NAION and GLP-1 Drugs: FDA's 2026 Vision-Loss Warning Explained

Key points

• On 2026-11-09 FDA announced it is requiring all makers of GLP-1 receptor agonist medicines to add a warning about the “potential increased risk” of NAION, a rare cause of sudden vision loss in one eye [1][2].
• Studies disagree on size. Meta-analyses of observational work put the relative risk at about two to two and a half times in people with type 2 diabetes, with “low” certainty in one of them; some large studies found no significant difference [10][11][14][15].
• In absolute terms the extra risk is small: FDA’s trial pooling gave 6.4 versus 3.8 cases per 100,000 person-years, and EMA described about one extra case per 10,000 person-years [1][3].
• FDA says the warning covers the whole class, tirzepatide included. For tirzepatide we found no cohort or trial estimate, only reporting signals and case reports [1][17][18].
• Research-use-only vials of semaglutide, tirzepatide and retatrutide have no FDA-approved labeling, so the new label text does not apply to them [1].

What did FDA announce on 2026-11-09?

FDA published a Drug Safety Communication titled “FDA Requires Safety Labeling Changes for Glucagon-Like Peptide-1 Receptor Agonists (GLP-1 RAs) Due to Potential Increased Risk of Rare But Serious Eye Condition,” and sent a MedWatch bulletin the same day [1][2]. It says the agency is requiring application holders “for all GLP-1 RA medications” to add information about the potential increased risk of NAION to the Warnings and Precautions section of labeling [1].

The list in FDA’s table is: exenatide (the Byetta application was withdrawn in 2025 for reasons other than safety; generics are marketed), Victoza and Saxenda (liraglutide), Trulicity (dulaglutide), Xultophy, Soliqua, Ozempic, Rybelsus and Wegovy (semaglutide), Mounjaro and Zepbound (tirzepatide, which also acts on the GIP receptor), and Foundayo (orforglipron, approved 2026) [1]. Retatrutide is not in the table [1].

An AP post on X on 2026-11-10 said the drugs “will carry a new warning” about a rare side effect. We read it through a mirror and could not open the AP article, so we cite only FDA for what was announced [1][6].

The label text itself was not yet public when we looked. The Ozempic label on DailyMed (Novo Nordisk, version published 2026-06-10) did not mention NAION on 2026-11-11, so the final wording had not appeared there [5].

What is NAION?

FDA describes non-arteritic anterior ischemic optic neuropathy as “a rare condition that can result in acute, painless, decreased vision in one eye, including permanent vision loss” [1]. It is caused by reduced blood flow to the front of the optic nerve, the optic disc, and FDA says the underlying cause is not completely understood [1]. Risk factors may include smoking, diabetes, high blood pressure, sleep apnea, anemia and kidney disease [1]. If it has occurred in one eye, the chance that it occurs in the other is higher [1].

FDA lists the symptoms as sudden painless vision loss in one eye (possibly noticed on waking), blurred or missing portions of vision, distorted colours and reduced central sharpness. It says patients “should contact their health care professional right away” if they have any of these [1]. That is the only reader guidance this article gives.

FDA cites published literature putting background incidence at 2.3 to 10.2 cases per 100,000 persons per year in adults older than 50 [1].

What does the evidence show?

FDA’s own review combined three strands: a pooled analysis of clinical trials, a cohort study in its Sentinel system, and published observational studies. It found animal data did not identify a mechanism, and called adverse-event reports hard to interpret because NAION risk factors overlap with conditions common in people treated with these drugs [1].

The table lists the studies we read. We read abstracts only, not full texts.

