Departments of *Psychiatry †Neurology §Medicine, University of Pittsburgh School of Medicine Departments of ‡Epidemiology ∥Biostatistics, University of Pittsburgh Graduate School of Public Health, Pittsburgh, PA.
Alzheimer Dis Assoc Disord. 2014 Apr-Jun;28(2):113-21. doi: 10.1097/WAD.0000000000000018.
The International Working Group (IWG) criteria for mild cognitive impairment have variable utility in predicting progression to dementia, partly depending on the setting. We explored an empiric approach to optimize the criteria and cutoff points in a population study.
In a cohort of adults aged 65 years or older, we identified 1129 individuals with normal or only mildly impaired cognition by cognitive classification, and 1146 individuals without dementia (Clinical Dementia Rating <1). Operationally defining the IWG criterion set, we examined its sensitivity and specificity for the development of severe cognitive impairment and dementia (Clinical Dementia Rating ≥1) over 4 years. We then disaggregated the criteria and used Classification and Regression Tree analyses to identify the optimal predictive model.
The operational IWG criteria had 49% sensitivity and 86% specificity for the outcome of severe cognitive impairment, and 40% sensitivity and 84% specificity for the outcome of dementia. Classification and Regression Tree modeling improved sensitivity to 82% for the cognitive outcome and 76% for the dementia outcome; specificity remained high. Memory scores were the most important predictors for both outcomes. The optimal cutoff points were around 1.0 SD below the age-education mean. The best fit was observed when prediction was modeled separately for each age-education group.
Objective cognitive measurements contributed more to the prediction of dementia than subjective and functional measures. Those with less education only required memory testing, whereas those with more education required assessment of several cognitive domains. In cases in which only overall norms are available, the appropriate threshold will vary according to the individual's age and education.
国际工作组（IWG）的轻度认知障碍标准在预测向痴呆进展方面的实用性各不相同，部分原因取决于环境。我们探索了一种经验方法来优化人群研究中的标准和截止值。
在一项 65 岁或以上成年人的队列研究中，我们通过认知分类确定了 1129 名认知正常或仅有轻度认知障碍的个体，以及 1146 名无痴呆（临床痴呆评定<1）的个体。根据操作性定义的 IWG 标准集，我们检查了其在 4 年内发展为严重认知障碍和痴呆（临床痴呆评定≥1）的敏感性和特异性。然后，我们对标准进行了细分，并使用分类和回归树分析来确定最佳预测模型。
操作性 IWG 标准对严重认知障碍的发生具有 49%的敏感性和 86%的特异性，对痴呆的发生具有 40%的敏感性和 84%的特异性。分类和回归树建模将敏感性提高到认知结果的 82%和痴呆结果的 76%；特异性仍然很高。记忆评分是这两种结果的最重要预测因素。最佳截止值约为年龄教育平均值的 1.0 个标准差以下。当按每个年龄教育组分别对预测进行建模时，观察到最佳拟合。
客观认知测量对痴呆的预测比主观和功能测量更有价值。受教育程度较低的人只需要进行记忆测试，而受教育程度较高的人需要评估几个认知领域。在仅提供总体规范的情况下，适当的阈值将根据个人的年龄和教育程度而有所不同。
