Predicting Glaucoma Before It Starts: How Close Are We to Genetic Risk Scores That Actually Change Patient Outcomes?
Glaucoma – a group of diseases damaging the optic nerve – is the leading cause of irreversible blindness worldwide (pmc.ncbi.nlm.nih.gov). Globally it affects tens of millions of people, a number expected to grow with aging populations (pmc.ncbi.nlm.nih.gov). The most common form, primary open-angle glaucoma (POAG), is often silent in its early stages. In fact, studies estimate roughly half of glaucoma cases remain undiagnosed until vision loss begins (bmcmedgenomics.biomedcentral.com) (pmc.ncbi.nlm.nih.gov). This is unfortunate because early detection matters: standard treatments (eye drops, laser or surgery to lower intraocular pressure) can effectively slow or halt progression when started early (bmcmedgenomics.biomedcentral.com) (pmc.ncbi.nlm.nih.gov). Glaucoma’s insidious onset but treatable nature makes it an ideal candidate for predictive screening. Genetics offers one promising avenue. POAG is highly heritable – first-degree relatives have about a 9-fold higher risk than average! (pmc.ncbi.nlm.nih.gov). Estimates put genetic heritability of POAG at roughly 70–80% (pmc.ncbi.nlm.nih.gov) (pmc.ncbi.nlm.nih.gov). These facts suggest that a person’s DNA contains valuable clues to their future glaucoma risk.
Early clinics have long tested for rare single-gene mutations (e.g. MYOC, OPTN) in families with juvenile or early-onset glaucoma (pmc.ncbi.nlm.nih.gov). But such Mendelian variants account for only a small minority of cases (pmc.ncbi.nlm.nih.gov) (pmc.ncbi.nlm.nih.gov). Most glaucoma is polygenic: influenced by many common genetic variants each contributing a small risk. Over the past decade, large genome-wide association studies (GWAS) have identified hundreds of genomic loci linked to glaucoma and related traits (www.nature.com) (pmc.ncbi.nlm.nih.gov). For example, a 2023 study (N > 600,000 Europeans plus multi-ancestry cohorts) found 263 independent risk loci, and further expanded this to 312 loci by including diverse populations (www.nature.com). These discoveries go beyond intraocular pressure genes – they include factors involved in optic nerve structure and even immune pathways. Such rich genetic data raise the question: can we summarize an individual’s inherited risk into a single score that meaningfully predicts future glaucoma?
Polygenic Risk Scores for Glaucoma
A polygenic risk score (PRS) does exactly that: it adds up the small effects of thousands of common genetic variants into one number (bmcmedgenomics.biomedcentral.com). In simple terms, a PRS estimates how a person’s DNA influences their chance of developing a disease. Importantly, a PRS is not a diagnosis – it is a probabilistic risk estimate (bmcmedgenomics.biomedcentral.com). For glaucoma, researchers have now built PRS using well-established risk variants and tested them in large cohorts. The results are encouraging: people in the highest percentiles of the glaucoma PRS are at substantially higher risk of the disease than those with average scores (pmc.ncbi.nlm.nih.gov) (pubmed.ncbi.nlm.nih.gov).
For example, one study in an Australian population used hundreds of variants related to eye pressure and optic nerve shape. Individuals in the top PRS decile had about 5–6 times the odds of developing glaucoma compared to those in the bottom decile (pmc.ncbi.nlm.nih.gov). Another comprehensive PRS (using thousands of SNPs for glaucoma and its related traits) showed an even larger effect: the top decile had roughly 10–20× the risk of glaucoma relative to the bottom decile (pmc.ncbi.nlm.nih.gov). In practical terms, adding a PRS to conventional risk factors boosts the accuracy of predicting who will get glaucoma. For instance, a recent analysis of four large European-ancestry cohorts found that a model with age, sex, high eye pressure, and family history had a concordance (C-statistic) of about 0.75. Adding the glaucoma PRS raised this to ~0.82 (pubmed.ncbi.nlm.nih.gov) – a substantial improvement. In the same study, patients in the highest PRS quintile were ~4–5 times more likely to develop glaucoma than those in the middle quintile (pubmed.ncbi.nlm.nih.gov). Notably, higher PRS scores also correlated with more severe disease: top-risk individuals were diagnosed younger, had larger optic nerves, and were more likely to need glaucoma surgery (pubmed.ncbi.nlm.nih.gov) (pmc.ncbi.nlm.nih.gov).
