GLP-1s and NAION: What the Evidence Actually Shows

NAION has become one of the more talked-about safety concerns surrounding GLP-1 therapy. The overall evidence is reassuring, with appropriate caution.

There is a credible safety signal, particularly with semaglutide in people with type 2 diabetes. The absolute risk, however, appears to be very small. Results have also been inconsistent across studies, and the randomized clinical trial data available so far have not shown an increase in NAION.

The quick take

What is NAION?

Non-arteritic anterior ischemic optic neuropathy, or NAION, occurs when blood flow to the front portion of the optic nerve becomes inadequate.

The usual presentation is sudden, painless vision loss or a new visual-field defect in one eye, often first noticed on awakening.

NAION can cause permanent visual impairment, so it deserves to be taken seriously. At the same time, describing every case simply as “blindness” paints a more dramatic picture than the usual clinical presentation. In one modern population cohort, median visual acuity was approximately 20/40 at diagnosis and 20/50 at final follow-up, although significant permanent visual-field loss can occur. ^[13]

NAION also occurs in people who have never taken a GLP-1 medication.

In one U.S. population cohort, overall incidence was approximately 4 per 100,000 people per year, rising to about 10 per 100,000 among adults aged 50 or older. Estimates vary by population. Baseline risk is considerably higher in people with diabetes and other vascular risk factors. ^[13]

What have the studies actually found?

The literature makes much more sense when the studies are viewed together instead of focusing on the largest hazard ratio.

There are additional studies on both sides. Hsu and colleagues found a longer-term association emerging after the first year of semaglutide exposure, with a stronger signal among patients with hypertension. Other large matched analyses found no significant association at all.

That variability is important when interpreting causality. ^[5]

The pooled randomized analysis evaluated completed Novo Nordisk trials of liraglutide and semaglutide, rather than all GLP-1 based drugs. Exposure differed between groups: 119,393 participant-years with active therapy and 86,384 with placebo, corresponding to approximately 3 versus 6 confirmed cases per 100,000 participant-years. Five placebo cases occurred in four participants. The analysis included manufacturer employees as authors. With only eight confirmed cases, it cannot exclude a small increase in risk. ^[9]

Relative risk can sound much scarier than absolute risk

This is probably the most useful way to put the concern into perspective.

Consider three of the positive studies:

A headline can accurately say that risk was doubled or nearly tripled.

A patient hearing that may reasonably think the complication is common.

The absolute numbers show something very different. In the population studies that find an association, we are generally talking about tens of additional events per 100,000 patient-years.

The risk is not zero. It is very small.

Why does diabetes keep showing up?

This is one of the more interesting patterns in the data.

Most of the positive studies involve people with type 2 diabetes. The obesity-only evidence has been considerably less convincing. ^[12]

That may be because patients with diabetes already have many of the characteristics associated with NAION:

• Diabetes itself
• Hypertension
• Dyslipidemia
• Obstructive sleep apnea
• Obesity
• Cardiovascular and vascular disease
• Older age
• Smoking
• Crowded optic-nerve anatomy

In a population-based NAION cohort, approximately 80% had hypertension, 74% had hyperlipidemia, 39% had diabetes, and 23% had obstructive sleep apnea. ^[13]

In one study of adults older than 67, diabetes itself was associated with roughly a 40% increase in NAION risk. ^[14]

This makes observational GLP-1 research difficult. The patient most likely to receive semaglutide may already be more likely to develop NAION because of their underlying metabolic and vascular disease.

Statistical matching helps, although it cannot account for every difference between patients.

What about tirzepatide?

I would be cautious about making molecule-to-molecule comparisons right now.

The largest study involving tirzepatide grouped it together with semaglutide. Case reports and pharmacovigilance databases have also identified events with tirzepatide, but those types of data are useful for detecting signals rather than calculating incidence.

We simply do not have the same depth of tirzepatide-specific epidemiologic evidence that we have for semaglutide. ^{[7] [15]}

Current data do not establish tirzepatide as safer or NAION as a class effect.

What might be causing it?

We do not know.

Several possibilities are being studied:

Baseline vascular disease

This is probably the simplest contributor. Patients with diabetes, hypertension, dyslipidemia, OSA, and vascular disease already have more vulnerable optic-nerve circulation.

Rapid improvement in glycemia

The 2026 UK analysis found a stronger NAION association in patients whose A1C fell by approximately 1 percentage point or more.

