早发冠心病风险预测模型的构建与验证  

Construction and verification of risk prediction model for premature coronary artery disease

作　　者：蒋玉娇 王一华 胡淑文 门冰欣 胡娜娜 张亚苹 张锦[2] JIANG Yu-jiao;WANG Yi-hua;HU Shu-wen;MEN Bing-xin;HU Na-na;ZHANG Ya-ping;ZHANG Jin(The First Clinical College of Medicine,Lanzhou University,Lanzhou 730000,China;Department of Cardiology,The First Hospital of Lanzhou University,Lanzhou 730000,China)

机构地区：[1]兰州大学第一临床医学院,甘肃省兰州市730000 [2]兰州大学第一医院心内科

出　　处：《中国心血管病研究》2024年第12期1084-1089,共6页Chinese Journal of Cardiovascular Research

基　　金：甘肃省科技计划项目(22JR11RA018)。

摘　　要：目的针对现有冠心病预测模型在年轻人群中适用性不佳的问题,探讨早发冠心病(PCHD)的独立危险因素,构建PCHD风险预测模型并加以验证。方法连续收集兰州大学第一医院2023年1月~2023年12月疑诊为冠心病的年轻患者共1305例,根据冠状动脉造影结果分为PCHD组(745例)和非冠心病(NCHD)组(560例),以8∶2比例随机纳入训练集与验证集。比较两组基线资料,使用LASSO回归初筛危险因素,在训练集中进行logistic回归分析筛选PCHD的独立危险因素,继而构建PCHD风险预测模型。绘制受试者工作特征(ROC)曲线并计算曲线下面积(AUC),绘制校准曲线及决策曲线(DCA)进行模型验证及评价。结果LASSO回归共筛选出22个非零系数,纳入多因素logistic回归分析结果显示,男性、年龄、高血压病史、糖尿病史、高血压家族史、低水平载脂蛋白AI(ApoAI)、脂蛋白a[Lp(a)]、随机葡萄糖(RG)及天冬氨酸氨基转移酶(AST)是PCHD的独立危险因素(P<0.05)。基于此构建的PCHD列线图模型,ROC曲线显示训练集AUC=0.786,(95%CI 0.758~0.813),验证集AUC=0.796,(95%CI 0.742~0.849)。Hosmer-Lemeshow显示,训练集(P=0.795)和验证集(P=0.558)的校准曲线与理想曲线拟合良好。DCA显示使用该模型进行临床决策时,当训练集和验证集阈概率处于8%~72%和8%~60%时,患者净获益>0。结论在有冠心病相关症状的年轻人群中,基于男性、年龄、高血压病史、糖尿病史、高血压家族史、ApoAI、Lp(a)、RG及AST构建的PCHD风险预测模型在区分度、较准度及临床适用性方面均表现良好。Objective To investigate the independent risk factors of premature coronary heart disease(PCHD)and constructed and verified the PCHD risk prediction model to improve the poor applicability of the existing prediction models in young people.Methods A total of 1305 young patients suspected of coronary heart disease from the First Hospital of Lanzhou University from January to December 2023 were continuously collected.According to the coronary angiography results,the patients were divided into the PCHD group(745 cases)and non coronary heart disease(NCHD)group(560 cases),and the training set and validation set were randomly included in the ratio of 8∶2.The baseline data of the two groups were compared.LASSO regression was used for preliminary screening of risk factors,and logistic regression analysis was performed in the training set to screen the independent risk factors for PCHD;and then the PCHD risk prediction model was constructed.Receiver operating characteristic curve(ROC)was drawn and area under the curve(AUC)was calculated.Calibration curve and decision curve analysis(DCA)were drawn for model verification and evaluation.Results A total of 22 non-zero coefficients were selected by LASSO regression.Multivariate logistic regression analysis showed that male,age,history of hypertension,diabetes,family history of hypertension,low level apolipoprotein AI(ApoAI),lipoprotein a[Lp(a)],random glucose(RG)and aspartate aminotransferase(AST)were the independent risk factors for PCHD(P<0.05).Based on the constructed PCHD nomogram model,the ROC curve showed that the training set AUC=0.786(95%CI 0.758-0.813)and the verification set AUC=0.796(95%CI 0.742-0.849).The Hosmer-Lemeshow shows that the calibration curves of the training set(P=0.795)and the validation set(P=0.558)fitted the ideal curve well.DCA showed that when using this model for clinical decision making,the net benefit for patients was>0 when the training set and validation set threshold probabilities were 8%-72%and 8%-60%.Conclusion In the group of young people

关 键 词：早发冠心病 心血管疾病 年轻人群 风险预测模型 

分 类 号：R541.4[医药卫生—心血管疾病]