prospective multicenter validation of preoperative plasma ceramides as novel predictive biomarkers for clinically relevant post-hepatectomy liver failure.

Background: Circulating ceramides (CERs) are associated with liver diseases and dysfunction. The utility of plasma CERs in predicting post-hepatectomy liver failure (PHLF) remains unclear. This study aimed to evaluate the clinical utility of preoperative plasma CERs for predicting clinically relevant PHLF (CR-PHLF).

Methods: This study included 736 patients who underwent hepatectomy across four independent hospitals in China. The training set included 392 patients from 2019 to 2021, and the prospective internal and external validation sets included 195 patients from 2022 to 2024 and 149 patients from 2023 to 2025, respectively. Preoperative plasma CERs were measured using targeted lipidomics. Grade B/C PHLF was classified by the criteria of the International Study Group of Liver Surgery and defined as CR-PHLF. Predictors for CR-PHLF were identified by least absolute shrinkage and selection operator (LASSO) logistic regression and receiver operating characteristic (ROC) analysis. The study is registered on ClinicalTrials.gov (NCT********) and the Research Registry (******).

Results: Plasma CER(d18:1/20:1) demonstrated a positive correlation with preoperative liver dysfunction and superior predictive power for CR-PHLF, with an AUROC of 0.837 (95% CI: 0.782-0.892; P < 0.001). Major hepatectomy, direct bilirubin, and CER(d18:1/20:1) were identified as independent PHLF predictors and integrated into an innovative CR-PHLF prediction model (Hpx-CER model). The model exhibited commendable discrimination in the training set (AUROC = 0.896; 95% CI: 0.851-0.941; P < 0.001), and the prospective internal (AUROC = 0.907; 95% CI: 0.845-0.970; P < 0.001) and external validation sets (AUROC = 0.862; 95% CI: 0.707-1.000; P < 0.001). Across all high-risk subgroups of CR-PHLF, including major hepatectomy, difficult hepatectomy, cirrhosis, malignant liver diseases, and preoperative liver dysfunction, our model consistently outperformed conventional models in predicting CR-PHLF.

Conclusion: Plasma CER(d18:1/20:1) is a novel predictor of CR-PHLF. The Hpx-CER model performs commendably in predicting CR-PHLF and provides reliable preoperative risk estimations.