Objectives

Our aims were to 1) detect the occurrence of three mild cognitive impairment (MCI) subtypes in the general population; 2) identify cases of cognitive impairment which are not detected by current operational criteria for MCI and; 3) determine the predictive value of the subtypes for identifying future Alzheimer disease (AD).

Design

Three-year prospective study.

Setting

Population-based Swedish study, the Kungsholmen Project.

Participants

Three hundred seventy-nine nondemented older adults aged 75-95.

Measurements

Standard plus modified MCI criteria were applied at baseline. In the modified definitions, the requirement for normal general cognition was removed. A category for persons without MCI who had only global cognitive deficits was added. Three-year progression to AD was assessed (DSM-III-R criteria).

Results

Occurrence per 100 nondemented persons of MCI-amnestic, MCI-multidomains, and MCI-single-nonmemory was 2.1%, 1.8%, and 7.2%, respectively. When applying modified definitions for MCI-amnestic and MCI-multidomains, the occurrence almost doubled. Seven percent of the sample had impairment on a global cognitive task but performed at normal levels on all other domain-specific tasks. MCI-multidomains showed the highest progression to AD (hazard ratio [HR]: 23.6, 9.3-60.1). MCI-amnestic reached similar predictivity only when using the modified definition (HR: 17.9, 6.8-46.9). Even participants without MCI who had only global deficits had a ninefold risk of AD (HR: 9.1, 2.8-29.4).

Conclusions

Two-thirds of MCI-multidomains, but only half of MCI-amnestic progress to AD. The standard MCI criteria failed to identify those people with global cognitive deficits who have, however, a high risk of progressing to AD.