Neuropsychological prediction of decline to dementia in nondemented elderly.

This study examined whether baseline neuropsychological performance in elderly assessed at a research clinic could accurately predict subsequent decline to dementia. Logistic regression analyses were applied to (1) 213 nondemented elderly with a Global Deterioration Scale (GDS) score of 1, 2, or 3, of whom 74 (35%) subsequently declined to any diagnosis of dementia, and (2) a diagnostically more restricted subset of this sample (N = 179), of whom 56 (31%) declined to a diagnosis of probable Alzheimer's disease (AD). The mean follow-up intervals were 3.8 and 3.7 years, respectively. A small set of baseline neuropsychological measures (especially a Paragraph Delayed Recall Test) significantly differentiated decliners from nondecliners to dementia or AD, after accounting for the contribution of age, sex, education, follow-up interval, and the rating of global clinical status. When examined in combination with the other factors or alone, the cognitive tests produced reasonably high specificities (91%-97%) and sensitivities (73%-89%). Using the obtained regression model, a similar level of classification accuracy was replicated on an independent sample of 119 nondemented elderly. A subanalysis of the high-risk GDS 3 subgroup indicated that cut scores from the paragraph test distinguished nondecliners from decliners (overall accuracies 87%-91%), implying that this assessment may accurately predict future cognitive status in elderly with mild cognitive impairment.