CAC评分结合临床特征预测冠心病CCTA的价值研究

Study on the Value of CAC Score Combined with Clinical Characteristics in Predicting CCTA in Coronary Heart Disease

目的探讨冠状动脉钙化(CAC)评分对非典型胸痛患者冠状动脉CT血管成像(CCTA)结果的预测价值。方法纳入953名因非典型胸痛而接受CCTA和CAC扫描的患者数据,包括心血管危险因素、CAC评分等在内的63个变量被用来建立随机森林(RF)模型。参与者中70%作为训练模型,30%为验证模型。将RF模型的预测性能与两种传统的Logistic回归模型进行了比较。结果梗阻性冠心病的发生率为16.4%。射频模型的受试者特征下面积为0.841,CACS模型为0.746,临床模型为0.810。RF模型明显优于其他两种模型(P<0.05)。此外,校正曲线和Hosmer-Lemesow检验表明,RF模型具有良好的分类性能(P=0.556)。CAC评分、年龄、血糖、同型半胱氨酸和中性粒细胞是RF模型中最重要的五个变量。结论RF模型在预测梗阻性CAD方面优于传统模型。在临床实践中,RF模型可以改善风险分层,优化个体管理。

Objective To investigate the predictive value of coronary artery calcification(CAC)score for coronary CT angiography(CCTA)in patients with atypical chest pain.Methods Data from 953 patients undergoing CCTA and CAC scans for atypical chest pain were included.63 variables,including cardiovascular risk factors,CAC scores,etc.,were used to establish a random forest(RF)model.70%of the participants served as training models and 30%as validation models.The predictive performance of RF model was compared with two traditional Logistic regression models.Results The incidence of obstructive coronary heart disease was 16.4%.The subject area under characteristic was 0.841 in the radiofrequency model,0.746 in the CACS model,and 0.810 in the clinical model.RF model was significantly better than the other two models(P<0.05).In addition,calibration curve and Hosmer-Lemesow test show that the RF model has good classification performance(P=0.556).CAC score,age,blood glucose,homocysteine,and neutrophils were the five most important variables in the RF model.Conclusion RF model is better than traditional model in predicting obstructive CAD.In clinical practice,the RF model can improve risk stratification and optimize individual management.