Predicting postoperative pancreatic fistula after robotic pancreatoduodenectomy using International Study Group on Pancreatic Surgery and fistula risk scores: European multicentre retrospective cohort study

Background: Postoperative pancreatic fistula represents the leading cause of morbidity and mortality after robotic pancreatoduodenectomy. Various scores have been proposed to stratify patients based on their postoperative pancreatic fistula risk, including three fistula risk scores, and two International Study Group for Pancreatic Surgery scores. This study compares the performance of these scores in patients undergoing robotic pancreatoduodenectomy. Methods: This is a multicentre European retrospective study in consecutive patients receiving robotic pancreatoduodenectomy for all indications (April 2014 to December 2021). The performance of the International Study Group for Pancreatic Surgery 4-tier (A–D) risk score, and its 3-tier (A–C) modification (International Study Group for Pancreatic Surgery 3-tier), fistula risk scores, alternative-fistula risk scores and the updated alternative-fistula risk scores in postoperative pancreatic fistula grade B/C prediction were compared based on their discrimination (area under the curve), calibration and clinical utility, evaluated through decision curve analyses. Results: Overall, 919 patients undergoing robotic pancreatoduodenectomy were included. The rate of grade B/C postoperative pancreatic fistula was 22.2% (n = 204). The area under the curve for the five scores differed only slightly: International Study Group for Pancreatic Surgery 0.63 (95% confidence interval (c.i.) 0.58 to 0.67), International Study Group for Pancreatic Surgery 3-tier 0.63 (95% c.i. 0.58 to 0.67), fistula risk scores 0.65 (95% c.i. 0.61 to 0.69), alternative-fistula risk scores 0.64 (95% c.i. 0.60 to 0.68) and updated alternative-fistula risk scores 0.65 (95% c.i. 0.60 to 0.69). The International Study Group for Pancreatic Surgery, International Study Group for Pancreatic Surgery 3-tier, fistula risk scores and alternative-fistula risk scores underestimated the risk of postoperative pancreatic fistula. In contrast, the updated alternative-fistula risk score was well-calibrated at low predicted risks, but overestimated postoperative pancreatic fistula risk for high-risk patients. In decision curve analyses, the updated alternative-fistula risk score showed a higher clinical utility compared with the four other risk scores. Conclusion: The clinical utility of the updated alternative-fistula risk score for robotic pancreatoduodenectomy slightly outperformed the four other fistula risk scores, and might be used for patient counselling and patient stratification in clinical practice and research.