Semaglutide and NAION: What the New Correspondence Shows

Two years after the first large database analysis linked semaglutide prescribing to non-arteritic anterior ischemic optic neuropathy (NAION), the question is still being fought over in the pages of JAMA Ophthalmology — this week in the form of new correspondence [PMID 42658542, PMID 42658546].

NAION is a form of optic nerve ischemia that presents as sudden, usually painless vision loss. A 2024 retrospective database analysis reported that patients prescribed semaglutide had a higher rate of new NAION diagnoses than a matched group prescribed non-GLP-1 glucose-lowering drugs [PMID 38958939]. The analysis was large and influential; it was also observational, and it immediately raised the question that every database study of this kind faces: whether the association is real or an artifact of how the data were collected, coded, and timed.

What this week’s exchange is about

The new items are a letter and a reply. The letter, by a single author, is titled “Data and Temporal Risk Concerns of Semaglutide and NAION”; the response, from the authors of the original work, answers directly [PMID 42658542, PMID 42658546]. The titles alone mark the fault line. “Data” points at the quality of the underlying records — NAION is a clinical diagnosis, and in administrative or electronic health record data it lives or dies by coding accuracy. “Temporal risk” points at the harder problem: whether the exposure window examined truly precedes the outcome, and whether the timing of a prescription in a database is a trustworthy proxy for actual drug exposure at the time of an event.

Both concerns are structural to this literature, not specific complaints about one paper. NAION is rare enough that prospective trials will never have the statistical power to address it directly; the evidence will come from retrospective cohorts. That makes the methodological arguments in exchanges like this one the actual frontier of the safety question. A retrospective signal survives only if it survives scrutiny of coding, confounding, and time.

The honest limit

A correspondence exchange does not settle anything. It does, however, serve a real function in a public debate: it forces the original authors to defend their analytical choices under the journal’s scrutiny, and it gives readers a compact record of where the disagreement actually sits. For the semaglutide–NAION question, the disagreement sits exactly where the 2024 report left it — on whether the observed association reflects biology or the shape of the data.

The practical reading for researchers and prescribers is unchanged. NAION remains a rare outcome, the evidence remains observational, and the risk discussion belongs in the clinical conversation around a class of drugs prescribed at enormous scale. This exchange adds a layer of methodological detail, not a new verdict — and that is the correct frame until either mechanism studies or much larger cohorts move the question forward.

For a compound-level reference record covering semaglutide’s preparation and administration, see the Semaglutide reference record on ResearchProtocols.