GLP-1 Agonists and Retinal Health: Help or Harm?

Peptide Therapies for Diabetic Eye Disease

3.0% DR complications

In the SUSTAIN-6 cardiovascular outcomes trial, 3.0% of patients receiving semaglutide developed diabetic retinopathy complications versus 1.8% on placebo, generating one of the most debated safety signals in the GLP-1 class.

Marso et al., New England Journal of Medicine, 2016

Marso et al., New England Journal of Medicine, 2016

View as image

When the SUSTAIN-6 cardiovascular outcomes trial reported that semaglutide was associated with more diabetic retinopathy (DR) complications than placebo, it raised immediate questions about the safety of GLP-1 receptor agonists for the eyes of patients with diabetes.^[1] Five years of subsequent research has complicated rather than resolved this signal. Real-world evidence largely contradicts the trial finding. Preclinical data suggests GLP-1 receptor activation may actually protect retinal cells. And a separate concern about a rare optic nerve condition (NAION) has added a new dimension to the debate. For how this fits into the broader landscape of peptide-based eye therapies, see our pillar article on peptide therapies for diabetic eye disease.

The SUSTAIN-6 signal: what the trial actually showed

The semaglutide retinopathy concern originated from SUSTAIN-6, a cardiovascular outcomes trial that randomized 3,297 patients with type 2 diabetes and high cardiovascular risk to semaglutide or placebo for 2 years. Marso et al. (2016) reported that 3.0% of semaglutide patients developed DR complications (vitreous hemorrhage, DR-related blindness, or need for photocoagulation/intravitreal agents) compared to 1.8% on placebo (HR 1.76).^[1]

Several features of this signal are important for context:

• Patients who developed DR complications overwhelmingly had pre-existing retinopathy at baseline
• The semaglutide group experienced larger, faster HbA1c reductions than placebo
• The "early worsening" phenomenon with rapid glycemic improvement was already known from insulin trials
• SUSTAIN-6 was powered for cardiovascular events, not retinal outcomes, so the DR finding was based on a small number of events

The FDA added a warning about DR worsening to semaglutide labels (Ozempic, Wegovy), making the retinal question permanently visible to clinicians and patients.

The rapid glycemic improvement hypothesis

The leading explanation for DR worsening with GLP-1RAs is not a direct retinal toxicity but rather a hemodynamic consequence of rapid glucose normalization.

In patients with long-standing poorly controlled diabetes, retinal blood vessels have adapted to a high-glucose environment. When glucose drops rapidly, these compromised vessels cannot adapt quickly enough. Retinal ischemia worsens transiently, triggering neovascularization and hemorrhage. This phenomenon was first documented with insulin initiation decades before GLP-1RAs existed.

Jalleh et al. (2026) addressed this mechanism in their comprehensive JCI safety review, noting that DR worsening appears most likely in patients with pre-existing retinopathy and poor baseline glycemic control who experience large, rapid HbA1c reductions.^[2] Patients without pre-existing retinopathy or with well-controlled diabetes appear to be at minimal risk. This suggests the signal is a class effect of effective glucose lowering rather than a GLP-1-specific retinal toxicity.

Meta-analysis evidence: early worsening, late protection

Kapoor et al. (2023) published a meta-analysis in Survey of Ophthalmology that pooled DR outcomes from randomized clinical trials of GLP-1RAs. Their analysis revealed a nuanced pattern: GLP-1RA use was associated with increased risk of early-stage DR compared to placebo, but when compared to insulin (which also produces rapid glucose lowering), GLP-1RAs showed protection against late-stage DR.^[3]

This comparison matters because insulin is the most relevant clinical comparator. If GLP-1RAs cause more DR than placebo but less than insulin in the long term, the signal may reflect the glucose-lowering effect rather than a unique drug toxicity. The net retinal effect over years of treatment may actually favor GLP-1RAs over insulin intensification.

Real-world evidence: the SUSTAIN-6 signal fades

Barkmeier et al. (2026) published a comparative effectiveness study in Ophthalmology Retina examining the risk of sight-threatening DR across different GLP-1RA agents (semaglutide, dulaglutide, liraglutide, exenatide) in routine clinical practice. Using target trial emulation methodology, they found no significant difference in sight-threatening DR risk between agents.^[4]

This finding directly contradicts the idea that semaglutide carries a unique retinal risk within the GLP-1RA class. If the SUSTAIN-6 signal reflected a semaglutide-specific toxicity, semaglutide should show higher DR rates than other GLP-1RAs in real-world data. It does not.

Hong et al. (2026) extended this comparison to tirzepatide (a dual GIP/GLP-1 receptor agonist). Their study in Ophthalmology Retina compared ocular outcomes between tirzepatide and GLP-1RAs in patients with type 2 diabetes. Despite tirzepatide producing greater glycemic and metabolic improvements than GLP-1RAs, ocular outcomes did not differ significantly between groups.^[5] This adds further evidence that the degree of glucose lowering, not the specific drug mechanism, determines retinal risk.

