Semaglutide and non-arteritic anterior ischaemic optic neuropathy: a 121-fold reporting signal and a bounded genetic null
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background
A cohort study on 3 July 2024 linked semaglutide to non-arteritic anterior ischaemic optic neuropathy (NAION), and in June 2025 the European Medicines Agency (EMA) accepted NAION as a very rare side effect, about one additional case per 10,000 person-years.
Objective
To measure how much of the signal the reporting system generates itself, test whether three reporting systems are independent, and let a genetic instrument adjudicate rather than corroborate.
Methods
FAERS (12,305,041 reports, 2018Q1 to 2026Q2), JADER (1,054,239 reports), and VigiAccess counts were analysed with reporting odds ratios, covariate-adjusted regression, and pre-specified probes. Six cis-eQTL variants for GLP1R expression instrumented anterior ischaemic optic neuropathy (AION, 244 cases) and five AION-spectrum traits, with colocalisation.
Results
Semaglutide carried 879 NAION reports and a reporting odds ratio of 120.8 (95% CI 110.6 to 132.0), about sixty times the relative risk the EMA accepted, and the odds that a report named NAION rose 48.6-fold afterwards. VigiBase held 871 such reports where FAERS held 879, and the two systems give opposite accounts of the GLP-1 class. Of fifteen observations separating toxicity from reporting dynamics, one favoured reporting dynamics, two favoured toxicity weakly, two bounded an expression-mediated mechanism, and ten decided nothing. GLP1R expression was unassociated with AION (0.56, 0.21 to 1.49), a null powered only above fourfold, while glaucoma was bounded (1.05, 0.87 to 1.25) and no relevant ocular trait colocalised.
Conclusions
Semaglutide’s NAION signal is amplified by the reporting system, and the databases that appear to corroborate it hold the same records, so the causal question stays open in both directions.
