Mohammad AskarPublic working copy · medical review pending

Translational glaucoma research · 176

Translational glaucoma research: from the bench to the clinic

Glaucoma research reaches practice slowly. Separate a risk score from a diagnosis, and a research model from a validated clinical pathway.

Public working copy · medical review pending. Source, translation and medical checks remain open. Read the limitations in the text and references. These educational materials do not replace individual medical assessment. Published: 2026-11-18T19:45:52+01:00. Internal project references lead to this notice. They are not publicly accessible primary sources; the source limitations remain stated in the text.

A polygenic risk score estimates a probability, not a disease. Lowering intraocular pressure remains the treatment we can act on; genetics and artificial intelligence support clinical judgement without replacing it.

Terminology and draft status

Terminology review is not complete. No new WHO UMD entry verification is claimed for this copy; pending entries and limits are recorded in terminology.md.

01Why translation is slow

Glaucoma is a multifactorial neurodegenerative optic neuropathy: retinal ganglion cells and their axons are lost, the optic disc and retinal nerve fibre layer change, and visual field damage follows. Lowering intraocular pressure (IOP) is the part of treatment we can modify, and it remains the basis of therapy. Some patients still progress while their measured pressure looks normal, which is why the export asks for additional approaches rather than a replacement for pressure control. 1

The export lists five barriers between laboratory work and the clinic. Transfer is slow; animal models such as the mouse eye differ in size, angle anatomy, aqueous outflow and immune response; regulation and ethics add requirements; the steps needed for translation are poorly funded; and progress needs several disciplines at one table. These are qualitative statements, not measured failure rates, and a barrier is not proof that a particular treatment failed. 1

In clinic the same gap appears as a test that is precise but not validated, or a risk model that does not change what you do. Ask three questions before adopting a new tool: which decision does it change, what is the reference standard, and what happens to the patient when the result is wrong. 1

02Polygenic architecture and rare monogenic exceptions

Most glaucoma is polygenic: many common variants, each with a small effect, together shift susceptibility. No single glaucoma gene explains the common disease, so a negative single-gene test does not clear a patient, and family history remains an aggregate marker of risk. 1

Rare monogenic disease needs a precise phenotype and interpretation of pathogenicity, inheritance and zygosity. MYOC is relevant to juvenile/early-onset open-angle disease, CYP1B1 to congenital glaucoma; not every variant is disease-causing and naming a gene is not a diagnosis. The exact variant and pedigree require genetic counselling. 1

The export names IOP, age, nutrition, adiposity and diabetes as interacting factors. IOP and age are established risk factors. Nutrition, body weight and diabetes come largely from observational association, where more frequent examination and shared vascular risk can confound the finding. They are not treatment instructions: no diet, supplement or weight change has been shown here to prevent glaucoma, and lifestyle advice must never displace pressure control and scheduled follow-up. 1

Genetics has not produced a curative therapy. It has produced a better understanding of mechanism and, as the next chapter shows, probabilistic risk tools whose clinical value is still being defined. 1

03What a polygenic risk score is, and what it is not

A polygenic risk score (PRS) is a weighted sum of the effects of many risk variants identified in genome-wide association studies (GWAS), usually reported as a percentile within a reference population. The export gives the UK Biobank as an example of a discovery resource. A PRS is not a diagnostic test: it cannot confirm or exclude glaucoma, and the diagnosis still comes from the disc, the visual field and the pressure history of that patient. 1

The export says comparable biobanks are absent in Germany. NAKO shows that biobanking does exist: its 2022 cohort profile reports 205415 adults aged 19–74 recruited at 18 centres during 2014–2019 with biomaterial collection. Its separate ophthalmic analysis reports 48460 participants. These reports do not establish equivalence to UK Biobank or current clinical availability of a German glaucoma PRS; scale, linkage and access require separate comparison. 1 5 6

Even a well-built score is fragile. Its value depends on the discovery sample, variant selection, thresholding, imputation and the ancestry of the person in front of you; different pipelines can place the same patient in different percentiles. A percentile is a position in a distribution, not the probability that this person will develop glaucoma, and quoting high risk without an absolute risk and a time horizon misleads. 1

Clinical usefulness is a higher bar than statistical association. A score has to add to age, IOP, optic disc and OCT findings, visual field and family history, stay calibrated in the population where it is used, and be shown to improve decisions or outcomes, not only to sort patients into groups. 1 4

