关键词: 食管癌, 吻合口漏, 预测模型, Cox回归, 风险评分模型

Abstract: Objective To construct a predictive model for postoperative anastomotic leakage in esophageal cancer patients based on Cox regression combined with risk scoring.Methods A retrospective analysis was conducted on the data of 380 esophageal cancer patients who underwent surgery at our hospital from February 2023 to June 2024.The patients were divided into a training set (n=304) and a validation set (n=76) in an 8∶2 ratio.Based on the training set data,the patients were divided into an observation group (stomal leak,n=38) and a control group (no anastomotic leakage,n=266) according to the occurrence of postoperative anastomotic leakage.The demographic data,surgical methods,time of anastomotic fistula,time of gastrointestinal decompression,and number of lymph node dissection were collected from both groups.Univariate and Cox regression analysis was used to determine the relationship between each variable and postoperative anastomotic fistula,and a nomogram model was constructed using R software for internal verification to determine the discrimination and calibration for anastomotic leakage within 3 days and 5 days after the operation.Results A total of 49 patients (12.9%) with anastomotic leakage after surgery were included in this study.There were 38 cases (12.5%) of postoperative anastomotic leakage in the training set and 11 cases (14.5%) in the validation set.Univariate analysis showed that the differences in age,diabetes comorbidity,tumor diameter,pulmonary infection,preoperative chemoradiotherapy,preoperative targeted drug therapy,duration of gastrointestinal decompression,and number of lymph nodes dissected between the observation group and the control group were statistically significant (P＜0.05).Cox regression analysis revealed that pulmonary infection,age,preoperative targeted therapy,preoperative chemoradiotherapy,diabetes,duration of gastrointestinal decompression,and tumor diameter were influencing factors for postoperative anastomotic leak in esophageal cancer patients (P＜0.05).The area under the ROC curve of the training set prediction model was 0.94 (0.91~0.98),and the validation set was 0.97 (0.92~0.99),The two datasets prediction models show good model discrimination and calibration.Conclusion We constructed and validated the risk prediction model of anastomotic fistula after esophageal cancer based on Cox regression analysis,and systematically considered several clinically relevant factors including pulmonary infection,age,targeted therapy before surgery,preoperative chemoradiotherapy,diabetes,time of gastrointestinal decompression,and tumor diameter.The intuitive presentation of the nomogram model makes the risk assessment more understandable to both doctors and patients,thus contributing to the clinical decision making.

Key words: esophageal cancer, anastomotic fistula, prediction model, Cox regression, risk score model