An early predictive model for the progression of postoperative pancreatic fistula to grade C following pancreaticoduodenectomy based on early postoperative imaging features.

BACKGROUND: Grade C postoperative pancreatic fistula (POPF) represents the most severe complication following pancreaticoduodenectomy (PD). Although its incidence is relatively low, ranging from 1.8 % to 2.6 %, the mortality rate is alarmingly high, reaching up to 35 %. At present, there is a paucity of models for the early prediction of grade C POPF. This study aims to develop an early prediction model based on imaging and laboratory data collected within the first postoperative week in patients diagnosed with POPF. METHODS: 2732 patients who underwent PD, extended PD, or laparoscopic PD and developed POPF were collected from September 2011 to May 2022 retrospectively. Additionally, 659 cases were obtained from a prospective study on the occurrence of delayed abdominal hemorrhage in patients with POPF after PD, which took place from June 2022 to July 2023. The imaging and laboratory data within the first postoperative week were collected and analyzed. Seventy percent of the retrospectively gathered cases were designated as the model establishment dataset. The remaining 30 % of the retrospectively collected cases were combined with the retrospectively gathered prospective data to form a validation dataset. A nomogram was established based on the independent risk factors. RESULTS: Multivariable analysis identified the positive bubble sign, complete dehiscence of the pancreaticojejunostomy anastomosis, positive bacterial culture of the drainage fluid, suspected sepsis and bile leakage as independent risk factors of grade C POPF. A nomogram was constructed, with an area under the curve (AUC) of 0.890. A calibration curve was plotted, with an average absolute error of 0.04. In the validation set, the average absolute error of the calibration curve was 0.058. CONCLUSIONS: Positive bubble sign, complete dehiscence of the pancreaticojejunostomy anastomosis, positive bacterial culture of the drainage fluid, suspected sepsis, and bile leakage are identified as independent risk factors for grade C POPF. The model developed on this basis demonstrates reasonable reliability and stability. This nomogram could be employed to assist surgeons identifying high-risk patients with grade C POPF at an early stage.