A large-scale study published this week in The BMJ has delivered what may be the most consequential safety signal yet for the blockbuster class of GLP-1 receptor agonist drugs — medications like semaglutide (sold as Ozempic and Wegovy) and tirzepatide (Mounjaro and Zepbound) that have become the fastest-growing pharmaceutical category in history. The finding: patients using these drugs face a meaningfully elevated risk of a condition called non-arteritic anterior ischemic optic neuropathy, or NAION, a form of sudden vision loss that has no proven treatment.
Not a rare curiosity. A genuine clinical concern.
The study, led by researchers at Harvard Medical School and Massachusetts Eye and Ear, analyzed data from more than 100 million patients across two major U.S. medical records databases. After adjusting for diabetes status, obesity, cardiovascular risk factors, and other confounders, the researchers found that patients prescribed semaglutide had roughly two to three times the odds of developing NAION compared with matched controls. Tirzepatide showed a similar, though somewhat less pronounced, signal. The association held across both diabetic and non-diabetic populations — a critical distinction, since millions of people now take these drugs purely for weight loss, without any underlying metabolic disease.
NAION occurs when blood flow to the optic nerve is suddenly disrupted, causing painless but often permanent vision loss in one eye. It’s the most common acute optic neuropathy in people over 50, yet it remains poorly understood. There’s no FDA-approved therapy. No way to reverse the damage once it occurs. And while the absolute incidence is low in the general population, the relative risk increase identified in this study is large enough to warrant serious attention — particularly given that an estimated 40 million or more prescriptions for GLP-1 drugs were written in the United States last year alone.
The Harvard team, led by ophthalmologist Joseph Rizzo III, had first flagged this association in a smaller retrospective study published in 2024. That earlier work, which appeared in JAMA Ophthalmology, examined records from a single institution and found a striking increase in NAION diagnoses among patients on semaglutide. But it was limited in scope — roughly 17,000 patients — and critics rightly pointed out the possibility of confounding, referral bias, and the inherent limitations of single-center data. The new BMJ study was designed to address those objections head-on.
And it does. The scale is orders of magnitude larger. The researchers used both the TriNetX Diamond network and a separate federated database, capturing patients from hundreds of health systems across the country. They employed active comparator designs — pitting GLP-1 users against patients taking other diabetes or obesity medications — to reduce confounding by indication. They ran sensitivity analyses, subgroup analyses, and time-stratified analyses. The signal persisted through all of them.
The hazard ratios varied depending on the specific comparison group and database, but the pattern was consistent. In the primary analysis comparing semaglutide users with non-users matched on key clinical characteristics, the adjusted odds ratio for NAION ranged from approximately 2.1 to 3.6. For tirzepatide, the estimates were somewhat lower but still elevated. The risk appeared highest in the first year after initiating therapy, suggesting a possible acute mechanism rather than a slow cumulative effect.
What might explain the link? The authors propose several biological hypotheses. GLP-1 receptor agonists cause rapid weight loss, which can alter hemodynamics and potentially affect perfusion to vulnerable vascular beds like the optic nerve head. The drugs also lower blood pressure, sometimes substantially, which could reduce perfusion pressure in an already compromised optic disc. There’s also emerging evidence that GLP-1 receptors are expressed in retinal and optic nerve tissue, raising the possibility of a direct pharmacological effect — though this remains speculative.
Another possibility: the so-called “disc at risk.” Patients with small, crowded optic nerve heads — a common anatomical variant — are known to be predisposed to NAION. If GLP-1 drugs alter vascular dynamics even modestly, they could push these anatomically vulnerable individuals over a threshold. The study couldn’t assess optic disc anatomy, which the authors acknowledge as a limitation.
Novo Nordisk, the manufacturer of Ozempic and Wegovy, has previously stated that its clinical trial data do not show an increased risk of NAION and that the company takes all safety signals seriously. Eli Lilly, which makes Mounjaro and Zepbound, has made similar statements. Neither company has issued a formal response to the new BMJ publication as of this writing. But the pressure is mounting. The FDA has been aware of the earlier JAMA Ophthalmology findings and is reportedly monitoring post-marketing safety data related to ocular events.
