“FDA performed a meta-analysis of clinical trials across GLP-1 RA development programs because of the small number or absence of cases of NAION observed in individual trials.”
The FDA on Monday ordered every maker of a GLP-1 drug to add a warning about a rare eye condition, and the number that carries the order is smaller than the headlines suggest: fifteen. That is how many probable cases of NAION, a sudden, painless loss of vision in one eye that can be permanent, turned up when the agency pooled 126 randomized trials covering 134,253 patients, according to its drug safety communication. Ten were in people taking a GLP-1 drug, five were in people on placebo, and none were in people on other diabetes medicines. The warning goes into the Warnings and Precautions section of the label for the whole class, from Ozempic and Wegovy to Mounjaro, Zepbound and the newly approved Foundayo.
Here is how to read those fifteen, because almost nobody has walked through them. The 10 cases came over 156,947 person-years of GLP-1 exposure and the 5 over 131,932 person-years of comparator exposure, which the FDA states as 6.4 cases per 100,000 person-years against 3.8. Subtract and you get 2.6 extra cases per 100,000 person-years, or roughly one extra case for every 38,000 to 39,000 years of treatment. If 100,000 people stayed on one of these drugs for a year, the trials point to two or three more cases of NAION than you would otherwise see.
The FDA did not publish a confidence interval for that trial comparison, though it did for its other study, so we ran one from the counts and person-years it did publish. We held the 15 events fixed, split them in proportion to exposure time, and used an exact test, which assumes a constant rate in both groups. The rate ratio comes out at 1.68, with a 95 percent interval of 0.52 to 6.27 and a two-sided p-value of 0.44. Put plainly, the trials alone cannot tell a real near-doubling from the luck of a small draw. That is a property of rare events, and it is the reason the agency pooled 126 trials in the first place.
Where the Warning Actually Comes From
It comes from the observational data, and the FDA’s own summary says so. Its Sentinel study followed adults with type 2 diabetes who started a GLP-1 drug, an SGLT-2 inhibitor or a DPP-4 inhibitor between 2019 and 2024. Rates ran from 26 to 32 per 100,000 person-years, far above the trial rates, though the populations and methods differ and the two sets should not be compared directly. Against SGLT-2 drugs the GLP-1 hazard ratio was 1.19 with an interval of 0.91 to 1.54, which is not significant. Against DPP-4 drugs it was 1.43, interval 1.02 to 2.01, which is.
The agency added that “methodologically rigorous” published studies consistently found higher risk against those same comparators, and that cases have been reported with every marketed GLP-1 product. It also found no biological mechanism in animal data. Europe got here first: the EMA’s safety committee concluded in June 2025 that NAION is a “very rare” side effect of semaglutide, affecting up to 1 in 10,000 people, or 10 per 100,000 at the outer limit.
Hold one more number next to those. The FDA says NAION already occurs in 2.3 to 10.2 of every 100,000 adults over 50 each year, and the GLP-1 trial rate of 6.4 falls inside that range. Diabetes, high blood pressure, sleep apnea and smoking are all risk factors, and they overlap heavily with who gets these drugs. The agency admitted that is why the adverse event reports it received were “challenging to interpret.”
Now the deal-room view. Novo Nordisk, which makes Ozempic and Wegovy, said in a statement that “Novo continues to believe the data do not establish a cause-and-effect relationship.” Eli Lilly noted the FDA cited no findings specific to its medicines, then said the communication applies to everything with GLP-1 activity. Both statements are accurate and neither changes the terms: the class-wide order means no maker can sell against another on eye safety, since each label now carries the same sentence. That is our read, not the FDA’s.
Our position is that the warning is right and the way it was presented is not. The label is justified by the observational evidence and by a European regulator reaching the same conclusion a year and a half ago, not by the trials, and the FDA should have said so and published the interval. Novo’s line would be true of many safety warnings, which is why “potential increased risk” is the FDA’s phrase and not “proven cause.” Nobody should stop an injection because of a headline. If you notice sudden vision loss in one eye, blurred or missing patches, or distorted color, contact a health care professional right away. The FDA says that if you have had an event, you and your prescriber should decide whether to continue.
The open question is the one the agency could not answer: why. Until someone finds a mechanism, the warning rests on correlation across datasets and on fifteen events spread over nearly 289,000 person-years that no one can yet read cleanly.