Across 126 randomized trials, FDA reviewers counted 15 probable cases of a rare optic nerve stroke. Ten turned up in people taking a GLP-1 drug, five in people on placebo, and none at all among patients on other diabetes medicines, according to the FDA Drug Safety Communication of November 9, 2026. On the strength of that count and two other bodies of evidence, gathered over the more than two years since a Boston study first raised the question, the agency now requires every drug in the class to carry a warning about non-arteritic anterior ischemic optic neuropathy, or NAION, in the Warnings and Precautions section of its label.
Headlines about drug risk tend to arrive as ratios: four times, seven times, a 19% increase. A ratio says nothing about how many people are affected until it is multiplied by how often the thing happens in the first place, and NAION does not happen often. The number a ratio leaves out is extra cases per 100,000 person-years, which is one hundred thousand people taking the drug for one year, or fifty thousand for two.
The label change, drug by drug
Every GLP-1 on the US market, plus one that also hits a second hormone receptor
The FDA order covers the whole class. Pharmacy Times lists the products: exenatide, liraglutide (Victoza, Saxenda, and Xultophy with insulin degludec), dulaglutide (Trulicity), lixisenatide with insulin glargine (Soliqua), semaglutide (Ozempic, Rybelsus, Wegovy), tirzepatide (Mounjaro, Zepbound), and orforglipron (Foundayo), approved this year; the FDA communication’s own table gives the same list and flags tirzepatide as a GIP and GLP-1 receptor agonist. NAION cases have been reported for every GLP-1 drug currently sold, the same article notes.
The agency’s own description of the condition is short. It calls NAION a rare condition that can bring acute, painless, decreased vision in one eye, including permanent vision loss. The name spells out the event: ischemic, starved of blood, at the anterior end of the optic nerve. When that loss is permanent, the nerve behaves like a blown fuse in a panel with no spare.
FDA’s verdict is carefully worded. It concluded there is “a potential increased risk of rare NAION events” in patients taking GLP-1 drugs, and its review of animal data turned up no mechanism that would explain one. The label went ahead without one.
Background on how semaglutide and its relatives work is in an earlier x9m8 explainer. As of November 9, the eye question is still open and is now printed on the label.

A Boston clinic and two large hazard ratios
The 2024 study that started the review measured something real, in a room full of eye patients
In July 2024, a team at Massachusetts Eye and Ear published a retrospective matched cohort study in JAMA Ophthalmology drawn from a single neuro-ophthalmology practice, covering December 2017 through November 2023. FDA cites it as the paper that prompted further assessment. Among 710 patients with type 2 diabetes, 194 were prescribed semaglutide; among 979 patients who were overweight or obese, 361 were. The comparison groups took other diabetes or weight-loss drugs.
The hazard ratios were large. For the diabetes group, 4.28 (95% CI, 1.62 to 11.29); for the weight group, 7.64 (95% CI, 2.21 to 26.36). Those ratios rest on small counts: 17 NAION cases against 6 in the diabetes comparison, 20 against 3 in the weight comparison.
Converted to absolute terms, the Boston numbers look enormous. Cumulative incidence at 36 months was 8.9% for semaglutide versus 1.8% in the diabetes group, and 6.7% versus 0.8% in the weight group. That is a gap of roughly 7,100 and 5,900 cases per 100,000 patients over three years.
Those figures cannot be carried out of the building. Every patient in the study had already been sent to a neuro-ophthalmologist, which is a little like estimating how often furnaces fail by surveying people who already called the boiler repairman. The authors say as much: one tertiary care institution that specializes in ophthalmology, findings that may not generalize, a retrospective design that cannot probe bias in which patients were prescribed semaglutide, and small event counts that produced wide confidence intervals. They also reviewed charts by hand and found that 40% of cases coded as ischemic optic neuropathy were not actually NAION, a warning they aimed at large databases that cannot be checked that way. Their conclusion is restrained. “This study’s findings suggest an association between semaglutide and NAION,” they wrote, adding that an observational study cannot establish causality.
FDA says it identified “several limitations” in the paper and judged that further assessment was warranted. The agency then went looking in two places with far more patients.
Fifteen events in 288,879 person-years
Randomized trials give the cleanest comparison and the thinnest count
The FDA meta-analysis pooled 126 randomized controlled trials from GLP-1 development programs: 134,253 patients, 78,217 on a GLP-1 drug and 56,036 on a comparator, with 156,947 person-years of GLP-1 exposure and 131,932 on comparators. The reported incidence was 6.4 probable NAION events per 100,000 person-years on GLP-1 drugs and 3.8 on comparators.
