最小绝对收缩和选择算子回归筛选急性缺血性脑卒中溶栓患者症状性脑内出血预测因素分析

Analysis of predictive factors for symptomatic intracranial hemorrhage in acute ischemic stroke patients treated with thrombolysis screened by least absolute shrinkage and selection operator regression

目的探讨基于最小绝对收缩和选择算子(LASSO)回归筛选急性缺血性脑卒中溶栓患者症状性脑内出血(sICH)发生的预测因素。方法选取2019年6月至2022年6月于鄂州市中心医院收治的428例急性缺血性脑卒中患者作为研究对象,静脉溶栓后24 h采用CT观察脑内出血情况,根据有无sICH分为sICH组36例和无sICH组392例。记录两组人口社会学资料、溶栓前实验室检查资料、影像学相关资料,采用LASSO回归分析筛选变量初步筛选重要预测因素,将非零回归系数的变量纳入多因素Logistic回归分析,确定影响急性缺血性脑卒中溶栓后发生sICH的独立预测因素。结果36例(8.41%)患者在溶栓后24 h内出现sICH归为sICH组,其余392例(91.59%)未出现sICH归为无sICH组。两组患者发病至溶栓时间、糖尿病、心房颤动、入院时美国国立卫生研究院卒中量表(NIHSS)评分、入院时血糖、C反应蛋白、CT显示大动脉高密度征象比较差异有统计学意义(t/χ^(2)=2.163、3.911、4.092、3.087、3.156、3.937、8.553,均P<0.05);LASSO回归算法共筛选出入院NIHSS评分、入院时血糖、C反应蛋白、大动脉高密度征象4个具有非零系数的显著相关指标;进一步多因素Logistic回归分析显示:入院NIHSS评分、入院时血糖、C反应蛋白、大动脉高密度征象为急性缺血性脑卒中溶栓后sICH的危险因素[OR(95%CI)=1.123(1.029~1.226)、1.391(1.086~1.781)、1.901(1.362~2.652)、3.112(1.364~7.104),均P<0.05];ROC曲线显示,基于LASSO-Logistic构建的模型预测急性缺血性脑卒中溶栓后sICH发生的AUC为0.802(95%CI=0.722~0.882),特异度为61.20%,敏感度为88.90%。结论基于LASSO回归筛选出的入院NIHSS评分、入院时血糖水平、C反应蛋白、大动脉高密度征象为急性缺血性脑卒中溶栓后sICH的预测因素。

Objective To investigate the predictive factors for symptomatic intracerebral hemorrhage(sICH)in acute ischemic stroke patients treated with thrombolysis screened by the least absolute schrinkage and selection operation(LASSO)regression.Methods A total of 428 acute ischemic stroke patients admitted to Ezhou Central Hospital from June 2019 to June 2022 were selected as the research subjects.CT was used to observe intracranial hemorrhage 24 hours after intravenous thrombolysis,and patients were divided into the sICH group and the non-sICH group based on the presence or absence of sICH.The sociodemographic data,pre-thrombolysis laboratory results,and relavant imaging data of the two groups were recorded.LASSO regression analysis was used to initially screen important predictive factors.Multivariable logistic regression analysis was performed on the selected variables with non-zero regression coefficients to determine the independent predictive factors influencing the occurrence of sICH after thrombolysis in acute ischemic stroke.Results A total of 36 patients(8.41%)developed sICH within 24 hours after thrombolysis were assigned to the sICH group,while the remaining 392 patients(91.59%)who did not develop sICH were assigned to the non-sICH group.There were statistically significant differences in the time from symptom onset to thrombolysis,diabetes,atrial fibrillation,National Institute of Health Stroke Scale(NIHSS)score at admission,blood glucose at admission,C-reactive protein,and CT-detected hyperdense large artery sign betweent the two groups(t/χ^(2)=2.163,3.911,4.092,3.087,3.156,3.937,8.553;all P<0.05).The LASSO regression algorithm screened out four significant correlation indexes with non-zero coefficients,including NIHSS score at admission,blood glucose at admission,C-reactive protein,and hyperdense large artery sign.Further multivariate Logistic regression analysis showed that NIHSS score at admission,blood glucose at admission,C-reactive protein and hyperdense large artery sign were risk factors for sICH after thrombolysis in acute ischemic stroke[OR(95%CI)=1.123(1.029-1.226),1.391(1.086-1.781),1.901(1.362-2.652),3.112(1.364-7.104);all P<0.05].The ROC curve showed that the model based on LASSO-Logistic had an AUC of 0.802(95%CI=0.722-0.882),with a specificity of 61.20%and a sensitivity of 88.90%for predicting the occurrence of sICH after thrombolysis in acute ischemic stroke.Conclusion The NIHSS score at admission,blood glucose at admission,C-reactive protein,and hyperdense large artery sign identified based on LASSO regression are predictive factors for sICH after thrombolysis in acute ischemic stroke.