A new study published in PLOS Digital Health suggests that an AI-derived retinal “biological age” marker correlates with reduced bone mineral density (BMD), elevated fracture risk, and future osteoporosis development.
The international research team, led by investigators from the Singapore Eye Research Institute and the National University of Singapore, evaluated whether retinal aging could serve as a non-invasive biomarker for skeletal health.
Using a deep learning algorithm known as “RetiAGE,” the researchers analyzed retinal photographs from two major cohorts: the Singapore-based PIONEER study and the UK Biobank. The algorithm estimates the probability that an individual is biologically older than 65 years based solely on fundus images.
In the cross-sectional PIONEER cohort, which included 1,965 older adults who underwent both retinal imaging and DEXA (Dual-energy X-ray Absorptiometry) scanning, higher RetiAGE scores were consistently associated with lower BMD and worse T-scores across multiple femoral regions. The associations remained significant even after adjusting for conventional osteoporosis risk factors, including age, sex, diabetes, smoking, physical activity, BMI, and glucocorticoid use.
Notably, participants with accelerated retinal aging also demonstrated higher major osteoporotic fracture and hip fracture risk scores, as calculated using FRAX (Fracture Risk Assessment Tool). The investigators reported that each standard deviation increase in RetiAGE was associated with significantly increased fracture risk metrics.
The longitudinal findings from the UK Biobank cohort strengthened the association further. Among 43,938 participants without osteoporosis at baseline, 1,492 individuals developed osteoporosis during a mean follow-up period of 12.2 years. Higher retinal biological age independently predicted incident osteoporosis, even after multivariable adjustment.
Participants in the highest RetiAGE quartile had a 40 percent higher risk of developing osteoporosis compared with those in the lowest quartile. Importantly, the relationship persisted in subgroup analyses for both women and men, and remained robust after excluding participants with ocular disease that could potentially alter retinal appearance.
The study authors argue that retinal imaging may provide an attractive opportunity for opportunistic osteoporosis screening. While DEXA remains the current gold standard for diagnosis, access limitations and cost mean that many patients are only diagnosed after already sustaining fractures. Retinal photography, by contrast, is rapid, non-invasive, and increasingly widespread in both ophthalmic and primary care settings.
The retina’s growing role as a “window” into systemic health is also reinforced by previous work linking retinal aging biomarkers to cardiovascular disease, stroke, Parkinson’s disease, and chronic kidney disease. The study authors suggest osteoporosis may now be added to that expanding list.
Still, the investigators acknowledge important limitations with their study. The RetiAGE algorithm was originally trained in a Korean population, and calibration differences across ethnicities and imaging systems may affect generalizability. The study also focused on association rather than real-world screening implementation.
Nonetheless, the findings add to mounting evidence that AI-based retinal biomarkers could eventually support broader systemic disease detection beyond ophthalmology itself – potentially transforming the humble fundus photograph into a multisystem screening tool for a wide variety of diseases.
