Skin-to-portal junction distance as a simple CT-based predictor of pancreatic fistula after pancreaticoduodenectomy: a validated model.

Abstract

Clinically relevant postoperative pancreatic fistula (CR-POPF) remains a significant concern after pancreaticoduodenectomy (PD), increasing morbidity and mortality. This retrospective study evaluated the skin-to-portal junction distance (SJD), a simple CT-based metric, as a predictor of CR-POPF and developed a validated predictive model for risk stratification. The study included 305 patients undergoing open PD, divided into training (n = 214) and validation (n = 91) cohorts. Sex-specific SJD cutoffs were determined using receiver operating characteristic analysis. Logistic regression identified CR-POPF risk factors and built the model, assessed by area under the curve (AUC), calibration plots, and decision curve analysis (DCA). CR-POPF occurred in 23.9% of patients overall, with incidences of 23.4% and 25.3% in the training and validation cohorts, respectively. High SJD (≥ 78 mm for males, ≥ 64 mm for females; odds ratio [OR] 5.140), main pancreatic duct diameter ≤ 3 mm (OR 2.720), non-pancreatic ductal adenocarcinoma (OR 3.820), and soft pancreatic texture (OR 2.660) were independent predictors. The model achieved AUCs of 0.832 in the training cohort and 0.806 in the validation cohort, with good calibration and clinical utility as shown by DCA. These results compared favorably with the Fistula Risk Score (FRS) and alternative-FRS. The high SJD group had a higher incidence of CR-POPF (36.3% vs. 14.1%, p < 0.001), longer operative time (p = 0.004), greater blood loss (p = 0.047), and more major complications of Clavien-Dindo grade ≥ III (p < 0.001). SJD is a simple, objective predictor of CR-POPF. The validated model facilitates perioperative risk stratification; however, external validation is needed.