两种脑电图监测技术预测CPR后昏迷患者不良预后的准确性比较

Comparison of the accuracy of predicting poor outcome of coma after cardiopulmonary resuscitation with two kinds of electroencephalogram techniques

目的探讨脑电图（EEG）分级与振幅整合脑电图（aEEG）模式分级对心肺复苏（CPR）后昏迷患者不良预后的预测准确性。 方法采用回顾性研究方法，选择2010年3月至2017年6月首都医科大学附属北京同仁医院重症加强治疗病房（ICU）收治的7 d内非低温治疗且完成EEG监测的CPR后昏迷患者。收集患者的一般资料、格拉斯哥昏迷评分（GCS）、EEG分级和aEEG模式分级。根据发病后3个月格拉斯哥预后评分（GOS）将患者分为预后不良组（GOS 1~2级）和预后良好组（GOS 3-5级），比较两组相关指标的差异；绘制受试者工作特征曲线（ROC），评价aEEG模式分级和EEG分级对不良预后的预测能力。 结果共纳入54例患者，其中男性31例，女性23例；年龄（53.9±19.3）岁；EEG Young分级中，1级17例（占31.5%），2-5级4例（占7.4%），6级33例（占61.1%）；aEEG模式分级中，慢波增多模式（1级）26例（占48.1%），全面抑制模式（4级）23例（占42.6%），癫痫持续状态模式（2级）4例（占7.4%），爆发-抑制模式（3级）1例（占1.9%）。发病后3个月预后不良36例，其中死亡26例，持续植物状态10例；预后良好18例，其中严重神经功能残疾16例，中度神经功能残疾2例。不同预后两组患者性别、年龄、缺氧持续时间比较差异无统计学意义；预后不良组患者意识障碍程度重于预后良好组〔GCS（分）：4.1±1.7比5.0±2.1，P〈0.05〕。一致性检验显示，不同医师对EEG分级及aEEG模式分级的一致性较好（Kappa值分别为0.917和0.932）。ROC曲线分析显示，aEEG和EEG分级预测CPR后昏迷患者不良预后的ROC曲线下面积（AUC）分别为0.815和0.720（均P〈0.01）。aEEG模式分级最佳截断值为2.5时，敏感度为79.3%，特异度为77.4%，阳性似然比为3.508，阴性似然比为0.267；EEG分级最佳截断值为4.5时，敏感度为82.8%，特异度为61.3%，阳性似然比为2.140，阴性似然比为0.281。 结论aEEG模式分级较EEG分级能够更准确地预测CPR后昏迷患者早期不良预后，且操作简单，适宜ICU应用。

ObjectiveTo compare the accuracy of electroencephalography （EEG） grading scale with amplitude-integrated electroencephalography （aEEG） in predicting poor outcomes （3-month）, who sustained coma after cardiopulmonary resuscitation （CPR） in adults.MethodsA retrospective study was conducted. The patients with post-anoxic coma admitted to intensive care unit （ICU） of Tongren Hospital, Capital Medical University from March 2010 to June 2017 were enrolled. EEG was registered and recorded at least once within 7 days of coma after CPR, while not being subjected to therapeutic hypothermia. General data, Glasgow coma scale （GCS）, EEG grading and aEEG model were collected. According to Glasgow prognosis score （GOS） of 3-month outcome, patients were divided into poor prognosis group （GOS 1-2） and good prognosis group （GOS 3-5）, and the differences of related indexes between the two groups were compared. The predictive ability of aEEG model and EEG grading for brain function prognosis was evaluated by receiver operating characteristic （ROC） curve.ResultsFifty-four patients were included, with 31 males and 23 females, and age of （53.9±19.3） years. Among the EEG Young grades, 17 cases （31.5%） were grade 1, 4 cases （7.4%） were grade 2-5, and 33 cases （61.1%） were grade 6. Among the aEEG model grades, 26 cases （48.1%） had slow wave pattern grade 1, 23 cases （42.6%） had suppressed mode grade 4, 4 cases （7.4%） had status epilepticus mode grade 2, and 1 case （1.9%） had burst suppression mode grade 3. Thirty-six patients had poor prognosis 3-month after onset, 26 of them died and 10 had persistent vegetative state. The prognosis was good in 18 cases, including 16 cases with severe neurological disability and 2 cases with moderate neurological disability. There was no significant difference in gender, age, anoxic time between two groups with different prognosis, while the degree of consciousness disorder in poor prognosis group was more severe than that in good prognosis group （GCS score： 4.1±1.7 vs. 5.0±2.1, P 〈 0.05）. The consistency test showed that different physicians had good consistency in EEG grading and aEEG model （Kappa values were 0.917 and 0.932, respectively）. It was shown by ROC curve analysis that the area under ROC curve （AUC） of aEEG model and EEG grading for predicting poor prognosis of coma patients after CPR were 0.815 and 0.720, respectively （both P 〈 0.01）; when the cut-off value of aEEG was 2.5, the sensitivity was 79.3%, the specificity was 77.4%, the positive likelihood ratios （PLR） was 3.508, and the negative likelihood ratios （NLR） was 0.267; when the cut-off value of EEG grading was 4.5, the sensitivity was 82.8%, the specificity was 61.3%, the PLR was 2.140, and NLR was 0.281.ConclusionsaEEG model was more accurate in prognosticating poor outcomes （3-month） in patients with post-anoxic coma, when compared to EEG grading. Its operation was simple, so aEEG is very suitable in ICU.