No study has compared a GLP-1 user against an untreated person for glaucoma. The one large comparison that exists ranks two of these drugs against each other. It covered 51,540 matched adults aged 60 and over. Incident open-angle glaucoma came in at a hazard ratio of 0.65 (95% CI 0.44 to 0.96) on tirzepatide against other GLP-1 drugs [1]. Eye safety has followed this class since the first reports of optic nerve damage, examined here in the optic nerve evidence by indication.
The cohort covered adults aged 60 and over with type 2 diabetes, newly prescribed tirzepatide or another GLP-1 drug between 2022 and 2024. Matching was one to one on age, sex, race, smoking history, HbA1c, BMI, prior glucose-lowering medication, corticosteroid use and comorbidities. Anyone with existing glaucoma or prior glaucoma medication was excluded, as was anyone using both drug types during follow-up. Follow-up ran two years, and the result survived several sensitivity analyses including an intention-to-treat design.
The separate optic nerve literature runs the other way and should not be folded in. Pooling eight cohorts, semaglutide carried a relative effect of 1.93 for non-arteritic anterior ischemic optic neuropathy (95% CI 1.22 to 3.08) [2]. That is a different structure, a different nerve and a different comparison. Glaucoma is not NAION.
The interval is worth reading rather than rounding. An upper bound of 0.96 clears 1 by four hundredths. That makes the result real by the usual convention and fragile under any reasonable amount of unmeasured confounding. Two years is also a short window for a disease that develops over decades. And a diagnosis of open-angle glaucoma depends on somebody having an eye examination at all, which is the same detection problem that shapes the retinopathy evidence.
What would make this more than a curiosity is a mechanism, and the study supplies none. Intraocular pressure, blood sugar variability and weight change are all plausible routes. None was measured. Until an untreated comparison exists, the practical use of this finding is narrow. It is one more small difference between two drugs usually discussed as interchangeable, alongside the carpal tunnel comparison and the network comparison.