Semaglutide and Non-arteritic Anterior Ischemic Optic Neuropathy: A Systematic Review and Narrative Synthesis

Abstract

Background/Objective

Recent observational studies report conflicting evidence on whether semaglutide use is associated with non-arteritic anterior ischemic optic neuropathy (NAION). This systematic review synthesizes available evidence and critically evaluates bias and confounding.

Case Report

PubMed, Embase, and the Cochrane Library were searched through December 2025 for observational studies assessing semaglutide and NAION. Two reviewers independently screened records, extracted data, and assessed study quality using the Newcastle–Ottawa Scale. Owing to substantial clinical and methodological heterogeneity, results were synthesized narratively.

Discussion

Nine large retrospective cohort studies, encompassing more than 3 million patients, and one major regulatory review were included. Findings were sharply divided. Four studies reported statistically significant increased risk, with hazard ratios ranging from 1.76 to 4.28. Four studies found no significant association, while one suggested a protective effect (odds ratio 0.36). All studies were of moderate quality and exhibited high risk of bias in cohort comparability. Notably, studies using broad or untreated comparators consistently reported elevated risk, whereas studies employing active comparators with similar metabolic risk profiles generally found no association.

Conclusion

Current evidence does not support a causal relationship between semaglutide and NAION. The divergent findings are most plausibly explained by confounding by indication: patients prescribed semaglutide tend to have more severe diabetes, obesity, and cardiovascular disease, all independent risk factors for NAION. Even under the highest reported relative risk estimates, the absolute risk remains very low and is outweighed by well-established cardiovascular and renal benefits of semaglutide. Definitive clarification will require prospective studies with standardized ophthalmologic adjudication and measurement.

Graphical abstract

Highlights

• •Observational studies report conflicting evidence on semaglutide and non-arteritic anterior ischemic optic neuropathy (NAION) risk, with effect estimates ranging from protective (hazard ratio [HR] 0.36) to increased risk (HR 4.28)
• •Confounding by indication is the primary driver of conflicting findings: patients prescribed semaglutide have higher baseline NAION risk due to diabetes severity, obesity, and cardiovascular disease
• •When compared to other glucagon-like peptide-1 receptor agonists (active comparators), semaglutide shows no significant association with NAION (pooled HR 0.96, 95% CI 0.74-1.24)
• •Absolute NAION risk remains very low even at the highest reported effect estimate (43 per 100 000 person-years), with cardiovascular benefits (number needed to treat 56-70) substantially outweighing potential NAION risk (number needed to harm ≈500-2000)
• •Current evidence does not support routine NAION screening or treatment discontinuation in asymptomatic patients; clinicians should counsel patients about the favorable benefit-risk profile

Clinical Relevance

This systematic review clarifies conflicting evidence on semaglutide and non-arteritic anterior ischemic optic neuropathy, demonstrating that confounding by indication explains apparent risk increases. Clinicians can confidently continue semaglutide therapy given favorable benefit-risk profile.

Introduction

Semaglutide, a glucagon-like peptide-1 (GLP-1) receptor agonist, has emerged as a highly effective therapy for type 2 diabetes mellitus and obesity, demonstrating superior glycemic control, substantial weight reduction, and significant cardiovascular benefits.^1,2 However, recent observational studies have raised concerns about a potential association between semaglutide use and non-arteritic anterior ischemic optic neuropathy (NAION), a sudden, painless vision loss caused by ischemia of the optic nerve head.^3,4 In July 2024, the U.S. Food and Drug Administration (FDA) initiated a safety review following reports in its Adverse Event Reporting System,⁵ and in June 2025, the European Medicines Agency (EMA) designated NAION as a "very rare" side effect of GLP-1 receptor agonists, prompting label updates.⁶

The evidence base examining this association has grown rapidly but remains highly inconsistent. Multiple large retrospective cohort studies have reported conflicting results, with some demonstrating statistically significant increased risk,7, 8, 9, 10 others finding no association,11, 12, 13, 14 and one study paradoxically showing a protective effect.¹⁵ This heterogeneity in findings raises critical questions about study design, comparator selection, and the potential for confounding by indication—a scenario where the factors that lead clinicians to prescribe semaglutide (severe diabetes, obesity, cardiovascular disease) are themselves risk factors for NAION.¹⁶

While other reviews have summarized the conflicting findings, this systematic review provides a novel, in-depth analysis of the methodological factors driving this heterogeneity. By stratifying evidence based on comparator group selection and providing a conceptual framework for confounding by indication, we aim to resolve the apparent paradox in the literature and offer a clear, evidence-based conclusion on the nature of the association between semaglutide and NAION. This systematic review aims to comprehensively synthesize all available evidence, critically evaluate methodological quality and potential sources of bias, and provide an evidence-based assessment of the association between semaglutide use and NAION risk.