The 2022 NAKO profile reports 205415 recruited, whereas the 2026 ophthalmic abstract describes its module within 205053 enrolled. The 48460 ophthalmic participants are a subset; these distinct reported denominators are not silently harmonised or declared a source erratum. Neither report validates a glaucoma PRS clinical pathway. 5 6

04Ancestry, prediction and the ethics of probability

Most GWAS of glaucoma have been performed in populations of European ancestry, while the disease is more prevalent and more severe in people of African ancestry. A score transfers only if the variants, their effect sizes and the correlation structure between variants are similar in the new population, which is often not the case. 1 4

A 2025 cross-sectional analysis tested three published PRS in 11673 cases and 66432 controls across seven ancestral groups. With the Gharahkhani score, top-quintile odds ratios ranged from 1.68 (95% CI 1.17 to 2.43) in Ghanaians to 7.05 (95% CI 2.73 to 19.6) in a South African multiple-ancestry group; with the Han score from 2.27 to 7.24, and with the Craig score from 1.51 to 6.31. Despite these odds ratios, the gain in AUROC and R² above baseline was lower in every African-ancestry group than in European-ancestry groups. 3

Two lessons follow. An odds ratio in the top quintile is not a diagnostic accuracy, and a lower AUROC means poorer discrimination between affected and unaffected people. A score validated elsewhere is therefore a research instrument until it is calibrated locally. Skin colour is a poor proxy for genetic ancestry, and self-reported ethnicity cannot replace genotyping and local validation. 3 4

The export also raises ethics and cost. Communicating a non-deterministic risk affects families, insurance and employment expectations, and requires counselling plus a plan for what will be done differently. The cost-effectiveness of population-wide genetic testing is unknown, and no screening programme in this material has been shown to prevent blindness by using a PRS. 1

These are highest-versus-lowest quintile odds ratios in a cross-sectional case-control analysis, not prospective absolute risks. Mean age was 76.9 years and 95.1% were male; case/control selection and the study mix limit generalisation. Improvement in discrimination does not establish calibration, decision benefit or prevention of blindness. 3

05The table: hypothetical scenarios and what the trial data support

The table preserves hypothetical original scenarios, not a validated PRS-driven follow-up schedule. Suspicious clinical findings guide referral regardless of the score. A high PRS with normal findings motivates research into closer surveillance, but neither the interval nor outcome benefit is established here. A low score must not dismiss a suspicious eye or replace clinically indicated follow-up. 1

A post hoc OHTS analysis examined 1010 of 1636 original participants with eligible genetic data. PRS below the 48th percentile was associated with adjusted odds of disease-free status at 20 years of 1.49 (95% CI 1.04–2.15), not a 1.49-fold absolute probability. In the highest clinical-risk tertile, initial-observation conversion rates were 61.1% with high versus 23.8% with low PRS; the abstract reports a contrast CI −63.0 to −11.6. Early-treatment rates were 37.3% versus 24.1%, with contrast CI −35.6 to 9.3 and p=0.32. The negative intervals describe a contrast, not confidence intervals for either individual percentage; exact estimands require the full report. 1 2

This is a selected genetic subset and a post hoc, threshold-dependent subgroup analysis with wide intervals. Initial observation does not mean 20 years without treatment: all participants were offered topical IOP lowering from 2002. OHTS originally supported pressure lowering at 60 months, with estimated glaucoma incidence 9.5% versus 4.4% (HR 0.40, 95% CI 0.27–0.59). The later PRS association is not a randomised comparison of PRS-guided versus clinical care and does not validate a PRS treatment threshold. 2 7

Explain the numbers as absolute risks over a defined period, state the uncertainty, and say what will change in this patient's care. If nothing changes, the test has added cost and anxiety without benefit. The export’s proposal to use the score as an adjunct to screening, to identify high-risk people earlier and monitor them more closely, is exactly the claim that still needs outcome evidence. 1

The table: hypothetical scenarios and what the trial data support
Polygenic risk score (PRS)Clinical findingsApproach in the original scenario
HighSuspicious findingsEarlier referral to the clinic, shorter intervals — hypothetical original proposal, not a validated schedule
HighNormal findingsCloser monitoring (genetic glaucoma suspicion) — hypothetical original proposal, not a validated schedule
LowNormal findingsRoutine follow-up — hypothetical original proposal, not a validated schedule

06Artificial intelligence and the integrated pathway

Artificial intelligence in glaucoma covers different tasks with different reference standards: classifying a fundus or optic disc photograph, segmenting the rim, the retinal nerve fibre layer or OCT compartments, and predicting progression from longitudinal data. Research accuracy on a curated dataset is not clinical benefit, and the export's statement that algorithms already support early detection describes developmental and screening research rather than an established standard of care. 1

