直肠癌前切除术后吻合口漏预测与临床决策分析

Prediction and clinical decision-making of anastomotic leakage after anterior resection of rectal cancer

目的:探索直肠癌前切除术术后吻合口漏的独立影响因素并建立吻合口漏的风险预测模型,进一步运用决策曲线分析将其与传统经验模型比较,寻找最优模型,指导术中预防性造口的临床决策,减少不必要的预防性造口。方法:回顾性搜集2017年10月至2019年12月我院404例行直肠癌前切除术患者的临床资料,运用χ^(2)检验及Lasso分析筛选变量后,行Logistic回归寻找吻合口漏的独立影响因素,建立预测模型并绘制列线图。再分别依据糖尿病、术前放化疗、血红蛋白、白蛋白、肿瘤距肛缘距离及本研究预测模型建立决策模型,绘制决策曲线,量化比较各决策模型及传统经验性预防造口对应阈值范围内的净获益率,评估临床获益情况。结果:男性(OR=4.490,95%CI:1.491~13.518)、输血(OR=5.822,95%CI:1.939~17.475)及肿瘤距肛缘的距离≤7 cm(OR=2.385,95%CI:1.086~5.237)为吻合口漏的独立危险因素,术前机械性肠道准备为独立保护因素(OR=0.360,95%CI:0.165~0.785)。预测模型一致性指数为0.735。受试者工作特征曲线下面积为0.755(95%CI:0.669~0.841)。模型在0.05~1.00阈值范围内最优,净获益率为1%~6%。结论:男性、输血、肿瘤距肛缘的距离≤7 cm、机械性肠道准备为直肠癌前切除术术后吻合口漏的独立影响因素。本文预测模型区分度及准确度较高,在0.1~0.2阈值范围内优于其它决策模型及临床经验性判断,临床获益率最高,可用于指导术中造口决策。

Objective:To explore the independent risk factors for anastomotic leakage after anterior resection of rectal cancer and establish an anastomotic leakage risk prediction model.Then,using decision curve analysis to compare it with traditional empirical models,find the optimal model to guide the intraoperative decision on defunctioning stoma,minimize unnecessary defunctioning stoma.Methods:The clinical data of 404 patients undergoing anterior resection of rectal cancer in our hospital from October 2017 to December 2019 were retrospectively collected.Firstly,we screened the effective variables by χ^(2) test and Lasso analysis,picked independent influence factors by Logistic regression analysis,then used R(4.0.5)to establish a predictive model and draw a nomogram.After that,we established decision models based respectively on diabetes,preoperative radiotherapy and chemotherapy,hemoglobin,albumin,the distance between the tumor and the anal margin,and the prediction model of this study.Finally,a decision curve was drawn through R(4.0.5).Those models on the level of net benefit in different threshold ranges were compared.Results:Male(OR=4.490,95%CI:1.491~13.518),blood transfusion(OR=5.822,95%CI:1.939~17.475)and the distance between the tumor and the anal margin≤7 cm(OR=2.385,95%CI:1.086~5.237)were independent risk factors for anastomotic leakage,and preoperative mechanical bowel preparation was an independent protective factor(OR=0.360,95%CI:0.165~0.785).The concordance index of the prediction model was 0.735.The area under the receiver operating characteristic curve was 0.755(95%CI:0.669~0.841).Our model was optimal within the threshold range of 0.05~1.00,and the net benefit rate was 1%~6%.Conclusion:Male,blood transfusion,the distance between the tumor and the anal margin≤7 cm,and mechanical bowel preparation are independent risk factors for anastomotic leakage after anterior resection of rectal cancer.Our prediction model is well discriminated and accurate,and is superior to other traditional empirical models in terms of guiding the intraoperative decision on defunctioning stoma within the threshold range of 0.1~0.2,can also achieve the highest clinical benefit.That is to say,using our predictive model to guide intraoperative stoma decision-making is reasonable.