Objective:
To assess whether AI-estimated retinal age from fundus photographs reflects biological aging and is associated with health outcomes, including sex-specific patterns.
Approach:
- Age prediction: Researchers fine-tuned RETFound using 130,360 fundus images from 71,343 UK Biobank participants to predict chronological age.
- Retinal age gap: They defined the retinal age gap (RAG) as predicted retinal age minus chronological age and corrected it for regression to the mean.
- Health and sex analyses: They examined associations with health indicators and diseases, followed outcomes for approximately 15 years, and assessed sex-specific patterns, model features, and genetic associations.
Key Findings:
- Participant-level age predictions had a mean absolute error of 2.85 years and correlated with chronological age at 0.90.
- Higher RAG was associated with adverse health indicators and conditions including cataract, glaucoma, macular degeneration, diabetes, metabolic syndrome, cancer, and hypertension.
- Over follow-up, higher RAG was associated with all-cause mortality and incident cardiovascular disease, COPD, stroke, dementia, thrombosis, ischemic heart disease, and cancer after adjustment for established risk factors and prior disease events.
- RAG ranked among the five largest contributors to prediction models for mortality, COPD, stroke, dementia, and cancer.
- Associations differed by sex: metabolic risk-factor relationships were generally stronger in males; postmenopausal women had higher RAG than premenopausal women, with cardiometabolic and inflammatory associations more similar to those in men.
- Model maps highlighted the macula, temporal optic disc, and retinal vasculature. Genetic analyses identified loci related to metabolism, longevity, retinal vascular traits, neurodegeneration, and age-related eye disease.
Interpretation:
The authors report that fundus photographs contain information associated with systemic health and age-related disease, with patterns that differ by sex. They do not present RAG as a ready-made clinical risk score.
Limitations:
- Both study cohorts were predominantly White European and relatively healthy, limiting generalizability.
- The authors note that RAG may reflect normal aging, disease-related retinal changes, and non-aging image characteristics.
Conclusion:
The study suggests that AI-derived retinal age gap is associated with multiple health outcomes and sex-specific patterns, but further validation is needed before clinical use.
Sources:
This content is an AI-generated, fully rewritten summary based on a published scholarly article. It does not reproduce the original text and is not a substitute for the original publication. Readers are encouraged to consult the source for full context, data, and methodology.