Validation of clinical scoring systems ART and ABCR after transarterial chemoembolization of hepatocellular carcinoma.

e15593 Background: Transarterial Chemoembolization (TACE) is the standard of care for intermediate stage hepatocellular carcinoma (HCC). It remains challenging to decide whether to repeat treatment. We performed external validations of two scoring systems recently developed to evaluate TACE: the ART (assessment for retreatment with TACE) and the ABCR (alpha fetoprotein [AFP], Barcelona clinic liver cancer [BCLC] stage, Child-Pugh score, and treatment response). We also compared their prognostic value. Methods: From 2000 to 2015, 871 patients with HCC underwent TACE at our tertiary referral hospital. We acquired baseline data on BCLC-stages and AFP levels, and data measured before the second TACE on Child-Pugh scores, aspartate aminotransferase levels, and radiologic tumor responses. Overall survival was calculated. Both Scores were validated and compared with Harrell´s C-index, the integrated Brier score (IBS), and prediction error curves. Furthermore, multivariate analyses including all variables used in each score were performed in order to identify independent predictors of survival. Results: 176 patients were included. Low and high ART scores predicted median survivals of 20.8 and 15.3 months, respectively. Low, intermediate, and high ABCR scores predicted median survivals of 24.5, 15.8, and 5.8 months, respectively. Harrell´s C-indexes were 0.572 and 0.608 and IBS were 0.135 and 0.128 for ART and ABCR, respectively. For both scores, an increase in Child Pugh ≥ 2 points and a radiological response independently predicted survival. Independent predictors of survival in both scores were an increase in Child Pugh ≥ 2 points and radiological response. In the ABCR score additionally the BCLC stage proved significant (all p < 0.05). Conclusions: Both ART and ABCR scores provided some indication of which patients had a dismal prognosis, and were unlikely to benefit from repeated TACE. ABCR had slightly greater predictive value. However, neither score was sufficient to support clear-cut clinical decisions. Further effort is necessary to determine criteria for valid predictions.