Assessing Correlation between Surrogate Endpoints and Overall Survival for Oncology Clinical Trials.

Abstract

Surrogate endpoints, such as progression-free survival (PFS) and objective response rate (ORR), are increasingly used in oncology trials to expedite drug development and decision making. This paper evaluates the effectiveness of these surrogate endpoints by analyzing their correlations with overall survival (OS) across different cancer types, treatments, and therapy lines at both the patient and trial levels using an integrated dataset from Bristol Myers Squibb. At the patient level, correlation between OS and PFS was consistently stronger than those between OS and best overall response (BOR), suggesting that PFS may serve as a more reliable surrogate for OS. Melanoma patients exhibited the highest correlation between OS and BOR, and immune-oncology (IO) therapy patients showed stronger correlations than those treated with chemotherapy. First-line therapy patients demonstrated stronger correlations between BOR, PFS, and OS compared with second-line or third-line patients. At the trial level, the correlation between PFS hazard ratio (HR) and difference in ORR (∆ORR) was stronger than that between other endpoints. Melanoma studies exhibited strong correlations with significant P-values. IO therapy studies exhibited more consistent correlations.