Millions of Americans taking GLP-1 receptor agonists for type 2 diabetes or weight loss now face a new safety question. A nationwide study of 588,168 veterans found that patients who started GLP-1 drugs had a roughly 35 percent higher three-year risk of nonarteritic anterior ischemic optic neuropathy, or NAION, compared with those who began a different class of diabetes medication. The finding, drawn from one of the largest real-world datasets ever applied to this question, sharpens a debate that has been building since smaller studies first flagged an optic-nerve signal tied to semaglutide.
Why the VA’s 588,168-patient dataset changes the GLP-1 safety debate
NAION is the most common cause of sudden optic-nerve damage in adults over 50. It cuts blood supply to the optic nerve head, often producing permanent vision loss in one eye. The condition is rare in absolute terms, but it strikes without warning and has no proven treatment. Any drug that shifts its incidence even modestly affects a large number of people when tens of millions of prescriptions are written each year.
The new study, authored by Choi, Al-Aly, and Xie and published in JAMA Network Open, used a target-trial emulation design applied to Veterans Affairs electronic health records spanning January 2017 through December 2024. It compared 139,546 veterans who initiated GLP-1 receptor agonists against 448,622 veterans who initiated SGLT2 inhibitors, a separate class of glucose-lowering drug. The three-year cumulative incidence of NAION reached approximately 39 cases per 100,000 person-years in the GLP-1 group versus roughly 29 per 100,000 in the SGLT2 group.
Importantly, the researchers attempted to balance the two groups on a wide range of baseline factors, including age, sex, cardiovascular history, kidney function, and prior eye disease, to approximate a randomized trial. Even after this adjustment, the excess NAION risk associated with GLP-1 therapy persisted. Because the comparison drug class, SGLT2 inhibitors, is itself widely used and not known to affect the optic nerve, the analysis offers a clinically relevant benchmark rather than a placebo-like control.
One hypothesis worth testing is whether the excess risk stems not from the GLP-1 molecule itself but from the speed at which these drugs lower blood sugar. Rapid reductions in HbA1c have long been suspected of stressing small blood vessels, including those feeding the optic nerve. If patients on insulin or sulfonylureas who achieve equally fast glycemic drops show a similar bump in NAION, the signal would point toward a metabolic mechanism rather than a drug-specific one. The VA study does not report patient-level HbA1c trajectories, leaving this question open.
Another unresolved issue is whether the risk is uniform across all GLP-1 agents or concentrated in specific molecules, doses, or delivery routes. The VA dataset spans a period when multiple GLP-1 drugs were on the market and being used at varying intensities for both diabetes and weight management. Without granular drug-by-drug analyses, clinicians cannot yet tell whether certain formulations might carry lower optic-nerve risk while preserving the well-established cardiovascular and renal benefits of the class.
How earlier semaglutide and meta-analysis findings align with the VA data
The VA results did not arrive in a vacuum. An earlier observational analysis focused specifically on semaglutide and NAION had already reported elevated odds among users of that particular GLP-1 agent, using propensity-score methods and detailed covariate adjustment. That study helped trigger the broader safety conversation and prompted calls for larger confirmatory analyses capable of distinguishing drug effects from background risk.
Separately, a real-world evidence assessment published in the Journal of Diabetes Science and Technology ran multiple retrospective cohort analyses and found that the strength of the NAION signal was sensitive to how tightly researchers matched patients and controlled for comorbidities. In some analytic configurations, the elevated risk attenuated substantially. That sensitivity matters because veterans in the VA system tend to carry heavier burdens of obesity, sleep apnea, and cardiovascular disease, all of which are independent risk factors for NAION.
A broader systematic review and meta-analysis of GLP-1 receptor agonists and ocular disorders, pooling data from several observational databases, showed inconsistent results across studies and highlighted significant heterogeneity in effect sizes. Some datasets suggested a modest increase in optic-nerve events, while others showed no meaningful difference compared with alternative diabetes therapies. A separate critical appraisal in Clinical and Experimental Ophthalmology examined the full chain of evidence for a causal link between GLP-1 drugs and NAION, concluding that observational designs alone cannot rule out confounding and detection bias. Patients on GLP-1 drugs may simply receive more frequent eye exams, raising the chance that NAION gets diagnosed rather than missed.
Viewed together, these strands of evidence paint a nuanced picture. The semaglutide-focused work points toward a possible molecule-specific effect, the VA target-trial emulation suggests a class-level signal in a large, relatively homogeneous population, and the meta-analytic literature underscores how fragile the association can be to study design choices. None of these studies, however, offers the kind of randomized, prospectively adjudicated data that regulators typically rely on to declare a definitive drug safety risk.
Unanswered questions about GLP-1 drugs and optic nerve damage
Several gaps in the evidence remain. The VA study reports outcomes only through three years of follow-up. Whether the risk difference widens, stabilizes, or narrows over longer periods is unknown. The full de-identified dataset and analytic code from the target-trial emulation have not been released for independent re-analysis, limiting the ability of outside researchers to probe residual confounding.
Patient-level details on which specific GLP-1 agents were prescribed, how quickly HbA1c fell after initiation, and why patients were censored from the study are available only in aggregate. Without individual-level data on glycemic trajectories, the rapid-reduction hypothesis cannot be formally tested through mediation analysis. The study also does not include direct author statements addressing how much of the observed difference might reflect detection bias rather than a true biological effect. For now, that leaves clinicians to interpret the findings against a backdrop of incomplete mechanistic understanding.
For patients and clinicians, the practical question is one of absolute versus relative risk. A 35 percent relative increase sounds alarming, but the absolute difference amounts to roughly 10 additional NAION cases per 100,000 person-years compared with SGLT2 therapy in the VA population. Put differently, more than 9,900 out of 10,000 GLP-1 users in the study did not experience NAION over three years. At the same time, NAION’s potential for irreversible vision loss means even small absolute risk changes warrant careful attention, especially in people who already have crowded optic discs, sleep apnea, or prior events in the fellow eye.
Balancing these risks against the proven benefits of GLP-1 drugs is complex. These medications lower blood glucose, promote substantial weight loss, and reduce major cardiovascular events in high-risk patients. For many, those advantages will far outweigh a small, uncertain increase in a rare eye condition. But for individuals at particularly high baseline risk of NAION, such as those with previous optic-nerve ischemia, poorly controlled sleep apnea, or severe nocturnal hypotension, the calculus may tilt differently, and alternative agents or slower titration strategies could be reasonable.
In the near term, experts are likely to focus on pragmatic steps rather than sweeping practice changes. That could include counseling patients about early warning signs of NAION-such as sudden, painless vision loss or dark areas in the visual field-and encouraging prompt ophthalmologic evaluation if they occur. Primary care clinicians and endocrinologists may also consider documenting baseline eye status more systematically before starting GLP-1 therapy in higher-risk individuals, while avoiding unnecessary alarm for the broader population of users.
Ultimately, resolving the GLP-1–NAION question will require more than additional retrospective analyses. Prospective registries with standardized eye examinations, linkage to pharmacy records, and rigorous event adjudication could clarify both incidence and timing. Randomized trials specifically powered for rare ocular outcomes are unlikely, but existing cardiovascular-outcome studies might still yield insights if their imaging and adverse-event data are revisited with NAION in mind. Until then, the VA study adds weight to a growing but still imperfect evidence base, nudging clinicians toward heightened vigilance without yet justifying wholesale abandonment of a transformative class of metabolic drugs.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
