Semaglutide and NAION: the data debate

A letter published on 27 August 2026 in JAMA Ophthalmology questions the strength of the data behind semaglutide's association with nonarteritic anterior ischemic optic neuropathy (NAION), a form of vision loss caused by reduced blood flow to the optic nerve. The letter, by Yu CH^[1], arrives about 14 months after the EMA's Pharmacovigilance Risk Assessment Committee concluded that NAION is a very rare side effect of semaglutide products (Ozempic, Rybelsus, Wegovy) and triggered a label update.

The core concern is methodological: do the available observational datasets establish a true causal relationship, or are there data-quality and timing problems that inflate the apparent risk?

What NAION is and why it matters

NAION is the most common acute optic nerve disorder in adults over 50. The optic disc loses blood supply, usually overnight, resulting in sudden painless vision loss, often a segment of the visual field. Most cases are unilateral. There is no proven treatment; the loss is typically permanent.

Background incidence is roughly 2 to 10 cases per 100,000 person-years in the general population. Several vascular risk factors are well established: hypertension, diabetes, dyslipidaemia, sleep apnoea, a structurally crowded (small cup-to-disc ratio) optic disc. The population most likely to be prescribed semaglutide carries many of these same risk factors, which is central to the confounding debate.

The signal that started the regulatory review

The NAION concern traces to a July 2024 retrospective cohort study published in JAMA Ophthalmology by Hathaway and colleagues at Mass Eye and Ear. The study used a large US insurance claims database and found that semaglutide-prescribed patients had roughly 4-fold higher odds of NAION compared to non-GLP-1 users with the same underlying condition, whether that condition was type-2 diabetes or overweight/obesity. A 2026 systematic review and meta-analysis^[2] covering multiple observational datasets similarly found elevated odds across studies.

None of the primary studies are randomised trials. Every result comes from claims data, electronic health records, or pharmacovigilance databases. This is exactly the point Yu's letter presses.

The data concerns Yu raises

The letter to JAMA Ophthalmology^[1] identifies two overlapping problems. The first is confounding by indication. Semaglutide is prescribed for diabetes and obesity, conditions that independently carry elevated NAION risk through shared vascular pathways. Isolating the drug's contribution from the underlying disease burden requires careful matching or adjustment that is difficult in claims data, where the exposure and the confounders often appear in the same billing codes.

The second concern is temporal. NAION has no ICD code of its own in many older coding systems; it is captured under broader optic neuropathy or ischemic optic neuropathy categories. If semaglutide patients receive more ophthalmic follow-up simply because they are more actively managed, NAION will be detected at higher rates in that group regardless of whether the drug causes it. The detection bias is hard to distinguish from a real pharmacological signal in retrospective data.

A separate 2026 cohort analysis

Ziegler and colleagues reported a hazard ratio of 2.33 (95% CI 1.54 to 3.54) for NAION in semaglutide initiators^[3] with an incidence rate of 123 cases per 100,000 person-years in exposed patients. That absolute rate is much higher than background incidence in the general population, but the enrolled population was enriched for diabetes and cardiovascular disease, conditions where NAION baseline risk is already elevated. Whether the HR reflects the drug or the patients taking the drug is the open question.

The EMA's regulatory response

The EMA's Pharmacovigilance Risk Assessment Committee reviewed the available evidence and in June 2025 concluded that NAION is a very rare side effect of semaglutide medicines^[4]. Very rare in EMA terminology means fewer than 1 in 10,000 patients. The label for Ozempic, Rybelsus, and Wegovy was updated to include NAION in the adverse reaction table at that frequency. The product information for Ozempic was last revised in August 2026.

The label update is a precautionary step consistent with how regulators handle emerging pharmacovigilance signals: the standard is not proof of causation but a plausible association that warrants transparency with prescribers. Yu's letter does not argue the label update was wrong; it argues the underlying evidence base deserves closer scrutiny before the association is treated as settled.

What the reply from the original authors says

The JAMA Ophthalmology correspondence includes a reply by Heberer, Bress, and Lee^[5] (the authors of the original cohort work). The reply defends the analytical approach and argues that the active comparator designs used, comparing semaglutide initiators to initiators of other drugs for the same indication, substantially reduce confounding by indication. They maintain that the consistency across multiple independent datasets, using different data sources and methods, strengthens the case for a real association.

The disagreement is not about whether semaglutide is harmful but about what the available observational data can and cannot prove. Both sides agree that a prospective study designed specifically to test the NAION hypothesis would settle the question better than retrospective record linkage.

What this means for patients and prescribers

NAION is now listed on the semaglutide label as a very rare adverse reaction, which is the current regulatory reality. Prescribers are expected to tell patients about the risk; patients who notice sudden vision changes while on semaglutide should seek urgent ophthalmological review. That clinical guidance holds regardless of the methodological debate.

The letter's significance is for the research agenda, not for individual clinical decisions today. If the association is partly or wholly explained by detection bias or confounding, future studies with better designs will show that. If it is real, better studies will quantify it more precisely. Neither outcome changes the current label.

What we do not yet know

No prospective ophthalmological monitoring study exists for semaglutide at scale. No randomised trial has NAION as a pre-specified endpoint. The mechanistic story is also incomplete: semaglutide's effect on blood pressure, platelet aggregation, and microvascular perfusion are all plausible pathways, but none has been traced to optic disc ischemia in controlled conditions. The absolute risk, even if real, may be small enough that the cardiovascular, metabolic, and mortality benefits of semaglutide outweigh it for most patients. That calculation requires better numbers than the field currently has.