Adjusting survival analysis for the presence of unadjudicated study events

In the Coumadin Aspirin Reinfarction Study (CARS), the primary (efficacy) endpoint was time to the first occurrence of reinfarction, ischemic stroke, or cardiovascular death. Study events likely to be a component of the primary endpoint were reviewed by a blinded Events Classification Committee (ECC). To accommodate the inherent delay imposed by the adjudication process, we used the results of prior adjudication to predict classifications for selected unadjudicated study events in order to conduct "up-to-date" interim analyses for the Data and Safety Monitoring Committee (DSMC). Rates at which previously adjudicated study events were confirmed as primary event components were used to adjust the Kaplan-Meier estimates of survival and the log-rank statistics. Multiple events for a given subject were weighted according to the probability that prior events would be confirmed. Once all study events were adjudicated, final analyses confirmed the results of the adjusted interim analysis. For monitoring clinical trials with adjudicated endpoint events, models based on prior adjudication histories can be reliably used to predict event classifications for unadjudicated events and support accurate, contemporary interim analyses.