Why effect sizes are systematically larger for progression-free survival than overall survival in cancer drug trials: Prognostic scores as a way forward

Abstract

Cancer drugs have accumulated the most approvals over the past years. Overall survival (OS) is considered the gold standard for cancer trial outcomes. However, its use has declined over the past years, in favor of surrogate endpoints, such as progression-free survival (PFS). PFS allows to assess outcomes earlier and, thus, accelerates approval of cancer drugs. Previous studies have demonstrated a poor correlation between PFS and OS. Using simulation models, we examined why PFS usually overestimates survival benefit. We created a publicly accessible web application that allows users to run the simulations with different parameter settings. Based on the findings, we propose that assessment of preliminary evidence should be based on a combination of OS result and prognostic scores that reflect the health status of surviving patients.