P339 - RETINAL VASCULAR COMPLEXITY AS A PREDICTOR OF INCIDENT CARDIOVASCULAR EVENTS: A REPORT FROM THE CANADIAN LONGITUDINAL STUDY ON AGING

Background: Retinal fractal dimension (Df), a measure of vascular complexity derived from retinal imaging via deep learning, reflects systemic microvascular health and may serve as a non-invasive biomarker for cardiovascular disease (CVD) risk. This study examined the association between retinal Df and incident CVD, and explored potential underlying biological pathways using metabolomic profiling.

METHODS AND RESULTS: Our analyses included 30,097 participants (mean age 62.96 ± 10.25 years; 51% female) from the comprehensive cohort of the Canadian Longitudinal Study on Aging (CLSA), followed for up to 6 years. High-resolution retinal images were acquired using digital fundus photography, and three Df measures (total, arterial, and venular) were quantified from 50,957 images using Vessel Assessment and Measurement Platform for Images of the Retina (VAMPIRE) software. A higher Df (given per 1 standard deviation (SD)) indicates higher vascular branching. A subset of 9,785 participants also had serum levels of 812 common metabolites (i.e., amino acids, lipids, carbohydrates, choline, polyamines, cofactors, and vitamins) measured via Metabolon. We used Cox regression to assess the associations between Df measures and incident CVD, defined as myocardial infarction, angina, peripheral vascular disease, or stroke. Linear regression evaluated associations between Df and metabolite levels, adjusted for age and sex. Predictive performance of Df measures was compared to a multivariable clinical model including age, sex, ethnicity, smoking, body mass index, waist-hip ratio, hypertension, diabetes, dyslipidemia, using area under the receiver operating characteristic curve (AUC-ROC).

During follow-up, 2,870 participants (9.54%) experienced a cardiovascular event. Higher Df values were significantly associated with reduced CVD risk: total Df (unadjusted HR per 1 SD Df increase = 0.61; 95% CI: 0.52–0.70), arterial Df (HR = 0.75; 95% CI: 0.68–0.83), and venular Df (HR = 0.71; 95% CI: 0.64–0.78). Thirty-seven metabolites were nominally associated with incident CVD, two of which, 1-palmitoyl-2-linoleoyl-GPI (16:0/18:2), 1-linolenoyl-GPC (18:3), were also linked to arterial Df. Higher levels of these two metabolites were associated with higher arterial Df and protective effects on CVD risk. These metabolites are involved in lipid-related pathways including phosphatidylinositol and lysophospholipid metabolism that may regulate inflammation and endothelium function. Finally, the AUC for the Df-based prediction model for CVD was 0.58, compared to 0.68 for the clinical model.

Conclusion: Retinal vascular complexity, as measured by fractal dimension, is significantly associated with incident cardiovascular events but provides limited predictive value beyond established clinical risk factors. Nonetheless, Df analysis highlights novel lipid metabolic pathways as potential mediators between microvascular changes and cardiovascular disease risk.