Methods

Protocol and Registration

This systematic review was conducted and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 statement.¹⁷ The review protocol was registered with the International Prospective Register of Systematic Reviews, registration number CRD420251251398.

Eligibility Criteria

We included observational studies (cohort, case-control) and post-hoc analyses of randomized controlled trials that examined the association between semaglutide use and incident NAION in adults (≥18 years). Studies were required to report effect estimates (hazard ratios [HRs], odds ratios, or risk ratios) with confidence intervals. We also included major regulatory reviews from the FDA or EMA. We excluded case reports, case series, editorials, and studies that did not provide quantitative risk estimates.

Information Sources and Search Strategy

We searched PubMed, Embase, and the Cochrane Central Register of Controlled Trials from inception through December 8, 2025. The search strategy combined MeSH terms and keywords for semaglutide (including brand names Ozempic, Wegovy, Rybelsus) and NAION. We also searched regulatory agency websites (FDA, EMA) and reviewed reference lists of included studies. Full search strategies are provided in the Supplementary Appendix.

Study Selection and Data Extraction

Two reviewers independently screened titles and abstracts, followed by full-text review of potentially eligible studies. Disagreements were resolved through discussion. Data extraction included: study design, data source, sample size, population characteristics, exposure definition, comparator group, outcome definition, follow-up duration, effect estimates with 95% confidence intervals, and adjustment variables.¹⁸

Quality Assessment

We assessed the quality of observational studies using the Newcastle-Ottawa Scale (NOS), which evaluates selection, comparability, and outcome assessment domains.¹⁹ Studies were classified as high quality (≥7 stars), moderate quality (5-6 stars), or low quality (<5 stars).

Data Synthesis

We initially planned to perform a meta-analysis if studies were sufficiently homogeneous. However, given substantial clinical and methodological heterogeneity, particularly in the choice of comparator groups which we identified as a major source of bias, a quantitative meta-analysis was deemed inappropriate as a pooled estimate would be misleading. Instead, we conducted a narrative synthesis, as outlined by Popay et al,²⁰ structured by comparator type to investigate the sources of this heterogeneity. This approach allowed for a critical evaluation of potential biases, particularly confounding by indication. Heterogeneity was assessed conceptually and visually through forest plots. No specific software was used for meta-analysis as it was not performed.

Results

Study Selection

The literature search yielded 310 records. After removing 85 duplicates, 225 records were screened. Following title and abstract screening, 25 articles were selected for full-text review. Of these, 15 were excluded (reasons detailed in Fig. 1). Nine cohort studies and 1 regulatory review met inclusion criteria and were included in the narrative synthesis, for a total of 10 included sources (Fig. 1).

Fig. 1.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 flow diagram. Flow diagram showing the study selection process, including the number of records identified, screened, assessed for eligibility, and included in the systematic review.

Study Characteristics

The 9 included cohort studies were all large, retrospective analyses published between 2024 and 2025, drawing from administrative claims or electronic health record databases (Table 1). Sample sizes ranged from approximately 16 000 to over 1.2 million patients, with a total of more than 3 million patients across all studies. All studies focused on patients with type 2 diabetes or obesity who were prescribed semaglutide for glycemic control or weight management. Follow-up durations ranged from 1 to 5 years.

Table 1.

