Highlights
A retrospective multi-center cohort analysis using TriNetX data from 2015 to 2024 found that patients prescribed NSAIDs had a lower subsequent risk of age-related macular degeneration (AMD) than matched non-users.
The association was consistent across multiple follow-up windows, with hazard ratios ranging from 0.31 at 6 months to 0.48 at 5 years, and an overall study-period hazard ratio of 0.58.
Risk reduction signals were also reported for both non-exudative AMD and exudative AMD, and in subgroup analyses of aspirin and non-selective COX inhibitors.
Despite the very large sample size and matching strategy, the study remains observational and cannot establish causality; confounding by indication, exposure misclassification, and surveillance bias remain important concerns.
Background
Age-related macular degeneration is a leading cause of irreversible central vision loss in older adults. Its public health burden is substantial because AMD directly affects reading, driving, facial recognition, and independence. Clinically, AMD spans a broad continuum from early drusen and retinal pigment epithelium abnormalities to late-stage non-exudative geographic atrophy and exudative, neovascular disease. Although intravitreal anti-VEGF therapy has transformed outcomes for neovascular AMD, preventive strategies remain limited, and the search for modifiable risk factors continues.
Biologically, the hypothesis that nonsteroidal anti-inflammatory drugs might influence AMD risk is plausible. Inflammation, complement activation, oxidative stress, and dysregulated lipid handling all contribute to AMD pathogenesis. Cyclooxygenase-mediated inflammatory signaling has also been implicated in retinal disease, making anti-inflammatory drug exposure an attractive epidemiologic question. At the same time, prior literature has been mixed, particularly for aspirin, where some observational studies have suggested harm, neutrality, or context-specific effects rather than clear protection. This inconsistency makes large real-world datasets useful, but also underscores the need for careful interpretation.
Study Design and Methods
Hsu and colleagues conducted a retrospective cohort study using the TriNetX deidentified multi-center electronic medical records network. The analysis covered the period from January 1, 2015, through December 31, 2024. The investigators identified patients with prescriptions for NSAIDs and compared them with individuals who had never received NSAID prescriptions and had no prior history of AMD.
The study enrolled 634,794 patients in the NSAID-exposed cohort and 634,794 patients in the non-exposed comparator cohort after 1:1 propensity score matching. Mean age was 59.93 ± 11.95 years in NSAID users and 59.68 ± 12.15 years in non-users. Matching variables included age, sex, race, smoking, and selected adult comorbidities considered relevant to future AMD risk.
The main outcome was incident AMD after the index date. The authors reported cumulative incidence and hazard ratios at prespecified time points including 6 months, 1 year, 3 years, and 5 years, as well as across the total study period. Secondary analyses examined aspirin exposure, non-selective COX-2 inhibitor exposure as described by the authors, and AMD subtype outcomes including non-exudative and exudative AMD.
Key Results
Overall association between NSAID prescriptions and AMD risk
The principal finding was a statistically significant association between NSAID prescription exposure and lower future AMD incidence compared with non-use. The effect estimate was strongest early and attenuated over time, although it remained below 1.0 throughout follow-up:
At 6 months, the hazard ratio for AMD was 0.31 (95% CI, 0.27-0.36). At 1 year, the hazard ratio was 0.36 (95% CI, 0.33-0.39). At 3 years, the hazard ratio was 0.42 (95% CI, 0.40-0.44). At 5 years, the hazard ratio was 0.48 (95% CI, 0.47-0.50). Across the entire study duration, the hazard ratio was 0.58 (95% CI, 0.56-0.59).
Several observations deserve attention. First, the consistency of the direction of association across time points is notable. Second, the attenuation over time may reflect diminishing treatment separation, changing exposure over follow-up, or survivor and selection effects. Third, because the cohort is extremely large, precision is high and confidence intervals are narrow; however, statistical precision does not eliminate systematic bias.
Drug subgroup analyses
The authors also reported a protective association for specific NSAID categories. During the full study duration, patients prescribed aspirin had a lower AMD risk than non-users, with a hazard ratio of 0.72 (95% CI, 0.69-0.75). For patients prescribed non-selective COX-2 inhibitors, the reported hazard ratio was 0.41 (95% CI, 0.36-0.46) compared with respective non-users.
