摘要
目的 构建非瓣膜性心房颤动(non-valvular atrial fibrillation,NVAF)患者左心房血栓(left atrial thrombus,LAT)或自发显影(spontaneous echo contrast,SEC)的随机森林模型,并分析LAT/SEC患者抗凝治疗的临床结局。方法 回顾性分析2016年1月至2021年11月河北省沧州市中心医院接受经食管超声心动图检查的NVAF患者的临床资料。构建随机森林模型,通过受试者工作特征曲线(receiver operating characteristic curve,ROC)评价随机森林模型预测LAT/SEC的曲线下面积(area under the curve,AUC),并与CHADS2和CHA2DS2-VASc评分进行对比。结果 386例NVAF患者中,38例(9.8%)发生LAT/SEC。影响LAT/SEC发生的7个最重要的变量是左心室舒张末期内径、估算肾小球滤过率、左心室射血分数、非阵发性心房颤动、左心房内径、未抗凝治疗和脑卒中史。随机森林模型的AUC为0.93(95%CI:0.88~0.99),CHADS2评分的AUC为0.73(95%CI:0.63~0.83),CHA2DS2-VASc评分的AUC为0.72(95%CI:0.62~0.83),随机森林模型的AUC高于CHADS2(Z=3.42,P<0.001)和CHA2DS2-VASc(Z=3.05,P<0.001)评分。随机森林模型预测LAT/SEC的敏感度为92.6%,特异度为89.8%。38例LAT/SEC患者中,55.3%(21/38)患者经抗凝治疗后复查经食管超声心电图,66.7%(14/21)患者LAT/SEC溶解。17例直接口服抗凝剂治疗的患者中,4例LAT和8例SEC溶解;4例华法林治疗的患者中,1例LAT和1例SEC溶解。结论 随机森林模型可用于预测NVAF患者LAT/SEC,其诊断效能优于CHADS2及CHA2DS2-VASc评分。三分之二的LAT/SEC患者抗凝治疗后LAT/SEC溶解。
Objectives To construct a random forest model for left atrial thrombus(LAT)or spontaneous echo contrast(SEC)in patients with non-valvular atrial fibrillation(NVAF),and analyze the clinical outcome of anticoagulation ther-apy in patients with LAT/SEC.Methods The clinical data of patients diagnosed with NVAF who underwent transoesoph-ageal echocardiography in Cangzhou Central Hospital from January 2016 to November 2021 were retrospectively analyzed.The random forest model was constructed,the area under the curve(AUC)of random forest model were assessed using the receiver operating characteristic curve(ROC),and compared with CHADS2 and CHA2DS2-VASc scores.Results Among 386 patients with NVAF,38(9.8%)had LAT/SEC.The 7 most important variables affecting LAT/SEC were left ventricular end diastolic diameter,estimated glomerular filtration rate,left ventricular ejection fraction,non-paroxysmal atrial fibrillation,left atrial diameter,absence of anticoagulation therapy,and history of stroke.The random forest achieved an AUC of 0.93(95%CI:0.88-0.99),CHADS2 score achieved an AUC of 0.73(95%CI:0.63-0.83),and CHA2DS2-VASc score achieved an AUC of 0.72(95%CI:0.62-0.83).By comparison,the AUC of random forest model was higher than that of CHADS2 score(Z=3.42,P<0.001)and CHA2DS2-VASc score(Z=3.05,P<0.001).The sensitivity and specificity of the random forest model for predicting LAT/SEC were 92.6%and 89.8%,respectively.Among the 38 patients with LAT/SEC,55.3%(21/38)underwent repeated transoesophageal echocardiography,and 66.7%(14/21)LAT/SEC disappeared.LAT/SEC disappeared in 12(4 LAT and 8 SEC)of 17 patients treated with direct oral anticoagulant and in 2(1 LAT and 1 SEC)of 4 patients treated with warfarin.Conclusions The random forest mode can predict LAT/SEC in patients with NVAF,and the diagnostic performance of random forest model is superior to CHADS2 score and CHA2DS2-VASc score.LAT/SEC dissolve after anticoagulation in two-thirds of patients with LAT/SEC.
作者
郑楠
刘冰
闫洪伟
李凤鹏
于恺
张军
ZHENG Nan;LIU Bing;YAN Hong-wei;LI Feng-peng;YU Kai;ZHANG Jun(Fifth Department of Cardiology,Cangzhou Central Hospital,Cangzhou,Hebei 061000,China)
出处
《岭南心血管病杂志》
CAS
2024年第1期22-27,56,共7页
South China Journal of Cardiovascular Diseases
基金
沧州市科学计划项目(项目编号:213106002)。
关键词
心房颤动
左心房血栓
自发显影
随机森林
atrial fibrillation
left atrial thrombus
spontaneous echo contrast
random forest