Subtract one from the other and the excess is about 2.6 cases per 100,000 person-years. Turned around, that is one additional probable case for roughly every 38,700 person-years of treatment. A city of 100,000 people, all on the drug for a full year, would by this estimate see two or three extra cases beyond what it would have seen anyway.
The crude ratio between the two rates works out to about 1.7. FDA’s communication gives no confidence interval for it, so this post calculated one itself from the published counts (method in the notes, and the figure is this post’s arithmetic, not FDA’s): an exact 95% interval of roughly 0.5 to 6.3. That range runs from a protective effect to a sixfold harm. With ten events on one side and five on the other, the trials can point a direction and nothing more, a limitation that sits in the arithmetic, not in anyone’s motives.
Randomization is the reason this evidence matters anyway. Trial participants were assigned their drug by chance, so the smoking, sleep apnea, blood pressure and diabetes severity that plague observational comparisons are, on average, balanced between arms.

Sentinel, pharmacy claims and the comparator problem
Against one diabetes drug the signal vanishes; against another it holds
The third source was Sentinel, FDA’s network of insurance-claims data. The analysis followed adults with type 2 diabetes who started a GLP-1 drug, an SGLT-2 inhibitor or a DPP-4 inhibitor between 2019 and 2024. NAION incidence ranged from 26 to 32 events per 100,000 person-years: 30 to 32 for GLP-1 drugs, 31 for SGLT-2 inhibitors, 26 for DPP-4 inhibitors. FDA does not explain on the page why the GLP-1 figure is a range.
Compared with SGLT-2 inhibitors, the hazard ratio was 1.19 (95% CI, 0.91 to 1.54), not statistically significant. Compared with DPP-4 inhibitors, it was 1.43 (95% CI, 1.02 to 2.01), an increased risk.
Rates in Sentinel’s diabetic adults run several times higher than in the trials. The FDA page does not say why. The comparator, meanwhile, decides the headline. The raw gap between GLP-1 drugs and SGLT-2 inhibitors is somewhere between minus one and plus one case per 100,000 person-years. Against DPP-4 inhibitors it is four to six.
Applying the adjusted hazard ratios to the comparator rates gives a rough absolute excess. FDA did not publish this step; it is this post’s own back-of-the-envelope arithmetic, and it treats a hazard ratio as a rate ratio. Against DPP-4 inhibitors: about 11 extra cases per 100,000 person-years, with an interval from under 1 to about 26. Against SGLT-2 inhibitors: about 6, with an interval that includes zero.
Europe reached a similar scale through a different door. In June 2025, the European Medicines Agency’s safety committee classed NAION as a very rare side effect of semaglutide, a frequency category meaning it may affect up to 1 in 10,000 people taking the drug. EMA reported that several large epidemiological studies in adults with type 2 diabetes found an approximately two-fold increase in risk, which it put at roughly one extra case per 10,000 person-years of treatment. That is 10 per 100,000, close to the Sentinel estimate against DPP-4 inhibitors.
Three sources, one table
Same question, three different rooms, three different answers
| Source | Population | Comparison | Relative measure | Absolute rate or gap | Main weakness |
|---|---|---|---|---|---|
| Hathaway et al., JAMA Ophthalmology, 2024 | 1,689 patients at one Boston neuro-ophthalmology practice, 2017–2023 | Semaglutide vs other diabetes or weight-loss drugs | HR 4.28 (diabetes); HR 7.64 (overweight/obese) | 36-month cumulative incidence 8.9% vs 1.8% and 6.7% vs 0.8%; gap of about 7,100 and 5,900 per 100,000 over three years | Referral population, retrospective, 23 events per analysis |
| FDA meta-analysis of 126 RCTs | 134,253 trial participants, 288,879 person-years | GLP-1 drugs vs placebo or other drugs | Crude rate ratio about 1.7 (this post’s calculation; exact 95% CI about 0.5 to 6.3) | 6.4 vs 3.8 per 100,000 person-years; excess about 2.6 | Only 15 events in total |
| FDA Sentinel claims study | Adults with type 2 diabetes starting treatment, 2019–2024 | GLP-1 drugs vs SGLT-2 inhibitors; vs DPP-4 inhibitors | HR 1.19 (0.91–1.54); HR 1.43 (1.02–2.01) | 30–32 vs 31 and 26 per 100,000 person-years; HR-implied excess about 6 and 11 (this post’s calculation) | Observational, claims-coded diagnoses, result depends on comparator |
Even divided across its three years, the Boston gap is hundreds of times larger than anything in the bottom two rows. Nobody’s arithmetic is wrong. The Boston study counted eye-clinic patients over three years; the trials and Sentinel counted ordinary patients one year at a time.
