GLP-1 + Diabetic Retinopathy: SUSTAIN-6 Signal + Monitoring

SUSTAIN-6 (Marso 2016 NEJM^[1]) reported a hazard ratio of 1.76 (95% CI 1.11–2.78) for diabetic-retinopathy complications with semaglutide vs placebo — a 76% increase. The signal was confined to patients with pre-existing retinopathy at baseline, and the dominant hypothesis is rapid HbA1c lowering producing the same early-worsening pattern observed in the DCCT (1998 Archives of Ophthalmology^[6]) and the ACCORD Eye sub-study (Chew 2010 NEJM^[5]). Liraglutide (LEADER^[3]) and dulaglutide (REWIND^[4]) did not reproduce the signal; the FDA carries a Section 5.6 warning on the Ozempic and Wegovy labels but not on Trulicity. This article walks through what the data actually says and the practical ophthalmology screening pathway that follows from it.

The honest summary

• SUSTAIN-6 showed a real signal, in a specific subgroup. Marso 2016^[1] reported diabetic-retinopathy complications in 50 of 1,648 semaglutide patients (3.0%) vs 29 of 1,649 placebo (1.8%), HR 1.76 (95% CI 1.11–2.78, p = 0.02). Vilsboll 2018^[2] showed the signal was concentrated in patients with pre-existing retinopathy at baseline and correlated with the magnitude of HbA1c drop in the first 16 weeks.
• It is the rate of glucose lowering, not the GLP-1 drug itself. The DCCT (1998^[6]) and the ACCORD Eye sub-study (Chew 2010^[5]) showed the same early-worsening pattern with intensive insulin in T1D and intensive multi-drug therapy in T2D. Rapid normalization of chronic hyperglycemia is the proximate trigger.
• Other GLP-1s did not reproduce it. LEADER (liraglutide^[3]) and REWIND (dulaglutide^[4]) showed no excess retinopathy signal; semaglutide produces the largest and fastest HbA1c drop of the three, consistent with the rate-of-lowering hypothesis.
• Obesity-only patients without T2D are not in the risk group. The FDA Wegovy and Ozempic Section 5.6 warnings apply to T2D patients with established retinopathy; routine pre-treatment ophthalmology screening is not required for obesity-only patients without diabetes.

What SUSTAIN-6 actually reported

SUSTAIN-6 (Marso 2016 NEJM^[1]) randomized 3,297 patients with T2D and high cardiovascular risk to semaglutide 0.5 mg or 1.0 mg weekly vs placebo for a median 2.1 years. The primary composite (cardiovascular death, nonfatal MI, nonfatal stroke) favored semaglutide (HR 0.74, p < 0.001 for noninferiority, p = 0.02 for superiority). The retinopathy-complications endpoint — a pre-specified secondary composite of vitreous hemorrhage, blindness, or photocoagulation or anti-VEGF therapy — went the wrong way:

• Semaglutide: 50 events of 1,648 patients (3.0%)
• Placebo: 29 events of 1,649 patients (1.8%)
• Hazard ratio: 1.76 (95% CI 1.11–2.78, p = 0.02)

Vilsboll 2018 (Diabetes Obesity & Metabolism^[2]) ran a pre-specified post-hoc analysis and found three things that explain the signal. First, the excess risk was concentrated in patients with pre-existing diabetic retinopathy at randomization (the absolute risk roughly doubled in this subgroup; the no-retinopathy subgroup showed no signal). Second, the excess was concentrated in patients with the largest HbA1c drop in the first 16 weeks of treatment. Third, the magnitude of HbA1c reduction correlated linearly with event rate — the steeper the drop, the higher the risk. Verma 2020^[7] confirmed the microvascular-burden signal in a pooled LEADER + SUSTAIN-6 analysis: patients with pre-existing microvascular disease carried the bulk of the adverse outcomes.

Mechanism: rapid glycemic lowering, not the molecule

The pathophysiology is borrowed from a literature that predates GLP-1s by twenty years. In chronic hyperglycemia, the retinal vasculature adapts to a high-glucose milieu with capillary dropout, microaneurysms, and a fragile collateral network. When systemic glucose drops rapidly — whether from intensive insulin in the DCCT (1998^[6]), multi-drug intensive therapy in ACCORD (Chew 2010^[5]), or a potent GLP-1 in SUSTAIN-6^[1] — the retinal capillary bed transitions from glucose-saturated to relative ischemia. That ischemia drives VEGF expression and, in eyes with pre-existing damage, accelerates progression to vitreous hemorrhage, macular edema, or proliferative disease. Patients with healthy retinas have the vascular reserve to accommodate the transition; patients with damaged retinas do not.