In summary, current glaucoma PRS models can stratify risk. Those in the top few percent of PRS have many-fold higher odds of disease than average. These findings have been reproduced by independent groups: for example, MacGregor et al. found a ~5.6× risk for top-decile individuals (pmc.ncbi.nlm.nih.gov), and Gao et al. reported 10–20× risk for extreme deciles (pmc.ncbi.nlm.nih.gov). Adding PRS to simple clinical factors consistently improves risk prediction models (pubmed.ncbi.nlm.nih.gov) (pmc.ncbi.nlm.nih.gov). In practice, this means we might one day screen a person’s DNA to decide how aggressively to search for glaucoma.
Performance Across Populations
Most PRS development to date has been in people of European ancestry, which poses challenges for broader use. When tested in other groups, European-based scores still detect some risk but with reduced accuracy. For example, a PRS derived from UK Biobank data gave an AUC of ~0.79 in Europeans but only ~0.76 in South Asians (pmc.ncbi.nlm.nih.gov). That is, it worked, but slightly worse. In African-ancestry populations, published PRS show even more limited performance. A recent JAMA Ophthalmology analysis of nearly 80,000 subjects from Africa and Europe found that the highest PRS quintile in African-descended groups did have higher glaucoma risk, but the overall predictive power (AUC) was much lower than in Europeans (pmc.ncbi.nlm.nih.gov). In short, PRS transfer across ancestries but incompletely (pmc.ncbi.nlm.nih.gov) (pmc.ncbi.nlm.nih.gov). This underscores the need for larger, more diverse genetic studies. Efforts are underway (e.g. global consortia, Biobank collaborations) to include Asians, Africans, Latinos, and others, which should yield improved scores for everyone.
AI and Integrated Prediction
Beyond genetics alone, many groups are using artificial intelligence (AI) and machine learning (ML) to build richer glaucoma risk tools. AI can digest complex data – clinical information, imaging, and genetics – and detect patterns humans cannot. Recent reviews note that ML models incorporating conventional risk factors (age, intraocular pressure, optic nerve/retinal nerve fiber measurements, family history) alongside imaging and genomic data achieve strong accuracy (www.sciencedirect.com) (pmc.ncbi.nlm.nih.gov). For example, deep learning algorithms have been trained on standard eye exams (like color fundus photographs or optical coherence tomography scans) to predict future glaucoma. One notable study used baseline fundus photos from people with ocular hypertension and achieved about 0.88 accuracy in predicting who would develop glaucoma 1–3 years later (pmc.ncbi.nlm.nih.gov). An external validation reported AUC ~0.88–0.89 for such predictions (pmc.ncbi.nlm.nih.gov). These models even learned to estimate retinal nerve fiber thickness from photos; eyes with thinner predicted nerve fiber layer at baseline had significantly higher future risk (pmc.ncbi.nlm.nih.gov).
Meanwhile, ML on electronic health records (EHR) is also promising. In one large multicenter study, algorithms using diagnoses, medications, lab values and demographics identified patients at high risk of glaucoma 1 year before onset with AUC ≥0.81 (pmc.ncbi.nlm.nih.gov). Another deep-learning model (Ha et al.) combined fundus images with clinical data and achieved AUCs of 0.98–0.99 for predicting normal-tension glaucoma development among at-risk patients (pmc.ncbi.nlm.nih.gov). Importantly, these AI tools often flag known risk features (e.g. higher baseline eye pressure or thinner nerve fiber layers) as the most predictive inputs.
To date, most AI prediction has focused on eyeball images and clinical data rather than raw genetics. But future models could integrate a person’s PRS as yet another input. In other fields (like cardiology and cancer), hybrid models combining PRS, lifestyle, and imaging show the best results. In glaucoma, this approach is only beginning. A recent narrative review highlights ML’s potential, noting that modern algorithms (random forests, support-vector machines, etc.) can handle multi-modal inputs for both population-wide risk assessments and personalized predictions (www.sciencedirect.com). Such tools could eventually tailor screening intensity, follow-up intervals, or even preventive treatments based on a person’s overall risk profile.
