Background and Purpose—We sought to determine the predictive value of magnetic resonance imaging measures of vascular disease (white matter hyperintensities [WMHs], lacunes, microbleeds, and infarcts) compared with atrophy on the progression of mild cognitive impairment to dementia.

Methods—We included 152 consecutive patients with mild cognitive impairment. Baseline magnetic resonance imaging was used to determine the presence of medial temporal lobe atrophy and vascular disease (presence of lacunes, microbleeds, and infarcts was determined, and WMHs were rated on a semiquantitative scale). Patients were followed up for 2±1 years.

Results—Seventy-two (47%) patients progressed to dementia during follow-up. Of these, 56 (37%) patients were diagnosed with Alzheimer's disease, and 16 (10%) patients were diagnosed with a non-Alzheimer dementia (including vascular dementia, frontotemporal lobar degeneration, and Parkinson dementia). Converters were older and had a lower Mini-Mental State Examination score at baseline. On baseline magnetic resonance imaging, patients who progressed to a non-Alzheimer dementia showed more severe WMHs and had a higher prevalence of lacunes in the basal ganglia and microbleeds compared with nonconverters. Cox proportional-hazard models showed that, adjusted for age and sex, baseline medial temporal lobe atrophy (hazard ratio=2.9; 95% CI, 1.7 to 5.3), but not vascular disease, was associated with progression to Alzheimer's disease. By contrast, deep WMHs (hazard ratio=5.7; 95% CI, 1.2 to 26.7) and periventricular hyperintensities (hazard ratio=6.5; 95% CI, 1.4 to 29.8) predicted progression to non-Alzheimer dementia. Furthermore, microbleeds (hazard ratio=2.6; 95% CI, 0.9 to 7.5) yielded a >2-fold increased, though nonsignificant, risk of non-Alzheimer dementia.

Conclusion—Medial temporal lobe atrophy and markers of cerebrovascular disease predict the development of different types of dementia in mild cognitive impairment patients.