FDA orders GLP-1 eye-risk warnings: what the numbers mean

The condition is non-arteritic anterior ischemic optic neuropathy, or NAION. The FDA’s decision is consequential, but narrower than a declaration that these medicines cause blindness: it requires information in the Warnings and Precautions section of prescribing labels, not a recall or a new boxed warning. The agency has not specified a deadline, so the announcement does not mean every label has already changed.⁠4

For patients, the clearest instruction is also the most urgent: sudden, painless vision loss needs immediate contact with a health care professional. Blind spots, tunnel vision, distorted colors or blurry central vision also warrant contact right away. Do not wait for a routine prescription appointment. The warning is not an instruction for everyone taking these medicines to stop.⁠4

NAION occurs when blood flow falls at the front of the optic nerve, where it meets the eye. Its underlying cause is not completely understood. Someone may first notice lost vision in one eye on waking; the damage can be permanent, and there is no known treatment for the condition. That combination—rarity but potentially irreversible harm—makes both alarmism and dismissal unhelpful.⁠4⁠1

The FDA pooled 126 randomized controlled trials because individual studies had few or no cases. Together, they included 134,253 participants: 78,217 received a GLP-1 medicine and 56,036 received placebo or another glucose-lowering medicine. The following rates describe probable NAION events per 100,000 person-years of follow-up, not percentages of patients losing vision.⁠4

A person-year means one person followed for one year. The GLP-1 groups contributed 156,947 person-years and had 10 probable events; comparison groups contributed 131,932 person-years and had five. All five comparison events occurred among placebo recipients. Importantly, however, the reported 3.8 rate uses follow-up from the entire comparison group, including people receiving other diabetes medicines—not just placebo recipients.⁠4

The crude difference is about 2.6 additional probable events per 100,000 person-years in the GLP-1 groups. That is a descriptive subtraction, not a count of cases proven to have been caused by treatment. With only 15 probable events altogether, these data cannot supply a precise personal risk or establish that every ingredient, dose and use carries the same risk.⁠4

For background, the FDA cites estimates of 2.3 to 10.2 NAION cases per 100,000 people annually among adults older than 50. That range is not a matched control group for these trials. Setting it beside the trial rates and declaring the medicines either safe or dangerous would ask the comparison to do something it cannot.⁠4

The early semaglutide study reported an approximately fourfold association among patients with diabetes, according to NBC News’s account of the FDA review. That historical finding helped trigger the investigation; it is not today’s class-wide risk estimate. The FDA identified several limitations and expanded its assessment to clinical trials, animal research, observational studies and reports collected after medicines reached patients.⁠1⁠4

A separate FDA Sentinel analysis illustrates why one dramatic relative-risk figure cannot settle the issue. It followed adults with type 2 diabetes starting a GLP-1 medicine or one of two other diabetes drug types—SGLT-2 inhibitors and DPP-4 inhibitors—between 2019 and 2024. NAION rates were 30 to 32 events per 100,000 person-years among GLP-1 recipients, 31 with SGLT-2 inhibitors and 26 with DPP-4 inhibitors. These are results from a different population and study design, not substitutes for the randomized-trial rates.⁠4

After adjustment for differences between patient groups, the comparison with SGLT-2 inhibitors produced a hazard ratio of 1.19, with a 95 percent confidence interval of 0.91 to 1.54. That interval includes no difference, so the increase was not statistically significant. Against DPP-4 inhibitors, the hazard ratio was 1.43, with an interval of 1.02 to 2.01—a statistically elevated association. Reporting only the second result would conceal a material part of the evidence. Neither comparison proves causation.⁠4

The FDA also reviewed other rigorous observational studies that found elevated associations against those diabetes treatments. But reports of individual cases remain difficult to interpret: diabetes, high blood pressure, smoking, sleep apnea, anemia and kidney disease may themselves raise NAION risk, and several are common among people prescribed these medicines. Cases have been reported across all currently marketed GLP-1 products; the animal-data review did not identify a mechanism explaining how the drugs might increase risk.⁠4

The evidence adds up to a potential risk worth warning about—not a demonstrated explanation for every reported case. A class-wide label requirement is a regulatory response to that concern, not proof of identical risk across the class.

The scope extends well beyond the four brands most familiar from weight-loss headlines. Novo Nordisk’s semaglutide products Ozempic, Wegovy and Rybelsus are covered. So are Eli Lilly’s Mounjaro and Zepbound, which contain tirzepatide. Although tirzepatide acts on both the GLP-1 receptor and another hormone receptor, GIP, it is included in this action. Foundayo, the orforglipron medicine approved in 2026, is also listed.⁠4

Older medicines and insulin combinations are included too: Victoza and Saxenda, containing liraglutide; Trulicity, containing dulaglutide; Xultophy, combining liraglutide with insulin degludec; Soliqua, combining lixisenatide with insulin glargine; and marketed generic exenatide. The FDA table also lists the historical exenatide brand Byetta, approved in 2005. Its application was voluntarily withdrawn after it stopped being marketed—not for safety or effectiveness reasons—under an August 2025 notice. That withdrawal should not be confused with this warning.⁠4

These medicines lower blood sugar after meals and act on brain pathways controlling appetite and food intake. Beyond type 2 diabetes and long-term weight management, particular products have approved benefits involving cardiovascular events, kidney disease, obstructive sleep apnea or metabolic liver disease with moderate-to-advanced scarring. Those uses depend on the product and the patient; they are not interchangeable benefits of every GLP-1 medicine.⁠4

A previous NAION event raises the risk of recurrence, including in the other eye. Patients who have had one should discuss continuing or stopping treatment with their clinician. When NAION is suspected, the FDA advises clinicians to arrange prompt ophthalmology evaluation and to consider discontinuing the medicine unless its benefits clearly outweigh the risk.⁠4

That leaves two decisions on different clocks. A treatment decision requires an individualized weighing of benefits and harms. A sudden change in vision requires action right away. The uncertainty about causation is real; it is not a reason to delay having symptoms assessed.