Semaglutide and optic neuritis in the FDA adverse event reporting system: a comparator-based, time-calibrated pharmacovigilance analysis

Background Glucagon-like peptide-1 receptor agonists and dual glucose-dependent insulinotropic polypeptide/glucagon-like peptide-1 receptor agonists are increasingly used for glycemic control, obesity management, cardiovascular risk reduction, and kidney protection. Ocular safety research has focused mainly on diabetic retinopathy and nonarteritic anterior ischemic optic neuropathy. Whether inflammatory neuro-ophthalmic events, particularly optic neuritis, show a reproducible postmarketing reporting association remains uncertain. Methods We performed a case/non-case pharmacovigilance study using FAERS reports from 2004Q1 to 2026Q1. Reports were deduplicated by CASEID/CASEVERSION. Primary-suspect GLP-1/GIP exposures were linked to MedDRA-defined inflammatory neuro-ophthalmic outcomes. The primary comparison was semaglutide versus SGLT2 inhibitors/metformin, with secondary DPP-4 inhibitor, strict non-GLP-1 weight-loss, separate SGLT2 inhibitor, separate metformin, and indication-restricted analyses. We used ROR, PRR, transparently labeled approximate shrinkage proxies, BH-FDR adjustment, exact report-year stratification, exact-year frequency matching, reporter-channel sensitivity, and Firth-type adjusted logistic regression. Results After latest-version deduplication, 59,790 semaglutide and 178,670 pooled SGLT2 inhibitor/metformin reports were included. The crude semaglutide-optic neuritis ROR was 11.67 (60 vs 15 events; 95% CI 6.68-20.40; BH q=1.99e-27). Given the different report-year distributions (semaglutide: 62.7% in 2024-2026; pooled comparator: 17.5%), the principal calendar-time-calibrated estimate was the exact report-year Mantel-Haenszel common OR of 5.06 (95% CI 2.64-9.72), supported by exact-year 1:1 frequency matching (OR 4.64, 95% CI 2.42-8.90). Separate comparator analyses remained positive for SGLT2 inhibitors (ROR 15.68, 95% CI 6.55-37.53) and metformin (ROR 8.53, 95% CI 4.43-16.42), whereas the obesity/weight-management restriction was sparse and not supported. Estimates were positive in health-professional and consumer/legal reports. The Firth-type adjusted OR was 16.76 (95% CI 8.51-33.00). Among 12 reports with valid TTO, the median was 62.5 days; TTO was analyzed descriptively without hazard modeling and did not support an early-failure or temporal causal inference. Conclusion FAERS data support a semaglutide-optic neuritis reporting association that persists after calendar-time calibration and comparator expansion, but the magnitude is sensitive to reporting era and unmeasured clinical heterogeneity. The finding is hypothesis-generating, cannot establish incidence, relative risk, or causality, and should not be interpreted as complete separation from ischemic optic neuropathy/NAION. The sparse TTO subset was descriptive only and did not support an early-failure or temporal causal interpretation.