ABSTRACT
Recently, the albumin-bilirubin (ALBI) score, a continuous index consisting of only albumin and bilirubin, has been developed for objectively assessing liver function in patients with hepatocellular carcinoma (HCC). However, the ALBI score was arbitrarily categorized into three ALBI grades based on two artificially predetermined cutoff points with no explanation and statistical grounds, causing a considerable loss of discriminatory ability. This study aims to propose a modified ALBI (mALBI) grade for offering a detailed evaluation of hepatic reserve and specify its role during clinical practice in the HCC setting. The study population comprised 3540 HCC patients treated with mainstream therapies including hepatectomy (n=2056), thermal ablation (n=550), and transcatheter arterial chemoembolization (n=934) from 2002 to 2017. The ALBI score was stratified into four mALBI grades through a recently proposed statistical method aiming to select the optimal cutoff points of a continuous predictive variable by maximizing the discriminative ability in a multivariable Cox regression model. The mALBI grade had an overall better discriminatory ability than the ALBI grade in predicting overall survival through Harrell's C-index (0.614 vs. 0.598, P<0.001). Both visual inspections of Kaplan-Meier curves and calculation of hazard ratios displayed a more subtle evaluation of liver function using the mALBI grade. Moreover, the newly identified cut-point (ALBI score = -2.29) between the mALBI grade 2a and 2b was much closer to a 30% retention rate of indocyanine green at 15 minutes, an indicator for the performance of a subsegmentectomy. The newly proposed mALBI grade provides a more subtle assessment of liver function to guide clinical decision-making and predicts the prognosis of HCC patients more accurately than the original ALBI grade.