A retrospective cohort study published online ahead of print in the American Journal of Ophthalmology ↗ on April 21 found that GLP-1 receptor agonists were associated with lower rates of major heart, kidney, and retinal complications in patients with type 2 diabetes who already had diabetic retinopathy. The study used the TriNetX research network and covered 173,216 adults between 2015 and 2022.
The headline numbers concern the eye outcomes. Patients on GLP-1 drugs (the class includes semaglutide ↗, dulaglutide ↗, liraglutide ↗, exenatide, tirzepatide ↗, and lixisenatide) progressed to proliferative diabetic retinopathy at 22% lower rates than matched controls (hazard ratio 0.78). They had 30% fewer retinal vein occlusions (HR 0.70). They developed neovascular glaucoma, the late-stage complication where new blood vessels grow into the iris and choke off drainage, 35% less often (HR 0.65). Retinal artery occlusion and non-arteritic ischemic optic neuropathy (NAION) showed no significant association either direction.
That last finding is the one that matters most for the regulatory conversation. NAION is the eye condition that became a public concern around semaglutide in 2024, after a Mass Eye and Ear cohort study reported a substantial increase in NAION risk among type 2 diabetes patients on the drug. The new TriNetX analysis, in a much larger retinopathy-positive cohort, finds NAION HR 0.88 with a confidence interval (0.54 to 1.44) that crosses one. That is not a refutation of the 2024 signal, but it is a much wider denominator and a cleanly null result in the population most likely to develop ocular complications anyway.
The systemic benefits are large. Patients on GLP-1 drugs had 35% fewer myocardial infarctions, 22% fewer ischemic strokes, 22% fewer lower-extremity amputations, 25% fewer coronary revascularization procedures, and 22% fewer heart failure exacerbations. The kidney numbers are striking even by the standards of GLP-1 cardiovascular outcomes literature: 32% lower acute kidney injury risk (HR 0.68) and 60% lower need for renal replacement therapy (HR 0.40).
The study design has limits worth naming. Retrospective cohort studies, even with propensity score matching, cannot match patients on factors that were not measured. These include adherence to other medications, lifestyle changes that often accompany a new GLP-1 prescription, socioeconomic status, and access to specialty ophthalmology care. The 30,613-patient matched cohort is large, but the comparator group is patients with the same diagnosis who did not get the drug, and the reasons they did not get the drug are not random. Propensity score matching adjusts for measured confounders. It cannot adjust for the unmeasured ones.
Diabetic retinopathy is the population the study targets on purpose. Patients with pre-existing DR are usually excluded from prospective GLP-1 cardiovascular outcome trials, because their retinopathy risk profile makes them a complication concern. SUSTAIN-6, the original semaglutide cardiovascular outcomes trial, reported a numerical increase in DR worsening on the drug versus placebo, which has shaped the prescriber conversation since 2016. Subsequent analyses attributed the SUSTAIN-6 signal largely to rapid HbA1c lowering rather than to a direct drug effect, but the regulatory caution remained. The TriNetX study targets exactly the patient group prescribers are most cautious about, and finds the opposite of what the 2016 signal would predict.
Future-proofing the finding requires a prospective trial in a DR-positive population. The authors call for one in the conclusion. Until that trial reads out, the regulatory conversation has a new data point. The patients most likely to be flagged for GLP-1 retinopathy concern were, in this cohort, the patients who benefited most from the drug class.