Ozempic, Wegovy and vision: what the research actually says

No, there is no reason to stop Ozempic or Wegovy because of what you may have read about the eyes. A signal does exist: it concerns a rare disorder of the optic nerve called NAION, and the medicines agencies are taking it seriously. But it remains very rare in absolute terms, it justifies neither systematic screening beforehand nor stopping treatment as a precaution, and what really protects you comes down to one thing: knowing how to recognise the signs that mean you must be seen without delay.

THE STARTING POINT

Where does the concern about Ozempic and Wegovy come from?

In July 2024, an American neuro-ophthalmology team published in JAMA Ophthalmology an analysis linking semaglutide, the molecule in Ozempic and Wegovy, to an excess of NAION cases (PMID 38958939). The news travelled around the world within a few days. Since then the question has come up in consultation almost every week, often put like this: can my treatment make me lose my sight?

NAION, or non-arteritic anterior ischaemic optic neuropathy, is a sudden failure of the blood supply to the head of the optic nerve. The nerve suffers for want of enough blood, rather like a heart attack, but on a scale of a few millimetres. The consequence is a loss of vision in one eye, sudden in onset, often partial, and most often permanent. There is at present no treatment of proven effectiveness.

This condition is not new and did not wait for these medicines in order to exist. It occurs mainly after the age of 50, in people who have several vascular risk factors together: high blood pressure, diabetes, sleep apnoea, smoking, along with an anatomical feature, a small optic nerve head with little cupping, which ophthalmologists call a disc at risk.

That is where the whole difficulty of the exercise lies. Semaglutide is now widely prescribed, including in France, where its use has been the subject of a national study (PMID 39930059). When a medicine is given to hundreds of thousands of people who already have the risk factors for a disease, a few cases will necessarily occur. The whole question is whether more of them occur than the underlying profile explains on its own.

THE CONTROVERSY

Why was this study questioned by its peers?

The scientific community’s response was swift and public. As early as December 2024, the same journal published seven critical letters simultaneously, together with the authors’ reply (PMID 39480403, 39480405, 39480407, 39480409, 39480440, 39480446, 39480447, 39480455). Seven commentaries on a single paper is unusual: the subject was judged important, and the method open to question.

The objections bear on specific points, and it is useful to know them in order to read press headlines at the right distance.

• The population studied is that of a specialist referral centre, where disorders of the optic nerve are concentrated by construction: such a recruitment does not represent the general population.
• The number of events observed remains small, which makes the risk estimate unstable.
• The comparison group chosen and the definition of exposure to the medicine were judged open to challenge.
• Not all the vascular factors are taken into account: obesity, sleep apnoea and diabetes are both reasons for prescribing and risk factors for NAION.
• The referral pathway that brought patients to this centre may have favoured the inclusion of cases exposed to the treatment.

Among these letters, one is signed by an ophthalmologist practising at the Hôpital National des Quinze-Vingts, in Paris. The authors of the original study replied, maintained that a signal exists, and acknowledged that the size of the effect measured in their centre could not be transposed as it stands to the general population.

This point deserves to be said plainly, because it is almost always missing from alarmist articles: this publication is not the conclusion of the matter, it is its starting point. It did its job, which was to raise the alarm. The next job was to check, and it has been done.

THE STATE OF THE EVIDENCE

What do the studies published since then show?

Two years on, the picture is mixed, and it is mixed in both directions. Several large databases find a comparable signal (PMID 39491755, 40146102, 40788646). Other work, carried out on other populations and with other methods, finds no significant excess risk (PMID 40098249, 40015592, 40742732). Neither camp has a monopoly on the data.

What the meta-analyses say

The syntheses published since then converge on an order of magnitude: a relative risk close to 2, that is to say roughly doubled compared with people not on the treatment (PMID 42166479, 42060879, 41475544, 41692115). This convergence is real, and it would be dishonest to play it down.

But those same meta-analyses declare a low to moderate certainty of evidence. The reason is simple: most of them pool observational studies, not randomised trials designed to answer this question. An observational study cannot rule out that part of the risk attributed to the medicine in fact belongs to the state of health that led to it being prescribed. For its part, the meta-analysis that brings together 78 randomised trials, that is more than 73,000 participants, has too few cases of NAION to reach a conclusion: the association does appear in it, but with a very wide confidence interval (PMID 40810985).

What is still awaited

A dedicated trial, planned to follow its participants over five years, is under way. Its results have not been published to date. Until they are, the honest way to put it remains this: the signal is probable, its size remains uncertain, and its translation into a number of cases remains small.

ORDERS OF MAGNITUDE

A doubled risk, yes, but doubled from what?

A relative risk means nothing as long as the starting risk is unknown. After the age of 50, the incidence of NAION is estimated at around 10 cases per 100,000 people per year in the available population studies (PMID 38358828, 9186104). It is therefore a rare disease, and that is the figure to have in mind before reading anything at all about a doubled risk.

