1426-P: Comparative Performance of a Retinal Imaging–Derived Biomarker and Traditional ASCVD Risk Scores in a Diabetic Cohort from a Community Endocrine Practice in Long Island City, New York

Introduction and Objective: Adults with Diabetes have an increased risk of atherosclerotic cardiovascular disease (ASCVD). Because retinal blood vessels can be visualized non-invasively, retinal imaging offers a unique approach to cardiovascular disease (CVD) risk assessment in diabetes. AI-based tools, such as the Dr. Noon CVD platform, enable automated cardiovascular risk estimation from retinal images. Our objective was to evaluate how retinal image-derived biomarkers align with established ASCVD risk prediction tools in diabetic populations. Methods: In a cohort of 50 adults with diabetes, CVD risk was assessed using a continuous retinal imaging-derived biomarker generated by Dr. Noon CVD and compared with three ASCVD risk tools: Framingham, Pooled Cohort Equations (PCE), and PREVENT. Risk was analyzed in continuous and categorical forms, with borderline, intermediate, and high risk classified as non-low risk. Associations were evaluated using unadjusted logistic regression. Discrimination was assessed using receiver operating characteristic (ROC) analysis with area under the curve (AUC) and DeLong confidence intervals. Results: The retinal biomarker showed substantial variability (mean 0.53 ± 0.27). A high proportion of individuals were classified as non-low risk by all calculators, most frequently by Framingham. Higher retinal biomarker values were associated with increased odds of non-low CVD risk classification by PCE (OR 1.12; 95% CI 1.01-1.26) and PREVENT (OR 1.12; 95% CI 1.00-1.25), but not Framingham. ROC analysis showed acceptable discrimination for PCE (AUC 0.71; 95% CI 0.50-0.92) and PREVENT (AUC 0.70; 95% CI 0.49-0.91), with a sensitivity-favoring profile. Conclusion: In this diabetic cohort, Dr. Noon CVD showed moderate alignment with contemporary ASCVD risk models, with strongest concordance for PCE and PREVENT. Therefore, retinal imaging may provide complementary information beyond traditional risk calculators.