The NAKO ophthalmic data show the practical constraint: of 48460 examined participants, self-reported glaucoma was 2.0%, and four deep-learning quality models classified 68.2% of fundus images as gradable. Image quality, ungradable cases and self-report instead of examination all limit what a screening algorithm can achieve in a general population. 1 6

The export names data bias, poor reproducibility and hesitant clinical acceptance. Add dataset shift between centres and cameras, label noise, unclear responsibility for a missed case, workflow fit and the risk of deskilling. No product-specific glaucoma clearance or outcome trial is verified in this material, so no algorithm should be presented to a patient as an approved diagnostic replacement. The export’s own framing remains correct: decision support, not replacement. 1

The productive future is integrated: genetics, genomics, proteomics, imaging, artificial intelligence, neurobiology and clinical data in one pathway, built by laboratory scientists, clinicians and data scientists together. The export's image for this is “Bench <> Bedside <> Data Science”: laboratory, clinic and data science in one loop. The export's four requirements remain the test of any such pathway: solid basic research, genuinely multidisciplinary teams, patient-centred clinical validation with outcomes that matter, and a clinician who stays responsible as diagnostician, translator of new technology and advocate for the patient. Its closing thought is the right limit: genetic testing reflects a probability rather than a certainty, and clinicians still shape the pathway. 1

What are the main learning points of this lecture?

Translation matters only when a finding changes a decision. A polygenic risk score estimates probability, not diagnosis, and needs local calibration. Lowering intraocular pressure remains the treatment with randomised support, and artificial intelligence supports clinical judgement rather than replacing it. Check every statement against the individual clinical findings.

Which findings require urgent assessment?

Rapid visual deterioration, severe pain, marked inflammation, a pressure rise or suspected infectious or neurological disease require urgent assessment.

References and access status

  1. 1
    Immutable original export; complete lecture export; export attribution unverified

    Complete export read; the export's source list names an Arabic-titled lecture ('004_رؤى حديثة في أبحاث الجلوكوما الانتقالية_ من المختبر إلى العيادة') and states that the original URL is not traceable in the source export; original attribution, named speaker, date and any recording or transcript not independently verified.

  2. 2
    Sekimitsu et al. 2024, PMID 39509108: POAG polygenic risk score and disease onset, post hoc OHTS analysis

    Abstract text read 16 November 2026 in the PubMed search-result abstract display; full article, supplement, genotype and threshold methods unread. Post hoc design and subgroup limits stated in the text. Parent independently read the actual cached PubMed abstract display supplied by the delegate on 16 November 2026; no fresh direct full-text access claimed.

  3. 3
    Chang-Wolf et al. 2025, PMID 39541127: PRS performance in populations of African descent

    Abstract text read 16 November 2026 in the PubMed search-result abstract display; full article, supplement and per-model derivation details unread. Parent independently read the actual cached PubMed abstract display supplied by the delegate on 16 November 2026; no fresh direct full-text access claimed.

  4. 4
    Cooke Bailey et al. 2022, PMID 36672852: diversity in polygenic risk of primary open-angle glaucoma

    Abstract text read 16 November 2026 in the PubMed search-result abstract display; complete review text, included study list and all effect estimates unread. Parent independently read the actual cached PubMed abstract display supplied by the delegate on 16 November 2026; no fresh direct full-text access claimed.

  5. 5
    Peters et al. 2022, PMID 36260190: German National Cohort (NAKO) framework and baseline examination

    Abstract text read 16 November 2026 in the PubMed search-result abstract display; full cohort profile, response rates and biobank access procedures unread. Parent independently read the actual cached PubMed abstract display supplied by the delegate on 16 November 2026; no fresh direct full-text access claimed.

  6. 6
    NAKO ophthalmic baseline 2026, PMID 42894708, Ophthalmol Sci 6(11):101408

    Abstract text read 16 November 2026 in the PubMed search-result abstract display; full baseline analysis, imaging protocol and quality-model details unread. Parent independently read the actual cached PubMed abstract display supplied by the delegate on 16 November 2026; no fresh direct full-text access claimed.

  7. 7
    OHTS 2002, PMID 12049574: ocular hypertension treatment trial

    Historical full-text appraisal from review 005, 2 to 6 October 2026, reused through review 161, 15 November 2026; not newly read. Open treatment and masked endpoint assessment preserved.