Callahan Brandy L, Ramirez Joel, Berezuk Courtney, Duchesne Simon, Black Sandra E
LC Campbell Cognitive Neurology Research Unit, Sunnybrook Health Sciences Centre, 2075 Bayview Avenue, Rm A4 21, Toronto, Ontario, M4N 3 M5, Canada.
Heart & Stroke Foundation Canadian Partnership in Stroke Recovery, Sunnybrook Health Sciences Centre, Toronto, Canada.
Alzheimers Res Ther. 2015 Nov 5;7(1):68. doi: 10.1186/s13195-015-0152-z.
The definition of "objective cognitive impairment" in current criteria for mild cognitive impairment (MCI) varies considerably between research groups and clinics. This study aims to compare different methods of defining memory impairment to improve prediction models for the development of Alzheimer's disease (AD) from baseline to 24 months.
The sensitivity and specificity of six methods of defining episodic memory impairment (< -1, -1.5 or -2 standard deviations [SD] on one or two memory tests) were compared in 494 non-demented seniors from the Alzheimer's Disease Neuroimaging Initiative using the area under the curve (AUC) for receiver operating characteristic analysis. The added value of non-memory measures (language and executive function) and biomarkers (hippocampal and white-matter hyperintensity volume, brain parenchymal fraction [BPF], and APOEε4 status) was investigated using logistic regression.
Baseline scores < -1 SD on two memory tests predicted AD with 75.91 % accuracy (AUC = 0.80). Only APOE ε4 status further improved prediction (B = 1.10, SE = 0.45, p = .016). A < -1.5 SD cut-off on one test had 66.60 % accuracy (AUC = 0.77). Prediction was further improved using Trails B/A ratio (B = 0.27, SE = 0.13, p = .033), BPF (B = -15.97, SE = 7.58, p = .035), and APOEε4 status (B = 1.08, SE = 0.45, p = .017). A cut-off of < -2 SD on one memory test (AUC = 0.77, SE = 0.03, 95 % CI 0.72-0.82) had 76.52 % accuracy in predicting AD. Trails B/A ratio (B = 0.31, SE = 0.13, p = .017) and APOE ε4 status (B = 1.07, SE = 0.46, p = .019) improved predictive accuracy.
Episodic memory impairment in MCI should be defined as scores < -1 SD below normative references on at least two measures. Clinicians or researchers who administer a single test should opt for a more stringent cut-off and collect and analyze whole-brain volume. When feasible, ascertaining APOE ε4 status can further improve prediction.
目前轻度认知障碍（MCI）标准中“客观认知障碍”的定义在不同研究团队和临床机构之间存在很大差异。本研究旨在比较定义记忆障碍的不同方法，以改进从基线到24个月时阿尔茨海默病（AD）发展的预测模型。
在来自阿尔茨海默病神经影像学计划的494名非痴呆老年人中，使用接受者操作特征分析的曲线下面积（AUC）比较了六种定义情景记忆障碍方法（在一项或两项记忆测试中< -1、-1.5或-2个标准差[SD]）的敏感性和特异性。使用逻辑回归研究非记忆测量（语言和执行功能）和生物标志物（海马和白质高信号体积、脑实质分数[BPF]和APOEε4状态）的附加值。
两项记忆测试的基线分数< -1 SD预测AD的准确率为75.91%（AUC = 0.80）。只有APOE ε4状态进一步改善了预测（B = 1.10，SE = 0.45，p = 0.016）。一项测试中< -1.5 SD的临界值准确率为66.60%（AUC = 0.77）。使用连线测验B/A比值（B = 0.27，SE = 0.13，p = 0.033）、BPF（B = -15.97，SE = 7.58，p = 0.035）和APOEε4状态（B = 1.08，SE = 0.45，p = 0.017）可进一步提高预测能力。一项记忆测试中< -2 SD的临界值（AUC = 0.77，SE = 0.03，95%CI 0.72 - 0.82）预测AD的准确率为76.52%。连线测验B/A比值（B = 0.31，SE = 0.13，p = 0.017）和APOE ε4状态（B = 1.07，SE = 0.46，p = 0.019）提高了预测准确性。
MCI中的情景记忆障碍应定义为至少两项测量得分低于正常参考值-1 SD。进行单项测试的临床医生或研究人员应选择更严格的临界值，并收集和分析全脑体积。可行时，确定APOE ε4状态可进一步改善预测。
