Highlight
- Comparing glaucoma definitions using diagnosis alone versus diagnosis plus treatment in UK Biobank participants.
- Stronger odds ratios for established risk factors and biomarkers observed when treatment evidence is included in case definitions.
- Findings underscore importance of precise phenotyping in large biobank studies for glaucoma research and risk stratification.
Study Background
Glaucoma is a leading cause of irreversible blindness worldwide, characterized by progressive optic neuropathy and visual field loss. Early identification and accurate disease classification are central to managing glaucoma and elucidating its complex risk factors. Large-scale population biobanks offer unprecedented opportunities to study glaucoma epidemiology, genetics, and biomarkers, but remain challenged by heterogeneous disease definitions derived from diagnostic records or treatment data. Accurate phenotypic classification is crucial, as misclassification may dilute risk associations and hinder translational insights.
Study Design
This observational cohort study analyzed data from 503,325 participants enrolled in the UK Biobank, a major population-scale resource. Glaucoma case status was defined in two distinct ways: (1) “diagnosis-only” cases based on self-reported diagnosis, primary care records, or hospital inpatient data (n=16,154), and (2) “diagnosed-and-treated” cases which met the first criterion and had evidence of receiving glaucoma-specific treatment, including medication or surgery (n=7,012). Participants without glaucoma numbered 481,772.
Key exposures were established glaucoma risk factors, including corneal-compensated intraocular pressure (IOPcc), a polygenic risk score (PRS), and clinical biomarkers indicative of disease severity — macular retinal nerve fiber layer thickness (mRNFL) and ganglion cell layer thickness (mGCL). The primary outcome was the magnitude of association, expressed as odds ratios (ORs), of these factors with glaucoma status by case definition.
Key Findings
Across all risk factors and severity biomarkers, odds ratios were consistently stronger for the diagnosed-and-treated group compared to the diagnosis-only group, suggesting higher specificity and clinical relevance when treatment data informs case status.
For participants in the highest 5% of IOPcc, the odds of glaucoma were substantially elevated—OR 23.6 (95% CI: 21.3–26.2) in the diagnosed-and-treated cohort versus OR 11.4 (95% CI: 10.6–12.3) in the diagnosis-only cohort (p < 1×10⁻¹⁰ for difference). This doubling of OR magnitude highlights the precision afforded by including treatment evidence.
Similar patterns emerged for severity biomarkers. For the thinnest 5% of mRNFL thickness, ORs were 8.0 (6.9–9.2) with treatment inclusion versus 4.7 (4.2–5.3) diagnosis-only (p = 2.32×10⁻⁷). For mGCL thickness, ORs were 7.1 (6.1–8.3) vs. 4.2 (3.7–4.7) respectively (p = 2.53×10⁻⁷). The polygenic risk score showed parallel trends: OR 8.4 (7.9–8.8) in diagnosed-and-treated vs. 5.6 (5.4–5.9) diagnosis-only (p < 1×10⁻¹⁰).
These consistent differences were observed across all evaluated glaucoma risk factors, reinforcing the robustness of the findings.
Expert Commentary
This study elegantly demonstrates that incorporating treatment data into glaucoma case definitions within population biobanks substantially enhances phenotypic accuracy, reflected by stronger observed effect sizes for important risk factors and biomarkers. Given the heterogeneity and often incomplete clinical documentation in electronic health records, relying solely on diagnosis codes or self-report may misclassify individuals, including mild or suspect cases without established disease or treatment.
The higher odds ratios associated with treatment-based definitions suggest that these criteria better capture clinically relevant glaucoma cases, including those with more definitive disease requiring therapeutic intervention. Consequently, epidemiological and genetic studies leveraging biobank data should consider the inclusion of treatment information to mitigate misclassification bias, improve power for detecting true associations, and refine risk stratification models.
Nonetheless, some limitations merit mention. The study utilized cross-sectional and longitudinal data, but residual confounding from unmeasured factors remains possible. The UK Biobank population has known demographic biases toward healthier, middle-aged individuals, which may limit generalizability to populations with different ancestry or healthcare access. Also, treatment-based definitions may exclude early or untreated glaucoma cases, potentially biasing toward more advanced disease phenotypes.
Future work incorporating imaging data, functional testing, and longitudinal clinical follow-up would further refine glaucoma phenotyping in large cohorts. Integration of multimodal data could better discriminate glaucoma subtypes, track progression, and personalize risk prediction.
Conclusion
In conclusion, this large-scale UK Biobank study highlights that glaucoma definitions incorporating both diagnosis and treatment information yield stronger and presumably more accurate associations with established risk factors and severity biomarkers than diagnosis-based definitions alone. These findings carry important implications for research methodologies in glaucoma epidemiology, genetics, and biomarker discovery using population biobanks. Careful case definition including treatment evidence should be prioritized to enhance data validity, supporting improved translational insights and ultimately better patient outcomes.
Funding and Clinical Trials Registration
Details on funding sources and clinical trial registration are not specified in the current publication. Researchers utilizing biobank data should refer to UK Biobank resource documentation for relevant information.
References
- W S, A K, E A, As B, J D, Hv DM. Glaucoma in UK Biobank: A Comparison of Diagnostic- Versus Treatment-based Definitions. Ophthalmology. 2026 Jun 17. PMID: 42309491.
- Tham YC, Li X, Wong TY, et al. Global prevalence of glaucoma and projections of glaucoma burden through 2040: a systematic review and meta-analysis. Ophthalmology. 2014;121(11):2081-2090.
- Quigley HA, Broman AT. The number of people with glaucoma worldwide in 2010 and 2020. Br J Ophthalmol. 2006;90(3):262-267.
- George P, Vijaya L. Genetic factors in glaucoma: role of genome-wide association studies. Curr Opin Ophthalmol. 2019;30(2):109-115.
