Presentation (CTI)

OBJECTIVES: The European Commission's PICO exercises identified Duration of Response (DoR) and Time to Response (TTR) as key outcomes in hepatocellular carcinoma (HCC). Traditional DoR and TTR analyses in the responder subgroup lack causal interpretation and compromise the intention-to-treat (ITT) principles. We assessed the applicability of randomization-preserving methods for DoR and TTR in HCC using IMbrave150 data.
METHODS: We applied randomization-preserving methods to updated data from the IMbrave150 study, independently of patients' response. For DoR, Restricted Mean DoR (RMDoR) was estimated over 23 months. For TTR, Cumulative Incidence Functions (CIF) was applied to estimate cumulative response rates at 3 months and time to reach a 10% response rate.
RESULTS: In the IMbrave150 trial, traditional median DoR among responders had been reported as 18.1 for atezolizumab plus bevacizumab (atezo+bev) vs 14.9 months for sorafenib (sora) (Kudo et al., 2023). The treatment difference in RMDoR among all patients was 3.02 months (95% CI: 1.88-4.14), indicating a benefit for atezo+bev of approximately 3 months over Sora. For TTR , the 3-month cumulative response rate for atezo+bev was 17% (13%-21%) vs. 5.6% (2.6%-10%) for Sora; time to 10% response was 2.6 months in the atezo+bev arm vs. 4.1 months in the sora arm.
CONCLUSIONS: The importance of DoR in HCC for EU HTA is underscored by its inclusion in the PICO exercise. Although traditional methods better describe DoR among responders, they lack causal interpretation. Randomization-preserving methods were able to show the benefit of atezo+bev in terms of longer DoR and faster TTR among all patients. Our analysis suggests that RMDoR may have utility in answering PICO questions while preserving the ITT principle. Prespecification of the used time intervals and thresholds is crucial.