For a sense of where the baseline sits, FDA puts NAION in the general population at 2.3 to 10.2 cases per 100,000 persons per year in adults older than 50.

The strongest case for worry
Small per patient is not the same as small
A reader could finish the table and file the whole thing under rounding error. FDA did not.
All three sources lean the same way. The trials, with randomization doing the heavy lifting, show more cases on the drug. Sentinel shows an elevated risk against DPP-4 inhibitors with an interval that clears 1. EMA’s epidemiological picture shows roughly a doubling. FDA adds that methodologically rigorous observational studies in patients with type 2 diabetes have consistently reported increased NAION risk with GLP-1 drugs compared with DPP-4 or SGLT-2 inhibitors, a pattern it calls consistent with its own findings. A signal that keeps pointing the same way across different methods earns more weight than any single interval suggests.
The outcome is also not a stomachache. NAION can take central vision in one eye for good, and FDA notes that patients who have had it face a higher risk of recurrence. A doubling of a rare event is still rare. For the person it happens to, the denominator stops mattering.
Then there is scale, which cuts against the “tiny” reading. These drugs are used by very large numbers of people. Multiply even a few extra cases per 100,000 person-years by that many person-years and the count of real people with real vision loss is no longer hypothetical.
Against all of that, NAION’s known risk factors, which Pharmacy Times lists as smoking, diabetes, hypertension, sleep apnea, anemia and kidney disease, overlap heavily with the conditions these drugs treat. The Sentinel comparison against SGLT-2 inhibitors found no significant difference. And all five comparator-arm cases in the trials came from placebo groups, with none among patients on other glucose-lowering drugs, a reminder of how few events the whole analysis stands on.
What the label asks of patients and prescribers
The instructions are about speed, and about who decides
The advice to patients is narrow. They should know about the risk and contact a health care professional right away for sudden, painless decreased vision in one eye. Anyone who has already had NAION should talk with their prescriber about whether to continue or stop.
For clinicians, the communication asks them to tell patients to report symptoms immediately, refer suspected cases promptly to an ophthalmologist, and consider discontinuing the drug in patients who develop NAION unless the benefits clearly outweigh the risk. Pharmacy Times’ counseling notes add that the vision loss may be noticed on waking and can show up as blind spots, tunnel vision, color distortion or hazy central vision, and that the counseling applies across the class, including when a patient switches from one drug to another.
None of that amounts to an instruction to stop a prescribed medicine, and nothing in this post is one either. The trade between a small, uncertain eye risk and the benefits a drug was prescribed for depends on a patient’s own history, which is a conversation for the prescriber, not a blog. Anyone weighing it should bring the questions to that person. Suspected side effects can be reported to FDA MedWatch.
The class has drawn regulatory attention before. Earlier x9m8 posts looked at why GLP-1 supply ran short and what FDA warning letters allege about compounded GLP-1 sellers, whose products sit outside the controlled trials behind every number above.
Two and a half extra cases per 100,000 person-years in the trials, about ten in Europe’s reading of the epidemiology, about eleven against one comparator in Sentinel and essentially none against another. If the label had to print one number instead of a paragraph, which should it be?
This post is for general information and is not medical advice. Questions about a specific prescription belong with the prescriber.
Sources
- U.S. Food and Drug Administration, Drug Safety Communication, “FDA requires safety labeling changes for glucagon-like peptide-1 receptor agonists (GLP-1 RAs) due to potential risk of NAION,” November 9, 2026: https://www.fda.gov/drugs/drug-safety-communications/fda-requires-safety-labeling-changes-glucagon-peptide-1-receptor-agonists-glp-1-ras-due-potential
- Pharmacy Times, “FDA Requires Class-Wide GLP-1 RA Labeling Changes for Potential NAION Risk,” November 9, 2026: https://www.pharmacytimes.com/view/fda-requires-class-wide-glp-1-ra-labeling-changes-for-potential-naion-risk
- Hathaway JT, Shah MP, Hathaway DB, et al., “Risk of Nonarteritic Anterior Ischemic Optic Neuropathy in Patients Prescribed Semaglutide,” JAMA Ophthalmology 2024;142(8):732-739 (PMC full text): https://pmc.ncbi.nlm.nih.gov/articles/PMC11223051/
- European Medicines Agency, “PRAC concludes eye condition NAION is a very rare side effect of semaglutide medicines Ozempic, Rybelsus and Wegovy,” June 6, 2025: https://www.ema.europa.eu/en/news/prac-concludes-eye-condition-naion-very-rare-side-effect-semaglutide-medicines-ozempic-rybelsus-wegovy
- FDA MedWatch adverse event reporting: https://www.accessdata.fda.gov/scripts/medwatch/