Toward Clinical Use: Screening and Early Intervention
If genetic risk scores (and AI tools) are so promising, when will they enter the clinic? Currently, routine genetic screening for glaucoma is not standard practice. Health systems generally do not screen the general public for glaucoma by any method (even clinical exams) because universal screening has not proven cost-effective (pmc.ncbi.nlm.nih.gov) (pmc.ncbi.nlm.nih.gov). Instead, many programs focus on high-risk groups: for example, people of African ancestry or those with a family history, who are known to have higher glaucoma prevalence. In Australia, current guidelines recommend that first-degree relatives of glaucoma patients start eye exams 5–10 years before the relative’s age of onset (pmc.ncbi.nlm.nih.gov). People of African descent are advised to begin screening around age 40, compared to age 50 for European ancestry (pmc.ncbi.nlm.nih.gov). Similar targeted screening guidelines exist elsewhere.
Introducing genetic screening (like a PRS test) into routine care will require multiple pieces to fall into place. Cost is one factor. Genotyping technology is becoming very affordable (cheap SNP microarrays or sequencing), but an overall screening program still has costs: lab processing, data analysis, and follow-up visits. Preliminary health-economic models suggest the idea could be cost-effective. For example, Liu et al. (2022) modeled PRS-based screening in the UK and Australia and estimated incremental cost-effectiveness ratios near £25,000–Ott$34,000 per quality-adjusted life year – within typical willingness-to-pay thresholds (pmc.ncbi.nlm.nih.gov). In their simulation, the targeted program using PRS had a roughly 60–80% chance of being deemed cost-effective in those countries. Similar analyses in other diseases have also reached optimistic conclusions. That said, these models depend on assumptions (test cost, glaucoma prevalence, treatment efficacy) that must be validated in the real world.
Feasibility and workflow are other hurdles. Eye clinics and primary care would need to collect DNA samples (e.g. saliva or blood), perform the genotyping, compute the PRS, and then interpret it. This requires infrastructure (labs, software) and trained personnel (genetic counselors, ophthalmologists knowledgeable about genomics). Importantly, doctors would need clear guidelines on how to act on PRS information. For instance, at what threshold of genetic risk should a patient be sent for more frequent eye exams? Early studies have shown that personalized risk reports help patients understand their PRS result and its implications (pmc.ncbi.nlm.nih.gov) (bmcmedgenomics.biomedcentral.com). One study even designed graphical report formats and found laypeople prefer absolute risk visuals with follow-up action advice (pmc.ncbi.nlm.nih.gov) (bmcmedgenomics.biomedcentral.com). Such work in risk communication will be vital before wide adoption.
Perhaps the biggest necessity is evidence that PRS-based screening actually improves outcomes. We know from older trials that treating people at high eye pressure reduces progression. For example, the classic Ocular Hypertension Treatment Study showed one in which lowering pressure in high-risk individuals cut the development of glaucoma roughly in half. However, that study focused on clinical risk factors (eye pressure), not genetics. We still need proof that telling someone they have a high genetic risk – and subsequently intervening – prevents vision loss. This would likely require controlled studies: e.g. randomize high-PRS individuals to earlier treatment vs standard care and follow vision outcomes. Such trials take years.
Ongoing Studies and Implementation Efforts
Fortunately, research groups are already tackling many of these questions. In Australia, the GRADE study (Genetic Risk Assessment of Degenerative Eye disease) is a prospective trial that began enrollment around 2023 (pmc.ncbi.nlm.nih.gov). About 1,000 unselected adults over age 50 will have DNA genotyped. Their glaucoma and AMD PRS will be calculated, and then the investigators will compare disease prevalence in the top, middle, and bottom PRS deciles (pmc.ncbi.nlm.nih.gov) (pmc.ncbi.nlm.nih.gov). If the high-PRS group indeed shows significantly more undiagnosed glaucoma compared to lower-PRS groups, that would be a strong proof-of-concept for clinical validity and for targeted screening. A complementary study (the INSiGHT trial) is assessing the psychological impact of giving people their glaucoma PRS results (pmc.ncbi.nlm.nih.gov) (pmc.ncbi.nlm.nih.gov). Researchers from the GRADE study will invite participants from the very high, very low, and average PRS groups to receive their risk result and then follow them with questionnaires. This will inform how patients react to genetic risk information – hopefully providing guidance on counseling and consent before rolling it out more broadly.
