Proposal for the use of progression-free survival in unblinded randomized trials

Abstract

Progression-free survival is an attractive end point for clinical trials when an overall survival end point may be confounded by additional treatments administered after progression. When a trial is performed in an unblinded manner, however, there is the potential for bias between the treatment arms because of the subjective aspects of the progression end point. We discuss the magnitude of this potential bias and suggest methods for lessening it. We propose the carrying forward of any progression information to two designated time points for the statistical analysis for trials that are not blinded. This proposal, possibly combined with central review of progression scans for these two time points, essentially eliminates any bias, with little risk of major efficiency loss compared with using the reported progression times.