Study	Design and population	Comparator	Events	Relative estimate	Absolute figures
FDA trial meta-analysis [1]	126 randomised trials, 134,253 patients, 288,879 person-years (PY)	Placebo or other glucose-lowering drugs	15 probable (10 GLP-1 RA, 5 placebo, 0 other drugs)	Not given	6.4 vs 3.8 per 100,000 PY
FDA Sentinel cohort [1]	Adults with type 2 diabetes starting a drug 2019-2024	SGLT-2 and DPP-4 inhibitors	Not given	HR 1.19 (0.91-1.54) vs SGLT-2i; 1.43 (1.02-2.01) vs DPP-4i	26-32 per 100,000 PY across groups
Hathaway 2024 [7]	One neuro-ophthalmology centre, matched cohorts	Patients not prescribed semaglutide	Type 2 diabetes: 17 vs 6; overweight/obese: 20 vs 3	HR 4.28 (1.62-11.29); HR 7.64 (2.21-26.36)	36-month incidence 8.9% vs 1.8%; 6.7% vs 0.8%
Grauslund 2024 [8]	All people with type 2 diabetes in Denmark, 2018-2024, 424,152 people	Not exposed to once-weekly semaglutide	218 (67 in exposed)	HR 2.19 (1.54-3.12)	0.228 vs 0.093 per 1,000 PY
Simonsen 2025 [9]	Denmark and Norway, 61,377 semaglutide users	SGLT-2i	32	Pooled HR 2.81 (1.67-4.75)	Rate difference +1.41 per 10,000 PY
Cai 2025 [10]	OHDSI network, 810,390 new semaglutide users	Empagliflozin, sitagliptin, glipizide	Not in abstract	Sensitive definition: not significant (HR 1.23-1.44); specific definition vs empagliflozin HR 2.27 (1.16-4.46)	14.5 per 100,000 PY
Chou 2025 [11]	TriNetX, 160 health care organisations in 21 countries, propensity-matched	Matched non-users (details not in abstract)	Not in abstract	No significant association in any group	Not given
Ueda 2026 [12]	Nationwide cohort, 107,518 GLP-1 RA vs 185,898 SGLT-2i users	SGLT-2i	62 vs 64	1-year RR 1.93 (1.00-3.73); 5-year RR 1.69 (0.95-3.01)	0.04% vs 0.02% at 1 year; 0.12% vs 0.07% at 5 years
Reynolds 2026 [13]	Target trial emulation	SGLT-2i; DPP-4i	Not in abstract	Not in abstract	8.5 vs 5.5 and 7.8 vs 4.2 per 10,000 at 18 months

Two pooled analyses of observational studies exist. A Bayesian meta-analysis in Neurology (5 studies, 1,593,554 patients) found a semaglutide relative risk of 2.52 (95% credible interval 1.56-4.72) versus people not on GLP-1 RAs [14]. A PLoS Medicine meta-analysis of five studies in type 2 diabetes reported a hazard ratio of 2.17 (1.73-2.74) with GRADE certainty “low” [15]. A 2026 consensus statement from the North American Neuro-Ophthalmology Society and the American Academy of Ophthalmology says the evidence is mostly retrospective EHR and claims data, that many studies report “a small possible increased risk,” that some find no correlation, and that the overall magnitude is low [16].

Relative versus absolute risk

A relative risk of 2 multiplies a very small number. FDA’s trial pooling gives 6.4 versus 3.8 events per 100,000 person-years. The difference is 2.6 per 100,000 PY by our arithmetic; FDA’s communication gives no confidence interval, and the comparison rests on 15 events [1].

EMA’s 2025 review described a roughly two-fold increase in adults with type 2 diabetes in large epidemiological studies, equal to “approximately one additional case of NAION per 10,000 person-years of treatment” [3]. Simonsen’s rate difference of 1.41 per 10,000 PY is in the same range, and the authors say the absolute risk remains low [9].

Reynolds and colleagues put the 18-month difference at 3.0 per 10,000 versus SGLT-2 inhibitors and 3.6 per 10,000 versus DPP-4 inhibitors, a number needed to harm of 3,333 and 2,778. They say the absolute risk is very low and may reflect residual confounding [13]. Ueda reports a risk difference of 0.02% at one year and 0.05% at five years [12].

These figures come from different populations, follow-up lengths and definitions, and none is an estimate for an individual.

Why do the studies disagree?

• Referral population. Hathaway drew from one neuro-ophthalmology service, so baseline NAION rates are far above population rates, and the event counts are small with wide intervals. The authors say causality is not established, and FDA said it “identified several limitations” [1][7].
• Outcome definitions. In Cai, a sensitive NAION definition gave no significant difference from comparators, while a specific definition gave a hazard ratio of 2.27 against empagliflozin. The authors call the increase “modest” and smaller than previously reported [10].
• Comparator choice. FDA’s Sentinel cohort found no significant difference against SGLT-2 inhibitors but an increase against DPP-4 inhibitors [1]. Ueda’s result attenuated in a metformin-baseline subgroup [12].
• Confounding. Diabetes, hypertension, sleep apnea and kidney disease are both NAION risk factors and reasons for prescribing, and Ueda and Reynolds both state that unmeasured confounding is possible [1][12][13].
• Few events. Ueda had 62 and 64 events, Simonsen 32 and FDA’s trial analysis 15 [1][9][12].
• Population. Ueda notes limited generalisability to people without diabetes [12]. Chou concluded there is no increased risk in the general population [11].

