Zhongguo cuzhong zazhi (Feb 2024)

急性穿支动脉脑梗死患者早期神经功能恶化列线图的建立与验证 Establishment and Verification of Nomogram of Early Neurological Deterioration in Patients with Acute Penetrating Artery Infarction

  • 步红静1,马娜2,张盼盼2,刘远洪2 (BU Hongjing1, MA Na2, ZHANG Panpan2, LIU Yuanhong2)

DOI
https://doi.org/10.3969/j.issn.1673-5765.2024.02.006
Journal volume & issue
Vol. 19, no. 2
pp. 158 – 166

Abstract

Read online

目的 探讨急性穿支动脉脑梗死(penetrating artery infarction,PAI)患者早期神经功能恶化(early neurological deterioration,END)的危险因素、建立列线图模型并对其进行评价。 方法 回顾性纳入2021年1月—2023年2月濮阳市人民医院急性外侧豆纹动脉及脑桥旁正中动脉脑梗死患者,入院后5 d内NIHSS评分增加≥2分的患者纳入急性脑梗死后END组;入院后5 d内NIHSS评分增加<2分的患者纳入非END组。按7∶3的比例将数据集随机划分为训练集和测试集,测试集用来评估模型性能。在训练集中,用R(4.2.3)软件行单因素分析,对于P<0.10的变量采用最小绝对收缩和选择运算(least absolute shrinkage and selection operator,LASSO)回归、logistic回归分析筛选出PAI患者发生END的独立危险因素,最后构建列线图预测模型。分别对训练集和测试集采用ROC曲线及其AUC评估模型的区分度,采用决策曲线分析(decision curve analysis,DCA)评估模型的临床实用性,采用校准图评估模型准确度。 结果 共纳入400例急性外侧豆纹动脉或脑桥旁正中动脉脑梗死患者,其中男性261例(65.25%),年龄64(56~70)岁;END组135例(33.75%),非END组265例(66.25%)。训练集中急性PAI患者 280例,94例(33.57%)发生END;测试集120例,41例(34.17%)发生END。训练集中11个变量 (P<0.10)进入LASSO回归,筛选出5个变量:入院时舒张压、糖尿病病史、吸烟史、中性粒细胞与淋巴细胞比值、梗死灶最大直径。多因素logistic回归分析显示,中性粒细胞与淋巴细胞比值(OR 40.85,95%CI 13.34~196.43,P<0.01)、糖尿病病史(OR 24.10,95%CI 6.92~106.30,P<0.01)、吸烟史(OR 6.16,95%CI 1.54~28.39,P=0.01)、梗死灶最大直径(OR 4.93,95%CI 1.35~19.82,P=0.02)是PAI患者发生END的独立危险因素,纳入列线图。采用Bootstrap法进行内部验证,分别绘制训练集及测试集的ROC曲线、校准曲线、DCA曲线。训练集和测试集ROC的AUC分别为0.88、0.87;校准图预测值与实际值一致性较好、DCA曲线显示预测模型临床实用性较高。 结论 中性粒细胞与淋巴细胞比值、吸烟史、糖尿病病史、梗死灶最大直径是急性PAI患者发生END的独立危险因素,列线图预测模型具有一定的临床实用价值。 Abstract: Objective To investigate the risk factors for early neurological deterioration (END) in patients with acute penetrating artery infarction (PAI) and to develop and evaluate a nomogram model. Methods Patients with acute cerebral infarction of lateral lenticulostriate artery and paramedian pontine artery in Puyang people’s hospital from January 2021 to February 2023 were retrospectively included. Patients with NIHSS score increased by≥2 points within 5 days after admission were included in the END group. Patients with NIHSS score increased by≤2 points within 5 days after admission were included in the non-END group. The data set was randomly divided into a training set and a test set according to the ratio of 7∶3, and the test set was used to evaluate the model performance. In the training set, univariate analysis was performed with R software, and for variables with P<0.10, least absolute shrinkage and selection operator (LASSO) regression and logistic regression analysis were used to screen out the independent risk factors for the occurrence of END in PAI patients, and finally a nomogram prediction model was constructed. The ROC curve and AUC were used to evaluate the discrimination of the model for the training set and the test set, respectively. Decision curve analysis (DCA) was used to evaluate the clinical utility of the model, and calibration chart was used to evaluate the accuracy of the model. Results A total of 400 patients with acute cerebral infarction of lateral lenticulostriate artery and paramedian pontine artery were included, including 261 males (65.25%). The median age was 64 (56-70) years. There were 135 cases (33.75%) in the END group and 265 cases (66.25%) in the non-end group. Of 280 acute PAI patients in the training set, 94 cases (33.57%) had END. Of 120 patients in the test set, 41 cases (34.17%) had END. In the training set, 11 variables (P<0.10) were entered into LASSO regression and 5 variables were screened out: diastolic blood pressure at admission, history of diabetes, history of smoking, neutrophil to lymphocyte ratio (NLR), maximum diameter. Multivariate logistic regression analysis showed that NLR (OR 40.85, 95%CI 13.34-196.43, P<0.01), history of diabetes (OR 24.10, 95%CI 6.92-106.30, P<0.01), history of smoking (OR 6.16, 95%CI 1.54-28.39, P=0.01) and maximum diameter (OR 4.93, 95%CI 1.35-19.82, P=0.02) are independent risk factors for the occurrence of END in acute PAI patients, and they are included in the nomogram. Bootstrap method was used for internal verification, and ROC curve, calibration curve and DCA curve of training set and test set were drawn respectively. The AUC of ROC of the training set and the test set were 0.88 and 0.87, respectively. The predicted value of calibration chart is in good agreement with the actual value, and the DCA curve shows that the prediction model has high clinical practicability. Conclusions NLR, history of smoking, history of diabetes and maximum diameter are independent risk factors for the occurrence of END in acute PAI patients, and the nomogram prediction model has certain clinical practical value.

Keywords