Abstract
PURPOSE: Because of the increased number of sequential treatments used for advanced hepatocellular carcinoma (HCC), there is a need for surrogate endpoints of overall survival (OS). We analyze whether objective response (OR) is an independent predictor and surrogate endpoint of OS. PATIENTS AND METHODS: A systematic review of randomized clinical trials (RCT) in advanced HCC published between 2010 and 2020 was conducted to explore OS surrogacy of OR by Response Evaluation Criteria in Solid Tumors (RECIST) and modified RECIST (mRECIST). In parallel, RCTs exploring the impact of OR on OS in a time-dependent multivariate analysis were integrated in a meta-analysis. RESULTS: Of 65 RCTs identified in advanced HCC, we analyzed 34 studies including 14,056 patients that reported OS and OR by either RECIST (n = 23), mRECIST (n = 5), or both (n = 6). When exploring surrogacy, the trial-level correlation between OR odds ratio and OS HR was R = 0.677 by mRECIST and R = 0.532 by RECIST. Meta-analysis of five RCTs assessing predictors of survival in multivariate analysis found that patients with OR by mRECIST presented a pooled HR for OS of 0.44 (95% confidence interval, 0.27-0.70; P < 0.001) compared with nonresponders. Responses to atezolizumab-bevacizumab had a greater impact on OS than tyrosine kinase inhibitor responses. CONCLUSIONS: OR-mRECIST is an independent predictor of OS in patients with advanced HCC. Although correlation of OR-mRECIST and OS is better than with OR-RECIST, the level of surrogacy is modest. Thus, it can be used as endpoint in proof-of-concept phase II trials, but the data do not support its use as a primary endpoint of phase III investigations assessing systemic therapies.