chatGPT:
Summary
The paper tests whether retinal biological age, estimated from retinal photographs using an AI model called RetiAGE, is associated with osteoporosis risk. RetiAGE is a probability-like score reflecting whether the retina appears biologically “older,” trained to estimate the likelihood that a person is over 65. The rationale is that the retina may capture systemic ageing signals, including vascular, inflammatory, metabolic and lifestyle-related damage.
The authors use two cohorts:
- PIONEER, Singapore: a cross-sectional cohort of 1,965 older adults with both retinal photographs and DEXA bone-density measurements.
- UK Biobank: a prospective cohort of 43,938 participants with retinal photographs and no osteoporosis at baseline, followed for incident osteoporosis over about 12.2 years.
In PIONEER, higher RetiAGE was associated with lower bone mineral density, especially in femoral/hip regions, and lower T-scores. After adjustment for age, sex, calcium intake, diabetes, hypertension, smoking, physical activity and glucocorticoid use, several associations remained statistically significant.
RetiAGE was also associated with osteoporosis diagnosis and fracture-risk estimates. In the adjusted PIONEER analysis, each SD increase in RetiAGE was associated with higher odds of osteoporosis and higher FRAX-derived major osteoporotic fracture and hip fracture scores.
In UK Biobank, higher RetiAGE predicted future osteoporosis. The main adjusted result was HR 1.12 per SD increase in RetiAGE, with a dose-response pattern across quartiles; the highest quartile had HR 1.40 versus the lowest quartile. The association persisted in women and men, with HR 1.10 in women and 1.25 in men in subgroup analyses.
The paper also reports that adding RetiAGE to an osteoporosis self-assessment model improved discrimination for osteoporotic fracture risk, with the C-index rising from 0.585 to 0.635.
Claimed novelty
The main novelty is not that the retina can reflect systemic disease — that is already an active field — but that the authors link an AI-derived retinal ageing biomarker specifically to bone mineral density, osteoporosis diagnosis, fracture-risk scores and future osteoporosis incidence.
More specifically, the novel elements are:
-
Retinal biological age as a skeletal ageing marker
The paper proposes that RetiAGE may capture systemic biological ageing processes relevant to bone loss, not merely ocular ageing. -
Combination of cross-sectional DEXA and longitudinal outcome data
The PIONEER cohort provides direct DEXA-derived BMD and T-score associations, while UK Biobank provides prospective incident osteoporosis prediction. That strengthens the paper compared with a purely cross-sectional retinal-marker study. -
Potential opportunistic screening route
Retinal photography is relatively non-invasive and often already used in eye screening. The authors suggest it might help identify people who should receive further osteoporosis evaluation, especially where DEXA access is limited. -
Incremental value beyond simple osteoporosis risk tools
The claim that RetiAGE improves prediction beyond the Osteoporosis Self-assessment Tool is clinically relevant, though still exploratory. -
Biological plausibility via shared ageing mechanisms
The discussion links retinal ageing and skeletal ageing through inflammation, oxidative stress, microvascular dysfunction, immune regulation and possible genetic links such as IRF4. This is plausible, though not proven mechanistically in this paper.
Critique
The study is interesting and potentially useful, but the clinical claim should be treated cautiously.
The strongest part is the two-cohort design: direct DEXA data in an Asian older cohort plus prospective UK Biobank follow-up. The consistency across different populations makes the association more credible than a single-cohort observation.
However, the effect size is modest. In UK Biobank, the main hazard ratio is 1.12 per SD increase in RetiAGE. That is statistically significant, but not large. It suggests RetiAGE may be a weak-to-moderate risk marker rather than a stand-alone screening tool. Age, sex and BMI remain much stronger predictors.
A second issue is outcome definition. In PIONEER, osteoporosis is DEXA-defined, which is robust. In UK Biobank, incident osteoporosis is based on recorded diagnoses from healthcare data, not systematic repeat DEXA scans. That means the UK Biobank outcome may partly reflect ascertainment bias: people with more medical contact, fractures, imaging, or comorbidities may be more likely to receive an osteoporosis code.
A third concern is confounding. Retinal ageing may reflect smoking, vascular disease, inflammation, socioeconomic status, nutrition, sunlight exposure, physical activity, kidney function, frailty or healthcare access. The authors adjust for several variables, but residual confounding is likely. The paper shows association, not that retinal ageing is mechanistically upstream of bone loss.
There is also a calibration/generalisation issue. RetiAGE was trained in a Korean health-screening population and applied to Singaporean and UK Biobank cohorts. The authors acknowledge that imaging devices, acquisition protocols, ethnicity, pigmentation and cohort characteristics may shift score distributions. That matters if the tool is proposed for real-world screening.
The FRAX analysis is also somewhat circular or at least hard to interpret. FRAX already includes age, sex, weight and femoral-neck BMD. Showing RetiAGE correlates with FRAX risk may partly reflect correlation with age or BMD, rather than independent fracture biology.
The GWAS element is interesting but feels underdeveloped. The reported loci, such as IRF4 and CTNNB1, are biologically suggestive, but the paper does not establish a causal genetic pathway from retinal ageing to osteoporosis. It is more hypothesis-generating than mechanistic proof.
Bottom line
This is a worthwhile biomarker paper. It supports the idea that AI-derived retinal age captures some systemic ageing signal relevant to bone health. The strongest finding is that RetiAGE is associated with lower BMD in DEXA data and prospectively predicts osteoporosis in UK Biobank.
But the clinical implication should be modest: RetiAGE is not a replacement for DEXA, and probably not yet a stand-alone osteoporosis screening test. Its best potential use is as an opportunistic triage marker: someone undergoing retinal imaging for another reason might be flagged for conventional osteoporosis risk assessment or DEXA if their retinal biological age appears high.