Background: Response Evaluation Criteria in Solid Tumors (RECIST) has been shown to be a poor surrogate for survival benefit with targeted therapy in advanced hepatocellular cancer (HCC). We investigated whether response evaluated using modified RECIST (mRECIST) predicted overall survival (OS) using comparable data from two Phase II trials conducted to evaluate the efficacy/safety of nintedanib versus sorafenib as first-line treatment of mainly Caucasian patients (study 1199.37; NCT01004003) or Asian patients (study 1199.39; NCT00987935) with advanced HCC. Methods: Data from patients (n=188) treated with either nintedanib or sorafenib were pooled for analyses. Cox regression and Kaplan–Meier survival analyses were used to explore differences in OS between the responders and non-responders according to RECIST 1.0 and mRECIST criteria. Two multivariate Cox proportional hazards selection models were conducted to identify baseline variables that predict survival and if RECIST response (analysis 1) and mRECIST response (analysis 2) predict survival at the 0.2 level of significance. Results: Response rates were 4.3% (n=8) by RECIST and 14.9% (n=28) by mRECIST. Discordance between RECIST/mRECIST evaluation was most common for assessment of partial response (n=22; 11.7%) and stable disease (n=24; 12.8%). OS was significantly longer in patients with response compared to patients without response; RECIST: hazard ratio (HR) 0.32 (95% CI 0.13–0.81), p=0.0113 and mRECIST: 0.53 (95% CI 0.33–0.85), p=0.0079. In both multivariate models, presence of macrovascular invasion (MVI) and presence of extrahepatic spread (EHS) at baseline were associated with worse OS. RECIST response and mRECIST response were both significant predictors of survival (see table). Conclusions: Objective mRECIST response is a surrogate for OS and as such should be considered a valid endpoint for use in HCC trials. Clinical trial information: NCT01004003 and NCT00987935