The potential protective side: preclinical evidence

While the clinical debate focuses on harm, preclinical research tells a different story. GLP-1 receptors are expressed on retinal cells, and GLP-1 receptor activation appears to have protective effects.

Huang et al. (2026) published a review in Biochemical Pharmacology examining GLP-1RA effects on endoplasmic reticulum (ER) stress in diabetic retinopathy. Chronic hyperglycemia induces ER stress in retinal neurons and vascular endothelial cells, contributing to cell death and vascular dysfunction. GLP-1 receptor activation reduces ER stress markers and inhibits apoptosis in retinal cell lines and animal models of DR.^[6]

Gong and Orge (2026) reviewed the broader evidence on GLP-1RAs and ocular disease, covering mechanisms relevant to glaucoma, DR, and age-related macular degeneration. They identified anti-inflammatory, anti-oxidant, and neuroprotective mechanisms of GLP-1 receptor activation in retinal tissue, operating through pathways independent of glucose lowering.^[7]

These preclinical findings create a paradox: GLP-1 receptor activation may protect retinal cells while rapid glucose normalization (an indirect consequence of GLP-1RA treatment) may harm retinal vasculature in vulnerable patients. The net clinical effect likely depends on the patient's baseline retinal status and the speed of glycemic improvement.

The specific protective pathways identified in preclinical models include:

• ER stress reduction: GLP-1R activation decreases CHOP, GRP78, and other ER stress markers in retinal pigment epithelial cells exposed to high glucose conditions
• Anti-inflammatory effects: suppression of NF-kB signaling reduces retinal vascular inflammation, which drives DR progression
• Neuroprotection: GLP-1R activation on retinal ganglion cells activates PI3K/Akt survival signaling, reducing neuronal apoptosis
• Vascular stabilization: in animal models, GLP-1R signaling reduces vascular leakage and inhibits pathological neovascularization

Whether these effects are clinically meaningful at the systemic GLP-1RA doses used in diabetes and obesity treatment, versus the concentrations achievable with intravitreal delivery, is unknown. Several research groups are investigating topical or intravitreal GLP-1 analogs specifically designed for retinal delivery, which could separate the protective retinal effects from the systemic glucose-lowering effects that produce the early worsening risk.

Other GLP-1RA cardiovascular trials and retinal outcomes

The SUSTAIN-6 retinopathy signal was not replicated in other major GLP-1RA cardiovascular outcome trials. LEADER (liraglutide), PIONEER 6 (oral semaglutide), REWIND (dulaglutide), and EXSCEL (exenatide) did not report significant increases in DR complications. This inconsistency across trials has been interpreted in multiple ways:

• SUSTAIN-6 enrolled a higher proportion of patients with pre-existing retinopathy
• Semaglutide produces larger, faster HbA1c reductions than liraglutide, dulaglutide, or exenatide, potentially generating a larger early worsening effect
• The SUSTAIN-6 finding may represent statistical noise in a small number of events rather than a true signal

The absence of a consistent signal across trials is one reason that most diabetes guidelines have not added retinal screening requirements specifically for GLP-1RA initiation, though they emphasize the importance of baseline and regular retinal examinations for all patients with diabetes.

The NAION question: a new ocular concern

Non-arteritic anterior ischemic optic neuropathy (NAION) is a rare condition where blood supply to the optic nerve is suddenly compromised, causing acute vision loss. Reports of NAION in patients taking semaglutide generated attention separate from the DR signal.

Abdelaal et al. (2026) published a systematic review in Ophthalmology examining whether semaglutide use is associated with NAION risk. Their analysis found uncertain evidence for an association, with major confounders complicating interpretation: obesity, diabetes, hypertension, and sleep apnea are all independent risk factors for NAION and are also conditions for which semaglutide is prescribed.^[8]

Chen et al. (2026) conducted a separate meta-analysis specifically asking whether semaglutide increases NAION risk. Their pooled analysis also found insufficient evidence to confirm an association, noting that the absolute number of NAION cases in both semaglutide and comparator groups was very small, limiting statistical power.^[9]

Cheng et al. (2026) approached the question through multi-database pharmacovigilance, analyzing GLP-1RA-related ophthalmic reports across the FAERS database and international reporting systems. They found disproportionate reporting of ophthalmic events with GLP-1RAs but could not establish causation given the inherent limitations of spontaneous reporting data.^[10]

The NAION signal remains under investigation. No study has established a causal mechanism, and the overlap between NAION risk factors and GLP-1RA indications makes isolating a drug effect methodologically difficult. NAION has an incidence of approximately 2-10 per 100,000 people per year in the general population, and the condition disproportionately affects individuals with the same cardiometabolic risk profile (obesity, diabetes, hypertension, hyperlipidemia) that drives GLP-1RA prescribing. Disentangling a drug signal from this baseline risk in an observational setting requires very large sample sizes and careful confounder adjustment, which current studies have not fully achieved.

