Retinal age gap, or the difference between retinal age predicted by a deep learning (DL) model and chronological age, may be a biomarker that can predict the risk of patient mortality, according to a study published in the British Journal of Ophthalmology.
Researchers conducted a study to assess whether DL can utilize retinal fundus imaging to identify biological age. They analyzed retinal fundus and optical coherence tomography (OCT) imaging data consisting of 80,169 images from 46,969 participants.
A subset of individuals who were considered healthy at baseline (n=11,052, mean age 52.6±7.97 years, 53.7% women, 19,200 fundus images) comprised the dataset on which the DL model predicted chronological age. Mean absolute error (MAE) was 3.55 years.
Remaining participants (n=35,913) with mortality data formed the analysis dataset. “Retinal age gap,” the difference between predicted retinal age and chronological age, was positive if it appeared “older.”
The mean retinal age gap was -1.31±4.82. Among participants with a positive retinal age gap, 51% had gaps of more than 3 years. Retinal age gaps of 5 and 10 years were present among 27.6% and 4.34% of those with positive gaps, respectively.
After adjusting for confounders, researchers noted a 2% increase in mortality risk per 1-year increase in retinal age gap (HR 1.02, P =.020). Mortality risk was comparable among individuals with retinal age gaps in the second (and lowest) quantile.
Individuals in the third and fourth quantiles had greater positive retinal age gaps and significantly higher mortality risk (HR 1.21, P =.014; HR 1.35, P =.003, respectively). They had 49% to 67% higher risks of mortality not linked to CVD or cancer after adjustments (HR 1.49 and 1.67, respectively; P =.005). Multivariate adjustments did not indicate an association between retinal age gap and death due to CVD or cancer.
“The fast, non-invasive, and cost-effective nature of fundus imaging enables it to be an accessible screening tool to identify individuals at an increased risk of mortality,” according to investigators. “This risk stratification will assist tailored healthcare decision-making, as well as targeting and monitoring of interventions. Given the rising burden of non-communicable diseases and population aging globally, the early identification and delivery of personalized healthcare might have tremendous public health benefits.”
Limitations of the study include volunteer bias, an underrepresentation of unhealthy individuals, and a lack of ethnic diversity among participants.
Zhu Z, Shi D, Guankai P, et al. Retinal age gap as a predictive biomarker for mortality risk. Br J Ophthalmol. Published online January 18, 2022. doi:10.1136/bjophthalmol-2021-319807