BACKGROUND Recently, albumin-bilirubin (ALBI) grade has been suggested as a better surrogate for hepatic functional reserve for patients with hepatocellular carcinoma (HCC).

AIMS We developed and validated a novel prediction model to predict outcome for HCC patients who underwent transcatheter arterial chemoembolization (TACE) as a first-line therapy.

METHODS From a multivariate Cox regression model for overall survival, five objective variables (ALBI grade), the Barcelona clinic liver cancer (BCLC) stage, response after the first TACE session, Alpha-fetoprotein level, and sex were chosen and the ABRAS score was developed from the derivation cohort (n = 476) and scored to generate an 8-point risk prediction model. The model's prognostic performance was assessed in the randomly assigned internal validation set (n = 475) and external validation set (n = 243).

RESULTS The ALBI grade was able to stratify patient survival within the same Child-Pugh class. The time-dependent area under receiver operating characteristics curves (AUROCs) for overall survival at 1 and 3 years were 0.78 and 0.73 in the training set, 0.78 and 0.71 in the internal validation set, and 0.70 and 0.65 in the external validation set, respectively. When stratified by BCLC stage, ABRAS score at a cutoff point of more than 3, 4, and 5 for BCLC stage 0/A, B, and C could identify subset of patients with dismal prognosis.

CONCLUSION ABRAS score was useful in estimating prognosis for patients who underwent TACE as a first-line therapy. This score can be useful in planning and guiding treatment strategies with TACE, which warrants prospective validation.