Real-world comparison of cabozantinib versus regorafenib as second-line therapy for advanced hepatocellular carcinoma after first-line immune checkpoint inhibitors: A propensity score–matched analysis.

132

Background: Immune checkpoint inhibitor (ICI)–based regimens are the standard first-line therapy for advanced hepatocellular carcinoma (HCC). Upon progression, cabozantinib and regorafenib are commonly used, yet no head-to-head or real-world comparative data exist specifically in the post-ICI setting. Prior indirect comparisons were limited to post-sorafenib populations. We compared the real-world effectiveness and safety of cabozantinib versus regorafenib after first-line ICI. Methods: Using the TriNetX Global Collaborative Network (171 healthcare organizations), we identified adults with HCC (C22.0) who received first-line ICI (atezolizumab, durvalumab, or nivolumab) followed by cabozantinib (n=484) or regorafenib (n=360). Propensity score matching (1:1) balanced cohorts on age, sex, race/ethnicity, Charlson comorbidity components, liver-specific covariates (cirrhosis, portal hypertension, ascites, chronic hepatitis), and laboratories (albumin, bilirubin, INR, platelets, creatinine). The primary outcome was overall survival (OS). Safety endpoints included diarrhea, fatigue, thrombocytopenia, hepatotoxicity, GI bleeding, ascites worsening, hyponatremia, ED visits, and hospitalizations. Results: After matching, 216 patients remained per cohort with well-balanced characteristics (all standardized differences <0.11). Results are shown in Table 1. Conclusions: In this propensity score–matched real-world analysis, cabozantinib was associated with significantly longer OS than regorafenib in patients with advanced HCC progressing on first-line ICI. However, cabozantinib had higher rates of diarrhea, fatigue, and thrombocytopenia, while regorafenib had more ED visits. These findings provide the first real-world head-to-head evidence for these TKIs in the post-ICI setting and may inform treatment sequencing decisions.