Neurovascular retinomics for predicting heart diseases.

Area of Science:

• Ophthalmology and Cardiology
• Biomarker Development
• Medical Imaging Analysis

Background:

• Heart disease remains a leading cause of mortality globally.
• Current risk prediction models have limitations in accuracy and accessibility.
• Retinal imaging offers a potential non-invasive window into systemic vascular and neural health.

Purpose of the Study:

• To evaluate retinomics, a composite biomarker derived from retinal imaging, for predicting incident heart disease.
• To assess the performance of retinomics against the established WHO-CVD risk score.
• To identify specific retinomic features associated with heart disease risk.

Main Methods:

• Utilized LASSO regression for feature selection in a large cohort (39,450 participants).
• Employed Gradient Boosting Machine for predictive modeling and Cox regression for association analysis.
• Integrated vascular parameters from colour fundus photography (CFP) and neural parameters from optical coherence tomography (OCT).

Main Results:

• The retinomic model (with age and sex) achieved a C-index of 0.731 for overall prediction and 0.761 for 5-year events.
• Demonstrated superior discrimination compared to the WHO-CVD risk score (C-index 0.736 vs 0.708, p < 0.001).
• Identified 17 retinomic features significantly associated with heart disease risk, independent of traditional risk factors.

Conclusions:

• Neurovascular retinomics provides a powerful, non-invasive tool for heart disease risk prediction and stratification.
• The model's simplicity and low cost make it highly feasible for widespread screening, especially in resource-limited settings.
• Aligns screening intervals with routine eye examinations, enhancing timely cardiovascular risk assessment.