Anastomotic leakage is a dangerous postoperative complication of oesophageal surgery. The present study aimed to develop a simple and practical scoring system to predict the risk of anastomotic leakage after oesophageal resection. A consecutive series of 330 patients who underwent oesophageal cancer surgery from January 2016 to January 2018 at the Shanghai Chest Hospital were included to develop a prediction model. Anastomotic leakage was evaluated using oesophagography, computed tomography, or flexible endoscopy. Least absolute shrinkage and selection operator regression based on a generalized linear model was used to select variables for the anastomotic leakage risk model while avoiding overfitting. Multivariable logistic regression analysis was applied to build forest plots and a prediction model. The concordance index or the area under the curve was used to judge the discrimination, and calibration plots verified the consistency. Internal validation of the model was conducted, and the clinical usefulness and threshold screening of the model were evaluated by decision curve analysis. The factors included in the predictive nomogram included Sex, diabetes history, anastomotic type, reconstruction route, smoking history, CRP level and presence of cardiac arrhythmia. The model displayed a discrimination performance with a concordance index of 0.690 (95% confidence interval: 0.620–0.760) and good calibration. A concordance index value of 0.664 was maintained during the internal validation. The calibration curve showed good agreement between the actual observations and the predicted results. The present prediction model, which requires only seven variables and includes Sex, diabetes history, anastomotic type, reconstruction route, smoking history, CRP level and presence of cardiac arrhythmia, may be useful for predicting anastomotic leakage in patients after oesophagectomy.

中文翻译：

---

新颖的列线图可预测食管切除术后患者发生吻合口漏的风险。

吻合口漏是食管手术危险的术后并发症。本研究旨在开发一种简单实用的评分系统，以预测食管切除术后吻合口漏的风险。纳入了2016年1月至2018年1月在上海胸科医院接受食管癌手术的330例连续患者，以建立预测模型。使用食道造影，计算机断层扫描或柔性内窥镜评估吻合口漏。使用基于广义线性模型的最小绝对收缩和选择算子回归来选择吻合口漏风险模型的变量，同时避免过拟合。多变量逻辑回归分析被用于建立森林地块和预测模型。一致性指数或曲线下的面积用于判断判别，校准图验证了一致性。对模型进行内部验证，并通过决策曲线分析评估模型的临床有效性和阈值筛选。预测列线图中包含的因素包括性别，糖尿病史，吻合类型，重建途径，吸烟史，CRP水平和心律不齐。该模型显示出的辨别性能为0.690（95％置信区间：0.620–0.760），并且校准良好。在内部验证期间，一致性指数值为0.664。校正曲线显示出实际观察结果与预测结果之间的一致性。目前的预测模型