Introduction

Nonarteritic anterior ischemic optic neuropathy (NAION) is the second most common form of optic neuropathy and the second leading cause of blindness among adults.^1-3 It is characterized by sudden, painless vision loss, relatively afferent pupillary defect, decreased color perception, hyperemic optic disc swelling, and visual field defect.¹ It is likely caused by an impaired perfusion of the optic nerve. Since there is no treatment available and it often leads to irreversible visual loss, it is crucial to determine potential risk factors that may predict NAION.² Several systemic and anatomic factors have been identified, namely, sex (male), hypertension, hypercholesterolemia, diabetes, end‑organ damage, crowding of the optic nerve head, and medications, such as amiodarone and phosphodiesterase‑5 inhibitors.^3,4 Recently, there have been several reports^5-7 suggesting that the use of once‑weekly semaglutide, a glucagon‑like peptide‑1 receptor agonist (GLP‑1 RA), is associated with an increased risk of NAION. Other studies, however, did not confirm these findings.^8,9

Therefore, due to the controversial the evidence, we sought to further analyze the association of different GLP‑1 RAs and a broader set of outcomes that included blindness, NAION, papilledema, and optic neuritis in a large national cohort of patients observed for a prolonged time.

Patients and methods

This was a retrospective cohort study. Electronic health records were obtained from 300 medical clinics operated by Luxmed, the largest private health care provider in Poland, bewteen January 1, 2018, and June 30, 2024. To be enrolled in the study group (SG), the patients had to use a GLP‑1 RA for a minimum of 4 weeks and be over 18 years. This was an individual patient data analysis. Only the patients who were issued their first GLP‑1 RA prescription were included. We used frequency matching to form the control group (CG) which met the same criteria concerning age, sex, and primary diagnosis. The only difference was the absence of any past or current GLP‑1 RA use. The assessed outcomes included the incidence of adverse events, such as vision loss (International Classification of Diseases, Tenth Revision [ICD‑10] code H54 with all subcategories), NAION (ICD‑10 code H47.01), papilledema (ICD‑10 code H47.1), and optic neuritis (ICD‑10 code H46), evaluated during a maximum of 2‑year clinical observation. Adverse events were clinically driven based on the diagnosis made during an ophthalmologic examination. On the date of enrollment to our registry, both SG and CG patients were free of the outcome diagnosis.

This study followed the STROBE reporting guidelines.¹⁰ The study was approved by the local Bioethics Committee of Jan Kochanowski University in Kielce, Poland (62/2024), and was conducted in accordance with the Declaration of Helsinki.

Results

A total of 36 591 patients who met the inclusion criteria were identified. Forty‑six were excluded due to prior NAION, optic neuritis, blindness, or papilledema diagnosis. The SG comprised 36 545 patients, matched by age, sex, and primary diagnosis, with an equal number of patients (n = 36 545) in the CG. Baseline demographic characteristics and comorbidities for both groups are outlined in Supplementary material, Table S1. Regarding the GLP‑1 RA use, most patients were treated with semaglutide (n = 9934 subcutaneously and n = 4115 orally), followed by liraglutide (n = 12 516), dulaglutide (n = 4370), and tirzepatide (n = 984), while 4626 were on a combination of any GLP‑1‑RAs.

At 2‑year follow‑up, the incidence of outcome measures such as NAION, blindness, optic neuritis, and papilledema is shown in Supplementary material, Table S2. There was no difference in the incidence of any of them among the patients treated with a GLP‑1 RA in comparison to those who did not receive it.

The Kaplan–Meier curves for event‑free survival from vision loss and blindness are shown in Supplementary material, Figure S2. The median time to the first event (NAION, blindness, optic neuritis, and papilledema combined) in the SG was 246 days (95% CI, 175–317) vs 392 days (95% CI, 221–496) in the CG (P = 0.03).

