Incorporating Retinal Vascular Parameters Improves Stroke Risk Prediction

A study published in Heart found that retinal vascular analysis is superior at predicting stroke risk than traditional models.

The retinal vascular network has similar anatomical and physiological features as that of the brain, making it a potential noninvasive method for assessing vascular health of the brain.

Investigators from The Royal Victorian Eye and Ear Hospital in Australia and The Hong Kong Polytechnic University in Hong Kong sourced data for this study from the United Kingdom (UK) Biobank. Participants (N=45,161) who received an eye examination including fundus imaging were evaluated for incident stroke on the basis of retinal vascular parameters. To obtain retinal vascular parameters, fundus images were analyzed using the deep-learning model Retina-based Microvascular Health Assessment System (RMHAS).

The study population comprised 54.9% women, they had a mean age of 55.4±8.18 years, 91.2% were White, they had a BMI of 27.1±4.67, and 3.9% had diabetes.

During a median follow-up of 12.5 years, 749 incident strokes occurred. Individuals who had stroke were significantly older, more were men, they had higher BMI, fewer were never smokers, they had higher blood pressure, lower high-density lipoprotein cholesterol, and more had diabetes than individuals without stroke (all P <.001).

Stroke risk was increased with every 1-SD increase in length diameter ratio of arteries and vessels in the macular region (adjusted hazard ratio [aHR] range, 1.101-1.141; both P ≤.03), chord length of arteries and vessels in the macular region (aHR range, 1.122-1.134; both P =.01), and arc length of the arteries and vessels in the macular region and artery nonterminal and terminal points (aHR range, 1.098-1.144; all P ≤.03).

Conversely, stroke risk was increased with every 1-SD decrease in the following:

• Central retinal artery equivalent (aHR, 0.906; P =.04)
• Fractal dimension of the arteries and vessels (aHR range, 0.862-0.890; both P =.01)
• Number of arterial bifurcation, branching, and nonterminal points (HR range, 0.862-0.868; all P =.01)
• Number of arterial segments (aHR, 0.856; P =.01)
• Number of vessel segments in the macular region (aHR, 0.906; P =.04)
• Number of arterial terminal points (aHR, 0.851; P =.01)
• Artery bifurcation density (aHR, 0.903; P =.03)
• Branching density of the arteries, veins, and vessels in the macular region (aHR range, 0.892-0.910; all P £.04)
• Vessel area density of the vessels and arteries (aHR range, 0.860-0.896; all P ≤.02)
• Vessel skeleton density of the arteries (aHR range, 0.840-0.872; all P ≤.01)
• Inflection count tortuosity of the arteries (aHR, 0.908; P =.03)

Using traditional risk factors, stroke risk was predicted with an area under the receiver operating characteristic (AUROC) curve of 0.738. The predictive model was improved when traditional risk factors were combined with retinal vessel parameters (AUROC, 0.752; P <.001).

The major limitation of this study was the lack of diversity in the UK Biobank.

The study authors concluded, “…our study showed that this set of comprehensive retinal vascular parameters was of added predictive value for incident stroke, indicating its potential application as a non-invasive screening method for individuals with increased risk.”

This article originally appeared on The Cardiology Advisor