Development and validation of the albumin-bilirubin gamma-glutamyl transferase score for enhanced prognostic accuracy after hepatocellular carcinoma resection

Background: The albumin-bilirubin (ALBI) score, used for predicting outcomes after hepatocellular carcinoma (HCC) resection, does not directly capture liver cell damage or biliary obstruction. Gamma-glutamyl transferase (GGT), which reflects hepatic oxidative stress and inflammation, may complement the ALBI score. We sought to develop the ALBI-GGT score, a composite prognostic tool, and evaluate its performance to predict long-term outcomes among patients undergoing HCC resection. Methods: Patients undergoing curative-intent HCC resection (2000–2023) were identified from an international, multi-institutional database. The cohort was divided into training (65%) and testing cohorts (35%). Multivariable Cox analysis examined the association of ALBI-GGT score with overall survival (OS). Results: Among 759 patients, the median ALBI score was −2.78 (−3.02 to −2.48), and the median GGT was 55.0 U/L (31.0–93.0). On multivariable analysis, ALBI score (hazard ratio [HR], 1.473 [1.112–1.950]; P =.007) and GGT (HR, 1.007 [1.004–1.010]; P <.001) were predictors of overall mortality, alongside tumor burden score (HR, 1.051 [1.015–1.090]; P =.006) and American Society of Anesthesiologists class >2 (HR, 1.473 [1.005–2.161]; P =.047). There was a near-linear correlation between increasing ALBI scores and GGT and higher hazards of death. The ALBI-GGT score demonstrated the highest predictive accuracy in the testing set (concordance index, 0.68 [0.58–0.72]), outperforming the ALBI score (0.62 [0.56–0.69]) and GGT (0.65 [0.58–0.72]). The ALBI-GGT achieved the lowest Akaike and Bayesian information criteria. Time-dependent area under the curve (AUC) analysis demonstrated consistent superiority over 0 to 60 months. At 1-, 3-, and 5-years, the ALBI-GGT score had AUCs of 0.782, 0.725, and 0.688, respectively, outperforming ALBI score and GGT. The ALBI-GGT score was able to stratify patients into distinct prognostic groups (5-year OS, low ALBI-GGT [85.0%] vs intermediate ALBI-GGT [65.8%] vs high ALBI-GGT [56.8%]; P <.001). Conclusion: ALBI score alone may be insufficient to prognostically stratify patients with HCC. Combining ALBI score with GGT was a superior tool to stratify patients relative to long-term survival.