Genetic risk score improves glaucoma screening in large population cohort

Glaucoma, a silent thief of sight that affects roughly three percent of the elderly population, has long frustrated public health planners. The disease is treatable when caught early, yet population-wide screening has been judged economically unviable for decades, largely because the condition is too rare in the general population for current tests to be applied efficiently. A landmark study from the Netherlands now demonstrates that a simple genetic score can change that equation dramatically. In the first prospective, double-blind screening study of its kind, researchers showed that people in the top fifth of a genetic risk score distribution were more than seven times as likely to have glaucoma as those in the bottom fifth, turning the mathematics of screening on its head.

The findings, published in the European Journal of Epidemiology, come from EyeLife, an add-on study embedded in Lifelines, a large population-based cohort and biobank covering the Northern Netherlands. The team, led by Anna Neustaeter and Nomdo Jansonius of the University Medical Center Groningen together with epidemiologists Ilja Nolte and Harold Snieder, invited 1,829 Lifelines participants aged 55 and older. Crucially, invitations went only to individuals sitting in either the highest or lowest 20 percent of a glaucoma genetic risk score distribution, with age and gender balanced between the two groups. Of those invited, 1,022 participants ultimately completed the full ophthalmic work-up and were included in the analysis: 487 with high genetic risk and 535 with low genetic risk. Neither the researchers performing the screening nor those interpreting the results knew which group any participant belonged to, and the participants themselves did not know they had been selected on the basis of their DNA.

The genetic risk score at the heart of the study was built from 268 independent single nucleotide polymorphisms, or SNPs, drawn not only from genome-wide association studies of glaucoma itself but also from studies of closely related traits, or endophenotypes. These included intraocular pressure, the primary risk factor for the disease, as well as central corneal thickness and optic disc parameters. Each participant’s score was calculated as a weighted sum of risk alleles, with the weights derived from log-transformed odds ratios reported in the largest available meta-analyses. Scores were then adjusted and scaled separately for the two genotyping platforms used within Lifelines, the Illumina HumanCytoSNP-12v2.0 and the Infinium Global Screening Array, to account for differences in genotyping quality. The authors note that this design was deliberately curated, combining 49 SNPs from glaucoma studies with 176 SNPs linked to intraocular pressure, 26 linked to both, 13 linked to corneal thickness and 4 linked to disc parameters.

Participants underwent a battery of examinations at the Department of Ophthalmology in Groningen. Visual fields were tested using frequency doubling technology perimetry, with any abnormal test location retested to confirm reproducibility. Optical coherence tomography imaged the peripapillary retinal nerve fiber layer, the macula, and the anterior segment, allowing measurement of both structural nerve damage and central corneal thickness. Non-mydriatic fundus photography captured 45-degree images centered on the optic disc and macula, from which a glaucoma specialist graded glaucomatous optic neuropathy, defined as a vertical cup-disc ratio of at least 0.7, focal notching of the neural rim, or a peripapillary splinter hemorrhage. Intraocular pressure was measured, and participants with abnormalities were invited back for a full ophthalmic examination including gonioscopy to confirm that the angle of the eye was open.

The classification system, adapted from the International Society of Geographical and Epidemiological Ophthalmology framework and informed by the Northern Finland birth cohort eye study, used a “two-of-three” logic combining visual field, nerve fiber layer, and optic disc assessments. Participants were labeled as having definite glaucoma when functional and structural loss coexisted, probable glaucoma when either confirmed field loss or structural abnormalities with optic neuropathy were present, and possible glaucoma when structural loss appeared alone. The primary outcome combined definite and probable cases of primary open-angle glaucoma.

The results were striking. Fifty glaucoma cases emerged in the high genetic risk group, comprising 29 definite and 21 probable diagnoses, compared with only nine cases, two definite and seven probable, in the low genetic risk group. That corresponds to a prevalence of 10.3 percent in the high-risk quintile against just 1.7 percent in the low-risk quintile. After adjusting for age, sex, and genotyping platform using a log-link generalized linear model with a binomial family, the relative risk for glaucoma in the high genetic risk group was 7.4, with a 95 percent confidence interval of 3.7 to 14.7. When the genetic risk score was treated as a continuous variable, each one standard deviation increase carried an adjusted relative risk of 1.83 for glaucoma and 1.38 for possible glaucoma. Even for possible glaucoma, the milder structural category, the high-risk group showed a prevalence of 8.2 percent versus 3.7 percent in the low-risk group.

