Development and internal validation of a nomogram for predicting grade B/C hemorrhage after pancreaticoduodenectomy.

Abstract

Objective: Postpancreatectomy Hemorrhage (PPH) is a severe complication after Pancreaticoduodenectomy (PD). This study aims to investigate the predictors and develop a nomogram to predict grade B/C PPH.

Methods: Data were collected from patients who underwent PD at the Department of General Surgery of the First Affiliated Hospital of Nanchang University from January 2016 to December 2023. These patients were then randomly divided into a training set and a validation set. Logistic regression analysis was performed to investigate the predictors and develop a nomogram. The model's discriminative ability was evaluated using the Area Under Curve (AUC) of the receiver operating characteristic, calibration was evaluated using calibration curves, clinical utility was evaluated using decision curve analysis, and generalizability was evaluated by examining model's performance in the validation set. SHapley Additive exPlanations (SHAP) were used to interpret feature contributions and visualize the model's decision logic.

Results: Preoperative Prothrombin Time (PT), preoperative Total Bilirubin (TB) (≥171 μmol/L), Postoperative Pancreatic Fistula (POPF), and Postoperative Biliary Fistula (POBF) were independent predictors of grade B/C PPH. The AUC values were 0.848 and 0.721 in the training set and the validation set, respectively. The model demonstrated satisfactory discriminative ability, calibration, clinical utility, and generalizability. SHAP indicated that POPF was the most influential predictor.

Conclusions: Preoperative PT, preoperative TB (≥ 171 μmol/L), POPF, and POBF were identified as independent predictors of grade B/C PPH, and SHAP indicated the importance of POPF. The nomogram based on these predictors can help clinicians predict grade B/C PPH.