The occurrence of blindness, NAION, papilledema, and optic neuritis across different categories of GLP‑1 RA indications in both groups is shown in Table 1. The incidence of blindness varied significantly across different GLP‑1 RA types, with an unadjusted P value below 0.001 (oral semaglutide, n= 6 [0.15%]; subcutaneous semaglutide, n = 19 [0.19%]; liraglutide, n= 10 [0.08%]; dulaglutide, n = 17 [0.39%]; tirzepatide, n = 0; a combination of any, n = 11 [0.24%]). In the logistic regression model assessing independent predictors of blindness, only the use of dulaglutide was significant (P = 0.04). The results of multivariable analysis for independent correlates of combined outcome measures (vision loss, NAION, papilledema, optic neuritis) are presented in Supplementary material, Tables S3 and S4. GLP‑1 RA use was not shown to be an independent predictor of the outcome, also in the subgroup of patients with diabetes and diabetes and obesity, respectively (Supplementary material, Tables S5–S8).

Discussion

In this retrospective cohort study, we assessed the risk of blindness, NAION, papilledema, and optic neuritis in over 36 000 patients treated with different GLP‑1 RAs and the matched controls. The main finding was that GLP‑1 RA use was not associated with an increased risk of these outcomes in the overall patient population, irrespective of the indication for GLP‑1 RA use.

NAION is a major cause of blindness or severely impaired vision in adults, representing the most common cause of acute optic neuritis in patients over 50 years old, and the second most frequent form of all optic neuritis after glaucoma.¹¹ The pathophysiologic mechanisms underlying the onset of NAION remain poorly understood.^11-15 It is considered to be a multifactorial disorder in which the causative event, that is, the acute ischemia of the optic nerve head, could be triggered by different combinations of local and systemic risk factors in different patients. Risk factors for NAION include older age, male sex, white race, acute arterial hypotension and acute hypovolemic episodes, metabolic syndrome, chronic renal failure and dialysis, migraine, obstructive sleep apnea syndrome, and certain drugs, such as amiodarone, phosphodiesterase type‑5 inhibitors, interferon-α, oral contraceptives, and sumatriptan.¹¹ Because of the relationship between NAION and blindness, we sought to assess this outcome in that cohort. As NAION is diagnosed based on acute monocular, painless visual loss and the presence of sectorial or diffuse optic nerve edema, we also assessed papilledema as the outcome in that patient population. We decided to evaluate the incidence of optic neuritis, because it is one of differential diagnoses in individuals younger than 50 years, generally presenting as acute or subacute vision loss with variable levels of eye discomfort and disc edema.¹¹ We also know from previous studies that diabetes, consistently identified as the only significant predictor of outcome in our patient cohort, may cause vision loss and optic neuritis,^16,17 especially during long‑term observation.

This study has several strengths. The sample size was substantially larger than in the prior study that assessed the relation between GLP‑1 RA use and NAION. In addition, this is the only study that included patients treated with GLP‑1 RAs other than semaglutide, such as liraglutide, dulaglutide, and tirzetapide. The median follow‑up was long at approximately 2 years. The controls were matched to balance the cohorts for potential confounders and selection bias limitation.

The initial study that suggested a link between semaglutide and NAION was a retrospective propensity score matched cohort that included 16 827 individuals.⁵ The cumulative incidence of NAION for the semaglutide and non–GLP‑1 RA cohorts over 36 months was 8.9% (95% CI, 4.5%–13.1%) and 1.8% (95% CI, 0%–3.5%), respectively. The Cox proportional hazards regression model showed a higher risk of NAION for patients prescribed semaglutide (HR, 7.64; 95% CI; 2.21–26.36; P <⁠0.001).

A preprint of a cohort study conducted in Denmark and Norway⁶ assessed incidence rates of NAION among semaglutide and sodium glucose cotransporter‑2 inhibitor (SGLT‑2i) users, along with HRs and incidence rate differences. A total of 44 517 eligible users of semaglutide for the management of T2DM in Denmark and 16 860 in Norway were analyzed. An unadjusted incidence rate of NAION of 2.19 per 10 000 person‑years among Danish semaglutide initiators was observed, as compared with 1.18 among SGLT‑2i initiators. After adjustment, there was a 3‑fold increase risk of NAION in semaglutide users vs SGLT‑2i users.