This is why the SUSTAIN-6 signal does not appear in LEADER (liraglutide^[3]) or REWIND (dulaglutide^[4]): both produce a smaller and slower HbA1c reduction than semaglutide 1.0 mg. It is also why the same signal does not appear in the obesity trials of semaglutide (STEP) or tirzepatide (SURMOUNT) — those enrolled patients without T2D, who do not carry the pre-existing retinal damage that defines the risk subgroup.

FOCUS: the dedicated retinopathy trial

Novo Nordisk launched FOCUS (NCT03811561^[9]) as the regulatory follow-up to the SUSTAIN-6 signal. It is a five-year randomized double-blind placebo-controlled trial in approximately 1,500 T2D patients with established diabetic retinopathy, with a primary endpoint of a three-step or greater progression on the ETDRS Diabetic Retinopathy Severity Scale (or development of proliferative disease, or photocoagulation or anti-VEGF intervention, or vitreous hemorrhage or blindness). The trial began enrolling in 2019; primary results have not yet been published as of this writing in mid-2026. The expected readout will resolve whether the SUSTAIN-6 signal is a transient early-worsening phenomenon that resolves over years or a sustained adverse effect.

The other GLP-1 retinopathy data

Liraglutide. LEADER (Marso 2016 NEJM^[3]) randomized 9,340 T2D patients to liraglutide 1.8 mg daily or placebo. Retinopathy events occurred in 2.3% of liraglutide vs 2.0% of placebo — not statistically significant (p = 0.33). No FDA Section 5 retinopathy warning was added to the Victoza or Saxenda labels.

Dulaglutide. REWIND (Gerstein 2019^[4], microvascular sub-analysis) randomized 9,901 T2D patients and reported no excess retinopathy with dulaglutide 1.5 mg weekly. Kwan 2022^[8] confirmed the safety pattern across microvascular complication strata: HbA1c reduction was similar across patients with and without baseline retinopathy, and no excess events were observed.

Tirzepatide. The SURPASS T2D program and the SURMOUNT obesity program have not reported a comparable retinopathy signal. The SURPASS trials enrolled T2D patients but did not stratify enrollment by baseline retinopathy status, and the SURMOUNT trials enrolled obesity-only patients without T2D — the wrong risk subgroup to detect the signal. The FDA Mounjaro and Zepbound labels do not carry a Section 5 retinopathy warning.

What the FDA labels actually require

The Wegovy and Ozempic Section 5.6 (Diabetic Retinopathy Complications in Patients with Type 2 Diabetes Mellitus) labels say: rapid improvement in glucose control has been associated with a temporary worsening of diabetic retinopathy; in patients with a history of diabetic retinopathy, monitor for progression. The label does not mandate a specific screening interval, but the practical interpretation aligned with ADA Standards of Care is:

• T2D patient, no prior eye exam: dilated fundoscopy or non-mydriatic fundus photography before starting semaglutide, or within 30–60 days.
• T2D patient with documented retinopathy: ophthalmology consultation before initiation; baseline severity grading; ophthalmology re-exam every 3–6 months during titration and the first year of treatment.
• T2D patient with active proliferative disease or vision-threatening DME: defer GLP-1 initiation until laser photocoagulation or anti-VEGF therapy has stabilized the retina, then proceed with monthly ophtho monitoring for the first six months.
• Obesity-only patient without T2D: no routine ophthalmology screening required.

Magnitude: retinopathy progression rate by baseline severity

ETDRS staging for non-specialists

Ophthalmologists grade diabetic retinopathy on the Early Treatment Diabetic Retinopathy Study (ETDRS) severity scale. The clinically relevant strata for GLP-1 decision-making are:

• No retinopathy. Normal fundus. GLP-1 initiation does not require ophthalmology gating; routine annual eye exam continues.
• Mild non-proliferative DR (NPDR). Microaneurysms only. GLP-1 initiation is reasonable with ophthalmology aware; re-exam at 6 months.
• Moderate NPDR. Microaneurysms plus dot/blot hemorrhages or hard exudates. GLP-1 initiation with ophthalmology consultation; re-exam every 3–6 months.
• Severe NPDR. 4-2-1 rule met (severe hemorrhages in 4 quadrants, venous beading in 2 quadrants, or intraretinal microvascular abnormalities in 1 quadrant). Strong indication for anti-VEGF or pan-retinal photocoagulation consideration before GLP-1 escalation.
• Proliferative DR (PDR). Neovascularization of disc, retina, or iris; vitreous hemorrhage. Defer GLP-1 until retina is treated.
• Diabetic macular edema (DME). Center- involving DME requires anti-VEGF treatment (Lucentis, Eylea, or Vabysmo) before or in parallel with GLP-1; the agents are compatible.

