International validation of a natural-killer-cell–based model to predict recurrence-free survival in hepatocellular carcinoma

Background: Models estimating recurrence-free survival (RFS) after hepatectomy for hepatocellular carcinoma (HCC) rely on clinical variables and often overlook tumor immunobiology. The Liver Immune Status Index (LISI), derived from BMI, albumin, and Fibrosis-4 (FIB-4), reflects liver-resident natural killer (NK) cell activity. We validated the HISCO-HCC score, combining LISI, tumor burden score (TBS), and alpha-fetoprotein (AFP), using an international cohort. Methods: Patients undergoing curative-intent hepatectomy for HCC (2000–2023) were identified from an international database (median follow-up: 38.9 [14.9–67.5] months). RFS was the primary endpoint. LISI's predictive performance was compared with other liver-related indices. The original HISCO-HCC (oHISCO-HCC) was recalibrated via multivariable Cox regression in a training cohort (80 %) stratified by region, yielding a modified score (mHISCO-HCC). Validation was conducted in the testing cohort (20 %). Results: Among 1213 patients, LISI had the highest AUCs among liver-related indices for 1-/2-year RFS (0.60/0.60) and 1-/5-year OS (0.64/0.60). The formula: mHISCO-HCC = 0.49 × TBS + 0.41 × log(AFP) + 0.13 × LISI. In testing, mHISCO-HCC outperformed oHISCO-HCC and mHALT-HCC for 12-/36-/60-month RFS (AUCs: 0.73/0.71/0.66) with the lowest AIC. It also had the highest OS AUCs and stratified RFS and OS (p < 0.001). Conclusions: The mHISCO-HCC score, integrating tumor morphology, biology, and NK cell-based immunity, improves prediction of recurrence and survival. It may aid postoperative stratification.