急性缺血性脑卒中风痰证预测模型构建与验证

Construction and validation of a predictive model for wind-phlegm syndrome in acute ischemic stroke

目的基于LASSO回归变量筛选的Logistic回归模型(以下简称为“LASSO-Logistic方法”)构建急性缺血性脑卒中(acute ischemic stroke,AIS)患者风痰证的临床预测模型,为研究该病种中医临床诊治提供依据。方法纳入北京中医药大学东直门医院通州院区脑病科六区2017年6月至2022年6月收治的1073例AIS患者,依据证候判断标准分为风痰组和非风痰组,利用LASSO-Logistic方法进行多因素回归分析,构建AIS患者风痰证的临床预测模型,应用Bootstrap法进行1000次重抽样进行模型内部验证,采用一致性系数(concordance index,C-index)、校准曲线、决策曲线分析法(decision curve analysis,DCA)对模型的区分度、校准度和临床有效性进行评价。结果(1)通过LASSO回归模型筛选得到16个变量,进一步的Logistic回归分析显示性别、年龄、生活评分、合并高脂血症、纤维蛋白原定量(fibrinogen,FIB)、白细胞计数(white blood cell count,WBC)为风痰证预测模型的独立影响因素(P<0.05);(2)利用上述6个指标构建列线图模型,模型中各参数的系数分别为:性别(β=0.518)、年龄(β=-0.02)、生活评分(β=0.015)、合并高脂血症(β=1.42)、FIB(β=-0.199)和WBC(β=0.4);(3)模型评价方面,该模型的C-index为0.712(95%CI[0.680,0.745]),Hosmer-Lemeshow拟合优度检验提示本模型校准度良好(P>0.05);(4)DCA曲线显示,当阈值率>3%以及<93%时该模型对AIS风痰证进行预测可临床获益。结论本研究结合性别、年龄、生活评分、合并高脂血症、FIB、WBC共6种独立影响因素,初步构建AIS患者风痰证的临床预测模型。

Objective To construct a clinical nomogram model of wind-phlegm syndrome in acute ischemic stroke(AIS)patients based on Lasso-logistic method,and to provide a basis for studying the clinical diagnosis and treatment of AIS in traditional Chinese medicine.Methods The 1073 patients with AIS admitted to the Department of Encephalopathy,Tongzhou Hospital,Dongzhimen Hospital,Beijing University of Traditional Chinese Medicine from June 2017 to June 2022 were divided into wind phlegm group and non wind phlegm group according to the criteria for syndrome judgment.The logistic regression model(hereinafter referred to as“LASSO Logistic Method”)selected by LASSO regression variables was used for multivariate regression analysis to construct a clinical prediction model for wind phlegm syndrome in AIS patients,the Bootstrap method was used to conduct 1000 re samples for internal validation of the model.The consistency coefficient(C-index),calibration curve,and decision curve analysis(DCA)were used to evaluate the discrimination,calibration,and clinical effectiveness of the model.Results(1)16 variables were obtained through LASSO regression model screening.Further logistic regression analysis showed that gender,age,life score,hyperlipidemia,FIB,and WBC were independent influencing factors for the prediction model of wind phlegm syndrome(P<0.05).(2)Using the above six indicators to construct the nomograph model,the coefficients of each parameter in the model:gender(β=0.518),Age(β=-0.02),Life Score(β=0.015),combined with hyperlipidemia(β=1.42)、FIB(β=-0.199)and WBC(β=0.4).(3)In terms of model evaluation,the C-index of the model was 0.712(95%CI[0.680,0.745]),and the Hosmer-Lemeshow goodness of fit test indicates that the calibration of the model was good(P>0.05).(4)The DCA curve shows that when the threshold rate(Pt)was more than 3%and less than 93%,the model can predict the wind phlegm syndrome of AIS and have clinical benefits.Conclusion In this study,the independent influencing factors of gender,age,life score,combined with hyperlipidemia,FIB and WBC were included to preliminatively construct a clinical prediction model of wind-phlegm syndrome in AIS patients.