Development and external validation of DISPAIR fistula risk score for clinically relevant postoperative pancreatic fistula risk after distal pancreatectomy

This study describes the development and external validation of the DISPAIR score, a preoperative clinical prediction model to estimate the risk of postoperative pancreatic fistula after distal pancreatectomy. It is based on three variables measured before operation: pancreatic thickness, transection site, and history of diabetes. On external validation, it showed satisfactory discrimination (area under the curve 0.80) and calibration (slope 0.719, intercept 0.192) for predicting pancreatic fistula. Background Highly utilized risk scores for clinically relevant postoperative pancreatic fistula (CR-POPF) have guided clinical decision-making in pancreatoduodenectomy. However, none has been successfully developed for distal pancreatectomy. This study aimed to develop and validate a new fistula risk score for distal pancreatectomy. Methods Patients undergoing distal pancreatectomy at Helsinki University Hospital, Finland from 2013 to 2021, and at Karolinska University Hospital, Sweden, from 2010 to 2020, were included retrospectively. The outcome was CR-POPF, according to the 2016 International Study Group of Pancreatic Surgery definition. Preoperative clinical demographics and radiological parameters such as pancreatic thickness and duct diameter were measured. A logistic regression model was developed, internally validated with bootstrapping, and the performance assessed in an external validation cohort. Results Of 668 patients from Helsinki (266) and Stockholm (402), 173 (25.9 per cent) developed CR-POPF. The final model consisted of three variables assessed before surgery: transection site (neck versus body/tail), pancreatic thickness at transection site, and diabetes. The model had an area under the receiver operating characteristic curve (AUROC) of 0.904 (95 per cent c.i. 0.855 to 0.949) after internal validation, and 0.798 (0.748 to 0.848) after external validation. The calibration slope and intercept on external validation were 0.719 and 0.192 respectively. Four risk groups were defined in the validation cohort for clinical applicability: low (below 5 per cent), moderate (at least 5 but below 30 per cent), high (at least 30 but below 75 per cent), and extreme (75 per cent or more). The incidences in these groups were 8.7 per cent (11 of 126), 22.0 per cent (36 of 164), 63 per cent (57 of 91), and 81 per cent (17 of 21) respectively. Conclusion The DISPAIR score after distal pancreatectomy may guide decision-making and allow a risk-adjusted outcome comparison for CR-POPF.