• Starting point: around 10 cases per 100,000 people per year after the age of 50, or about 1 case per 10,000.
• Risk roughly doubled: that adds something in the order of 10 cases per 100,000 people treated per year.
• Put another way: about one additional case per 10,000 people treated for a year.
• The European Medicines Agency concluded on 6 June 2025 by adopting the frequency “very rare”, an order of magnitude compatible with this calculation.
• Out of 10,000 people treated for a year, the vast majority will develop no disorder of the optic nerve.

Three approaches arrive at the same order of magnitude: the epidemiology of the disease, the relative risk from the meta-analyses, and the classification adopted by the European regulator. These approaches are not independent — the regulator’s decision rests on the same data as the first two — but they overlap without contradicting one another, which makes the order of magnitude plausible.

“Doubled” and “very rare” therefore do not contradict each other. Doubling a very small quantity gives a quantity that stays very small. That is exactly the situation here, and it is what sensational headlines conjure away, because relative risk lends itself to a hook better than absolute risk does.

This risk, which is very rare, does not call into question the value of these treatments, whose indication is a matter for your doctor. What is a matter for the ophthalmologist is to recognise the signs that should lead you to be seen without delay, and to provide the retinal monitoring that is recommended in any case for anyone with diabetes.

Dr Moïse Tourabaly

ALERT

The signs that should send you for advice without waiting

This is the genuinely actionable part of the subject. NAION gives no warning and does not hurt. It shows itself as a sudden loss of vision in one eye, very often noticed on waking, with no redness and no pain in the eye.

• A sudden loss of vision in one eye, coming on within a few hours, frequently discovered on getting up.
• Loss of part of the visual field, often the lower half, described as a curtain or a fixed shadow.
• Vision veiled or darkened on one side only, while the other eye sees normally.
• An eye that is neither red nor painful: the absence of pain must on no account be taken as reassuring.
• Colours and contrasts that look washed out on one side only.

The simple reflex to remember: close one eye, then the other, in turn. Many one-sided disorders go unnoticed for several days because the healthy eye compensates in binocular vision. If the difference between the two eyes is clear, you need to be seen.

Faced with these signs, what to do is to seek an urgent ophthalmological opinion, the same day if possible, without waiting for the next scheduled appointment. Examining the fundus makes it possible to study the appearance of the optic nerve, often swollen on the affected side, and to rule out the other causes of sudden loss of vision, some of which call for urgent treatment. The main one is giant cell arteritis (Horton’s disease), which after the age of 50 produces the same picture — sudden loss, one eye, no pain — but is often accompanied by temporal headaches, pain in the jaw on chewing or a general deterioration in overall health. It calls for urgent corticosteroid treatment, failing which the second eye is at risk.

One important point to finish with: this message is not addressed only to people treated with GLP-1 analogues. Any sudden loss of vision in one eye warrants a prompt examination, whatever the treatment being taken.

MECHANISM

Is it known how these molecules might act on the optic nerve?

It has to be said plainly: the mechanism has not been elucidated. The hypothesis repeated everywhere assumes a direct action of the medicine on the optic nerve, by way of the GLP-1 receptor. Yet the work that has looked for this receptor in the human retina found it only in a small fraction of the ganglion cells, those whose axons form the optic nerve, and did not find it in advanced retinopathy (PMID 29707863).

In other words, the anatomical basis of the leading hypothesis is fragile. Other work has explored the role of GLP-1 in the retina without settling the matter, to the point that a review article framed the question as an alternative: ally or adversary (PMID 28533296, 41857480).

Other lines of explanation are discussed, with no demonstration to date: a rapid fall in blood sugar and weight, variations in the perfusion of the optic nerve head, a drop in blood pressure during the night. None is established.

Acknowledging that we do not know does not weaken the argument, it places it. An undemonstrated mechanism does not prove the absence of an effect, but it forbids asserting a direct causal link and presenting it to the public as settled.

THE OTHER SIDE

Do these treatments also have favourable effects in ophthalmology?

Reading this whole file as a case for the prosecution would be a mistake, because the same class of molecules has a documented favourable side. A randomised double-blind placebo-controlled trial has shown that exenatide, another GLP-1 analogue, lowers intracranial pressure in idiopathic intracranial hypertension (PMID 36907221, 38212401), a condition that also threatens the optic nerve and mainly affects young women who are overweight.

On the general side, the cardiovascular safety trial of semaglutide showed a benefit on major cardiovascular events in patients with type 2 diabetes (PMID 27633186). This trial measured no favourable eye effect at all: its only ophthalmological result is an increase in complications of retinopathy, described in the next paragraph. It is by another route that blood sugar control counts for the eyes — lasting blood sugar control is, in diabetes guidelines, a recognised protective factor for the retina.

