GLP-1 Eyesight Headlines: Association, Safety Signals and Causation

In June 2025, Europe's safety committee concluded that NAION, an eye condition that can cause loss of vision, is a very rare side effect of semaglutide. The studies behind that decision do not all point in the same direction — and the differences are where the useful information lives.

In mid-2025, two kinds of headline about the same medicine were circulating at once. One warned that semaglutide had been linked to a rare cause of sudden blindness. Another reported that a large multinational study had found no such link. Both described real, published evidence. Understanding how they fit together requires separating three things headlines often merge: an association observed in a dataset, a regulatory conclusion drawn from the totality of evidence, and proof of causation.

What the EMA actually decided in June 2025

On 6 June 2025, the European Medicines Agency announced that its safety committee, PRAC, had concluded a review of medicines containing semaglutide — the active substance in Ozempic, Rybelsus and Wegovy — following concerns about a possible increased risk of non-arteritic anterior ischemic optic neuropathy (NAION), an eye condition that may cause loss of vision.^[1] PRAC recommended that product information be updated to include NAION as a side effect with a frequency of "very rare," meaning it may affect up to 1 in 10,000 people.^[1]

Two clarifications keep this decision in proportion. First, the conclusion concerns semaglutide specifically; the announcement is not a blanket verdict on every GLP-1 receptor agonist. Second, "very rare" is a regulatory frequency band, not an annual rate from a particular study or a precise personal probability. The attached advice is direct: anyone who experiences a sudden loss of vision or rapidly worsening eyesight during treatment with semaglutide should contact their doctor without delay, and if NAION is confirmed, treatment with semaglutide should be stopped.^[1]

The referral-clinic study that raised the question

The signal that drew wider attention came from Hathaway and colleagues, who in 2024 reported a retrospective matched cohort study using a registry of patients evaluated by neuro-ophthalmologists at one academic institution between December 2017 and November 2023.^[2] Among patients with type 2 diabetes, 17 NAION events occurred in 194 patients prescribed semaglutide versus 6 in 516 prescribed other antidiabetic medicines; the adjusted hazard ratio was 4.28 (95% CI, 1.62–11.29). Among patients with overweight or obesity, 20 events occurred in 361 semaglutide users versus 3 in 618 users of other weight-loss medicines; the hazard ratio was 7.64 (95% CI, 2.21–26.36).^[2]

Those ratios are large, but the setting matters: everyone in the dataset had been referred to neuro-ophthalmologists at one academic centre — a selected population, not a cross-section of all users. The authors were careful about this: they reported an association and stated that, because the study was observational, future study is required to assess causality.^[2]

Two national registries: smaller relative risk, small absolute risk

A 2025 Danish–Norwegian study by Simonsen and colleagues tested the question in national health registries, comparing people with type 2 diabetes who started semaglutide against those who started SGLT-2 inhibitors, another class of glucose-lowering medicine.^[3] Across 44,517 semaglutide initiators in Denmark and 16,860 in Norway, only 32 NAION events were observed in total. The adjusted pooled hazard ratio was 2.81 (95% CI, 1.67–4.75), and the incidence rate difference was +1.41 per 10,000 person-years (95% CI, +0.53 to +2.29).^[3]

This pair of numbers shows why relative-risk headlines need an absolute anchor. The hazard ratio says the rate among semaglutide users was close to triple the comparator rate. The rate difference says that tripling added roughly one to two extra cases per 10,000 person-years of use, because the underlying rate is very low. Neither figure is the "true" one; they answer different questions. A per-person-year study rate and the EMA's "up to 1 in 10,000" frequency band are also built from different denominators, so they cannot simply be equated. The authors' summary captures both halves: semaglutide use was associated with increased NAION risk compared with SGLT-2 inhibitors, but the absolute risk remains low.^[3]

The multinational study that found no significant association

Chou and colleagues analysed a deidentified global electronic medical records database spanning 160 health care organizations across 21 countries, with enrollment from January 2017 to August 2023.^[4] After propensity matching, they compared semaglutide with non-GLP-1 medicines in three groups — type 2 diabetes only, obesity only, and both conditions — and found no statistically significant association with NAION at one, two or three years of follow-up in any group.^[4]

That result deserves equal care. In the diabetes-only group at one year, the hazard ratio was 2.32 with a 95% confidence interval of 0.60 to 8.97^[4] — consistent with anything from a meaningful decrease to an eightfold increase. "Not statistically significant" here means the data could not separate the effect from no effect at that precision; it is not proof of zero risk. The authors' own conclusion was measured: semaglutide may not be associated with increased NAION risk in the general population, and avoiding it solely over NAION concerns may not be warranted given its established benefits.^[4]

Source	What it found	What it cannot show
EMA PRAC, June 2025	NAION listed as a very rare semaglutide side effect	A personal risk figure for any individual
Hathaway 2024, one US centre	Hazard ratios of 4.28 and 7.64 by indication	Causation, or risk in unselected users
Simonsen 2025, Denmark–Norway	HR 2.81; about +1.4 cases per 10,000 person-years	A class-wide GLP-1 effect
Chou, 21 countries	No significant association; wide intervals	Proof that the risk is zero

Why an unproven mechanism does not settle the argument

These observational studies do not establish the biological mechanism behind the association. EMA nevertheless reached a safety conclusion after reviewing available evidence, including clinical trials, postmarketing surveillance and the medical literature.^[1] An unresolved mechanism does not erase that regulatory assessment; it limits the explanation of how the event may arise.

The exact named outcome matters. This article concerns NAION; its estimates should not be relabeled as a general rate for every possible eyesight problem.

Reading the next headline

A proposed checklist for the next scary or reassuring story: identify the exact drug and condition; ask whether the number is relative or absolute; ask who was studied — a referral clinic, a national registry, a records network; and check whether "no association" comes with intervals narrow enough to be informative. For patients, the operational guidance remains the EMA's: sudden vision loss or rapidly worsening eyesight during semaglutide treatment warrants contacting a doctor without delay, and a confirmed NAION diagnosis means stopping semaglutide under medical care — not an independent decision to discontinue.^[1] Nothing here is advice for any individual's treatment.

Finally, a scope note. These studies concern semaglutide use recorded in health systems. They do not authenticate the composition or clinical suitability of a research vial. Research materials should not be consumed or self-administered. The EMA page was reviewed in full for this article; the three studies were reviewed as their original published abstracts, and their full texts were not assessed. Within those limits, the reviewed evidence includes positive observational associations, a study without a statistically significant association, and EMA's conclusion that NAION is a very rare side effect of semaglutide.