Abstract

INTRODUCTION: Retinal vasculometry (RV) provides a neurovascular biomarker which may relate to cognitive status. However, the presence and form of association remains unclear and unexamined at scale. METHODS: Artificial intelligence-enabled RV measures from 66,350 UK Biobank study participants were related to combined cognition scores. Differences in RV were examined per standard deviation (SD) increase in cognitive score, using multilevel linear regression, adjusted for age, sex, measurement center, ethnicity, and within-person RV clustering. RESULTS: One hundred ten thousand two hundred eighty-two retinal images from 63,165 (95%) participants (mean age 56.6 years, 55.5% female) were analyzed. A one SD increase in cognition score was strongly associated with increased arteriolar width, arteriolar tortuosity, increased venular width particularly among those < 50 years and venular area among those > 50 years; also, inversely associated with venular tortuosity, and arteriolar and venular width variance. DISCUSSION: These easily accessible, affordable, and non-invasive RV measures should be evaluated further as an early predictor of future neurodegenerative disease. HIGHLIGHTS: How cognitive status relates to retinal vasculometry (RV) measures remains uncertain and unexamined at scale.Using data from a large population-based study (UK Biobank) we show strong graded associations between cognitive status and RV, which contrast with some RV associations observed with aging. Specifically, increased arteriolar tortuosity, arteriolar and venular width (at younger ages), and area are positively associated, and venular tortuosity and arteriolar and venular width variability are inversely associated with higher cognitive status, all showing strong, graded, precise relationships. These associations appeared to be strongest for fluid intelligence and prospective memory tests.These easily accessible, non-invasive RV measures provide a neurovascular marker indicative of cognitive status, which should be evaluated as early predictors of neurodegenerative disease.