Maturation of overall survival (OS) in TIVO-3 with long-term follow-up.

Abstract

4557 Background: Maturity of survival data is a consideration in the value assessment and confident clinical application of oncology drugs based on trial evidence. TIVO-3 supported FDA-approval of tivozanib (TIVO) in relapsed/refractory (R/R) advanced RCC, meeting the primary endpoint of significantly improved PFS over sorafenib (SOR) (HR: 0.73, 95% CI, 0.56-0.95). Long-term follow-up analyses demonstrate that the PFS rate at 3-years with TIVO is >5x higher than with SOR (12% vs 2%, respectively), yet a significant OS benefit for TIVO has not been observed to date. Here we report the contribution of event accumulation and data maturation on the stability of KM survival estimates. Methods: Intent-to-treat analyses of Cox proportional hazards and log-rank statistics were used to estimate the HR and 95% CI for OS in the TIVO-3 trial at prespecified (2-years after last-patient-in [LPI]; ≥251 events) and exploratory extended follow-up timepoints (≥270 events; database closure). Patients were followed for survival until death, consent withdrawal, or loss to follow-up. Results: 350 patients were randomized 1:1 to TIVO (n=175) or SOR (n=175). 2-years post LPI and a mean follow-up of 17.9 months (data cut-off August 2019), 65% of patients experienced an event (HR: 0.99, 95% CI, 0.76-1.29). Subsequent analyses are reported at 20.3 (May 2020), 21.9 (January 2021), and 22.8 (May 2021) months follow-up. Accumulation of events and HR over time is shown in the Table. After almost 23 months of follow-up and realization of 80% of events, OS HR has decreased to below 0.90, in favor of TIVO. Conclusions: Serial OS analyses using KM estimates are subject to increased curve reliability with decreased censoring and limited residual patients at risk for death. Long-term follow-up of TIVO-3 suggests early and consistent PFS benefit with TIVO over SOR is associated with an OS HR decline over-time with more events. Clinical trial information: NCT02627963. [Table: see text]