Age amplified the effect. Among those with high genetic risk, the prevalence of definite or probable glaucoma climbed from roughly four percent at ages 55 to 59 to about 30 percent at age 80 and beyond. Among those with low genetic risk, the same trajectory ran from under one percent to about five percent. The relative risk itself did not vary significantly across age strata, suggesting the genetic signal operates consistently across the aging spectrum. Perhaps most sobering was the awareness data: of the 59 glaucoma cases identified, only 20, or 34 percent, knew they had the disease. Awareness was 52 percent among definite cases but a mere 14 percent among probable cases, and there was no meaningful difference in awareness between the high and low genetic risk groups, undermining the assumption that family history alone directs high-risk individuals to eye care.

The study’s design carried several methodological safeguards worth noting. Response bias was assessed by comparing participants who attended with those who did not; no differences emerged in age or gender, and participation rates were nearly identical between genetic risk groups at 55 and 57 percent, a figure that compares favorably with typical Lifelines add-on response rates of 10 to 50 percent. The authors also tested whether restricting analysis to the extreme quintiles of the score distribution distorted effect estimates; simulation showed that logistic regression coefficients from extreme-range and full-range models correlated at above 0.9, indicating minimal impact.

What does this mean for screening policy? The argument turns on a simple probability calculation. Population-based glaucoma screening has historically failed cost-effectiveness tests because disease prevalence sits around three percent, causing even highly specific tests to generate unacceptable rates of false positives. The Wilson and Junger criteria, laid out at the World Health Organization in 1968, demand that screening be directed at conditions where early detection is beneficial and economical. Glaucoma meets the clinical criteria: it is chronic, irreversible if untreated, largely asymptomatic in early stages yet detectable through optic nerve head changes and retinal nerve fiber layer thinning, and responsive to treatment that slows progression considerably. What has been missing is a way to raise the prior probability of disease in the screened population to an economically viable threshold.

The EyeLife data suggest genetic pre-screening can do exactly that. If one conservatively assumes the unmeasured middle quintiles behave like the lowest quintile, the overall prevalence implied by the study is 3.4 percent, closely matching known population figures. Under that assumption, more than half of all glaucoma cases would be concentrated in the highest genetic risk quintile. If a minimum prevalence of five percent is required for cost-effective screening, as the authors tentatively propose, individuals with high genetic risk should begin screening at around age 60, while those with low genetic risk might never need routine screening, or perhaps a single check near age 80. Genetic pre-screening may also mitigate length bias, the tendency of periodic screening programs to over-capture slow-progressing disease variants, since fast-progressing cases would be preferentially identified among genetically susceptible individuals monitored more intensively.

Important caveats remain. The study was conducted in a population of Northwestern European ancestry, and glaucoma’s genetic architecture differs across ethnicities, so any real-world deployment would require scores tailored to the population being screened. The score itself was designed in 2018 using only genome-wide significant variants available at that time; newer polygenic risk score methods that incorporate functional genomic annotations and far larger genome-wide association samples, such as those now emerging through the UK Biobank and consortia analyzing hundreds of risk loci, would likely push predictive power further. The authors explicitly frame their results as a lower bound of what genetic pre-screening can achieve. Questions also remain about how genetic risk information should be disclosed to patients, and about the ethical handling of genetic data, privacy, and stewardship, all of which require careful protocol development before clinical implementation.

Still, the core message is unambiguous. A single blood or saliva sample, genotyped once, could determine who should be walked through the visual field machine and who could safely wait. With roughly half of glaucoma cases undiagnosed worldwide and advanced disease at diagnosis remaining a major determinant of lifelong blindness, a tool that more than doubles the efficiency of case finding among older adults represents a genuine inflection point. The EyeLife study provides the first prospective evidence that DNA, rather than family anecdotes or incidental eye exams, can identify the people most likely to be losing their sight without knowing it, and in doing so brings cost-effective, population-level glaucoma screening within reach for the first time.