Semaglutide and wet AMD: signal vs proof

A pharmacovigilance paper published in the journal Eye in June 2026 named five medicines that turn up disproportionately often in adverse-event reports mentioning neovascular age-related macular degeneration, the wet form of the disease. Semaglutide, the GLP-1 receptor agonist sold as Ozempic, Rybelsus and Wegovy, is one of them^[1]. The other four are apixaban, carbamazepine, latanoprost and rituximab. Lists like that travel fast, so it is worth being exact about what this one measures. A reporting imbalance is a prompt to go and look. It is not a measured risk.

What the researchers actually did

The team works in the pharmacy department of the Eye and ENT Hospital at Fudan University in Shanghai. They began with the US Food and Drug Administration's spontaneous reporting database, known for years as FAERS and now renamed the Adverse Event Monitoring System, and searched for drugs whose reports mentioned wet AMD out of proportion to the rest of the database^[1]. Five came out. They then mapped those drugs onto the proteins they act on and ran summary-based Mendelian randomisation and colocalisation analyses against cis-pQTL data, which is a genetic way of asking whether a protein sits on the path to a disease or merely near it.

Six genes emerged: IGFBP6, MAPKAPK2, NFKB1, RGMA, RNASE1 and WARS1. Only WARS1 cleared the colocalisation test, which is the stricter of the two^[1]. Single-cell sequencing from patients with wet AMD placed most of the candidates in vascular remodelling endothelial cells, with WARS1 also specific to monocytes, and pathway analysis pointed at Toll-like receptor, TNF and NF-kappa B signalling. The authors call the result a potential association and a set of testable hypotheses for later experimental validation. That is their own wording, and it is the right size for the finding.

What a reporting signal can and cannot show

The FDA states the limits of this kind of data plainly. Submission of a report does not mean the information in it has been medically confirmed. For any given report, there is no certainty that a suspected drug caused the event. The information cannot be used to estimate how often the event actually happens. Duplicate and incomplete reports sit in the system^[2].

Three consequences follow for any study built on that foundation. There is no denominator, so a medicine taken by millions produces more reports of everything than one taken by a few thousand. Reporting also answers to attention, because once a drug and an organ have been linked in the press, symptoms that were previously put down to old age start arriving as formal reports. And the patients who receive these five medicines tend to be older and medically complicated, which describes the population that develops wet AMD anyway.

Wet AMD is common in the same people

The National Eye Institute puts age-related macular degeneration at 11 million people in the United States^[3]. The wet form, which the institute describes as abnormal blood vessels growing in the back of the eye and damaging the macula, is the less common late stage, and it usually causes faster vision loss. Risk climbs with age, is higher from 55 onwards, and is higher again for people with a family history and for smokers^[3]. The available treatments are anti-VEGF injections into the eye and photodynamic therapy, and the stated aim of both is to stop further vision loss.

What a finished safety review looks like

There is a useful comparison close at hand. The European Medicines Agency's safety committee examined a different eye condition, NAION, in people taking semaglutide, and worked through non-clinical studies, clinical trials, post-marketing surveillance and the medical literature before reaching a conclusion. NAION is a very rare side effect, meaning it may affect up to 1 in 10,000 people taking semaglutide, and the epidemiology points to roughly a doubling of risk in adults with type 2 diabetes, which works out at about one extra case per 10,000 person-years of treatment^[4].

That is what a completed assessment delivers: a named frequency plus an absolute number and a change to the product information. Nothing of the sort exists for wet AMD. The other documented eye finding is different again, because SUSTAIN-6 recorded more diabetic retinopathy complications on semaglutide than on placebo, at a hazard ratio of 1.76^[5]. Diabetic retinopathy is a separate disease from wet AMD, and NAION is separate from both. Our semaglutide safety record keeps the three apart for that reason.

What we don't yet know

Almost everything a patient would want to know. The full text of the Eye paper sits behind a subscription, so the reporting odds ratios, the confidence intervals and the drug-by-drug gene mapping are not visible in the public abstract^[1]. Nobody has published an absolute risk figure for semaglutide and wet AMD. Confounding by indication has not been tested here. The genetic arm describes biology that may well matter in wet AMD, but it does not show that any of the five drugs switches that biology on. A cohort study with a real denominator would settle far more than this design can, and none has been done.

What this means if you take semaglutide

Nothing about a prescription changes on the strength of a reporting signal. Changes in central vision, including blurring or straight lines that look wavy, deserve an eye examination whatever medicine you are on, and the National Eye Institute advises seeing an eye doctor right away about wavy lines^[3]. Country rules and the wider picture sit on our semaglutide page and its regulation section. This article is for educational and journalistic purposes only and does not constitute medical advice. Any decision about starting or stopping semaglutide belongs with a qualified healthcare professional.