Beyond Australia, several international groups are active. The Trøndelag Health Study (HUNT) in Norway is evaluating PRS in their population. The UK’s U.K. Biobank and other cohorts have generated PRS models (as seen in the above studies). Private companies and clinics in some countries offer genetic eye disease panels (usually focusing on monogenic genes), and some may include PRS for common eye diseases as an experimental add-on. However, to our knowledge no professional body currently recommends routine PRS testing for glaucoma in asymptomatic individuals.
What Patients and Eye Care Providers Need to Know
For patients with a family history of glaucoma, the current actionable advice remains: inform your eye doctor and consider earlier and more frequent eye examinations. The fact that POAG is hereditary means your risk is elevated, but genetics is just one piece of the puzzle. No single gene test can tell you definitively if you will get glaucoma. In cases of very early-onset glaucoma, genetic testing for Mendelian genes (e.g. MYOC mutations) is available and may be recommended (pmc.ncbi.nlm.nih.gov). For typical adult-onset glaucoma, PRS tests will likely become available (some direct-to-consumer genetic tests now report a glaucoma score), but keep in mind that these scores are still experimental. If you use them, do so under the guidance of a doctor or genetic counselor. High PRS should lead to action – ordinarily, that means more vigilant eye screening and risk-factor control (targeting even lower eye pressure, checking optic nerves carefully). For those with low PRS, it is tempting to relax, but clinical risk factors still matter. A low genetic risk does not guarantee you won’t develop glaucoma if, for example, you have high eye pressure or other risk factors. Thus PRS should supplement, not replace, standard care.
For researchers and clinicians, the roadmap is clear but challenging. Key focus areas include:
- Diversify the data. We must build large GWAS and biobank datasets that include non-European populations so that PRS can be equitable.
- Refine the scores. Multi-trait and multi-ancestry methods (like the recent Nature Genetics GWAS (www.nature.com)) can yield more powerful scores. Specialized scores (e.g. focusing on normal-tension vs high-tension glaucoma) may also emerge.
- Validate in clinics. We need trials or observational studies (like GRADE) showing that PRS-guided interventions actually improve patient outcomes (better vision preservation) compared to usual care.
- Integrate with other technologies. Combining PRS with AI models based on eye scans or EHR data could yield next-generation risk tools.
- Address ethical and economic issues. Work out the best way to offer PRS testing (cost-sharing, consent, return of results), and confirm that targeted screening is truly cost-effective in real health systems.
In sum, polygenic risk scoring for glaucoma is a maturing field. Recent large-scale genetic studies have unlocked hundreds of risk variants (www.nature.com). AI methods trained on eye images and health records are pushing the boundary of predicting glaucoma years before vision loss occurs (pmc.ncbi.nlm.nih.gov) (pmc.ncbi.nlm.nih.gov). Early data show PRS can identify groups at 4–20× higher glaucoma risk (pmc.ncbi.nlm.nih.gov) (pubmed.ncbi.nlm.nih.gov). But realizing the full potential – turning risk scores into better patient outcomes – will require more evidence and careful implementation. Ongoing trials like GRADE and INSiGHT will provide crucial insights (pmc.ncbi.nlm.nih.gov) (pmc.ncbi.nlm.nih.gov).
Conclusion: We are getting closer to being able to predict glaucoma before it starts, but we are not quite there yet. Current PRS models can point out who is genetically predisposed to glaucoma, and AI is helping make use of this information. However, for such tools to truly change patient care, we need to demonstrate in practice that identifying high-risk individuals (through genetics or AI) allows us to intervene earlier and reduce vision loss. That will likely come step-by-step: larger validation studies, followed by pilot screening programs, and eventually integration into ophthalmology guidelines. In the meantime, patients with a family history should continue regular eye check-ups and discuss any genetic testing or trials with their doctors. Though not routine today, genetic risk scoring for glaucoma is a real possibility in our near future – one that may finally tip the balance toward early detection and prevention of this sight-threatening disease.