The practical pre-GLP-1 screening pathway

1. Establish T2D status. Obesity-only patients without diabetes do not require routine retinopathy screening before GLP-1 initiation. T2D patients require documented fundus assessment within 12 months pre-treatment.
2. Dilated fundoscopy or non-mydriatic fundus photography. Either is acceptable for screening. Fundus photography with AI grading (IDx-DR or Eyenuk EyeArt) is FDA-cleared and increasingly used in primary care. Cost: roughly $50–100 per encounter.
3. If retinopathy is present, ophthalmology consult before GLP-1. The consult documents ETDRS severity and recommends the monitoring cadence. Patients with moderate or worse NPDR should not start a GLP-1 in a primary care vacuum.
4. During titration: ophthalmology exam at 3–6 months for any patient with baseline retinopathy. The first 16 weeks carry the highest risk per Vilsboll 2018^[2], so an early follow-up exam is the highest-yield touchpoint.
5. Treat active disease before escalating dose. Anti-VEGF injection (~$2,000–$4,000 per dose, covered by Medicare and most commercial plans) or pan-retinal photocoagulation should precede dose escalation in any patient with severe NPDR or PDR.
6. Continue annual exams after the first year. Long-term, the early-worsening hypothesis predicts that retinopathy risk normalizes once glycemic equilibrium is re-established. The FOCUS readout^[9] will confirm or refute this.

T1D patients on GLP-1: a separate calculation

GLP-1s are not FDA-approved for type 1 diabetes, but off-label use is common for the weight-loss and insulin-sparing benefits. The DCCT (1998^[6]) is directly relevant here: intensive insulin in T1D produced a 13% rate of early worsening retinopathy at 12 months, with most cases resolving by 36 months and long-term benefit dominating thereafter. For T1D patients with pre-existing retinopathy considering an off-label GLP-1, the same screening pathway applies, plus a baseline ophthalmology consult is mandatory rather than recommended.

What the data does not yet support

• A causal mechanism specific to semaglutide. The SUSTAIN-6 signal is most consistent with rate-of-glucose- lowering, not a direct retinal effect of the molecule. If FOCUS^[9] reads out positive, that interpretation may need to be revised.
• A safety call on tirzepatide retinopathy. The SURPASS T2D program did not pre-specify retinopathy stratification, and no dedicated retinopathy follow-up trial is registered. Patients with pre-existing retinopathy on tirzepatide should still receive the same baseline ophthalmology consult and monitoring cadence even though the label does not require it.
• A safety call on retatrutide retinopathy. The TRIUMPH program is enrolling but not yet reading out microvascular complication data. Same recommendation as tirzepatide: clinician-driven ophthalmology screening regardless of label status.

Related research

• Does semaglutide cause diabetic retinopathy? The SUSTAIN-6 evidence overview — the parallel article focused on the underlying risk evidence rather than the monitoring pathway
• SELECT cardiovascular benefits in non-diabetics — the obesity-only semaglutide CV trial, which does not enroll the retinopathy risk subgroup
• GLP-1 side effects timeline — what side effects appear in the first 16 weeks vs long-term
• GLP-1 in patients with pancreatitis history — another Section 5 warning that requires pre-treatment risk stratification
• Trulicity vs Ozempic vs Mounjaro for T2D — the head-to-head for T2D patients; relevant when retinopathy risk drives the agent choice

Important disclaimer. This article is educational and does not constitute medical advice. Decisions about GLP-1 initiation in patients with diabetic retinopathy should be made jointly by the prescribing clinician, the ophthalmologist or retinal specialist, and the patient. Patients with active proliferative retinopathy or vision-threatening macular edema should defer GLP-1 initiation until retinal treatment is complete. The cost figures cited are approximate and vary by region and insurance plan. PMIDs were verified live against the PubMed E-utilities API on 2026-05-29.

Last verified: 2026-05-29. Next review: every 12 months, or sooner if the FOCUS trial (NCT03811561) primary results publish.