FDA nonetheless concluded that “methodologically rigorous” observational studies consistently found increased risk against DPP-4 or SGLT-2 inhibitors and that these were generally consistent with its trial and Sentinel analyses [1].

What did European regulators say?

On 2025-06-06 EMA’s safety committee (PRAC) concluded that NAION is a “very rare side effect of semaglutide,” occurring in up to 1 in 10,000 people taking it. The conclusion covers semaglutide only: Ozempic, Rybelsus and Wegovy. Product information was to list NAION as very rare, and EMA said semaglutide should be stopped if NAION is confirmed. EMA described trial data as showing a “slightly higher risk” versus placebo [3]. The page was revised on 2025-06-13 to remove a reference to a legally binding decision [3]. WHO issued an alert on 2025-06-27 [4]. We did not check the UK (MHRA) or Germany (BfArM).

Tirzepatide and other GLP-1 drugs

FDA says NAION cases have been reported for all currently marketed GLP-1 RA products and that the labeling requirement covers the whole class [1]. For tirzepatide we found no cohort or trial estimate. A 2026 analysis of FAERS reports found a signal of 49 reports with a reporting odds ratio of 2.94 (2.20-3.93) [17]. A disproportionality ratio shows that a drug is reported more often with an event than other drugs are. It is not an incidence or a risk. Case reports exist [18].

Retatrutide

Retatrutide is not FDA-approved and is not in FDA’s table [1]. A PubMed search on 2026-11-11 for “retatrutide AND (NAION OR optic)” returned no records [19]. We therefore have no published data on its NAION risk, which is not evidence of safety or of risk. For its trial record see our retatrutide side-effects article.

What does this mean for research-peptide listings?

Peptilyze compares listings of research-use-only vials. On 2026-11-11 its public API showed semaglutide listed at 10 shops, tirzepatide at 10 and retatrutide at 24 [20][21][22]. Such vials are not FDA-approved medicines and carry no FDA-approved labeling, so the labeling requirement applies to the approved products in FDA’s table, not to these listings [1]. Peptilyze compares prices and does not test or endorse products. Compound pages: semaglutide, tirzepatide, retatrutide. Related: compounded GLP-1 rules and semaglutide versus tirzepatide.

Risks and unknowns

• Causation is unproven. FDA uses “potential increased risk,” and FDA’s animal review found no mechanism [1].
• Size is uncertain. Estimates range from no significant difference to a hazard ratio above 4, and the pooled observational estimate carries “low” certainty [7][11][15].
• Absolute risk is small in every dataset we read, and background incidence is not zero [1][9][12][13].
• Few trial events. FDA’s pooling rests on 15 probable events [1].
• People without diabetes are less well studied; Ueda cites limited generalisability [12].
• Other drugs in the class: we found no risk estimate for tirzepatide, orforglipron or retatrutide [1][17][19].
• Recurrence: FDA says people with a history of NAION are at increased risk of recurrence [1].
• Instructions to clinicians. FDA tells health care professionals to refer suspected NAION promptly to an ophthalmologist and to “consider discontinuing” the drug after an event unless benefits clearly outweigh the risk. That is regulator text for clinicians, not guidance from us [1].

Frequently asked questions

Did FDA say GLP-1 drugs cause NAION? No. FDA wrote of a “potential increased risk” and required a Warnings and Precautions entry across the class. Its animal review did not identify a mechanism [1].

How big is the risk? It depends on the study. Pooled observational studies give relative risks around 2 to 2.5, with low certainty in one of them [14][15]. In absolute terms, EMA and the Danish-Norwegian cohort put the excess at roughly 1 to 1.4 cases per 10,000 person-years, and a US target trial emulation at 3.0 to 3.6 per 10,000 people over 18 months [3][9][13].

Does the warning apply to tirzepatide? FDA’s requirement applies to all GLP-1 RA medicines, and Mounjaro and Zepbound are in its table [1]. We found no cohort or trial estimate specific to tirzepatide, only a reporting signal and case reports [17][18].

Does this apply to research-use-only products? Those products have no FDA-approved labeling, so the labeling requirement applies to approved products, not to them [1]. The studies here concern prescribed, approved drugs.

Which symptoms did FDA list? Sudden painless vision loss in one eye, blurred or missing vision, distorted colours and reduced central sharpness. FDA says to contact a health care professional right away [1].