A 5‑year longitudinal cohort study⁷ that assessed the risk of NAION associated with semaglutide in T2DM patients in Denmark included 424 152 participants exposed (n = 106 454) or unexposed (n = 317 698) to once‑weekly semaglutide. The median age was 65 years and half of the patients were men. Exposure to semaglutide was associated with a higher incidence of NAION (0.228 vs 0.093 per 1000 person‑years; P <⁠0.001) and the drug was found to be an independent NAION predictor (HR 2.19; 95% CI, 1.54–3.12) after adjustment. The median (IQR) time to NAION development from the first prescription was 22 (10.2−37.8) months. However, 2 other studies suggested no association between semaglutide use and NAION development^8,9. One of them⁸ included patient data obtained from 160 health care organizations across 21 countries. The final analysis comprised 37 314 participants with T2DM only, 129 690 participants with obesity only, and 130 216 participants with T2DM and aobesity. It showed that at 1‑year follow‑up the administration of semaglutide was not associated with the development of NAION in the T2DM‑only group (HR, 2.32; 95% CI, 0.6–8.97), the obesity‑only group (HR, 0.41; 95% CI, 0.08–2.09), and the T2DM with obesity group (HR, 0.81; 95% CI, 0.42–1.57).

In the other study,⁸ the authors performed 7 different retrospective real‑world cohort analyses on a 66 million patient electronic health records in adult users of GLP‑1 RAs for diabetes and obesity. The primary analysis included patients initiated on GLP‑1 RAs vs those not started on these drugs. The Cox proportional hazards model used to assess the time to develop NAION development showed a 3‑fold risk increase prior to adjustment; however, after controlling for confounders, the increased risk disappeared (HR, 1.45; 95% CI, 0.51–4.17; P = 0.49). The authors performed 5 sensitivity analyses that included the limitation to only semaglutide, only T2DM, only patients with obesity, and only patients who underwent a definitive ophthalmologic examination in the post‑treatment follow‑up. The authors concluded that there was no significant increase risk of NAION from semaglutide or any GLP‑1 RA use in any of the analyses after matching. The frequency of NAION ranged from 0.07% to 0.24%.

In the available studies, there are several differences that we would like to highlight. The size of the studies and GLP‑1 RA exposure differed significantly. Our study included more than 36 000 patients treated with different GLP‑1 RAs and a similar number of controls. In addition, the number of patients exposed to GLP‑1 RAs in our study was significantly higher than that of participants included in the prior ones. Taking into consideration only the patients treated with semaglutide, we included 25 times more participants (14 049 vs 555) exposed to this drug than the initial study of Hattaway et al.⁵ We also assessed exposure to other GLP‑1 RAs, such as liraglutide, dulaglutide, tirzepatide, and a combination of any. In terms of outcomes, all the other studies assessed only NAION, while we also considered other outcomes including blindness, papilledema, and optic neuritis. Regarding the patient population, our cohorts of diabetic patients and patients with obesity were similar to those described in the available literature in terms of age (with a median age of 60 years). The only difference was that in our study, only 30% of the participants were women, whereas in the majority of other studies, nearly 50%.

This study shows low incidence of NAION, blindness, papilledema, and optic neuritis in a cohort of middle‑aged (mostly) male participants with diabetes, with obesity, or with both, exposed to GLP‑1 RAs for a median time of 2 years. Interestingly, we observed a notable intergroup variability in blindness incidence, depending on the specific type of GLP‑1 RA used, and dulaglutide was found to be the only independent predictor of NAION.

Therefore, there is now conflicting evidence regarding the association of GLP‑1 RA and NAION, and further studies are needed to better understand this relationship, in particular due to the large number of patients worldwide that are on these medications.