BACKGROUND AND AIMS: Response to sorafenib is highly variable in hepatocellular carcinoma (HCC). Baseline inflammatory parameters and treatment toxicities may improve survival prediction in patients on sorafenib therapy. RESULTS: 442 patients with advanced stage HCC on sorafenib were recruited (follow-up 5096 person-months at risk). 88% had BCLC stage B or greater HCC and 72.3% had Child-Pugh A cirrhosis. On Cox multivariate regression, previously-treated HCC (HR 0.579, 95% CI 0.385-0.872, p=0.009), Cancer of Liver Italian Program (CLIP) score (HR 1.723, 95% CI 1.462-2.047, p<0.0001), baseline red cell distribution width (RDW; HR 1.234, 95% CI 1.115-1.290, p<0.0001) and neutrophil to lymphocyte ratio (NLR; HR 1.218, 95% CI 1.108-1.322, p<0.0001) were significant independent risks for shorter survival, whilst sorafenib-related diarrhoea was associated with prolonged survival (HR 0.533, 95% CI 0.373-0.763, p=0.001). The combination of RD-CLIP score (CLIP score multiplied by RDW) ≥ 70 and no treatment-related diarrhoea had good utility for predicting 3-month survival (AUC of 0.808 (95% CI 0.734-0.882), positive predictive value of 86.4% and negative predictive value of 83.3%), compared with CLIP (AUC=0.642) or BCLC score alone (AUC=0.579). RD-CLIP score ≥ 35 and no treatment-related diarrhoea had an AUC of 0.787 for predicting 12-month survival. METHODS: Patients with HCC were consecutively recruited from three tertiary centres (Japan, Italy and UK) and clinical data were prospectively collected. The primary study endpoint was overall survival (OS) after commencing sorafenib. CONCLUSION: The novel prognostic index of CLIP score combined with inflammatory marker RDW and treatment-related diarrhoea has good accuracy for predicting overall, 3 month and 12 month survival in patients on sorafenib.