Objective: To compare normative and empirical definitions of memory impairment in MCI. In clinical and research applications, the concept of Amnestic Mild Cognitive Impairment (aMCI) is used to identify individuals who do not meet criteria for dementia but who will develop diagnosable dementia at high rates over the next several years. A critical feature of aMCI is the presence of objective memory loss, defined by performance at or below a 1.5 standard deviation cut score for age and education matched controls on a memory test without reference to a specific memory test. Though this concept has been tremendously useful, we discuss some limitations of aMCI and propose to modify the construct. Because of the 1.5 standard deviation cut-score, the upper bound on the prevalence of aMCI at any age is 6.7 %. At age 65, where the prevalence of dementia is 1%, this definition identifies many individuals who do not develop dementia and has low specificity. At age 85, where the prevalence of dementia is 16% this definition misses many individuals who will develop dementia and has low sensitivity. These problems arise because memory impairment is defined solely based on the distribution of scores in the control population.
Design: We studied a systematic sample of 629 initially non-demented individuals recruited from the Medicare roles for the Bronx as part of the Einstein Aging Study (EAS). Using baseline testing, we compared the predictive validity of this empirical cut-score with a 1.5 standard deviation cut-score according to the definition of aMCI.
Materials and Methods: We used an empirical cut-score for memory impairment, based on its ability to predict the development of dementia over several years of follow-up. We selected the Free and Cued Selective Reminding because its controlled learning procedures reduce the influence of cognitive impairment in domains other than memory. An empirical cut-score of 24 on the Sum of Free Recall measure has been found to optimally predict incident dementia in two other independent samples.
Results: Over two years of follow-up 57 (8%) individuals developed DSM IV defined dementia. The aMCI cut predicted dementia with a sensitivity of 25% and specificity of 97%. The empirical cut-score predicted dementia with a sensitivity of 56% and specificity of 84%. Using a less restrictive definition of dementia, the sensitivity of the aMCI cut score fell to 15% while performance of the empirical cut-score was unchanged.
Conclusions: Though aMCI has lead to tremendous progress in dementia prevention, based both on theoretical grounds and practical results, an empirically derived cut-score may have important advantages. In several samples, we have seen that aMCI misses individuals who develop dementia and that this problem increases with increasing age. Using empirical cut-scores makes intuitive sense and improves the sensitive targeting of individuals for preventive intervention.
Back to S097 Age Associated Memory Impairment and Mild Cognitive Impairment: What is Our Next Direction?
Back to The Eleventh International Congress