GLP-1 study found no significant overall rise in sudden optic nerve vision loss, but counted only 29 cases

A study published on September 3, 2026 in Ophthalmology Retina looked at whether GLP-1 medicines raise the risk of a sudden, usually permanent form of vision loss. The headline result was not statistically significant. It was also based on only 29 cases, which leaves the question a long way from settled.

The condition is non-arteritic anterior ischemic optic neuropathy, shortened to NAION. It happens when blood flow to the front of the optic nerve drops and part of the nerve is damaged. People typically wake up with painless loss of vision in one eye, often in the upper or lower half of the visual field. It is uncommon, it usually does not fully recover, and there is no established treatment.

Concern about a possible link to GLP-1 medicines has been circulating for roughly two years. The authors of the new paper open by noting that published findings have been inconsistent, partly because earlier studies differed in why patients were taking the drug, what they were compared against, and how long they were followed.

The study compared new users of GLP-1 medicines against new users of two other diabetes drug classes

The team, led by Jonathan J. Lee at MedStar Georgetown University Hospital with colleagues at the University of Rhode Island College of Pharmacy and Mass Eye and Ear at Harvard Medical School, used administrative claims from United States private health plans covering 2012 through 2024.

They restricted the analysis to adults with type 2 diabetes and used what is called a target trial emulation. That means the study was designed to imitate the structure of a randomized trial as closely as claims data allow. Two features matter here. Only people newly starting a medicine were included, which avoids the distortion that comes from studying people who have already tolerated a drug for years. And the comparison group was people starting a different active diabetes medicine, specifically an SGLT-2 inhibitor or a DPP-4 inhibitor, rather than people on no treatment at all.

That comparison choice matters, because people who start any new diabetes medicine tend to differ from people who start none. The researchers then used propensity score methods to further balance the two groups on measured characteristics.

The overall hazard ratio was 1.87, with a confidence interval running from 0.85 to 4.12

The analysis included 19,505 adults. Of those, 9,213, or 47.2 percent, started a GLP-1 medicine and 10,292, or 52.8 percent, started an SGLT-2 inhibitor or a DPP-4 inhibitor. Across the whole group there were 29 cases of NAION.

Compared with the other two drug classes, GLP-1 use overall was not significantly associated with a higher risk of NAION. The adjusted hazard ratio was 1.87, with a 95 percent confidence interval of 0.85 to 4.12.

A confidence interval that stretches from 0.85 to 4.12 is doing most of the work in that sentence. Because the interval includes 1.0, the data are compatible with no increase in risk. They are also compatible with a fourfold increase. With 29 events spread across nearly 20,000 people, the study simply cannot distinguish between those possibilities. Reporting this as evidence that GLP-1 medicines are safe for the optic nerve would be as wrong as reporting it as evidence that they are dangerous.

The subgroup signal centered on liraglutide, an older medicine, and the numbers behind it are not public

The authors report that risk among people taking liraglutide was higher than in the comparison group within 12 or 18 months of follow-up, and that the elevated risk relative to the comparison drugs showed up in men and in older adults.

Those are subgroup findings, and subgroup findings drawn from 29 total events are fragile by construction. Splitting a small number of cases into smaller groups produces estimates that move a lot on very little data, and it raises the chance that something looks meaningful purely by accident. The published abstract does not give hazard ratios or confidence intervals for any of those subgroups, and the full paper is behind a subscription, so BariatricPal could not check those numbers against the source. The authors themselves conclude that further research is needed to confirm the findings.

Liraglutide, sold as Victoza for diabetes and Saxenda for weight management, is an older daily injection that has largely been displaced by weekly options. A signal concentrated in liraglutide users does not automatically carry over to semaglutide or to the newer medicines, and this study was not designed to answer that.

Claims data record billing codes, not eye examinations

Everything here rests on insurance claims. NAION was identified from diagnosis codes entered for billing, not from ophthalmologists reviewing charts, so both missed cases and miscoded ones are possible. The study covered people with type 2 diabetes on private insurance, which leaves out people taking these medicines for weight management alone, people on Medicare or Medicaid, and people paying cash.

Anyone who has a sudden change in vision in one eye should be evaluated promptly, whatever medicines they take. That is standard advice and it does not change because of this paper. What the paper does change is the state of the evidence, marginally: one more carefully designed study, with a result too imprecise to resolve the question either way.