Construction of a predictive model for postoperative pancreatic fistula following pancreaticoduodenectomy and an exploration of inflammatory biomarker associations.

Background: Clinically relevant postoperative pancreatic fistula (CR-POPF) remains a significant cause of morbidity after pancreaticoduodenectomy (PD). Existing predictive tools, such as the Fistula risk score (FRS), demonstrate inconsistent performance. This study aimed to develop a nomogram with superior predictive accuracy for CR-POPF. Methods: This retrospective cohort study analysed 1,026 PD patients (70% training, 30% validation). Least absolute shrinkage and selection operator (LASSO) and multivariate logistic regression were used to identify independent predictors and construct a dynamic nomogram. Model performance was assessed via the area under the curve (AUC), calibration, and decision curve analysis (DCA). Results: CR-POPF occurred in 18.4% of the cohort. Multivariable analysis identified five independent predictors: soft pancreatic texture, small pancreatic duct diameter, elevated body mass index (BMI), prolonged operation time, and absence of diabetes mellitus. Inflammatory markers were not retained in the final model. The nomogram showed robust discrimination with AUCs of 0.770 (training) and 0.759 (validation), significantly outperforming both the alternative FRS (a-FRS) (AUC =0.579) and the traditional FRS (AUC =0.529). Conclusions: We successfully developed and internally validated a superior nomogram integrating anatomical, metabolic, and procedural factors. This highly accurate tool facilitates precise preoperative and intraoperative risk stratification, guiding tailored preventive strategies to mitigate the burden of POPF.