摘要：

目的 对大脑半球大面积梗死（MCHI）患者发病急性期不同时间段脑电图变化的分析，确定脑电图预测预后的最佳时机。方法 共72 例MCHI 患者分别于发病第1 ~ 3 和4 ~ 7 天行脑电图监测，比较不同时间段脑电图模式和Synek 分级，预测预后之准确性，发病后3 个月采用Glasgow 预后分级（GOS）评估预后。结果 72 例患者发病后3 个月良好预后占62.50%（45/72）、不良预后37.50%（27/72）。发病第1 ~ 3天（RR = 0.357，95%CI：0.153 ~ 0.834，Mantel-Haenszel χ² = 6.147，P = 0.013）和第4～7天（RR = 0.240，95%CI：0.102 ~ 0.564，Mantel-Haenszel χ² = 13.601，P = 0.000）良性型脑电图模式为MCHI 患者预后之保护因素，其预测良好预后率为82.14%（23/28）~ 85.71%（30/35），准确度为62.50% ~ 72.22%；发病第4～7 天恶性型脑电图模式为MCHI 患者预后之危险因素（RR = 2.909，95%CI：1.611 ~ 5.253，Mantel-Haenszel χ² = 11.110，P = 0.001），其预测不良预后率为66.67%（16/24），准确度为73.63%。脑电图Synek 分级与GOS 评分在发病后第1 ~ 3（r_s = - 0.354，P = 0.002）和4 ~ 7 天（r_s = - 0.417，P = 0.000）均呈负相关，提示Synek 分级越高、预后越差。发病第4～7 天脑电图Synek 分级的预测准确度更高（83.33%对70.78%；χ² = 4.000，P = 0.039），与实际预后一致性更佳（Kappa 检验：κ = 0.639，95%CI：0.522 ~ 0.746 对κ = 0.406，95%CI：0.353 ~ 0.459；P = 0.001）、辨别力更强（受试者工作特征曲线下面积：0.86 ± 0.05，95%CI：0.761 ~ 0.958 对0.69 ± 0.07，95%CI：0.554 ~ 0.822；P = 0.002）。结论 MCHI 患者发病第1 ~ 3 天脑电图模式和Synek 分级可以作为脑损伤严重程度评估和指导治疗选择的参考指标，发病第4 ~ 7 天脑电图Synek分级预测预后的准确度可以作为长期预后预测和医疗决策的参考指标。

关键词: 脑梗死, 脑电描记术, 预后, 敏感性与特异性

Abstract:

Objective  To determine the optimal timing of electroencephalogram (EEG) in predicting the outcome in patients with massive cerebral hemispheric infarction (MCHI) by evaluating and comparing the EEG changes during different acute stages after onset.  Methods  A total of 72 MCHI patients were selected to be monitored by bedside EEG at two stages: 1-3 and 4-7 d after onset, respectively. The outcome after 3 months was assessed as good [Glasgow Outcome Scale (GOS) score 3-5] or poor (GOS score 1-2). Then the predictive accuracy of EEG patterns and gradings (the Synek scale) were calculated and compared between two stages to confirm the optimal timing of prediction.  Results  In 72 cases, 62.50% (45/72) had good and 37.50% (27/72) had poor outcome 3 months after onset. Benign EEG patterns were protective factors for the outcome of MCHI (1-3 d: RR = 0.357, 95%CI: 0.153-0.834, Mantel-Haenszel χ² = 6.147, P = 0.013; 4-7 d: RR = 0.240, 95%CI: 0.102-0.564, Mantel-Haenszel χ² = 13.601, P = 0.000) and they could predict good outcome with incidence rate of 82.14% (23/28)-85.71% (30/35) and the accuracy of 62.50%-72.22% . Malignant EEG patterns at 4-7 d were risk factors for the outcome of MCHI (RR = 2.909, 95%CI: 1.611-5.253, Mantel-Haenszel χ² = 11.110, P = 0.001), and they could predict poor outcome with incidence rate of 66.67% (16/24) and the accuracy of 73.63%. There was a significant negative correlation between the Synek scale and GOS score both during 1-3 d (r_s = - 0.354, P = 0.002) and 4-7 d (r_s = - 0.417, P = 0.000) after onset, indicating the higher the Synek scale was, the worse the outcome would be. The accuracy (83.33% vs 70.78%; χ² = 4.000, P = 0.039), consistency (Kappa test: κ = 0.639, 95%CI: 0.522-0.746 vs κ = 0.406, 95%CI: 0.353-0.459; P = 0.001) and area under curves (0.86 ± 0.05, 95%CI: 0.761-0.958 vs 0.69 ± 0.07, 95%CI: 0.554-0.822; P = 0.002) of the Synek scale for prediction of long-term outcome during 4-7 d after onset were significantly higher than those during 1-3 d after onset.  Conclusions  EEG patterns and the Synek scale during 1-3 d after MCHI onset could be used to evaluate the severity of brain injury and to further guide medical treatment, while the Synek scale during 4-7 d after onset could be used to evaluate the long-term outcome and treatment with a high accuracy.

Key words: Brain infarction, Electroencephalography, Prognosis, Sensitivity and specificity