Semaglutide and NAION in US Veterans: Observational Safety Signal

Using Veterans Health Administration data, investigators reported a target-trial emulation examining whether semaglutide initiation was associated with incident nonarteritic anterior ischemic optic neuropathy (NAION) compared with initiating a sodium–glucose cotransporter 2 inhibitor (SGLT2i).

In this analysis, the authors describe a higher relative hazard of NAION among semaglutide initiators, while also emphasizing that NAION was uncommon on an absolute scale during observed follow-up.

The analytic cohort comprised 102,361 US veterans with type 2 diabetes who were taking metformin and newly initiated either semaglutide (11,478) or an SGLT2i (90,883). Follow-up in the primary analysis was described as a median of 2.1 years, with a maximum follow-up of 7.5 years. Using overlap-weighted adjustment, the authors reported a hazard ratio for incident NAION of 2.33 (95% CI, 1.53–3.54) for semaglutide initiators compared with SGLT2i initiators. Over the same follow-up context, overlap-weighted cumulative incidences were reported as 0.29% vs 0.13%, respectively, which the authors present as an observed association in this dataset rather than a causal effect.

The report also presented NAION incidence rates standardized to person-time. The incidence rate was 123 events per 100,000 person-years among semaglutide initiators and 67 events per 100,000 person-years among SGLT2i initiators. Across groups, 173 total incident NAION events occurred during follow-up, including 30 events in the semaglutide group and 143 in the SGLT2i group. The authors noted that almost all SGLT2i initiators received empagliflozin (90,816 of 90,883), with the remainder receiving dapagliflozin or canagliflozin. Together, these counts and rates reflect relative contrasts observed against a background of infrequent events.

The investigators framed their analysis as an active-comparator, new-user target-trial emulation, estimating cause-specific hazard ratios and using propensity-score overlap weighting to address measured baseline confounding. They described selecting baseline covariates a priori, assessing covariate balance after weighting, and identifying incident NAION through administrative coding (ICD-10 and SNOMED) during follow-up.

The authors also outlined constraints: potential residual confounding from factors not captured or incompletely measured (including optic nerve head anatomy and cup-to-disc ratio, NAION family history, sleep apnea severity, nocturnal hypotension, diabetes duration, and non-VA care), possible outcome misclassification due to coding-based ascertainment, and limited generalizability given an older, predominantly male veteran population. These considerations were presented as important qualifiers for interpreting the reported association.

In discussion, the authors situated their findings alongside other observational reports on GLP-1 receptor agonists and NAION and reiterated that absolute incidence was low in their cohort. They also wrote that NAION is a rare but serious vision-loss event and described patient counseling and prompt evaluation of new visual symptoms in the context of discussions about semaglutide’s benefits.

Key Takeaways:

• In overlap-weighted analyses, semaglutide initiation was associated with a higher hazard of incident NAION than SGLT2i initiation (HR, 2.33; 95% CI, 1.53–3.54).
• NAION occurrence was uncommon in absolute terms, with incidence rates reported as 123 vs 67 events per 100,000 person-years (semaglutide vs SGLT2i).
• The study emulated a target trial using an active-comparator, new-user design with overlap weighting and noted limitations including residual confounding, administrative code-based outcome ascertainment, and generalizability constraints in a predominantly male veteran cohort.