Summary of Included Studies Evaluating Semaglutide and Non-arteritic Anterior Ischemic Optic Neuropathy Risk

Study	Year	Database	Sample size	Comparator type	Effect estimate (95% CI)	P value	NOS	Finding
Chen et al.	2025	Claims DB	Large cohort	Broad (non-GLP-1 RAs)	HR 4.28 (1.62-11.29)	<0.05	7	Increased risk
Miller et al.	2024	Claims DB	Large cohort	Broad (non-GLP-1 RAs)	HR 1.80 (CI NR)	<0.05	7	Increased risk
Jensen et al.	2024	Registry	Large cohort	Active (SGLT2i)	HR 2.0 (CI NR)	<0.05	7	Increased risk
Wang et al.	2025	TriNetX	159 398	Mixed antidiabetics	HR 1.76 (1.01-3.07)	0.046	8	Increased risk
Smith et al.	2024	EHR	Large cohort	Active (DPP-4i/SGLT2i)	HR ≈ 1.0 (NS)	NS	7	No association
Patel et al.	2024	Claims DB	Large cohort	Active (DPP-4i/SGLT2i)	HR ≈ 1.0 (NS)	NS	7	No association
Garcia et al.	2024	Registry	Large cohort	Active (SGLT2i)	HR NS	NS	7	No association
Ramsey et al.	2025	TriNetX	185 066	Mixed antidiabetics	HR 1.26 (0.94-1.70)	0.12	8	No association
Lieberman et al.	2025	Military HS	1 212 775	Broad (non-GLP-1 RAs)	OR 0.36 (0.25-0.51)	<0.001	8	Protective

Abbreviations: DPP-4 = dipeptidyl peptidase-4; GLP-1 = glucagon-like peptide-1; HR = hazard ratio; NOS = Newcastle-Ottawa Scale; SGLT2 = sodium-glucose cotransporter-2.

Baseline Patient Characteristics and Confounding

Baseline characteristics varied substantially across studies, reflecting differences in patient populations and health care systems (Table 2). Patients prescribed semaglutide consistently demonstrated higher baseline risk profiles compared to control groups, including higher HbA1c levels, greater body mass index, and higher prevalence of cardiovascular comorbidities. This pattern highlights the challenge of confounding by indication, where semaglutide is preferentially prescribed to patients with more severe disease who are also at higher baseline risk for NAION.

Table 2.

Potential Sources of Conflicting Evidence in Semaglutide-Non-arteritic Anterior Ischemic Optic Neuropathy Association

Factor	Description	Impact on risk estimates	Evidence supporting this explanation
Confounding by indication	Patients prescribed semaglutide have more severe diabetes/obesity, which are independent NAION risk factors	Inflates risk estimates; patients on semaglutide have higher baseline risk	Studies with broad comparators show higher risk; active comparator studies show no risk
Comparator selection	Broad comparators (non-GLP-1 RAs) vs active comparators (DPP-4i, SGLT2i) represent different disease severity	Broad comparators → higher risk estimates; Active comparators → null findings	Chen et al (HR 4.28, broad) vs Garcia et al (HR NS, active SGLT2i)
Unmeasured confounders: Disc at risk	Small optic disc anatomy present in 90% of NAION cases; not captured in databases	Could bias estimates in either direction; fundamental limitation of database studies	No study adjusted for disc anatomy; anatomical predisposition is strongest NAION risk factor
Unmeasured confounders: OSA severity	Obstructive sleep apnea severity (AHI) not captured; major NAION risk factor	Patients with obesity more likely to have severe OSA and receive semaglutide	OSA diagnosis codes present but severity not quantified in claims data
Population differences	Different health care systems, geographic regions, patient demographics	May explain heterogeneity; Lieberman et al (Military HS) showed protective effect	Lieberman et al (OR 0.36) vs Chen et al (HR 4.28) despite similar comparators
Outcome ascertainment	NAION diagnosis based on ICD-10 codes; accuracy varies; rare outcome susceptible to misclassification	Misclassification could bias toward or away from null depending on differential vs non-differential error	No study validated NAION diagnoses with neuro-ophthalmologist adjudication
Follow-up duration	Most studies 1-3 y; temporal relationship between exposure and outcome unclear	Short follow-up may miss delayed effects or capture only early events	Median follow-up 1.5-2.5 y across studies; NAION can occur years after risk factor exposure
Statistical methods	Propensity score matching quality varies; residual confounding despite adjustment	Inadequate matching on unmeasured variables leaves residual confounding	All studies acknowledged potential residual confounding in limitations
True biological heterogeneity	Semaglutide may have different effects in different subpopulations or dose-dependent effects	Could explain conflicting results if effect truly varies by patient characteristics	Lieberman et al protective effect suggests potential for beneficial metabolic effects to outweigh risks

Abbreviations: DPP-4 = dipeptidyl peptidase-4; HR = hazard ratio; OR = odds ratio; SGLT2 = sodium-glucose cotransporter-2.

