Statistical controversies in clinical research: end points other than overall survival are vital for regulatory approval of anticancer agents

Abstract

There is an ongoing debate about the relative merits of overall survival (OS) and other metrics that can be used as primary end points in cancer clinical trials. Although survival time is arguably the most objective metric for assessing the efficacy of anticancer treatment, OS as a clinical-trial end point needs to be conceptually distinguished from increased survival time as a goal desired by patients, clinicians and public-health policy makers. OS presents several drawbacks as a primary end point that threatens to hamper further drug development, including the increase in the number of patients and the much longer follow-up required in a clinical trial. In many settings of first-line therapy for metastatic disease, median OS is currently two to four times longer than median progression-free survival. As a result, the analysis of OS may be increasingly confounded by the effect of salvage therapies used after disease progression. In this review, we use straightforward statistical reasoning and examples from the oncology literature to argue that OS should no longer be the primary end point of most future phase III cancer clinical trials that aim at assessing the efficacy of novel therapies in the setting of metastatic disease.

Keywords: disease-free survival; end point determination; neoplasms; survival; survival analysis.