甘油三酯-葡萄糖指数联合血清学指标对胰十二指肠切除术后胰瘘的预测作用

Predictive value of the triglyceride-glucose index combined with serological indicators for pancreatic fistula after pancreaticoduodenectomy

摘要:

背景与目的 术后胰瘘（POPF）是胰十二指肠切除术（PD）术后最严重及最常见的并发症之一，也是PD术后死亡的主要原因。引起PD-POPF的危险因素较多，因此建立有效的预测模型具有重要的临床价值。本研究探讨甘油三酯-葡萄糖（TyG）指数联合血清学指标对PD后POPF的预测效能。方法 回顾性收集2019年1月—2024年6月兰州大学第二医院普通外科291例接受PD患者的术前一般资料、术前1周内的实验室指标及术后并发症资料。通过计算机随机数法按照7∶3比例随机分为建模组（203例）和验证组（88例），对建模组数据进行单因素Logistic回归及二元Logistic回归（Back-Wald法）进行多因素分析，并基于回归分析结果构建模型并利用列线图进行模型可视化。绘制受试者工作特征（ROC）曲线下面积（AUC）评价该列线图模型的区分度，绘制校准曲线评价模型预测概率、实际概率之间的曲线，决策曲线分析预测模型在实际临床决策中的应用价值。对可能存在的影响结局变量的影响因素进行亚组分析。结果 291例患者中共有70例患者发生POPF，其中建模组发生POPF 49例，验证组21例，建模组与验证组两组数据集之间的差异无统计学意义（均P>0.05），建模组单因素分析结果显示，体质量指数（BMI）、甘油三酯、TyG指数、白蛋白（ALB）、血小板计数（PLT）、淋巴细胞绝对值（LYM）、中性粒细胞绝对值（NEUT）与POPF明显有关（均P<0.05），多因素分析结果显示，BMI、TyG指数、ALB、PLT、LYM、NEUT是POPF的独立影响因素（均P<0.05）。根据多因素分析结果构建PD-POPF风险预测模型及列线图，模型AUC为0.80（0.73~0.86），将验证组数据带入预测模型并绘制ROC曲线，结果显示，验证组AUC为0.80（0.70~0.90），建模组、验证组模型校准曲线接近标准曲线。亚组分析显示，肿瘤性质与肿瘤分期对PD-POPF的影响因素的影响不大，本预测模型的稳定性较好。结论 TyG指数以及BMI、PLT、NEUT、ALB、LYM与PD-POPF的发生密切相关，基于TyG指数及上述影响因素建立的风险预测模型具有良好的预测效能，对指导临床进行早期干预具有重要意义。

Abstract:

Background and Aims Postoperative pancreatic fistula (POPF) is one of the most severe and common complications following pancreaticoduodenectomy (PD) and is a major cause of mortality after PD. Given the multiple risk factors associated with PD-POPF, developing an effective predictive model is of significant clinical importance. This study was conducted to explore the predictive performance of the triglyceride-glucose (TyG) index combined with serological indicators for POPF following PD.Methods The preoperative general data, laboratory indicators within one week before surgery, and postoperative complication data of 291 patients who underwent PD at the Department of General Surgery, Second Hospital of Lanzhou University, from January 2019 to June 2024, were retrospectively collected. Patients were randomly divided into a modeling group (203 cases) and a validation group (88 cases) using a computer-generated random number method at a 7∶3 ratio. Univariate Logistic regression and multivariate binary Logistic regression (Back-Wald method) were performed on the modeling group data. Based on regression analysis results, a predictive model was constructed and visualized using a nomogram. The discriminative ability of the nomogram model was evaluated by the area under the receiver operating characteristic (ROC) curve (AUC). A calibration curve was used to assess the agreement between predicted and actual probabilities, and a decision curve analysis was conducted to evaluate the clinical application value of the model. Subgroup analysis was performed on potential factors influencing the outcome variables.Results Among the 291 patients, 70 developed POPF, with 49 cases in the modeling group and 21 in the validation group. There was no statistically significant difference between the two groups (all P>0.05). Univariate analysis in the modeling group identified body mass index (BMI), triglycerides, TyG index, albumin (ALB), platelet count (PLT), absolute lymphocyte count (LYM), and absolute neutrophil count (NEUT) as significant factors associated with POPF (all P<0.05). Multivariate analysis revealed that BMI, TyG index, ALB, PLT, LYM, and NEUT were independent influencing factors for POPF (all P<0.05). A PD-POPF risk prediction model and nomogram were constructed based on these results. The model achieved an AUC of 0.80 (0.73-0.86), and when applied to the validation group, the ROC analysis yielded an AUC of 0.80 (0.70-0.90). The calibration curves of both the modeling and validation groups closely aligned with the standard curve. Subgroup analysis indicated that tumor nature and tumor stage had minimal impact on PD-POPF risk factors, demonstrating good model stability.Conclusion The TyG index, along with BMI, PLT, NEUT, ALB, and LYM, is closely associated with PD-POPF occurrence. The risk prediction model based on the TyG index and these influencing factors exhibits good predictive performance and holds significant clinical value for guiding early intervention.