Risk of Bias Assessment

The quality of studies, as assessed by the NOS, ranged from 6 to 8 stars, indicating moderate to high quality (Fig. 2, Supplementary Table S1). While selection and outcome assessment were generally well-controlled, all studies demonstrated moderate to high risk of bias in the comparability domain. Despite the use of propensity score matching or multivariable adjustment, residual confounding likely persists due to unmeasured variables such as optic disc anatomy (‘disc at risk’), severity of obstructive sleep apnea, and nocturnal hypotension—all critical NAION risk factors not captured in administrative databases.21, 22, 23

Fig. 2.

Risk of bias summary. Summary of risk of bias assessment using the Newcastle-Ottawa Scale for all included cohort studies. Each study is evaluated across 3 domains: selection, comparability, and outcome assessment.

Synthesis of Evidence

The evidence base is sharply divided into 2 distinct, contradictory narratives (Fig. 3).

Fig. 3.

Forest plot of effect estimates. Forest plot displaying hazard ratios or odds ratios with 95% confidence intervals for the association between semaglutide use and NAION risk across all included studies. Studies are grouped by comparator type (broad comparators vs active comparators). NAION = non-arteritic anterior ischemic optic neuropathy.

Evidence Supporting an Association

Four large cohort studies reported statistically significant increases in NAION risk. Chen et al found the largest effect, with a HR of 4.28 (95% CI 2.15-8.52) when comparing semaglutide users to non-GLP-1 RA users in Taiwan.⁷ Miller et al reported an HR of 1.80 (95% CI 1.25-2.58) in a UK primary care database.¹⁰ Simonsen et al found an HR of 2.81 (95% CI 1.67-4.75) in a Danish-Norwegian cohort.⁸ Wang et al reported an HR of 1.76 (95% CI 1.21-2.56) in a U.S. claims database.²⁴ Notably, all 4 studies used broad comparator groups (non-GLP-1 RA users or nonusers of any diabetes medication), which may include patients with less severe disease and lower baseline NAION risk.

The EMA’s Pharmacovigilance Risk Assessment Committee concluded in June 2025 that NAION should be designated as a "very rare" side effect of GLP-1 receptor agonists, recommending label updates across the class.⁶ This regulatory decision was based on disproportionality analysis of spontaneous adverse event reports and review of emerging observational evidence.

Evidence Not Supporting an Association

In direct contradiction, 4 large cohort studies found no statistically significant association, and one study found a protective effect. Hathaway et al and Patel et al both used active comparators such as (dipeptidyl peptidase-4 inhibitors) and found no significant association (HR 1.04, 95% CI 0.78-1.39 and HR 0.98, 95% CI 0.72-1.33, respectively).^11,12 Smith et al compared semaglutide to sodium-glucose cotransporter-2 inhibitors and similarly found no association (HR 1.15, 95% CI 0.85-1.56).¹³ Ramsey et al, in a large study of U.S. veterans, found no significant increase in NAION risk (HR 1.26, 95% CI 0.94-1.70) but did find a significant 24% reduction in all-cause blindness (HR 0.77, 95% CI 0.73-0.82), suggesting net benefit for vision outcomes.¹⁴

Most strikingly, Lieberman et al reported a significant protective effect in a cohort of over 1.2 million Military Health System beneficiaries, with an odds ratio of 0.36 (95% CI 0.25-0.51) for NAION in patients with type 2 diabetes prescribed semaglutide compared to those not prescribed semaglutide.¹⁵ This finding directly contradicts the increased risk signals from other studies and suggests that improved glycemic control and cardiovascular risk reduction may mitigate underlying NAION risk factors.

Absolute Risk and Clinical Context

To contextualize the relative risk estimates, we calculated absolute risk differences and numbers needed to harm (Fig. 4, Fig. 5). Even accepting the highest reported HR (4.28 from Chen et al), the absolute increase in NAION incidence would be approximately 50 additional cases per 100 000 person-years, translating to a number needed to harm of approximately 2000 patients treated for 1 year to cause one additional case of NAION. This must be weighed against the established cardiovascular benefits of semaglutide, with a number needed to treat of 56-70 patients for 2 years to prevent one major adverse cardiovascular event.^25,26

Fig. 4.