Sources

1. U.S. Food and Drug Administration, “FDA Requires Safety Labeling Changes for Glucagon-Like Peptide-1 Receptor Agonists (GLP-1 RAs) Due to Potential Increased Risk of Rare But Serious Eye Condition”, Drug Safety Communication (2026-11-09). https://www.fda.gov/drugs/drug-safety-communications/fda-requires-safety-labeling-changes-glucagon-peptide-1-receptor-agonists-glp-1-ras-due-potential
2. FDA MedWatch, email bulletin on the same communication (2026-11-09). https://content.govdelivery.com/accounts/USFDA/bulletins/4322d84
3. European Medicines Agency, “PRAC concludes eye condition NAION is a very rare side effect of semaglutide medicines Ozempic, Rybelsus and Wegovy” (2025-06-06, revised 2025-06-13). https://www.ema.europa.eu/en/news/prac-concludes-eye-condition-naion-very-rare-side-effect-semaglutide-medicines-ozempic-rybelsus-wegovy
4. World Health Organization, alert on semaglutide medicines and NAION (2025-06-27). https://www.who.int/news/item/27-06-2025-27-06-2025-semaglutide-medicines-naion/
5. DailyMed (NIH National Library of Medicine), Ozempic label, Novo Nordisk, version published 2026-06-10, checked by us 2026-11-11. https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=adec4fd2-6858-4c99-91d4-531f5f2a2d79
6. @AP, X post (2026-11-10), post text read through the fxtwitter mirror (X itself was not opened); the AP article returned 403 and was not used. https://x.com/AP/status/2120202143307100328
7. Hathaway JT et al., “Risk of Nonarteritic Anterior Ischemic Optic Neuropathy in Patients Prescribed Semaglutide”, JAMA Ophthalmol 142(8):732-739 (2024); abstract. https://pubmed.ncbi.nlm.nih.gov/38958939/
8. Grauslund J et al., Int J Retina Vitreous 10:97 (2024); abstract. https://pubmed.ncbi.nlm.nih.gov/39696569/
9. Simonsen E et al., Diabetes Obes Metab 27(6):3094-3103 (2025); abstract. https://pubmed.ncbi.nlm.nih.gov/40098249/
10. Cai CX et al., JAMA Ophthalmol 143(4):304-314 (2025); abstract. https://pubmed.ncbi.nlm.nih.gov/39976940/
11. Chou CC et al., Ophthalmology 132(4):381-388 (2025); abstract. https://pubmed.ncbi.nlm.nih.gov/39491755/
12. Ueda P et al., Ann Intern Med 179(9):1232-1243 (2026); abstract. https://pubmed.ncbi.nlm.nih.gov/42441962/
13. Reynolds KR et al., Ann Intern Med 179(9):1244-1253 (2026); abstract. https://pubmed.ncbi.nlm.nih.gov/42441967/
14. Dhivagaran T et al., Neurology 106(10):e214864 (2026); abstract. https://pubmed.ncbi.nlm.nih.gov/42060879/
15. Chrzanowski J et al., PLoS Med 23(5):e1005064 (2026); abstract. https://pubmed.ncbi.nlm.nih.gov/42166479/
16. DeParis SW et al., NANOS and American Academy of Ophthalmology consensus statement, Ophthalmology 133(8):1029-1035 (2026); abstract. https://pubmed.ncbi.nlm.nih.gov/42132708/
17. Kakkar AK et al., Diabetes Obes Metab (2026 Aug), FAERS disproportionality analysis; abstract. https://pubmed.ncbi.nlm.nih.gov/42259643/
18. Valášková J et al., Case Rep Ophthalmol (2026), case report; record only. https://pubmed.ncbi.nlm.nih.gov/42569780/
19. NCBI PubMed, search “retatrutide AND (NAION OR optic)”, run by us 2026-11-11 (0 records). https://pubmed.ncbi.nlm.nih.gov/?term=retatrutide+AND+%28NAION+OR+optic%29
20. Peptilyze public API, semaglutide (generatedAt 2026-11-11T10:15:23Z). https://peptilyze.com/api/v1/compounds/semaglutide
21. Peptilyze public API, tirzepatide (generatedAt 2026-11-11T10:15:23Z). https://peptilyze.com/api/v1/compounds/tirzepatide
22. Peptilyze public API, retatrutide (generatedAt 2026-11-11T10:15:23Z). https://peptilyze.com/api/v1/compounds/retatrutide

Not verified and not cited as fact: the AP article (returned 403); the full texts of all studies (abstracts only); numbers from the Silverii 2025 meta-analysis of randomised trials (abstract not available to us); the final FDA label wording; MHRA (UK) and BfArM (Germany) communications.

This article is information, not medical advice.