The same trial reported a significant increase in complications of diabetic retinopathy — vitreous haemorrhage, blindness, or the need for an intravitreal injection or photocoagulation — with a hazard ratio of 1.76 (95% confidence interval: 1.11 to 2.78) — this figure covers the trial as a whole. A post hoc analysis showed that the excess mainly concerned patients who already had retinal damage and whose blood sugar fell rapidly (PMID 29178519). This phenomenon of transient worsening has been known for a long time and is not peculiar to this molecule: it accompanies any rapid improvement in HbA1c, whatever the means used. It does not argue for giving up the treatment, it argues for monitoring the retina during the first months in people who are already affected.

IN PRACTICE

Is an eye examination needed before starting the treatment?

Not routinely: this risk does not, on its own, justify an eye examination before or during treatment. On stopping treatment, the American Academy of Ophthalmology and the North American Neuro-Ophthalmology Society (NANOS) write that they do not support any general rule for stopping it, and they point to a decision shared between the patient and their care team. The European Medicines Agency specifies the case in which stopping is required: if NAION is confirmed, semaglutide should be stopped.

In concrete terms, someone starting Ozempic or Wegovy does not have to ask for a specific eye examination on account of this risk. What does remain recommended, and already was before this controversy, is regular retinal monitoring for anyone with diabetes.

• If you have diabetes: fundus examination or retinal photography at the interval set by your ophthalmologist, as a general rule once a year.
• Diabetic retinopathy already known and a rapid fall in HbA1c: close monitoring during the first months, to be arranged with the team following you.
• Previous NAION in one eye, occurring before the treatment: mention it to the prescribing doctor and to the ophthalmologist; the decision is taken case by case.
• NAION confirmed during treatment: the European Medicines Agency asks for semaglutide to be stopped, a step to be arranged with the prescribing doctor.
• Sudden loss of vision in one eye: urgent ophthalmological opinion, without waiting for the next appointment.

As for the treatment decision itself, it belongs to the doctor who prescribes: general practitioner, diabetes specialist or endocrinologist. An ophthalmologist has neither the overall view of the metabolic file nor the standing to stop a treatment whose benefit plays out on the heart, the kidneys and life expectancy.

Frequently asked questions

Cachan practice · Tel. +33 1 45 47 08 11

Scientific sources

• Matched cohort study from a neuro-ophthalmology centre linking semaglutide and NAION, JAMA Ophthalmology, 2024. PMID 38958939
• Seven critical letters published simultaneously and the authors’ reply, JAMA Ophthalmology, December 2024. PMID 39480403, 39480405, 39480407, 39480409, 39480440, 39480446, 39480447, 39480455.
• Cohort studies finding a signal between GLP-1 analogues and NAION. PMID 39491755, 40146102, 40788646.
• Cohort studies finding no significant excess risk. PMID 40098249, 40015592, 40742732.
• Meta-analysis of the risk of NAION with GLP-1 receptor agonists. PLoS Med 2026. PMID 42166479
• Meta-analysis and systematic review, semaglutide and ischaemic optic neuropathy. Neurology 2026. PMID 42060879
• Meta-analysis of 78 randomised trials (73,640 participants), GLP-1 receptor agonists and NAION. JAMA Ophthalmol 2025. PMID 40810985
• European Medicines Agency, Pharmacovigilance Risk Assessment Committee (PRAC): NAION, a very rare side effect of semaglutide; treatment to be stopped if it is confirmed. Press release of 6 June 2025. ema.europa.eu
• Meta-analysis of observational studies, relative risk of NAION with GLP-1 analogues. Ophthalmology 2026. PMID 41475544
• Recent quantitative synthesis on the association between GLP-1 and NAION. Ophthalmology 2026. PMID 41692115
• Population-based incidence study of NAION (Olmsted County, Rochester), 2024. PMID 38358828
• Historical population-based incidence study of NAION, 1997. PMID 9186104
• Expression of the GLP-1 receptor in the human and animal retina, 2018. PMID 29707863
• Review of the role of GLP-1 in the retina, framed as an alternative, “ally or adversary”. Diabetes 2017. PMID 28533296
• Recent update on the retinal and neuro-ophthalmological effects of GLP-1 analogues. J Chin Med Assoc 2026. PMID 41857480
• Randomised double-blind placebo-controlled trial: exenatide and intracranial pressure in idiopathic intracranial hypertension. Brain 2023. PMID 36907221
• Further analysis on GLP-1 analogues and idiopathic intracranial hypertension. Eye (Lond) 2024. PMID 38212401
• Randomised cardiovascular safety trial of semaglutide (SUSTAIN-6), New England Journal of Medicine, 2016. PMID 27633186
• National study of semaglutide (Wegovy) use in France, 2025. PMID 39930059

This article is intended for information and does not replace a medical consultation. Indications, expected benefits and risks depend on each individual situation. For any question about your vision, consult an ophthalmologist.