支气管动脉栓塞介入治疗结核性大咯血后再咯血危险因素分析及相关预测模型构建

Risk factors analysis and related prediction model construction of hemoptysis after interventional treatment of tuberculous massive hemoptysis by bronchial artery embolization

目的探讨支气管动脉栓塞(bronchial artery emboliza-tion,BAE)介入治疗结核性大咯血后再咯血的危险因素,并构建相应的预测模型。方法选取2019年8月~2023年2月我院接受BAE治疗的240例肺结核大咯血患者为研究对象,根据术后7d再咯血情况,分为再咯血者组(n=64)和不再咯血组(n=176)。采用多因素Logistics回归分析BAE治疗结核性大咯血后再咯血的危险因素。将所有样本数据按照7∶3的比例随机分为训练集和验证集,训练集用于构建模型,验证集用于模型的内部验证。采用R语言构建BAE治疗结核性大咯血后再咯血的梯度提升决策树(Gradient Boosting Decision Tree,GBDT)预测模型和列线图模型,通过受试者操作特征(Receiver operating characteristic,ROC)曲线、校准曲线以及ROC曲线下面积(areaunder the curve,AUC)对比评价模型的效能。结果Logistic回归分析显示,病程≥90天、合并肺动脉瘘、合并其他体循环动脉、术前PLT≥229.5×10^(9)/L是支气管动脉栓塞介入治疗结核性大咯血后再咯血的危险因素(P<0.05)。基于以上指标构建BAE治疗结核性大咯血后再咯血的列线图预测模型和GBDT预测模型,其中GBDT模型训练集ROC曲线下面积为0.818(95%CI:0.744~0.891),列线图预测模型训练集ROC曲线下面积为0.732(95%CI:0.649~0.816),GBDT模型、列线图预测模型验证集ROC曲线下面积分别为0.677(95%CI:0.515~0.839)、0.611(95%CI:0.485~0.807)。两种模型训练集Delong检验显示GBDT模型AUC值高于列线图模型(P<0.05)。结论BAE介入治疗结核性大咯血后再咯血主要与病程≥90天、合并肺动脉瘘、合并其他体循环动脉等因素有关,以此构建的梯度提升决策树模型对BAE介入治疗结核性大咯血后再咯血的整体预测效能较好,可为临床决策提供参考。

Objective To analyze the risk factors of hemoptysis after interventional treatment of tuberculous massive hemoptysis by bronchial artery embolization(BAE),and to construct a corresponding prediction model to provide a basis for clinical work,so as to better manage this disease and its complications.Methods A total of 240 patients with massive hemoptysis of pulmonary tuberculosis who received BAE treatment in our hospital from August 2019 to February 2023 were selected as the research objects.According to the recurrence of hemoptysis 7 days after the operation,they were divided into the recurrent hemoptysis group(n=64)and non-hemoptysis group(n=176).Multivariate logistic regression was used to analyze the risk factors of hemoptysis after BAE treatment of tuberculous massive hemoptysis.All sample data were randomly divided into a training set and a verification set according to the ratio of 7:3.The training set was used to construct the model,and the verification set was used for internal verification of the model.The gradient boosting decision tree(GBDT)prediction model and nomogram model of hemoptysis after BAE treatment of tuberculous massive hemoptysis were constructed by R language.The efficacy of the model was evaluated by the receiver operating characteristic(ROC)curve,calibration curve,and area under the ROC curve(area under the curve,AUC).Results Logistic regression analysis showed that the course of disease≥90 days,combined with pulmonary artery fistula,combined with other systemic arteries and pre perative PLT≥229.5×10^(9)/L were risk factors for hemoptysis after interventional treatment of bronchial artery embolization(P<0.05).Based on the above indicators,a nomogram prediction model and a GBDT prediction model for hemoptysis after BAE treatment of tuberculous massive hemoptysis were constructed.The area under the ROC curve of the GBDT model training set was 0.818(95%CI:0.744-0.891),and the area under the ROC curve of the nomogram prediction model training set was 0.732(95%CI:0.649-0.816).The area under the ROC curve of the validation set of the GBDT model and the nomogram prediction model was 0.677(95%CI:0.515-0.839)and 0.611(95%CI:0.485-0.807),respectively.The Delong test showed that the AUC value of GBDT model was higher than that of the nomogram model(P<0.05).Conclusion Re-hemoptysis after BAE interventional treatment of tuberculous massive hemoptysis is mainly related to factors such as course of disease≥90 days,pulmonary artery fistula,and other systemic circulation arteries.The gradient-boosting decision tree model constructed on this basis has a good overall predictive effect on re-hemoptysis after BAE interventional treatment of tuberculous massive hemoptysis,which can provide a reference for clinical decision-making.