Evaluation of discordance measures in oncology studies with blinded independent central review of progression-free survival using an observational error model.

To confirm results obtained from local evaluation at investigational centers, many oncology studies utilize blinded independent central review (BICR) to make assessments of the primary endpoint, progression-free survival (PFS). The comparison of data often leads to large discordances between these observations, casting doubt on the reliability of the estimated treatment effects from these trials. Here we propose new statistics to measure discordance and evaluate their utility to detect bias in the local evaluation of progression relative to the standard measures of discordance. A new observational error model is proposed that can be used to describe the discordance in patient assessments between multiple readers.