Surrogate and modified endpoints for immunotherapy in advanced hepatocellular carcinoma

Background and Aims: Immunotherapies have altered the treatment paradigm in HCC. Surrogate and modified endpoints are used to assess early success in clinical studies and guide clinical practice. We sought to determine whether surrogate endpoints and modifications to the conventional criteria for tumor response (RECIST), including modified RECIST (mRECIST) and immune-modified RECIST (imRECIST), are valid measures to predict overall survival (OS) in HCC treated with immunotherapies. Approach and Results: We performed an individual-level post hoc analysis of patients treated with atezolizumab and bevacizumab in the IMbrave150 trial (N = 279) and a cross-sectional analysis of a multicenter real-world patient cohort treated with immunotherapy (N = 328). Landmark analyses showed that objective response rates by RECIST were associated with greater OS including among Child-Pugh A and B patients and among patients treated with immunotherapies in the first-or second-line setting (IMbrave150: HR 0.24, 95% CI, 0.17-0.33; RW: HR 0.25, 95% CI, 0.15-0.43). Objective response rates by mRECIST or imRECIST were not associated with the greater predictive power of OS benefit (mRECIST: HR 0.30, 95% CI, 0.22-0.42; imRECIST: HR 0.36, 95% CI, 0.30-0.51). Progression-free survival determined by RECIST was only moderately correlated with OS, and this association was not improved using mRECIST or imRECIST. Conclusions: Our results clarify the utility of surrogate and modified endpoints in HCC treated with immunotherapies and support the use of RECIST objective response rates as an appropriate signal-finding measure for the evaluation of emerging treatments. Contrary to their intended purpose, mRECIST and imRECIST did not provide meaningful improvements in predicting OS benefits.