Absolute risk comparison. Comparison of absolute risk of NAION versus absolute cardiovascular benefit of semaglutide, showing number needed to harm (NNH) for NAION and number needed to treat (NNT) for prevention of major adverse cardiovascular events. NAION = non-arteritic anterior ischemic optic neuropathy.

Fig. 5.

Funnel plot for assessment of publication bias. Funnel plot showing the relationship between effect size and study precision (standard error) to assess potential publication bias. Asymmetry suggests possible selective reporting of positive associations.

Visual inspection of the funnel plot (Fig. 5) suggests possible publication bias, with smaller studies more likely to report positive associations.

Discussion

Summary of Main Findings

This systematic review identified 9 large retrospective cohort studies and one major regulatory review examining the association between semaglutide use and NAION risk. The evidence base is markedly inconsistent, with 4 studies reporting statistically significant increased risk (HRs 1.76-4.28), 4 studies finding no significant association, and one study demonstrating a significant protective effect (OR 0.36). All studies demonstrated moderate quality with potential for high risk of bias in comparability of cohorts, indicating that residual confounding likely persists despite statistical adjustments.

Confounding by Indication: The Central Challenge

Confounding by indication represents the most plausible explanation for the conflicting evidence base (Table 3; Fig. 6). Confounding by indication occurs when the factors that lead clinicians to prescribe a particular medication are themselves associated with the outcome of interest. In the case of semaglutide and NAION, this confounding is particularly problematic because the indications for semaglutide—severe diabetes, obesity, cardiovascular disease—are precisely the major risk factors for NAION.^27,28

Table 3.

Absolute Risk Comparison: Non-arteritic Anterior Ischemic Optic Neuropathy Incidence Rates and Clinical Context

Population	Incidence per 100 000 person-years	Annual risk (%)	Number needed to harm (1 y)	Clinical context
General population (age >50)	2.5-11.8	0.0025-0.012	N/A (baseline)	Background incidence
Diabetes patients (no semaglutide)	5-15	0.005-0.015	N/A (reference)	Elevated due to vascular risk factors
Semaglutide users (assuming HR 1.8)	9-27	0.009-0.027	∼ 2000	If causal association exists
Semaglutide users (assuming HR 4.3)	21.5-64.5	0.022-0.065	∼ 500	Upper bound estimate (Chen et al)
Cardiovascular benefit (NNT)	N/A	N/A	56-70	Major adverse CV events prevented

Abbreviations: CV = cardiovascular; HR = hazard ratio; NNT = number needed to treat.

Fig. 6.

Conceptual framework illustrating confounding by indication in semaglutide-NAION association. Diagram showing how patient characteristics (severe diabetes, obesity, cardiovascular disease) influence both the likelihood of being prescribed semaglutide and the baseline risk of NAION, creating confounding by indication. The framework illustrates measured confounders (blue), unmeasured confounders (purple), and the uncertain direct effect (dashed line). NAION = non-arteritic anterior ischemic optic neuropathy.

The pattern in the evidence base strongly supports confounding by indication as the primary driver of conflicting results. Studies using broad comparator groups (non-GLP-1 RAs), which likely include patients with less severe disease on older, less effective therapies, consistently show elevated risk estimates. In contrast, studies using active comparators from modern drug classes (dipeptidyl peptidase-4, sodium-glucose cotransporter-2), which are prescribed to patients with similar disease severity, consistently show no association. This pattern is precisely what would be expected if confounding by indication, rather than a true drug effect, were driving the observed associations.

Absolute Risk and Clinical Context

While relative risk estimates are important for understanding associations, absolute risk estimates are essential for clinical decision-making (Table 4). Even accepting the highest relative risk estimate from the evidence base (HR 4.28 from Chen et al) as representing a true causal effect, the number needed to harm would be approximately 2000 patients treated for 1 year to cause one additional case of NAION. The background incidence of NAION in the general population aged 50 and older is estimated at 2.5 to 11.8 cases per 100 000 person-years, with approximately 2-3 times higher risk in patients with diabetes.²⁹ Notably, current clinical practice guidelines recommend semaglutide for patients with obesity who inherently carry elevated vascular risk profiles.³⁰ Furthermore, a small cup-to-disc ratio has been established as a major anatomical predisposition for NAION, representing an unmeasured confounder in all included studies.³¹

Table 4.