A published ophthalmic case series has also raised rapid metabolic correction as a possible contributor.

That is interesting, but still hypothesis-generating. ^{[8] [15]}

Falling blood pressure and optic-nerve perfusion

NAION has long been associated with impaired optic-nerve perfusion, and many episodes are discovered on awakening. Nocturnal hypotension has therefore been discussed in neuro-ophthalmology for years.

This has some practical relevance in obesity medicine.

A patient may start treatment with obesity, hypertension, diabetes, and several antihypertensive medications. After meaningful weight loss, their blood pressure may fall substantially while the antihypertensive regimen stays exactly the same.

A practical consideration during substantial weight loss is to reassess blood pressure treatment:

Don't let yesterday's appropriate antihypertensive regimen become tomorrow's over-treatment.

Rechecking home blood pressure and adjusting therapy when clinically appropriate is already good obesity-medicine care.

There is no evidence that GLP-1-related blood-pressure reduction or nocturnal hypotension causes NAION. Current evidence also does not support routinely changing the timing of antihypertensive medication to prevent it.

What is a “disc at risk”?

A large percentage of patients who develop NAION have a small, crowded optic nerve with little or no physiologic cup. Ophthalmologists often refer to this as a “disc at risk.”

It is a genuine susceptibility factor for NAION.

There is currently no evidence that every patient starting semaglutide or tirzepatide should undergo ophthalmologic screening for optic-disc anatomy.

A crowded disc is also not a GLP-1 contraindication.

For someone with prior NAION, significant optic-nerve disease, unexplained previous vision loss, or an unusual combination of ocular and systemic risk factors, involving ophthalmology or neuro-ophthalmology makes sense. ^[12]

Practical considerations in clinic

There is no established NAION-prevention protocol for GLP-1 users. Routine “disc-at-risk” screening, deliberately slowing weight loss, changing BP medication timing, or increasing water intake have not been shown to prevent GLP-1 associated NAION.

We should also keep the bigger vision risk in perspective

A balanced discussion of this rare potential medication complication should also account for the effects of diabetes itself on eye health.

Diabetic retinopathy remains the most frequent cause of new blindness among adults ages 20–74 in high-income countries. ^[11]

Chronic hyperglycemia is a well-established driver of diabetic retinopathy. High-quality randomized evidence has shown that improving glycemic control can prevent or delay its onset and progression. Blood-pressure and lipid management are important as well.

This does not establish that GLP-1 medication directly prevents blindness; the evidence is more nuanced.

GLP-1 based therapies can produce major improvements in glycemia, weight, blood pressure, cardiovascular risk, kidney outcomes, and overall cardiometabolic health. Those benefits matter when we discuss a rare potential adverse event.

There is also an important ophthalmologic consideration on the other side: rapid A1C improvement can temporarily worsen established diabetic retinopathy in some high-risk patients. ADA recommends assessing retinopathy status when glucose-lowering therapy is intensified in susceptible patients. ^[11]

Patients deserve a balanced discussion of both the potential risk and the established benefits.

Interpretation of the evidence

The NAION signal warrants attention. Multiple observational studies have found an association, particularly with semaglutide in people with type 2 diabetes. European regulators classify NAION as a very rare adverse effect of semaglutide. ^[10]

Multiple observational studies have found an association, especially with semaglutide in patients with type 2 diabetes. European regulators considered the evidence strong enough to classify NAION as a very rare adverse effect of semaglutide. ^[10]

At the same time:

• The absolute excess risk in population studies is generally very small.
• Several large observational studies have found no significant association.
• The strongest signal consistently appears in patients with diabetes and significant baseline vascular risk.
• Tirzepatide-specific evidence remains limited.
• Nearly 100,000 participants in pooled randomized trials showed 3 confirmed NAION cases with liraglutide or semaglutide versus 5 with placebo.
• A direct biological mechanism has not been established; European regulators nevertheless recognize NAION as a very rare semaglutide adverse effect.

For most appropriately selected patients, GLP-1 based therapies retain important benefits across weight, glycemia, cardiovascular and kidney outcomes. These benefits should be considered alongside the rare potential risk.

These medications are treating diseases that carry substantial cardiovascular, renal, metabolic, and ophthalmologic consequences of their own.

The goal should be to recognize the safety signal, understand who may be more vulnerable, monitor patients thoughtfully, and explain the numbers accurately.

A twofold increase in a very rare event can still be a very rare event.