Some people appear to age slower (or faster) than others, exhibiting vastly different age-related physical changes and disease risks. Recognition of this biological phenomenon has given rise to the concept of biological age – a measure of a person’s physiological and functional state. Scientists use a variety of means to gauge biological age, including methylation markers, gray matter volume, and facial aging. Findings from a recent study suggest that retinas provide useful biomarkers in determining a person’s biological age.

The retina is a thin, multicellular layer lining the rear, interior portion of the eye. It plays critical roles in the cascade of events involved in visual processing, converting the energy of photons of the visible light spectrum into biochemical signals and transmitting those signals to the brain. Poor retinal health is often an indicator of systemic illness, such as cardiovascular disease or nutritional deficiency.

The authors of the study viewed more than 80,000 retinal images collected from adults (average age, 55 years) participating in the UK Biobank Study. They also collected information about the participants' demographics, lifestyles, and overall health. The researchers used deep learning, a type of machine learning that mimics the way humans learn, to analyze images of the retinas and assign a biological age, which they referred to as “retinal age.” Then they calculated the retinal age gap – the difference between retinal age and chronological age. Having a positive retinal age gap was reflective of an older-appearing retina; having a negative retinal age gap was reflective of a younger-appearing retina. Finally, the researchers looked at links between retinal age gap and all causes of premature death.

They found that their machine learning model accurately predicted retinal age and chronological age to within 3.5 years. For every year of positive retinal age gap difference, the risk of premature death from any cause increased 2 percent. Having positive retinal age gaps greater than three years increased the risk of premature death from specific diseases (other than cardiovascular disease or cancer) by as much as 67 percent. These findings held true even after taking other factors into account, such as body weight, high blood pressure, or smoking.

These findings suggest that retinal age, as predicted via deep learning, is a powerful predictor of biological age and premature death risk. Collecting retinal images is a low-cost, non-invasive procedure that may be beneficial in identifying people at risk for premature disease and death. Learn about other strategies for predicting biological age in our overview article on epigenetic aging clocks.

• Link to study abstract.