Organ-specific role of neutrophil extracellular traps in shaping the postoperative thrombo-inflammatory niche in pancreatic ductal adenocarcinoma: a multi-model machine learning and mechanistic study

Background Neutrophil extracellular traps (NETs) play a crucial role in thrombosis formation and tumor progression. However, the association between NETs and portal vein thrombosis (PVT) is not well understood. Methods We retrospectively and prospectively developed and validated a predictive model for postoperative PVT following radical surgery for pancreatic ductal adenocarcinoma (PDAC) from Jan. 2011 to Dec. 2022 and Jan. 2023 to Dec. 2023. 30 clinicopathological variables, including the novel biomarker NETs, were evaluated. 11 ML algorithms were compared based on several evaluation indexes including AUC. SHapley Additive exPlanations (SHAP) analysis was utilized for feature ranking and interpretability. Results In the retrospective study, a total of 306 patients were included, among which 70 (22.9%) developed PVT. In the prospective study, a total of 64 patients were enrolled and the incidence of PVT was 25%. eXtreme gradient boosting (XGBoost) demonstrated superior discriminative performance among 11 evaluated ML-models. After feature reduction based on ranked feature importance, a final interpretable XGBoost model comprising 7 variables was established. This model accurately predicted postoperative PVT in both training (AUC = 0.906) and validation (AUC = 0.823) cohorts, and has been translated into a convenient tool to facilitate its utility in clinical settings. Conclusion In patients with PDAC, higher NETs levels are correlated with an increased risk of developing PVT and are indicative of poor prognosis. Assessment of Cit-H3 and D-dimer levels may aid clinicians in predicting PVT.