Incident ocular outcomes with tirzepatide versus GLP-1 receptor agonists and conventional therapies in diabetes and obesity: a retrospective cohort study from a global federated health research network

Background To evaluate incident ophthalmic complications in adults initiating tirzepatide for type 2 diabetes mellitus (T2DM) or overweight/obesity without diabetes, compared with glucagon-like peptide-1 receptor agonists (GLP-1RAs) and conventional therapies. Methods We performed a retrospective, propensity score–matched cohort study in the TriNetX global federated health research network (1 Jan 2022–31 Oct 2024). Adults (≥ 18 years) with T2DM or overweight/obesity who initiated tirzepatide were matched 1:1 to new users of GLP-1 receptor agonists and to non-GLP-1 antidiabetic or anti-obesity therapies, excluding patients with pre-existing ocular disease and treatment crossover. The primary outcome was the composite of incident ocular complications (visual impairment, diabetic retinopathy, or other retinal disorders); secondary outcomes were retinopathy and visual impairment. Results In T2DM, 1-year composite incidence was 3.5% of both tirzepatide and GLP-1RAs (HR 1.01, 95% CI 0.88–1.15). Versus non-GLP-1 antidiabetic therapies, tirzepatide reduced risk of the composite endpoints (3.5% vs. 5.6%; HR 0.61, 0.54–0.68), visual impairment (HR 0.54, 0.34–0.86), and retinopathy (HR 0.43, 0.32–0.56). In non-diabetic overweight/obesity, outcomes were similar to GLP-1RAs (composite 1.9% vs. 2.1%; HR 0.92, 0.81–1.03) but lower than non-GLP-1 anti-obesity drugs for the composite endpoint (1.9% vs. 2.3%; HR 0.82, 0.74–0.92) and visual impairment (HR 0.57, 0.38–0.85); retinopathy was rare with no significant between-group differences. Conclusions Tirzepatide was not associated with increased ophthalmic risk over 1 year in T2DM or non-diabetic overweight/obesity and was linked to lower risk versus conventional non-GLP-1 therapies, supporting short-term ocular safety. Longer follow-up is warranted. Trial registration Not applicable. This retrospective analysis of de-identified data did not involve a prospective health-care intervention requiring registration.