The PD-ROBOSCORE: A difficulty score for robotic pancreatoduodenectomy

Background: Difficulty scoring systems are important for the safe, stepwise implementation of new procedures. We designed a retrospective observational study for building a difficulty score for robotic pancreatoduodenectomy. Methods: The difficulty score (PD-ROBOSCORE) aims at predicting severe postoperative complications after robotic pancreatoduodenectomy. The PD-ROBOSCORE was developed in a training cohort of 198 robotic pancreatoduodenectomies and was validated in an international multicenter cohort of 686 robotic pancreatoduodenectomies. Finally, all centers tested the model during the early learning curve (n = 300). Growing difficulty levels (low, intermediate, high) were defined using cut-off values set at the 33rd and 66th percentile (NCT04662346). Results: Factors included in the final multivariate model were a body mass index of ≥25 kg/m2 for males and ≥30 kg/m2 for females (odds ratio:2.39; P < .0001), borderline resectable tumor (odd ratio:1.98; P < .0001), uncinate process tumor (odds ratio:1.69; P < .0001), pancreatic duct size <4 mm (odds ratio:1.59; P < .0001), American Society of Anesthesiologists class ≥3 (odds ratio:1.59; P < .0001), and hepatic artery originating from the superior mesenteric artery (odds ratio:1.43; P < .0001). In the training cohort, the absolute score value (odds ratio = 1.13; P = .0089) and difficulty groups (odds ratio = 2.35; P = .041) predicted severe postoperative complications. In the multicenter validation cohort, the absolute score value predicted severe postoperative complications (odds ratio = 1.16, P < .001), whereas the difficulty groups did not (odds ratio = 1.94, P = .082). In the learning curve cohort, both absolute score value (odds ratio:1.078, P = .04) and difficulty groups (odds ratio: 2.25, P = .017) predicted severe postoperative complications. Across all cohorts, a PD-ROBOSCORE of ≥12.51 doubled the risk of severe postoperative complications. The PD-ROBOSCORE score also predicted operative time, estimated blood loss, and vein resection. The PD-ROBOSCORE predicted postoperative pancreatic fistula, delayed gastric emptying, postpancreatectomy hemorrhage, and postoperative mortality in the learning curve cohort. Conclusion: The PD-ROBOSCORE predicts severe postoperative complications after robotic pancreatoduodenectomy. The score is readily available via www.pancreascalculator.com.