These subgroup findings may be clinically interesting, but they should be interpreted cautiously. Aspirin is often prescribed for cardiovascular prevention and may mark a patient population with different healthcare utilization, vascular risk management, or concurrent medication patterns. Similarly, prescription of non-aspirin NSAIDs may be linked to musculoskeletal disease, pain diagnoses, mobility, and health-system contact in ways that could affect AMD coding or ascertainment.
AMD subtype analyses
The association extended to both major AMD phenotypes. For non-exudative AMD, the hazard ratio after index NSAID prescription was 0.56 (95% CI, 0.55-0.58). For exudative AMD, the hazard ratio was 0.62 (95% CI, 0.59-0.65). This broad pattern could support a generalized anti-inflammatory signal, but it could also indicate non-specific differences in baseline risk between exposed and unexposed groups.
Clinical Interpretation
On first reading, the magnitude of association appears striking. Hazard ratios of this size, especially 0.31 at 6 months and 0.58 overall, would suggest a potentially important preventive effect if causal. Yet clinicians should resist the temptation to translate this immediately into prescribing behavior.
Why the caution? Because observational pharmacoepidemiology is highly vulnerable to hidden bias. AMD develops over years, while NSAID prescribing is often episodic and linked to unrelated conditions. If exposed patients differ from non-exposed patients in ways not fully captured by the matching algorithm, the estimated drug effect may be exaggerated. Important ophthalmic determinants of AMD risk, such as family history, detailed smoking intensity, nutritional status, sunlight exposure, baseline drusen burden, genetic polymorphisms including CFH and ARMS2, and frequency of eye examinations, are difficult to account for reliably in routine EMR data.
The especially strong early associations raise additional questions. A dramatic apparent benefit within 6 months is biologically possible only if NSAIDs alter near-term diagnosis, symptom recognition, or coding patterns, or if the groups were already meaningfully different at baseline despite matching. Since AMD is often indolent before diagnosis, a very rapid divergence may suggest ascertainment or selection effects rather than pure disease modification.
Strengths of the Study
The study has several important strengths. Its sample size is exceptionally large, which improves statistical power and allows analyses across multiple time horizons and disease subtypes. Use of a multi-center dataset enhances geographic and practice-setting diversity relative to a single-institution cohort. The exclusion of patients with prior AMD focuses the analysis on incident disease. Propensity score matching on demographic factors, smoking, and common comorbidities is an appropriate attempt to reduce measured confounding.
From a translational standpoint, the study is valuable because it revives interest in inflammation as a potentially modifiable pathway in AMD and provides a real-world signal worthy of further scrutiny.
Limitations and Methodological Caveats
The limitations are substantial and central to interpretation. First, exposure was defined by prescription rather than confirmed medication use. Many NSAIDs are also available over the counter, especially ibuprofen, naproxen, and aspirin, so the non-user group may still include substantial actual NSAID exposure. This can bias estimates in unpredictable ways.
Second, dose, duration, adherence, and treatment indication were not detailed in the abstract. These factors matter because short-term analgesic use may not have the same biologic implications as chronic anti-inflammatory exposure.
Third, AMD diagnosis in administrative or EMR datasets depends on coding fidelity and access to eye care. Patients with more frequent ophthalmic evaluation may be more likely to receive AMD diagnoses. Conversely, individuals engaged in care for chronic pain or cardiovascular prevention may differ in specialty access and diagnostic capture.
Fourth, residual confounding is likely. Even after matching, important variables may remain imbalanced or unmeasured. Confounding by healthy user behavior, by healthcare utilization, or by comorbidity severity can meaningfully distort results.
Fifth, the abstract does not report absolute risk differences, which are important for judging clinical relevance. Relative hazard reductions can appear large even when the underlying absolute event rate is low.
Finally, retrospective database studies cannot establish mechanism. A lower observed incidence of coded AMD does not necessarily mean NSAIDs prevented retinal degeneration.
How This Fits With Existing Evidence
The broader AMD literature supports inflammation as a key pathogenic axis, but the clinical evidence for NSAIDs as preventive agents remains unsettled. The Age-Related Eye Disease Study and Age-Related Eye Disease Study 2 established nutritional supplementation strategies for selected patients with intermediate AMD or late AMD in one eye, but did not identify NSAIDs as preventive therapy. Current ophthalmology guidelines do not recommend routine systemic NSAID use for AMD prevention.