How to interpret the evidence

The GLP-1 retinal safety story is more complex than either "safe" or "dangerous" captures:

What the evidence supports:

• Rapid glycemic improvement with any drug (including insulin) can worsen pre-existing retinopathy in the short term
• This risk is highest in patients with advanced baseline retinopathy and poor glycemic control
• Real-world data shows no semaglutide-specific retinal risk compared to other GLP-1RAs
• Preclinical evidence suggests direct retinal protective effects of GLP-1 receptor activation

What remains uncertain:

• Whether GLP-1RAs cause more early DR worsening than equivalent glycemic improvement from other drug classes
• Whether the protective preclinical findings translate to measurable clinical benefit in humans
• Whether NAION risk is genuinely elevated with GLP-1RA use
• Long-term retinal outcomes beyond 5 years of treatment

What the safety approach looks like:

• Kunutsor et al. (2026) recommend retinal screening before GLP-1RA initiation in patients with existing retinopathy, with gradual dose titration to avoid rapid glucose drops.^[11] This approach addresses the modifiable risk factor (speed of glycemic change) while allowing patients to benefit from GLP-1RA therapy.

For how somatostatin analogs represent a separate peptide approach to retinal protection, see the cluster sibling on somatostatin analogs for retinal neuroprotection.

The Bottom Line

The SUSTAIN-6 trial signal that semaglutide may worsen diabetic retinopathy is best understood as a consequence of rapid glycemic improvement in patients with pre-existing retinal disease, not a GLP-1-specific retinal toxicity. Real-world comparative data shows no difference in sight-threatening DR risk between semaglutide and other GLP-1RAs. A meta-analysis found GLP-1RAs actually protect against late-stage DR compared to insulin. Preclinical research identifies direct retinal protective mechanisms through GLP-1 receptor activation, including ER stress reduction and anti-inflammatory effects. The NAION association remains unconfirmed, with major confounders limiting causal conclusions. Retinal screening before treatment and gradual dose titration address the modifiable risk factor.

Frequently Asked Questions

Does semaglutide cause diabetic retinopathy?

The SUSTAIN-6 trial found 3.0% of semaglutide patients developed DR complications versus 1.8% on placebo (Marso et al., NEJM, 2016). However, affected patients overwhelmingly had pre-existing retinopathy. The leading hypothesis is that rapid blood sugar improvement, not semaglutide specifically, triggers early DR worsening in vulnerable retinas. Real-world data from Barkmeier et al. (2026) found no increased DR risk with semaglutide versus other GLP-1 drugs.

Can GLP-1 drugs protect the retina?

Preclinical evidence suggests yes. GLP-1 receptors are expressed on retinal cells, and GLP-1 receptor activation reduces ER stress, inhibits apoptosis, and decreases inflammation in retinal tissue (Huang et al., 2026; Gong and Orge, 2026). A meta-analysis found GLP-1RAs protect against late-stage DR compared to insulin (Kapoor et al., 2023). Whether these protective effects translate to measurable clinical benefit in humans is not yet confirmed.

Should I get an eye exam before starting a GLP-1 drug?

Kunutsor et al. (2026) recommend retinal screening before GLP-1RA initiation in patients with existing diabetic retinopathy, with gradual dose titration to minimize the risk of rapid glycemic changes. Patients without pre-existing retinopathy and with well-controlled diabetes appear to be at minimal risk of retinal complications from GLP-1RA therapy.

Is semaglutide linked to optic nerve damage (NAION)?

Reports of non-arteritic anterior ischemic optic neuropathy (NAION) in semaglutide users prompted investigation. Two systematic reviews (Abdelaal et al., 2026; Chen et al., 2026) found insufficient evidence to confirm a causal association. Obesity, diabetes, hypertension, and sleep apnea are independent NAION risk factors and are also conditions for which semaglutide is prescribed, making it difficult to isolate a drug-specific effect.

Do all GLP-1 drugs affect the eyes the same way?

Real-world comparative data suggests no significant difference in retinal risk between semaglutide, dulaglutide, liraglutide, and exenatide (Barkmeier et al., 2026). Tirzepatide, which produces even greater glycemic improvements, showed similar ocular outcomes to GLP-1RAs (Hong et al., 2026). This supports the hypothesis that glycemic change speed, not the specific drug, determines retinal risk.

Why does rapid blood sugar improvement worsen diabetic retinopathy?

In patients with long-standing poorly controlled diabetes, retinal blood vessels have adapted to a high-glucose environment. Rapid glucose normalization causes these compromised vessels to experience transient ischemia because they cannot adapt quickly enough. This triggers neovascularization and hemorrhage. The phenomenon was first documented with insulin treatment decades before GLP-1 drugs existed and appears to be a hemodynamic effect of glucose change rather than a drug-specific toxicity.