Background & Aims Sorafenib-related adverse events have been reported as clinical surrogates for treatment response in hepatocellular carcinoma (HCC); however, no consensus has been reached regarding the definition of responders. We evaluated the predictive abilities of different definitions for sorafenib response based on treatment-emergent adverse events, aiming to identify the most discriminatory one as a clinical marker. Methods From January 2010 to December 2014, 435 consecutive HCC patients treated with sorafenib were enrolled. Considering the type, severity and timing of adverse events, twelve different categories of sorafenib response were defined. By comparing their discriminatory abilities for survival, an indicative criterion was defined, the prognostic value of which was evaluated by time-dependent multivariate analysis, validated in various subsets and confirmed by landmark analysis. Results Using concordance (C)-index analysis and time-dependent receiver operating characteristic curves, the development of a hand-foot-skin reaction ≥ grade 2 within 60 days of sorafenib initiation (2HFSR60) showed the highest discriminating value. Based on this criterion, 161 (37.0%) sorafenib responders achieved decreased risk of death by 47% (adjusted HR 0.53, 95%CI 0.43-0.67, P < 0.001) and likelihood of progression by 26% (adjusted HR 0.74, 95%CI 0.58-0.96, P = 0.020) compared with non-responders. Notably, 2HFSR60 remained an effective discriminator among most subgroups and had superior predictive ability to previous definitions, even according to the landmark analysis. Conclusions Our study demonstrated that 2HFSR60, with the best discriminatory ability compared to currently available definitions of sorafenib-related adverse events, could be the optimal clinical marker to identify sorafenib responders with decreased risk of death by half.
Keywords: Adverse events; Hepatocellular carcinoma; Predictive value; Sorafenib.