The WPG score for predicting clinically relevant postoperative pancreatic fistula after Wirsung-pancreatogastrostomy.

Clinically relevant postoperative pancreatic fistula (CR-POPF) remains a major cause of morbidity after pancreaticoduodenectomy (PD). Most available prediction models were developed for pancreatojejunostomy (PJ), limiting their accuracy for Wirsung pancreatogastrostomy (WPG). This study developed and internally assessed a WPG-specific risk score (WPG Score) for predicting CR-POPF after PD. This bicentric retrospective study included 345 patients undergoing PD with WPG between 2014 and 2024 at two Italian high-volume centers. Independent predictors of CR-POPF (ISGPS grade B/C) were identified using multivariable logistic regression and incorporated into the WPG Score. Predictive accuracy was assessed through ROC analysis and compared with the original Fistula Risk Score (FRS) and the alternative FRS (a-FRS). Male sex (OR 3.81), BMI ≥ 25 (OR 3.02), and soft pancreatic texture (OR 4.40) were independently associated with CR-POPF. Ampullary/biliary pathology was retained in the final predictive model based on the pre-specified stepwise retention criterion and its biological plausibility. The WPG Score (0-6 points) stratified patients into low- (3.7%), intermediate- (18.9%), and high-risk groups (51.7%) (p < 0.0001). Apparent discrimination was good (AUC 0.779), with an optimism-corrected AUC of 0.757 on bootstrap internal validation. The model showed good calibration (Hosmer-Lemeshow p = 0.876; Brier score 0.118). Discriminative performance was superior to both FRS (AUC 0.779 vs. 0.613) and a-FRS (AUC 0.779 vs. 0.649) (p = 0.002). The WPG Score provides the first anastomosis-specific tool for predicting clinically relevant postoperative pancreatic fistula after pancreatogastrostomy. Although external validation is still required, the score demonstrated promising discriminative performance in this bicentric cohort.