A new study suggests that AI-powered retinal imaging may capture systemic aging signals associated with weak bones and provide a potential low-cost route to identify people who need formal osteoporosis evaluation before fractures occur.
Research: Retinal biological age correlates with bone density and fracture risk score and predicts osteoporosis development. Image credit: Steve Bottoms / Shutterstock
In a recent study published in the journal PLOS Digital HealthResearchers evaluated whether RetiAGE, an AI-derived probability score, could help address the limitations of current osteoporosis risk screening approaches. This model uses estimates of retinal biological age as a potential marker of systemic skeletal health.
This model was then used to predict skeletal health status in two large, ethnically distinct cohorts: the Singapore PIONEER study (n = 1,965) and the UK Biobank (n = 43,938). Model output showed that higher retinal biological age was inversely correlated with bone mineral density (BMD) and independently predicted the development of osteoporosis.
These findings suggest that RetiAGE is a non-invasive, low-cost candidate for “opportunistic” screening. This screening suggests that regular visits to the optometrist could ultimately also serve as an early warning system for bone health, helping to identify people who may need a formal osteoporosis evaluation before a fracture occurs.
Background of retinal aging and osteoporosis
Osteoporosis is a systemic skeletal disease characterized by progressive thinning of bone tissue, deterioration of bone microarchitecture, generalized bone loss, and increased risk of life-threatening fractures. According to current public health records, osteoporosis affects approximately 19.7% of the world’s population. However, researchers believe this value may significantly underestimate the true prevalence of the disease.
The clinical “gold standard” for diagnosis is accurate estimation of bone mineral density (BMD) using dual-energy X-ray absorptiometry (DEXA). However, DEXA screening is often limited to high-risk individuals due to high cost, limited access, low-dose radiation exposure, and the need for specialized equipment and personnel. Scientists argue that this leads to significant diagnostic gaps, with many patients discovering symptoms for the first time after a sentinel fracture, while many others’ symptoms remain ‘silent’.
As a result, there is a growing body of research aimed at identifying alternative screening tools that are low-cost and easily accessible, and the retina is a leading candidate. The retinal floor is the only site where the microvasculature and neural tissue can be directly and non-invasively imaged. Previous studies have linked retinal aging gaps to cardiovascular disease (CVD), Parkinson’s disease (PD), and chronic kidney disease (CKD).
RetiAGE osteoporosis research design
This study aims to investigate whether common biological processes link retinal degeneration and skeletal bone loss, allowing non-invasive assessment of the former to serve as a predictive surrogate for general condition, biological age, for the latter. Biological aging is a heterogeneous process that captures the degeneration of cells or tissues and often deviates from an individual’s chronological age or chronological age.
This study used a deep learning model to derive RetiAGE, a continuous probability score that estimates the likelihood that an individual is biologically 65 years of age or older. The model was developed using the Visual Geometry Group’s (VGG16) convolutional neural network (CNN) architecture and trained on 129,236 retinal images from 40,480 participants.
After training the model, we used RetiAGE to analyze data from two different populations. One was cross-sectional data (n = 1,965, mean age = 72.5) from the Singapore Population Health and Eye Diseases in Older Adults PROfilE (PIONEER) study (n = 1,965, mean age = 72.5), and prospective data from the UK Biobank (43,938 participants without baseline osteoporosis, mean age = 56.2, mean follow-up = 12.2). year).
The primary outcomes of this study include BMD T-score and 10-year fracture risk score calculated via the Fracture Risk Assessment Tool (FRAX). The study also conducted a genome-wide association study (GWAS) on 45,496 participants to identify the genetic drivers of RetiAGE.
Findings on retinal age and bone health
RetiAGE results show that the higher an individual’s retinal biological age score, compared to chronological and clinical risk factors, the weaker their bones tend to be.
The Singapore PIONEER study found that elevated RetiAGE scores were inversely associated with BMD and T scores in several thigh and hip regions, with some associations remaining significant even after adjusting for osteoporosis risk factors. Furthermore, for each standard deviation (SD) increase in retinal age, the risk score for major osteoporotic fractures increased by 0.48 and the risk score for proximal femoral fractures increased by 0.29.
In the UK Biobank longitudinal cohort, participants with older retinal age were significantly more likely to develop osteoporosis in the next 10 years, with a hazard ratio (HR) of 1.12 per SD (p < 0.001). Dividing the data into quartiles showed that those in the highest retinal age group had a 40% higher risk of osteoporosis than those in the lowest quartile (p = 0.003).
Additionally, the addition of RetiAGE to the traditional osteoporosis self-assessment tool (OST) significantly improved diagnostic performance. The concordance index (C-index) increased from 0.585 to 0.635, and the 10-year net reclassification index (NRI) was 2.5%.
Clinical significance of AI retinal screening
This study demonstrates that accelerated biological aging of the retina is an independent risk marker for loss of bone density and development of osteoporosis across multiple ethnicities. These findings suggest that retinal imaging may provide a scalable, low-cost, non-invasive method for opportunistic screening of osteoporosis risk in ophthalmology or primary care settings, but await further validation.
While DEXA remains the diagnostic standard, AI-driven retinal markers such as RetiAGE capture unique systemic aging signals and provide graded prognostic value beyond traditional demographic risk factors. However, the authors note that because RetiAGE was originally developed in a Korean population and applied without population-specific retraining, further calibration may be required across imaging devices, ethnic groups, and real-world clinical workflows before routine use.
Click here to download your PDF copy.
