Demaerschalk Bart M, Silver Brian, Wong Edward, Merino Jose G, Tamayo Arturo, Hachinski Vladimir
Mayo Clinic College of Medicine, Department of Neurology, Mayo Clinic Arizona, 13400 East Shea Boulevard, Scottsdale, AZ 85259, USA.
Can J Neurol Sci. 2006 May;33(2):200-4. doi: 10.1017/s0317167100004972.
To compare the inter-observer reliability of Alberta Stroke Programme Early CT Scoring (ASPECTS) with the ICE (Idealize-Close-Estimate) method of estimating > 1/3 middle cerebral artery territory (MCAT) infarction amongst stroke neurologists and to determine how well ASPECT Scoring predicts > 1/3 MCAT infarctions in acute ischemic stroke (AIS).
The European Cooperative Acute Stroke Study suggested that > 1/3 involvement of the MCAT on early CT scan was a risk factor for symptomatic intracerebral hemorrhage (SICH) following treatment with tissue plasminogen activator (tPA) for AIS but, in the absence of a systematic method of estimation had poor interobserver reliability (Kappa 0.49). The ICE method was developed to standardize the approach to estimating early MCAT infarct size and has very good interobserver reliability (Kappa 0.72). ASPECTS has comparable interobserver reliability and is reported to predict both neurological outcome and SICH.
Five stroke neurologists were tested with 40 AIS CT scans. Each performed blinded independent assessments of early ischemic changes with both ASPECTS and ICE. The reference standard was majority opinion of 1/3 MCAT determination of five neuroradiologists. A receiver operator curve (ROC) was constructed and likelihood ratios (LR) were calculated. Chance corrected agreement (kappa) and chance independent agreement (phi) were calculated for both methods, and analysis of variance was used to calculate reliability by intraclass correlation coefficient (ICC) for ASPECTS.
The LR for a positive test (> 1/3 MCAT) were extremely large and conclusive (approaching infinity) for ASPECTS of 0-3; were large and conclusive (30, 20, and 10) for ASPECTS of 4, 5, and 6 respectively; was an unhelpful 1 for ASPECTS of 7, and were again extremely large and conclusive (approaching zero) for ASPECTS of 8-10. A ROC plot supported an ASPECTS cutoff of < 7 as best for 1/3 MCAT estimation (94% sensitivity and 98% specificity). Kappa and Phi statistics were moderately good for both ASPECTS and ICE (0.7). ICC for ASPECTS was 0.8.
When experienced stroke neurologists utilize a formalized method of quantifying early ischemic changes on CT, either ASPECTS or ICE, the interobserver agreement and reliability are satisfactory. ASPECTS allows for a strong and conclusive estimation of the presence of 1/3 MCAT involvement and a cutoff point of < 7 results in best test performance.
比较阿尔伯塔卒中项目早期CT评分（ASPECTS）与理想化 - 近似 - 估计（ICE）方法在卒中神经科医生中对大脑中动脉供血区（MCAT）梗死面积超过1/3的评估者间可靠性，并确定ASPECT评分在急性缺血性卒中（AIS）中对超过1/3 MCAT梗死的预测效果。
欧洲急性卒中协作研究表明，早期CT扫描显示MCAT梗死面积超过1/3是急性缺血性卒中患者接受组织型纤溶酶原激活剂（tPA）治疗后发生症状性脑出血（SICH）的危险因素，但由于缺乏系统的评估方法，评估者间可靠性较差（kappa值为0.49）。ICE方法旨在规范早期MCAT梗死面积的评估方法，具有很好的评估者间可靠性（kappa值为0.72）。ASPECTS具有可比的评估者间可靠性，且据报道可预测神经功能结局和SICH。
5名卒中神经科医生对40例AIS患者的CT扫描进行测试。每位医生对早期缺血性改变分别采用ASPECTS和ICE方法进行盲法独立评估。参考标准为5名神经放射科医生对1/3 MCAT判定的多数意见。构建受试者工作特征曲线（ROC）并计算似然比（LR）。计算两种方法的校正机遇一致性（kappa）和机遇独立一致性（phi），并采用方差分析通过组内相关系数（ICC）计算ASPECTS的可靠性。
ASPECTS评分为0 - 3时，阳性检测（> 1/3 MCAT）的LR极大且具有决定性意义（接近无穷大）；评分为4、5和6时，LR分别为30、20和10，具有较大且决定性意义；评分为7时，LR为无帮助的1；评分为8 - 10时，LR再次极大且具有决定性意义（接近零）。ROC曲线表明，ASPECTS评分< 7作为1/3 MCAT评估的最佳截断值（敏感性为94%，特异性为98%）。ASPECTS和ICE的kappa和Phi统计值中等良好（0.7）。ASPECTS的ICC为0.8。
当经验丰富的卒中神经科医生采用正式的方法对CT上的早期缺血性改变进行量化时，无论是ASPECTS还是ICE，评估者间的一致性和可靠性都令人满意。ASPECTS能够对1/3 MCAT受累情况进行有力且决定性的评估，截断值< 7时测试性能最佳。
