Abstract

Source

Study at a Glance

Study Snapshot

Study Facts Table

What Researchers Actually Did

Grossman, Biousse, and Newman authored an invited editorial in Neurology contextualizing the companion systematic review and meta-analysis by Dhivagaran et al., which screened 144 publications and included 5 observational studies encompassing approximately 1.5 million individuals to assess the association between semaglutide and NAION. The editorialists synthesized this alongside the broader post-2024 literature, which by the time of writing included more than 50 publications, at least 20 studies, and regulatory and professional society statements from the EMA, NANOS, and the AAO. Study designs across this body of evidence included retrospective cohort series, ICD-10 code-based database searches, analyses of FDA Adverse Event Reporting System (FAERS) data, retrospective adverse event reviews from large GLP-1RA clinical trials, and multiple meta-analyses and systematic reviews.

The editorial does not report new primary data. Its purpose is to offer a calibrated interpretation of a contested and rapidly evolving evidence base, to highlight the methodological limitations of existing studies, and to provide actionable guidance for prescribers navigating shared decision-making with patients. The authors specifically flag that competing analyses have drawn opposing conclusions even when using identical data sources, most notably the TriNetX database, and that the absence of prospective randomized trials with NAION as a prespecified endpoint leaves the field reliant on observational evidence of variable quality.

Key Findings: Primary Outcomes

• The companion meta-analysis (Dhivagaran et al.) concludes with low-to-moderate certainty that semaglutide significantly increases NAION risk relative to non-GLP-1RA comparators by a factor of 2.52.
• The editorialists synthesize the broader literature and estimate that GLP-1RA exposure likely increases NAION risk approximately twofold, consistent with the companion meta-analysis.
• Hathaway et al. (2024), the index study that initiated this conversation, reported NAION risk greater than 4-fold higher in patients with diabetes on semaglutide and greater than 7-fold higher in overweight and obese individuals on semaglutide, with the authors speculating that greater dose exposure in the latter group may explain the difference.
• NAION itself occurs in up to 10 patients per 100,000 adults over age 50, making the absolute risk from any relative increase remain numerically small.

Key Findings: Secondary Outcomes and Subgroup Analyses

• The Dhivagaran meta-analysis excluded articles with overlapping populations and noted that many key studies used the TriNetX database. Only one TriNetX study was included in the primary analysis; however, sensitivity analyses incorporating all TriNetX-based studies reached similar conclusions.
• The editorial highlights patients with prior monocular NAION as a subgroup requiring especially careful consideration, given the risk of bilateral visual disability. The evidence base does not yet define the quantitative risk of second-eye involvement with continued GLP-1RA use.
• The EMA (June 2025) reviewed all available data and concluded that semaglutide should be stopped if NAION has occurred in one eye.
• The NANOS/AAO joint statement (July 2025) stated the proposed increased NAION risk is “much lower than the overall health risks” for individuals who need semaglutide, and explicitly did not recommend systematic GLP-1RA cessation in individuals with monocular NAION.
• Up to 40% of NAION patients lack classic vascular risk factors, a baseline reality that complicates attributing incident NAION to any single drug exposure in observational datasets.

Adverse Events and Safety Profile

NAION is the adverse event under active investigation in this literature. NAION is characterized by unilateral optic neuropathy with optic disc edema and a crowded optic disc anatomy (small cup-to-disc ratio). It is presumed to result from small vessel ischemia of the posterior ciliary arteries feeding the anterior optic nerve, producing a focal compartment syndrome with progressive ischemic injury. The condition is visually disabling and has no effective treatment. The editorial does not report a separate adverse event profile for GLP-1RAs beyond the NAION signal under discussion. Specific drugs previously associated with elevated NAION risk include amiodarone and phosphodiesterase-5 inhibitors; the editorial discusses whether GLP-1RAs warrant inclusion on this list.

Statistical Approach and Rigor

This editorial does not present original statistical analyses. The companion Dhivagaran meta-analysis pooled 5 observational studies encompassing approximately 1.5 million individuals and assigned a low-to-moderate certainty rating to its primary finding. The editorial explicitly acknowledges that the studies available are entirely observational, that prospective randomized trials with NAION as a prespecified endpoint do not exist, and that competing analyses using identical data sources have reached opposing conclusions. The methodological heterogeneity across included studies (ICD-10 code-based diagnoses, manually confirmed cohort series, FAERS pharmacovigilance data, and retrospective trial adverse event reviews) substantially limits the interpretability of any pooled estimate. The rapid pace of publication in this area also creates a practical problem: the Dhivagaran meta-analysis was submitted in August 2025, and by the editorial’s writing date of March 2026, additional relevant studies had already been published, raising concerns about completeness.

