NEW YORK (Reuters Health) – The difference between the biological age of a person’s retina and their actual age – the retinal age gap – may help estimate lifespan, a new study suggests.
Using data from more than 40,000 participants in the UK Biobank, researchers trained computers to predict life expectancy from retinal images and actual ages, they report in the British Journal of Ophthalmology.
“I would say the take-home message is: your eyes can tell your aging process,” said the study’s first author, Dr. Lisa-Zhuoting Zhu of the Centre for Eye Research Australia, in East Melbourne.
“The retina, the back of the eye, is a window to the vessel and neural systems of the body,” Dr. Zhu told Reuters Health by email. “Damage to the vascular system (for example, heart attack) and neural system (for example, stroke) would not only increase the risk of death but also leave marks on the retinal vessels and neurons. These aging footprints in the retina could be captured by an Artificial Intelligence model.”
To explore what the health of the retina might tell us about future lifespan, Dr. Zhu and her colleagues turned to the UK Biobank, a large-scale, population-based cohort of more than 500,000 UK residents aged 40 to 69. Detailed measurements were made and biological samples collected from all participants, who were also required to fill out comprehensive health questionnaires. The Biobank also collected paired retinal fundus and optical coherence tomography images for all participants.
The researchers included more than 80,000 fundus images taken from nearly 47,000 Biobank participants in their analysis. To train a computer to recognize possible associations between a person’s retinal age and their actual age, the team used data from 11,052 relatively healthy participants (19,200 images) with a mean age of 52.
Nearly all of the remaining participants had mortality data available and their information was used to explore the association between the “retinal age gap” and mortality. There was a strong correlation of 0.81 (P<0.001) between retinal age and chronological age, with a mean absolute error of 3.55 years.
When the team ran Cox regression models, they found that each one-year increase in the retinal age gap was associated with a 2% increase in the risk of all-cause mortality (hazard ratio, 1.02; P=0.020) and a 3% increase in non-cardiovascular and non-cancer mortality (HR, 1.03; P=0.041) after multivariable adjustments.
The researchers did not find any association between the retinal age gap and cardiovascular- or cancer-related mortality.
The new study is “very interesting,” said Dr. Jay Chhablani, an associate professor of ophthalmology at the University of Pittsburgh School of Medicine, in Pennsylvania, who was not involved in the study.
“The uniqueness of the study is to report non-(cardiovascular-disease)/non-cancer-related mortality in ‘healthy aging’ using retinal images,” Dr. Chhablani told Reuters Health by email. “‘Retinal age gap’ could be a biomarker for predicting mortality. However, similar studies need to be (performed) in different ethnic populations, along with the stringent quality check on retinal images. Applying this biomarker in addition to other biomarkers might expand its role in predicting mortality in the diseased population.”
Other experts were more critical of the study.
“It is well known that correlations of massive datasets can have good P-values, but may not mean so much,” said Dr. R. Theodore Smith, a professor of ophthalmology and neuroscience at the Icahn School of Medicine at Mount Sinai and an ophthalmologist at New York Eye and Ear Infirmary of Mount Sinai.
“The cloud of data in Figure 2 is a typical result,” Dr. Smith told Reuters Health by email. “The whole cloud follows or swarms along a straight line (high correlation, good P-value), but simply means that retinal age goes up with calendar age. Overall, big deal. However, the points on the y axis reliably fall only in a 10-year interval for each x.”
The key question, Dr. Smith said, is whether the model can predict calendar age with any accuracy. “That was not tested,” he added. “The answer is obviously, NO.”
SOURCE: https://bit.ly/3nQl8L1 The British Journal of Ophthalmology, online January 18, 2022.
Source: Read Full Article