Additive value of oncological resectability criteria for predicting post-hepatectomy liver failure in hepatocellular...

The recently introduced oncological resectability criteria (borderline resectable [BR] criteria) for hepatocellular carcinoma (HCC) have been validated for long-term outcomes; however, their relevance to short-term surgical morbidity, particularly post-hepatectomy liver failure (PHLF), remains uncertain. We aimed to evaluate whether these criteria provide additive predictive value beyond conventional liver function indices for PHLF. A retrospective analysis was performed on 732 patients who underwent initial hepatectomy for HCC between January 2005 and December 2024 at our institution. Patients were categorized as resectable, BR1, or BR2 according to the predefined BR criteria. Logistic regression models were constructed to predict PHLF, defined as International Study Group of Liver Surgery grade B/C. A baseline model using preoperative liver function parameters was developed, and tumor-related metrics-including BR criteria, Milan criteria, up-to-seven criteria, and tumor burden score (TBS)-were subsequently added to assess their additive predictive value. Model performance was assessed using area under the receiver operating characteristic curve (AUC), net reclassification improvement (NRI), integrated discrimination improvement (IDI), decision curve analysis (DCA), and calibration plots. PHLF occurred in 37 patients (5.1%). The baseline model incorporating albumin-bilirubin (ALBI) score and indocyanine green retention at 15 min (ICG-R15) showed moderate discrimination (AUC 0.71). Addition of the BR criteria significantly improved discrimination (AUC 0.81, p = 0.003) and reclassification by NRI (0.73, p < 0.001) and IDI (0.036, p = 0.025). By comparison, Milan and up-to-seven criteria led to smaller improvements, and TBS showed no meaningful effect. In DCA, the ALBI-ICG-BR model showed higher net benefit across relevant thresholds. The BR criteria provide additive value beyond established liver function indices for predicting PHLF in HCC, supporting more refined preoperative risk stratification.