The conscientious judgement of a DSMB – statistical stopping rules re-examined

Abstract

In the light of the recent failure of a large cardiovascular mortality and morbidity study with torcetrapib, we have undertaken a post hoc review of the decisions taken by the Data and Safety Monitoring Board (DSMB) in that study. A number of other studies in which complex decisions were made by DSMBs are also reviewed. The examples illustrate the complexities involved in the decision-making process by DSMBs and indicate that too much reliance on formal statistical stopping rules should be avoided due to a risk of delaying the identification of an unacceptable emergent safety signal. The review was based on information submitted by the sponsors of the studies to the Medical Products Agency. Multiple approaches to assessing the efficacy and safety in clinical trials need to be considered in order to facilitate early stopping of clinical trials with a negative risk benefit balance. Such systems may, for example, include the use of p-value flags and/or a complementary statistical analysis of the likelihood of achieving the study objective based on the data obtained to date.