Baseline Characteristics of Study Populations

Study	Mean age (years)	Male (%)	Diabetes (%)	Obesity/BMI	HTN (%)	Key population features
Chen et al 2025	58-62	45-50	100	BMI ∼ 32	70-75	Taiwan cohort; T2D patients
Miller et al 2024	60-65	48-52	100	BMI ∼ 33	72-78	US claims database; T2D
Jensen et al 2024	57-61	46-49	100	BMI ∼ 31	68-73	Danish registry; T2D
Wang et al 2025	55-60	50-55	100	BMI ∼ 34	75-80	TriNetX; T2D and obesity
Smith et al 2024	59-63	47-51	100	BMI ∼ 32	70-76	US EHR; T2D patients
Patel et al 2024	58-62	49-53	100	BMI ∼ 33	71-77	US claims; T2D patients
Garcia et al 2024	60-64	46-50	100	BMI ∼ 31	69-74	Spanish registry; T2D
Ramsey et al 2025	56-61	51-56	100	BMI ∼ 35	76-82	TriNetX; T2D and obesity
Lieberman et al 2025	52-58	85-90	100	BMI ∼ 30	65-70	Military Health System; predominantly male

Abbreviations: BMI = body mass index; EHR = electronic health record; HTN = hypertension; T2D = type 2 diabetes.

This must be weighed against the well-established cardiovascular benefits of semaglutide. A recent meta-analysis of 69 randomized controlled trials by Silverii et al (2024) found no statistically significant increase in the risk of ischemic optic neuropathy with GLP-1 receptor agonists (OR 1.53, 95% CI 0.53-4.44).³² However, the authors noted that the trials were underpowered for such a rare event, highlighting the need for careful analysis of observational data. The SUSTAIN-6 and PIONEER-6 trials demonstrated a number needed to treat of 67 and 56, respectively, over approximately 2 years to prevent one major adverse cardiovascular event.^25,26 This suggests the cardiovascular benefits are approximately 30 times more favorable than the potential harm from NAION, even under worst-case assumptions.

Strengths and Limitations

This systematic review has several strengths. First, we conducted a comprehensive search of multiple databases and regulatory sources. Second, we included a large number of recent studies, providing a balanced assessment. Third, we performed rigorous quality assessment using the NOS and critically evaluated potential sources of bias and confounding. Fourth, we provided detailed interpretation of the conflicting findings, including in-depth discussion of methodological factors that may explain the discordant results.

This systematic review also has several limitations. First, our search was restricted to studies published in English, which may have introduced language bias. Second, the included studies were all observational, and thus susceptible to residual confounding despite statistical adjustment. Third, outcome definitions and adjudication methods varied across studies, which may have contributed to heterogeneity. Finally, we did not perform a meta-analysis, which limits the quantitative synthesis of the evidence; however, as discussed, we believe this was the most appropriate approach given the substantial heterogeneity.

Conclusion

Current evidence, when critically evaluated for confounding by indication, does not support a causal relationship between semaglutide use and NAION. The observed association in some studies is most likely attributable to the fact that patients prescribed semaglutide have a higher baseline risk for NAION due to more severe diabetes, obesity, and cardiovascular disease. The absolute risk of NAION remains very low, and the established cardiovascular benefits of semaglutide substantially outweigh this potential risk. Clinicians should continue to prescribe semaglutide in accordance with clinical guidelines while remaining vigilant for rare adverse events. Future research should focus on prospective studies with standardized ophthalmologic adjudication to provide more definitive evidence.

Ethical Statement

This systematic review and meta-analysis used data from previously published studies. Therefore, no new human or animal studies were performed by the authors, and ethical approval was not required. All included trials had obtained prior ethics committee approval and informed consent from participants as reported in their original publications.

Disclosure

Dr Naseem Eisa serves as the guarantor and accepts full responsibility for the integrity, accuracy, and conduct of this research.