目的 建立可切除食管癌患者根治性切除术后发生吻合口瘘风险的列线图预测模型,并进行验证。 方法 采用回顾 性病例对照研究的方法,纳入 2015 年 8 月至 2020 年 8 月入青海大学附属医院确诊 cT1~ 3 / N0~ 1 期食管癌并接受根治性切除术 患者 588 例为建模组,根据食管癌切除术并发症共识小组(ECCG)提出的吻合口瘘定义诊断吻合口瘘 52 例和无吻合口瘘 536 例。 分别采用 Lasso 回归和多因素 Logistic 回归分析筛选危险因素,并由 R 软件构建列线图预测模型。 结果 建模组中吻合口 瘘患者吸烟指数≥400、美国麻醉师协会(ASA) 分级Ⅲ级比无吻合口瘘患者增加,肿瘤位置上部和中部增多,手术时间延 长,微创术式增多,线型闭合器和手工吻合增多,颈部吻合增多,空肠造口术和鼻肠道吻合增多,预后营养指数(PNI)下降,差 异有统计学意义(P<0. 05)。 Lasso 回归筛选出 5 个最具差异性的指标,经多因素 Logistic 回归分析显示,吸烟指数≥400(OR = 1. 659,95%CI = 1. 210~ 1. 969,P = 0. 006)、ASA 分级Ⅲ级(OR = 1. 968,95% CI = 1. 524 ~ 2. 326,P = 0. 002)、手工吻合(OR = 3. 002,95%CI = 2. 526 ~ 3. 425,P< 0. 001)、颈部吻合(OR = 2. 201,95% CI = 1. 625 ~ 2. 458,P = 0. 001)、PNI < 50(OR = 3. 652, 95%CI = 3. 120~ 3. 968,P<0. 001)是吻合口瘘发生的独立预测因素。 受试者操作特征(ROC) 曲线显示,模型的曲线下面积 (AUC)为 0. 899,明显高于单一危险因素的 AUC 值(均 P<0. 001)。 校正曲线显示预测结果与实际结果之间具有良好的一致 性(P= 0. 663)。 结论 可切除食管癌患者根治性切除术后吻合口瘘的发生与多个临床因素密切相关,构建定量列线图模型有 较好的预测效能,便于指导临床早期识别高危患者并进行针对性干预,有较好的临床应用价值。

Objective To establish a nomogram predictive model for anastomotic leakage in patients with resectable esophageal cancer after radical resection and verify it. Method A method of retrospective case - control study was conducted a total of 588 patients with cT1-3 / N0-1 esophageal cancer diagnosis undergoing radical resection into Affiliated Hospital of Qinghai University from August 2015 to August 2020 were included as the model group. In which 52 patients of anastomotic leakage and other 536 patients of non - anastomotic leakage were diagnosed according to the definition of anastomotic leakage proposed by the esophagectomy complications consensus group ECCG . Lasso regression and multivariate Logistic regression analysis were used to screen risk factors and R software was used to construct a nomogram model. Result In the model group the smoking index≥400 and ASA classification III of patients with anastomotic leakage were higher than those without anastomotic leakage tumor location of the upper and middle part was more surgery time was longer the minimally invasive method was more the linear stapled and Hand-sewn were more the neck anastomosis was more the jejunostomy and nasal-intestinal canal were more and prognostic nutritional index PNI was lower too P < 0. 05 . Lasso regression screening identified the 5 most differential indicators multivariate Logistic regression analysis showed that smoking index ≥ 400 OR = 1. 659 95% CI = 1. 210 - 1. 969 P = 0. 006 ASA grade Ⅲ OR = 1. 968 95% CI = 1. 524-2. 326 P = 0. 002 Hand-sewn anastomosis OR = 3. 002 95% CI = 2. 526- 3. 425 P < 0. 001 neck anastomosis OR = 2. 201 95% CI = 1. 625-2. 458 P = 0. 001 PNI < 50 OR = 3. 652 95% CI = 3. 120-3. 968 P < 0. 001 were the independent predictor of anastomotic leakage. The nomogram predictive model was constructed by R software. Receiver operating curve ROC showed that the area under the curve AUC was 0. 899 which was significantly higher than the AUC value of every single risk factor all P < 0. 001 . The corrective curve showed that the predictive result was in good agreement with actual result P = 0. 663 . Conclusion The occurrence of anastomotic leakage in patients with resectable esophageal cancer after radical resection is closely related to multiple clinical factors. The construction of a quantitative nomogram model has good predictive efficiency which is convenient to guide the early identification of high-risk patients and targeted intervention and has good clinical application value.