Evaluation of staging systems to predict prognosis in hepatocellular carcinoma patients treated with radioembolization

PURPOSE: To compare the prognostic accuracy of nine staging systems, some of which are well-known and some of which have only been more recently described, for patients with unresectable HCC treated with radioembolization (RE). MATERIALS AND METHODS: Individual scores or classes for the following staging systems were recorded or calculated for patients (n = 89) with unresectable HCC who underwent RE at a single tertiary care center from January 2008 to October 2016: Eastern Cooperative Oncology Group, Barcelona Clinic Liver Cancer, Hong Kong Liver Cancer, Okuda, Cancer of the Liver Italian Program (CLIP), Model for End Stage Liver Disease, Child-Pugh (CP) Categorical and Numeric, and Albumin-Bilirubin. For each staging system, a cox proportional hazards regression model was fit to the data and log-rank test statistics, concordance indices, Akaike Information Criteria (AIC) and other diagnostic statistics were calculated. RESULTS: Of the nine staging systems analyzed, the basic discriminatory ability assessed with the log-rank test (rejected at the α = .05-level) was significant for two of the systems: CP Numeric (p < .001) and CLIP (p < .05). Out of these two systems, CP Numeric system had a higher prognostic accuracy than CLIP with the lowest AIC (464.90), the highest optimism-corrected pseudo R(2) (0.16), and the highest estimated concordance index (0.64). CONCLUSION: As applied to our patient population, the CP Numeric system contained the most predictive prognostic information for patients with HCC undergoing radioembolization. However, all evaluated staging systems performed suboptimally, and the relative superiority of any of the systems remains unclear when ranking them according to common practice. Further evaluation of current ranking methodologies is recommended.