Importance: High-risk vascular diseases (HRVs) may remain undetected until catastrophe ensues. Detection from non-invasive retinal imaging would be highly significant.
Objective: To demonstrate that certain lesions of Age-Related Macular Degeneration (AMD) found on retinal imaging correlate with co-existing HRVs.
Design: Cross-sectional cohort study. Two years. Retinal image graders blinded to HRV status.
Setting: 2 retina referral clinics.
Participants: 151 consecutive AMD patients, ages 50-90, 97 females, 54 males, with lesions of soft drusen and/or subretinal drusenoid deposits (SDD). 12 others approached, 10 refused, 2 excluded.
Methods: Patients were classified by retinal imaging into SDD (SDD present, +/- drusen) or nonSDD (soft drusen only), and by history into HRV (cardiac pump defect (myocardial infarction (MI), coronary artery bypass grafting (CABG), congestive heart failure (CHF)), valve defect, and carotid stroke) or nonHRV, with serum risk factors and medical histories.
Main Outcome Measures: Correlations of HRV with SDD and other covariates (Univariate chi-square and multivariate regression). Performance of Machine Learning predicting HRV.
Results: 75 SDD subjects; 76 nonSDD subjects; HRV prevalence 19.2% (29/151).
1. High density lipoprotein (HDL) < 62 mg/Dl was found in 24/29 HRV, 42/122 nonHRV, OR 12.40, 95% Confidence Interval (CI) 5.125-30.014; p= 0.0002.
2. 15 Pump defects, 14/15 SDD, 8 Valve defects, 6/8 SDD (4 severe aortic stenosis), 6 carotid strokes, 5/6 SDD. Total HRVs 29, 25/29 SDD, OR 9.0, 95% CI 2.95-27.46; p= 0.000012.
3. Adjusted multivariate correlations. HRV with SDD (p= 0.000333). SDD and HDL < 62 with HRV (p= 0.000098 and 0.021).
4. Machine Learning prediction of HRVs from SDD status and HDL level: specificity 87.4%, sensitivity 77.4%, accuracy 84.9%; 95% CIs(%) 79.0-93.3, 58.0-90.4, 77.5-90.7, respectively.
Conclusions and Relevance: High-risk vascular diseases were accurately identified in a cohort of AMD patients from the presence of characteristic deposits (SDDs) on imaging and HDL levels. The SDDs are directly consequent to inadequate ocular perfusion resulting from the systemic vasculopathies. Further validation in larger cohorts of both vasculopathic and AMD subjects could bring this system into widespread medical practice, to reduce mortality and morbidity from vascular disease, particularly in women, where undiagnosed cardiac disease remains a serious issue.