Nomogram for predicting postoperative pancreatic fistula.

ABSTRACT

BACKGROUND: Previous studies analyzed risk factors for postoperative pancreatic fistula (POPF) and developed risk prediction tool using scoring system. However, no study has built a nomogram based on individual risk factors. This study aimed to evaluate individual risks of POPF and propose a nomogram for predicting POPF. METHODS: From 2007 to 2016, medical records of 1771 patients undergoing pancreaticoduodenctomy were reviewed retrospectively. Variables with p < 0.05 in multivariate logistic regression analysis were included in the nomogram. Internal performance validation was executed using a repeated cross validation method. RESULTS: Of 1771 patients, 222 (12.5%) experienced POPF. In multivariable analysis, sex (p = 0.004), body mass index (BMI) (p < 0.001), ASA score (p = 0.039), preoperative albumin (p = 0.035), pancreatic duct diameter (p = 0.002), and location of tumor (p < 0.001) were identified as independent predictors for POPF. Based on these six variables, a POPF nomogram was developed. The area under the curve (AUC) estimated from the receiver operating characteristic (ROC) graph was 0.709 in the train set and 0.652 in the test set. CONCLUSIONS: A POPF nomogram was developed. This nomogram may be useful for selecting patients who need more intensified therapy and establishing customized treatment strategy.