OBJECTIVE

To understand the neuropsychological basis of dementia risk among persons in the spectrum including cognitive normality and mild cognitive impairment.

METHODS

We quantitated risk of progression to dementia in elderly persons without dementia from 2 population-based studies, the Framingham Heart Study (FHS) and Mayo Clinic Study of Aging (MCSA), aged 70 to 89 years at enrollment. Baseline cognitive status was defined by performance in 4 domains derived from batteries of neuropsychological tests (that were similar but not identical for FHS and MCSA) at cut scores corresponding to SDs of ≤-0.5, -1, -1.5, and -2 from normative means. Participants were characterized as having no cognitive impairment (reference group), or single or multiple amnestic or nonamnestic profiles at each cut score. Incident dementia over the following 6 years was determined by consensus committee at each study separately.

RESULTS

The pattern of hazard ratios for incident dementia, rates of incident dementia and positive predictive values across cognitive test cut scores, and number of affected domains was similar although not identical across the FHS and MCSA. Dementia risks were higher for amnestic profiles than for nonamnestic profiles, and for multidomain compared with single-domain profiles.

CONCLUSIONS

Cognitive domain subtypes, defined by neuropsychologically derived cut scores and number of low-performing domains, differ substantially in prognosis in a conceptually logical manner that was consistent between FHS and MCSA. Neuropsychological characterization of elderly persons without dementia provides valuable information about prognosis. The heterogeneity of risk of dementia cannot be captured concisely with one test or a single definition or cutpoint.