胰十二指肠切除术术后出血列线图预测模型的建立和验证

目的 分析胰十二指肠切除术（PD）术后出血的危险因素，构建预测PD术后出血的列线图模型。方法 回顾性分析2017年1月—2023年1月494例在南昌大学第二附属医院行PD患者的临床资料。将2017年1月—2020年12月收集的376例患者作为训练集，2021年1月—2023年1月118例患者作为验证集。通过单因素分析、Lasso回归分析和多因素一般Logistic回归分析筛选预测因素并构建列线图预测模型。通过受试者工作特征曲线下面积（AUC）、校正曲线和决策曲线分析（DCA）评估模型的鉴别能力、一致性和临床效果。结果 多因素一般Logistic回归分析结果显示，血管重建、术后胰瘘、术后胆瘘、腹腔感染和白蛋白为PD术后出血的独立风险因素（P <0.05）。由上述因素构建列线图预测模型在训练集的AUC为0.870（95% CI：0.820，0.920），验证集AUC为0.799（95% CI：0.691，0.907），提示模型诊断效能较好，在训练集和验证集中绘制出的校正曲线与标准曲线较为接近，提示模型一致性较好。绘制的DCA曲线也表明了明显的正向净收益。结论 通过血管重建、术后胰瘘、术后胆瘘和腹腔感染和白蛋白构建的列线图预测模型能够很好识别出PD术后出血的高风险患者，具有很好的临床应用价值。

Development and validation of a nomogram predictive model for post-pancreaticoduodenectomy bleeding

Objective To analyze the risk factors for postoperative bleeding after pancreaticoduodenectomy (PD) and construct a nomogram predictive model for post-PD bleeding.Methods Clinical data of 494 patients who underwent PD at the Second Affiliated Hospital of Nanchang University from January 2017 to January 2023 were retrospectively analyzed. Among them, 376 patients collected from January 2017 to December 2020 were used as the training set, and 118 patients from January 2021 to January 2023 were used as the validation set. Predictive factors were selected through univariate and multivariate analyses, LASSO regression analysis, and logistic regression analysis, followed by the construction of a nomogram predictive model. The discriminative ability, consistency, and clinical utility of the model were evaluated using the area under the receiver operating characteristic (ROC) curve (AUC), calibration curve, and decision curve analysis (DCA).Results Logistic regression analysis showed that vascular reconstruction, postoperative pancreatic fistula, postoperative bile fistula, intra-abdominal infection, and albumin were independent risk factors for post-PD bleeding (P < 0.05). The constructed nomogram predictive model based on these factors had an AUC of 0.870 (95% CI: 0.820, 0.920) in the training set and an AUC of 0.799 (95% CI: 0.691, 0.907) in the validation set, indicating good diagnostic efficacy. The calibration curves in both the training and validation sets closely approximated the standard curve, indicating good consistency of the model. The DCA curves showed a clear positive net benefit.Conclusion The nomogram predictive model constructed based on vascular reconstruction, postoperative pancreatic fistula, postoperative bile fistula, intra-abdominal infection, and albumin can effectively identify high-risk patients for post-PD bleeding and has good clinical application value.