Aspirin has been particularly controversial. Some earlier observational studies suggested a possible association between aspirin use and neovascular AMD, whereas others found no meaningful increase in risk after adjustment for cardiovascular confounders. This new report points in the opposite direction, toward protection, highlighting just how sensitive these associations may be to study design, data source, and confounding structure.
Accordingly, the most responsible interpretation is not that NSAIDs prevent AMD, but that this large EMR-based analysis identifies a hypothesis-generating association that should be tested more rigorously.
Implications for Practice and Research
For practicing clinicians, the study should not change current prescribing standards. NSAIDs carry well-known risks, including gastrointestinal bleeding, renal toxicity, hypertension exacerbation, and cardiovascular concerns that vary by agent and patient profile. These established harms preclude any recommendation to initiate systemic NSAIDs solely for AMD risk reduction based on retrospective evidence.
However, the findings may inform future research in several ways. Prospective analyses should examine dose-response relationships, cumulative exposure, and treatment duration. Studies that use new-user designs, active comparators, and lag periods could better reduce bias. Integration of retinal imaging data rather than diagnosis codes alone would strengthen outcome validity. Mendelian randomization or mechanistic studies may help determine whether cyclooxygenase-related pathways are causally linked to AMD onset or progression.
Another worthwhile direction would be to distinguish systemic from topical ophthalmic NSAID exposure, and to explore whether any association differs by genetic risk strata, smoking burden, or baseline drusen status. Ultimately, randomized trials would be needed before preventive recommendations could be considered, although feasibility may be challenging given safety considerations and long latency to clinical AMD endpoints.
Conclusion
This large retrospective cohort study reports that patients prescribed NSAIDs had a lower subsequent risk of AMD than propensity-matched non-users, with consistent signals across follow-up intervals, drug subgroups, and both non-exudative and exudative AMD. The findings are intriguing and biologically plausible in light of inflammatory mechanisms in AMD, but they are not practice-changing. The magnitude of association, especially early after exposure, raises concern for residual confounding and other biases inherent to EMR-based observational research. For now, the study is best viewed as an important hypothesis-generating contribution that reinforces the need for better causal research on inflammation-targeted prevention in AMD.
Funding and ClinicalTrials.gov
The abstract provided does not report a funding source. No ClinicalTrials.gov registration number is listed, which is consistent with the retrospective observational design.
References
1. Hsu AY, Kuo HT, Liu DY, Hsia NY, Lai CT, Chiang CC, Lin CJ, Shao YC, Chen HS, Wu BQ, Tseng H, Hsu HC, Wang YH, Bair H, Tsai YY, Hsu MY, Cheng-Chung Wei J. NSAID use and the risk of AMD: Insights From a Multi-Center Cohort Analysis. Ophthalmology. 2026-05-01. PMID: 42070633. Available at: https://pubmed.ncbi.nlm.nih.gov/42070633/
2. Age-Related Eye Disease Study Research Group. A randomized, placebo-controlled, clinical trial of high-dose supplementation with vitamins C and E, beta carotene, and zinc for age-related macular degeneration and vision loss: AREDS report no. 8. Arch Ophthalmol. 2001;119(10):1417-1436. PMID: 11594942.
3. Age-Related Eye Disease Study 2 Research Group. Lutein + zeaxanthin and omega-3 fatty acids for age-related macular degeneration: the Age-Related Eye Disease Study 2 (AREDS2) randomized clinical trial. JAMA. 2013;309(19):2005-2015. PMID: 23644932.
4. Lim LS, Mitchell P, Seddon JM, Holz FG, Wong TY. Age-related macular degeneration. Lancet. 2012;379(9827):1728-1738. PMID: 22559899.
5. Wong WL, Su X, Li X, Cheung CMG, Klein R, Cheng CY, Wong TY. Global prevalence of age-related macular degeneration and disease burden projection for 2020 and 2040: a systematic review and meta-analysis. Lancet Glob Health. 2014;2(2):e106-e116. PMID: 25104651.
6. Cheung CMG, Lai TYY, Ruamviboonsuk P, Chen SJ, Chen Y, Freund KB, Gomi F, Koh A, Lee WK, Wong TY. Polypoidal choroidal vasculopathy: definition, pathogenesis, diagnosis, and management. Ophthalmology. 2018;125(5):708-724. PMID: 29110808.
