基于胱抑素C建立大动脉粥样硬化性卒中的诺莫图及其预测评价

A study on development and evaluation of a nomogram prediction model of cystatin C for the patients with stroke of large-artery atherosclerosis

目的探讨基于胱抑素C(Cys-C)及各项独立相关风险因素,建立发生大动脉粥样硬化性卒中(LAA)的诺莫图模型并评价其效能。方法对2019年1月-2022年1月徐州医科大学附属医院神经内科收治的800例急性缺血性卒中(AIS)患者的临床资料进行回顾性观察研究。将研究对象随机分配(2∶1的比例)到训练集(n=546)和验证集(n=254),运用Logistic回归分析相关指标,筛选LAA的独立危险因素,构建诺模图预测模型,并对该模型进行内部验证和外部验证。临床预测模型的区分能力,校准度和临床有效性评估分别采用受试者工作特征曲线(ROC),Calibration校准曲线和决策曲线分析(DCA)来评估。结果单因素和多因素Logistic回归分析筛选显示高血压、糖尿病、胱抑素C、低密度脂蛋白(LDL)、同型半胱氨酸(HCY)、超敏C反应蛋白(hs-CRP)为LAA的独立相关预测因素,以此构建的预测模型,并绘制ROC曲线,训练集曲线下面积为0.800(0.762~0.838),诊断阈值为0.6224,对应灵敏度为68.3%、特异度为79.2%,验证集曲线下面积为0.838(0.786~0.890),诊断阈值为0.5901,对应灵敏度为79.5%、特异度为78.5%,Calibration校准曲线拟合良好。结论该诺莫图模型具有良好的区分度和准确度,其效果评估比较理想,可以较为简便及直观地个体化识别LAA高危患者,对LAA患者具有一定的预测价值,可早期通过各项预防措施改善患者的预后情况。

Objective To establish a nomogram model based on cystatin C(Cys-C)to predict the risk of large-artery atherosclerosis(LAA)by analyzing the risk factors of LAA and to evaluate its effectiveness.Methods A retrospective observational study performed on clinical data of 800 patients with acute ischemic stroke(AIS)in the Department of Neurology,Affiliated Hospital of Xuzhou Medical University,from January 2019 to January 2022.Patients were randomly assigned(2∶1 ratio)to the training set(n=546)and the validation set(n=254).The independent risk factors of LAA were determined by logistic regression analysis.A nomogram model of the risk of LAA was established and the model was verified internally and externally.The nomogram was evaluated based on discrimination,calibration,and clinical efficacy using the concordance statistic(C-statistic),calibration plot and decision curve analysis(DCA),respectively.Results Univariate and multifactor logistic regression analysis screening showed hypertension,diabetes,cystatin c,low-density lipoprotein(LDL),homocysteine(HCY),and hypersensitive C-reactive protein(hs-CRP)as independent correlated predictors of LAA,and the prediction model constructed from this,and the ROC curve was plotted with an area under the training set curve of 0.800(0.762-0.838),with a diagnostic threshold of 0.6224,corresponding to a sensitivity of 68.3%and a specificity of 79.2%,and a validation set curve with an area under the curve of 0.838(0.786-0.890)and a diagnostic threshold of 0.5901,corresponding to a sensitivity of 79.5%and a specificity of 78.5%,with a well-fitted calibration curve.Conclusion The nomogram model established has a good degree of differentiation and accuracy.It has a good performance in predictive ability,which is simple,intuitive and individualized to screen high-risk groups and has a certain predictive value for the occurrence of LAA patients,and can improve their prognosis through various preventive measures at an early stage.