The timing is significant. GLP-1 drugs are no longer niche diabetes treatments. They’ve become cultural phenomena, prescribed to millions of otherwise healthy adults seeking weight loss. Celebrities tout them. Telehealth startups have built entire business models around prescribing them. Wall Street analysts project the GLP-1 market will exceed $100 billion in annual sales by the end of the decade. Novo Nordisk’s market capitalization briefly surpassed that of every company in Europe. The financial stakes are enormous — and so are the public health stakes if a serious adverse effect has been underappreciated.
The BMJ study doesn’t prove causation. The authors are careful to note this. Observational data, no matter how large or well-analyzed, can’t definitively establish that a drug causes a particular outcome. Residual confounding is always possible. Patients prescribed GLP-1 drugs may differ from controls in ways that aren’t captured in electronic health records — lifestyle factors, genetic predispositions, healthcare-seeking behavior. But the consistency of the signal across multiple databases, comparison groups, and analytic approaches makes it increasingly difficult to dismiss.
“This is no longer a hypothesis based on a handful of cases,” Rizzo told reporters. “The data are telling us something real.”
Some ophthalmologists have already begun adjusting their clinical practice. Doctors at major academic medical centers report asking patients about GLP-1 use during neuro-ophthalmology consultations. A few have started recommending baseline eye exams before initiating therapy, particularly for patients with known risk factors for NAION — small optic discs, sleep apnea, hypertension, diabetes. Whether such screening would actually prevent cases is unknown, but the impulse reflects genuine clinical unease.
The broader question is what regulators and drug manufacturers should do with this information. Updated labeling is one option. The FDA could require a warning about potential ocular risks, similar to the existing warnings about thyroid tumors (based on animal data) and pancreatitis. A more aggressive step would be mandating a post-marketing surveillance study specifically designed to assess NAION risk in a prospective cohort — something that would take years but could provide definitive answers.
Industry analysts have so far treated the NAION signal as a manageable concern rather than an existential threat to the GLP-1 franchise. The absolute risk remains small. Most patients taking these drugs will never develop NAION. But “small absolute risk” multiplied by tens of millions of users produces a non-trivial number of affected individuals. And unlike many drug side effects, NAION causes irreversible harm. You don’t recover from it.
There’s a historical parallel worth considering. When statins were first linked to muscle-related side effects, the initial response from manufacturers and some clinicians was dismissal — the benefits were too great, the absolute risk too small. Over time, the medical community came to recognize that statin-associated myopathy was real, clinically significant for a subset of patients, and worthy of informed consent discussions. The GLP-1/NAION question may follow a similar arc.
The BMJ paper also raises uncomfortable questions about the adequacy of pre-approval clinical trials for detecting rare but serious adverse events. Phase III trials for semaglutide and tirzepatide enrolled thousands of patients, but NAION is uncommon enough that even a doubled or tripled risk might not produce a statistically detectable signal in a trial population of 3,000 to 5,000. This is a well-known limitation of the clinical trial framework — one that post-marketing surveillance is supposed to address. Whether the current pharmacovigilance system is up to the task, given the speed and scale of GLP-1 adoption, is an open question.
Meanwhile, the drugs keep selling. Novo Nordisk reported semaglutide sales of approximately $21 billion in 2024. Eli Lilly’s tirzepatide franchise is growing even faster. Compounding pharmacies have rushed to fill supply gaps with versions of questionable quality, adding another layer of risk. The FDA recently moved to restrict compounded semaglutide, but enforcement remains spotty.
For patients, the calculus is personal. GLP-1 drugs offer real, substantial benefits — not just weight loss, but reductions in cardiovascular events, improvements in fatty liver disease, and emerging evidence of benefits in heart failure and kidney disease. These aren’t trivial. For many patients, the risk-benefit ratio will still favor treatment. But informed consent requires that patients know about potential risks, even uncommon ones. Especially irreversible ones.
So where does this leave clinicians? In a familiar but uncomfortable position: making decisions with incomplete information, weighing population-level data against individual patient circumstances, and hoping that the next study provides clarity rather than more ambiguity. The BMJ paper doesn’t close the book on GLP-1 drugs and eye safety. It opens a chapter that the medical community can no longer afford to skip.
The evidence is accumulating. The question now is whether anyone — regulators, manufacturers, prescribers — will act on it before the next large-scale study confirms what this one strongly suggests.
WebProNews is an iEntry Publication