BACKGROUND AND PURPOSE: Cerebral atrophy, detected as ventricular enlargement or sulcal widening on MRI, is recognized as a risk factor for vascular dementia or Alzheimer disease. However, its underlying pathophysiology is not known. We examined whether retinal microvascular assessment could provide predictive information on the risk of ventricular enlargement and sulcal widening on MRI. METHODS: A prospective, population-based study was conducted of 810 middle-aged persons without clinical stroke or MRI infarcts. All participants had a first cranial MRI and retinal photography in 1993 to 1995 and returned for a repeated MRI in 2004 to 2006 (median follow-up of 10.5 years). Retinal photographs were graded for presence of retinopathy and retinal microvascular abnormalities, and MRI images were graded for ventricular size and sulcal size according to standardized protocols. Ventricular enlargement and sulcal widening were defined as an increase in ventricular size or sulcal size of >or=3 of 10 grades between baseline and follow-up. RESULTS: After adjusting for age, gender, and cardiovascular risk factors, retinopathy and arteriovenous nicking at baseline were associated with 10-year ventricular enlargement (OR and 95% CI: 2.03, 1.20 to 4.42 for retinopathy and 2.19, 1.23 to 3.90 for arteriovenous nicking). Retinal signs were not associated with 10-year sulcal widening. CONCLUSIONS: Retinopathy and arteriovenous nicking are predictive of long-term risk of ventricular enlargement, but not of sulcal widening, independent of cardiovascular risk factors. These data support a microvascular etiology for subcortical but not cortical cerebral atrophy.