In Brief | Ophthalmology | Artificial Intelligence

Retinal age gap as a predictive biomarker for mortality risk

Originally Published in British Journal of Opthalmology, April 2023. This summary does not represent the original research, nor is it intended to replace the original research. Content Disclaimer
 

Traditionally chronological age has been used to determine age-related mortality and morbidity. However, there is increasing acknowledgement that biological rather than chronological age better indicates an individual's health status and the ageing process.^1(pg547) The issue until now has been how best to determine biological age reliably.
 

Researchers Zu et al. aimed to address this challenge by developing a deep-learning model to predict a patient's biological age using fundus imaging. Once the researchers had established the model, they investigated whether there was a correlation between the retinal gap—the difference between the biological age predicted by the model and the patient's chronological age—and an increased mortality rate.
 

The researchers used the UK Biobank study to extract and review 80000 images from 46,969 participants. The data from eleven thousand fifty-two participants (11052) with no prior medical history at the baseline examination were then used to train and validate the DL model for age prediction. 53.7% of the DL model group were female, and the mean age was 52.6 (+/-7.97years). 

The data from the DL model (retinal age) were then compared to the participants' chronological age. A strong correlation of more than 0.81 (p<0·001), with an absolute mean error of 3.55 years, was found between the two, reflecting the model's accuracy. The difference between the two ages was termed the retinal gap: a positive gap represented an 'older' retina and vice versa.
  

The researchers then reviewed the mortality data from the cohort.  At the 11-year follow-up period, 1871 (5.21%) of the participants had died from all causes. Deaths related to cardiovascular disease(CVD) and cancer were identified, and factors such as comorbidities, obesity, smoking and lack of physical activity were included as confounding factors in the analysis.

The researchers compared the DL model determinant of the retinal age gap in these patients with the available mortality data. Using a COX-proportional hazard regression method, they determined that patients in whom the retinal gap was in 3rd and fourth quantile had a significantly higher risk of mortality unrelated to CVD or cancer. Specifically, they determined that for every 1-year increase in the retinal gap, there was an associated 2% increase in all-cause mortality and a 3% increase in cause-specific mortality. They found no association between the retinal gap and mortality due to CVS or cancer.  
 

The authors note that this is the first study to propose a retinal age gap as a biomarker of ageing. Their findings showed a strong correlation between the retinal age, as determined by the DL model, and the participants' chronological age. Furthermore, the researchers concluded that a positive retinal age gap is significantly associated with an increased mortality risk. These findings indicate that retinal imaging may be a viable option for screening for risk stratification and delivery of tailored interventions.  

Access the original article 
Zhu Z, Shi D, Guankai P, et al Retinal age gap as a predictive biomarker for mortality risk. British Journal of Ophthalmology 2023;107:547-554.  https://bjo.bmj.com/content/107/4/547