The 30-second summary
- The study: a network meta-analysis in Diabetes, Obesity and Metabolism, published on 21 September 2026, compared GLP-1 medicines for their link to NAION, a rare stroke of the optic nerve.
- The ranking: semaglutide had the highest estimated risk, 1.54 times that of non-GLP-1 treatments. Liraglutide followed at 1.25 times.
- The context: NAION is very rare. Europe's medicines regulator puts it at up to 1 in 10,000 people on semaglutide.
- What to do: nothing new if your sight is fine. Sudden vision loss in one eye means a same-day call.
Where this fits
Our longer guide, GLP-1s and your eyes, walks through why three large studies of semaglutide and NAION reached three different answers. This new paper asks a narrower question: if there is a link, is it the same for every GLP-1 medicine?
NAION stands for nonarteritic anterior ischaemic optic neuropathy. Blood flow to the front of the optic nerve falls short, part of the nerve is damaged, and vision in that eye is usually affected for good.
What did the new study do?
The team, from the pharmacy department of the Eye and ENT Hospital of Fudan University in Shanghai, combined four methods:
- A network meta-analysis. This pools results from earlier studies and compares drugs with each other, even where no study tested them head to head.
- A check against real-world reports from the US FDA Adverse Event Reporting System, known as FAERS.
- Network toxicology, a computer method that looks for proteins and pathways a drug might act on.
- Molecular docking, a computer simulation of how well a drug binds to those proteins.
Only the first two are about people. The last two generate ideas about mechanism and prove nothing about risk.
What did it find?
| Medicine | NAION risk vs non-GLP-1 treatment |
|---|---|
| Semaglutide (Ozempic, Wegovy, Rybelsus) | HR 1.54 (95% CI 1.31 to 1.82) |
| Liraglutide (Victoza, Saxenda) | HR 1.25 (95% CI 1.08 to 1.45) |
In the FDA reports, semaglutide showed a strong reporting signal and liraglutide a weaker one. The computer work pointed to a protein called caspase-3, which is involved in cell death, as a possible link. The authors present all of this as a guide for future research, not a final answer.
How big is a 1.54 hazard ratio?
It sounds large. In absolute terms it is small, because the starting risk is small.
In June 2025, the European Medicines Agency's safety committee concluded that NAION is a very rare side effect of semaglutide, which in its own definition means it may affect up to 1 in 10,000 people. A risk that is half as large again as a very small number is still a very small number.
That is also why the size of the estimate matters. The first study to raise the alarm, from a single neuro-ophthalmology centre in Boston in 2024, found hazard ratios of 4.28 in people with type 2 diabetes and 7.64 in people with overweight or obesity (Hathaway et al.). A multinational study of nearly 300,000 people from 160 healthcare organisations in 21 countries found no clear association at all (Chou et al.). Pooled estimates like this new one sit between the two.
What this study cannot tell us
- The abstract does not say how many studies were pooled, or of what design, and the full paper is behind a paywall. We report only what the abstract states.
- Reporting signals are not rates. FAERS counts reports, not people. A drug that is widely used and widely discussed gathers more reports. The FAERS figure for semaglutide is a signal to investigate, not a measure of how often NAION happens.
- No figure for tirzepatide (Mounjaro, Zepbound) appears in the abstract, so this paper cannot tell you whether it differs.
- Docking and network methods are computer models. They suggest how a drug might act, not whether it does.
- Pooled observational data inherit their flaws. People prescribed semaglutide often have diabetes, sleep apnoea or high blood pressure, which carry their own NAION risk.
Does this change anything if you take Ozempic or Wegovy?
As of October 2026, the advice is the same as it was. The European Medicines Agency says anyone on semaglutide with sudden loss of vision or rapidly worsening eyesight should contact their doctor without delay, and that semaglutide should be stopped if NAION is confirmed.
Two groups have most reason to talk this through with their prescriber: people who have already had NAION in one eye, and people with other optic nerve risk factors. If you have diabetes, keep up with retinal screening; our eyes guide explains why that matters before starting or stepping up a dose.
Never stop a GLP-1 on your own because of this paper. If you are worried, take the question to your next appointment.
What to do with this
No food prevents NAION, and nothing in this paper changes what to eat. Food does help with the side effects you are far more likely to meet on a GLP-1, such as nausea and constipation, and regular meals with protein, fibre and enough fluid are the place to start.
Sources
- Sun H, Xin X, Yan J, Huang T. Comparative Risk of Nonarteritic Anterior Ischemic Optic Neuropathy Among Glucagon-Like Peptide-1 Receptor Agonists: An Integrated Safety and Mechanistic Analysis. Diabetes Obes Metab 2026. PubMed
- Hathaway JT, Shah MP, Hathaway DB, et al. Risk of Nonarteritic Anterior Ischemic Optic Neuropathy in Patients Prescribed Semaglutide. JAMA Ophthalmol 2024. PubMed
- Chou CC, Pan SY, Sheen YJ, et al. Association between Semaglutide and Nonarteritic Anterior Ischemic Optic Neuropathy: A Multinational Population-Based Study. Ophthalmology 2025. PubMed
- European Medicines Agency. Meeting highlights from the Pharmacovigilance Risk Assessment Committee (PRAC), 2 to 5 June 2025. EMA