of two medical treatments is 'better'. We point out that because many treatments affects several outcomes and because delay of death and reduction of hazard are not necessarily consonant with each other, the measure used to compare treatments may require a complex definition of superiority. The problem of defining 'better' is often a topic of intense discussion when a Data Monitoring Board grapples with the decision to recommend early termination of a randomized clinical trial. We discuss the statistical monitoring of such trials in the light of our our experience with two commonly used data monitoring procedures, stochastic curtailing and spending functions.