Semaglutide and NAION: Reading the Eye-Risk Numbers

EMA concluded that NAION, an optic-nerve condition that can cause vision loss, is a very rare side effect of semaglutide. That is a specific safety finding, not proof that most people taking the medicine will develop eye damage. Sudden loss of vision or rapidly worsening eyesight needs prompt medical attention¹.

The risk discussion becomes confusing when three different things are treated as interchangeable: an eye condition, a relative-risk estimate and an individual person's chance of experiencing it. Separating them makes the evidence useful without minimizing a serious outcome or making a rare one sound common.

What the regulator concluded

In its June 2025 review, EMA's safety committee considered clinical, observational, laboratory and postmarketing evidence. It recommended adding non-arteritic anterior ischemic optic neuropathy, or NAION, to semaglutide product information. EMA describes the frequency category as very rare¹.

Its summary says studies in adults with type 2 diabetes suggest approximately twice the risk compared with people not taking semaglutide. It translates that into roughly one additional case per 10,000 person-years of treatment¹. These are the regulator's approximate figures, not a prediction generated for an individual reader.

A person-year counts time under observation. Many people each followed for a shorter period can contribute the same total person-years as fewer people followed longer. It is not identical to a statement that a fixed fraction of everyone who ever starts treatment will eventually have the event.

Why relative and absolute risk sound different

Imagine a rare event with a hypothetical baseline rate of one case per 10,000 people over a specified period. Doubling that rate would produce two cases per 10,000, an increase of one case. The relative increase sounds large; the absolute increase remains small. This example illustrates the arithmetic and is not a replacement estimate for NAION.

For a common event, the same relative increase would produce many more additional cases. That is why a headline about risk “doubling” is incomplete without the baseline and observation period. It may accurately report a ratio while leaving out what most readers need to judge the size of the change.

The converse mistake is to use rarity to dismiss severity. An outcome can be uncommon and still deserve a clear warning because its consequences are substantial. A sensible account gives the frequency and explains what a person should do if symptoms occur.

What observational evidence can and cannot show

In an observational study, people are not assigned at random to start a drug. The groups may differ in health, reasons for treatment, access to care and how often an eye problem is detected. Researchers can try to account for measured differences, but adjustment is not the same as random assignment.

That does not make observational evidence disposable. Rare adverse events may be difficult to study in a trial designed for weight loss or blood-sugar control. A much larger real-world dataset can identify a signal that deserves investigation. The question is how consistent and credible that signal is, not whether the study fits a simple hierarchy label.

Regulatory assessment also draws on more than one study. EMA's conclusion should therefore not be described as the result of one alarming case report. Equally, a study with no statistically clear signal would not by itself erase the regulator's broader review.

When reading a new paper, look for the actual comparison and exclusions before reading the conclusion. A study of people with diabetes answers a different direct question from a study of people without diabetes using treatment solely for weight management. Combining both under “all users” can imply more precision than the evidence provides.

NAION is not the same as diabetic retinopathy

The name identifies an optic-nerve condition. Diabetic retinopathy concerns damage in the retina related to diabetes. Calling both “eye damage” may be understandable in a conversation, but it is too broad for interpreting a drug safety claim.

The National Eye Institute explains that diabetic retinopathy can cause vision loss and may initially have no symptoms. Diabetes eye care includes appropriate eye examinations and management of the underlying condition². Our diabetic-retinopathy evidence review explains that separate trial signal. A discussion of retinopathy should not be used as if it were direct evidence about NAION.

This distinction also prevents an unsafe shortcut in the other direction. A person with a visual change should not decide that it must be a known, harmless treatment effect. The symptom needs clinical assessment; an article cannot identify which eye condition is responsible.

What to do with a sudden change in vision

EMA advises people with sudden vision loss or rapidly worsening eyesight during semaglutide treatment to contact a doctor without delay. Its advice is to stop semaglutide if NAION is confirmed¹. Confirmation and treatment decisions require medical assessment, not a self-diagnosis from a symptom list.

Tell the clinician what medicine and formulation you take, when you started and when the visual change appeared. Bring the prescription or product label if available. Those details help the clinical team evaluate the event without assuming that timing alone establishes its cause.

Do not wait for a routine follow-up because an online discussion calls the event rare. Rarity describes a population. It does not make an active symptom less important for the person experiencing it. Conversely, the warning is not an instruction for every symptom-free reader to abandon prescribed treatment without discussing their situation.

Does the finding apply identically to every GLP-1 drug?

The EMA communication reviewed semaglutide medicines. A class-wide claim needs class-wide evidence. It is not accurate to transfer its exact risk estimate to tirzepatide, a research peptide or a compounded mixture simply because all appear in weight-loss discussions.

A different formulation also does not acquire a clean safety record from an absence of dedicated studies. For compounded drugs, FDA explains that it does not perform the same premarket review of safety, effectiveness and quality as for approved products³. “Not measured” is not equivalent to “no risk.”

Our semaglutide evidence page covers the broader treatment record, including benefits. The right risk-benefit discussion connects that record to the person's medical history. This page's narrower job is to keep a rare but serious eye finding from being exaggerated, blurred into a different diagnosis or dismissed.

Questions for the prescribing visit

Ask whether your eye history changes the treatment discussion, what visual symptoms should trigger urgent assessment and how routine eye care fits with the rest of your care. If another drug is proposed, ask which evidence supports the comparison rather than assuming all GLP-1 medicines have identical data.

For evaluating a news story, ask whether it names the regulator and condition, provides an absolute estimate and identifies the studied population. Those details make a short explanation more reliable than a dramatic headline followed by an unrelated list of common side effects.