Clinical Takeaway

Why This Matters Clinically

With over 40 million Americans already prescribed GLP-1RAs and 136 million potentially eligible, even a rare adverse event at the population level translates to a meaningful number of affected individuals. NAION causes permanent, often severe, visual field loss in the affected eye, and no treatment reverses it. The neuro-ophthalmic community has spent the past 20 months trying to characterize a signal that emerged from a single clinical observation, and the literature remains genuinely unsettled. The divergence between the EMA’s recommendation to stop semaglutide after a NAION event and the NANOS/AAO position that cessation should not be automatic reflects the absence of high-quality prospective data. Clinicians cannot wait for perfect evidence — they must counsel patients today. Understanding the approximate magnitude of the signal, its absolute rarity, and the stakes of stopping therapy equips prescribers to have that conversation rigorously.

CED Clinical Relevance

At CED Clinic, patients using cannabinoids for metabolic, cardiovascular, or inflammatory indications frequently carry comorbidities — type 2 diabetes, obesity, hypertension, obstructive sleep apnea — that are themselves independent NAION risk factors and are also the primary indications driving GLP-1RA prescriptions. When a patient presents managing both a GLP-1RA and a cannabis-based regimen, the clinical picture requires careful vascular risk stratification. Patients with a small cup-to-disc ratio on prior ophthalmic examination, a history of one NAION event, or multiple concurrent NAION risk factors should prompt direct coordination with their ophthalmologist or neuro-ophthalmologist before continuing or initiating GLP-1RA therapy. This paper provides the evidentiary basis for that conversation: a probable twofold relative risk increase against a rare baseline event rate, with no high-certainty prospective data to refine that estimate further.

Clinical Insight

Fits What We Already Know

The association between specific drugs and NAION is not new. Amiodarone and phosphodiesterase-5 inhibitors have previously been linked to elevated NAION risk, establishing a precedent for drug-induced vascular optic neuropathy. Classic systemic NAION risk factors — diabetes mellitus, hypertension, obstructive sleep apnea, and smoking — have long been recognized, and the association between small vessel ischemia and optic nerve injury is the foundational pathophysiologic mechanism (Biousse and Newman, N Engl J Med, 2015). The index publication by Hathaway et al. (JAMA Ophthalmol, 2024) introduced semaglutide into this risk landscape, and the cardiovascular benefits of semaglutide in type 2 diabetes had been established in the SUSTAIN-6 trial (Marso et al., N Engl J Med, 2016), providing the therapeutic context against which the NAION signal must be weighed. All references in this section are drawn exclusively from the editorial’s own citation list.

What This Study Teaches Us

This editorial teaches that a biologically plausible, statistically detected adverse signal can persist in the literature as genuinely contested when the underlying evidence consists entirely of observational studies with heterogeneous methods. The approximately twofold NAION risk estimate is the current best synthesis across more than 50 publications, but the certainty is low-to-moderate at best. The field also learns something about the limits of rapid pharmacovigilance: when more than 20 studies, some using the same data sources, reach opposing conclusions, the quality of the primary data and the methodological choices applied to it matter more than the volume of publications. The existence of divergent regulatory and professional society guidance from the EMA versus NANOS/AAO demonstrates that even expert bodies reviewing the same evidence can reach different actionable conclusions — underscoring that this is a genuinely uncertain clinical question, not a settled one dressed up as a controversy.

What It Does Not Show

• This editorial does not establish causality between GLP-1RA use and NAION. The editorialists explicitly state that causality is “far from established.”
• It does not quantify the risk of second-eye NAION in patients who continue GLP-1RAs after a unilateral event.
• It does not differentiate risk across individual GLP-1RA agents (liraglutide, tirzepatide, dulaglutide) beyond semaglutide, which is the primary focus.
• It does not define dose-response relationships with sufficient precision, though the Hathaway data suggest higher exposure in the overweight/obese cohort may correspond to higher risk.
• It does not provide prospective, randomized evidence; no such trial with NAION as a prespecified endpoint exists in the literature at the time of publication.
• It does not resolve the methodological disagreements between studies that used identical data sources and reached opposing conclusions.
• It does not address whether GLP-1RAs affect the severity or visual recovery trajectory of NAION once it has occurred.

Fits the Broader Conversation

This editorial arrives at a moment when GLP-1RA prescribing is accelerating globally while post-marketing pharmacovigilance is struggling to keep pace. The NAION signal, if confirmed causally, would represent the first addition of a blockbuster weight-loss and diabetes medication to the list of drugs associated with optic nerve injury. More broadly, this paper moves the field by synthesizing a fractured and methodologically inconsistent literature into a working clinical estimate — approximately twofold increased relative risk — that prescribers can actually use, while being transparent about the evidence ceiling. The divergence in regulatory response between the EMA and North American professional societies also advances a conversation about how much uncertainty is acceptable before clinical guidance changes. For neuro-ophthalmology specifically, this episode demonstrates the urgency of designing prospective studies with NAION as a prespecified endpoint in large GLP-1RA trials, a gap the field has not yet closed.

What This Means for Families

What This Means for Clinicians

What the Cautious Reader Should Hold Loosely