TOPIC: This systematic review and meta-analysis evaluates the rate and risk of non-arteritic anterior ischemic optic neuropathy (NAION) in adults receiving semaglutide for diabetes or weight loss.
CLINICAL RELEVANCE: Semaglutide is a transformative treatment for diabetes and obesity. NAION is a serious cause of sudden, often permanent, vision loss. A comprehensive synthesis of available and conflicting evidence of a potential link is needed to inform clinical practice.
METHODS: Our comprehensive search included MEDLINE, Embase, CENTRAL, clinical trial registries, and direct contact with trial sponsors to identify unpublished data from inception to March 2025. We included randomised-controlled trials (RCTs) and observational studies evaluating semaglutide versus placebo or standard therapy in adults. Incident NAION was the primary outcome. Pooled incidence rates, hazard ratios (HRs) and risk ratios (RRs) with 95% confidence intervals (CIs) were calculated using random-effects models, uniquely stratified by both clinical indication (diabetes/weight loss) and study design.
RESULTS: Our search identified eight RCTs (31,174 patients, including three RCTs not previously synthesised in meta-analyses) and eight observational studies (1,611,278 patients) were included. Observational studies for diabetes showed a pooled NAION incidence of 26.7 per 100,000 person-years in those on semaglutide versus 18.9 per 100,000 person-years in those not on semaglutide (HR1.85, 95% CI 1.20-2.85). RCTs for diabetes demonstrated five ischaemic optic neuropathy events in those on semaglutide versus one in those not, with no significant evidence of increased risk (RR 1.76, 95% CI 0.43-7.25). Observational studies of semaglutide for weight loss showed no evidence of increased NAION hazards (HR 1.57, 95% CI 0.69-3.59), and RCTs (four NAION events vs one) found no evidence of increased risk (RR 2.18, 95% CI 0.33-14.34).
CONCLUSIONS: The absolute rate of NAION in semaglutide users is low. Our novel, stratified analysis provides a more nuanced interpretation than previous pooled reviews, isolating a potential increased risk signal to observational data in the diabetes population only, a finding not corroborated by RCTs. The comprehensiveness of our search and rigorous sensitivity analyses to account for patient overlap in databases strengthen the validity of these conclusions. This synthesis highlights the need for continued pharmacovigilance while contextualising the low absolute risk.
