Six Per 100,000: The FDA's GLP-1 Eye Label Is a Payer and Litigation Problem, Not a Demand Shock

Read precisely, the order is narrower than the headline and more consequential in a different place. The FDA reviewed trial data, animal data and post-marketing data including observational studies and case reports, concluded that GLP-1 receptor agonists carry a potential increased risk of NAION, and required that information be added to the Warnings and Precautions section of labeling for all of them, including the oral small molecule orforglipron approved in April 2026 under the national priority voucher program. The FDA noted that NAION cases have been reported for all currently marketed GLP-1 products, which is why this is a class action rather than a molecule-specific one.

What is absent is as important as what is present: no boxed warning, no REMS and no screening mandate. Prescribers are told to instruct patients to report sudden, painless vision loss or visual field deficits, to refer suspected cases promptly to an ophthalmologist, and to consider discontinuation only where benefits do not clearly outweigh risk. The agency also flags that a patient with prior NAION is at higher risk of recurrence and that the fellow eye is at risk.

The American Academy of Ophthalmology and NANOS reached the same conclusion in a May 2026 clinical statement, and went further than the FDA on what a prescriber should do. Their review of 78 semaglutide trials found a point estimate of odds ratio 3.92 for ischemic optic neuropathy with a confidence interval of 1.02 to 15.02, wide enough to include no effect; a separate meta-analysis of 69 GLP-1 trials across all approved indications found no association at all, odds ratio 1.53 with a confidence interval of 0.53 to 4.44. Causality was not established, and the two societies declined to support the EMA's immediate-stop recommendation, favouring shared decision-making with an ophthalmologist instead.

The same statement identified the practical risk factors that matter for case selection: a small, crowded optic disc, the so-called disc at risk, plus diabetes, cardiovascular disease, hypertension, metabolic syndrome and obstructive sleep apnea, with up to one quarter of affected patients developing NAION in the fellow eye. Those are the patients a payer or a prescriber can actually identify, and they are the reason the label's incidence is more consequential in coverage policy than in prescriptions.

Take ten million patient-years of GLP-1 exposure. On the FDA's randomized-trial data that produces about 2,600 total NAION events and roughly 260 excess events against comparators; on the UK's regulatory estimate it produces about 1,000 excess events, four times as many, because the UK's number comes from observational data rather than trial data. That spread between two estimators is the real story about the liability tail: the size of the number depends on which evidence base a court or a payer adopts, and the trial-based number is the one written into the label itself.

For context on the base rates, general-population NAION incidence runs 2.3 to 10.2 per 100,000 people per year in adults over 50, and the FDA cites a background range of 11.4 to 82 per 100,000 person-years in type 2 diabetes. A person who takes a GLP-1 drug and never develops NAION is therefore still far more likely never to develop it than to develop it.

The federal multidistrict litigation over NAION claims was consolidated in December 2025 with 21 actions pending across three districts and nine related actions notified, before any US label change existed. The claims are failure to warn, design defect and breach of warranties across Ozempic, Wegovy, Saxenda and Trulicity, and the case schedule shows the docket is overwhelmingly Novo-shaped: seventeen actions against Novo Nordisk entities in the District of New Jersey, plus individual cases against the company in the Eastern District of Pennsylvania and the Northern District of Texas, against two actions naming Eli Lilly. Transferee Judge Karen S. Marston will run pretrial proceedings in Philadelphia.

A label does not create liability, but it re-prices it, and it does so asymmetrically. Because the FDA asked prescribers to weigh benefits against risks rather than to stop treatment, defendants gain a state-of-the-art defense authored by the regulator itself. Because the same document names the complication, the referral pathway and the risk factors, it also hands plaintiffs the exact roadmap their failure-to-warn theory needs. The label is simultaneously the best defence exhibit and the best complaint template, and the first four out-of-court payouts have already landed: Danish authorities granted compensation to four patients who suffered vision loss after using Novo Nordisk's GLP-1 medicines.

Here is where the label has teeth that a script forecast will miss. Medicare's GLP-1 Bridge began on July 1, 2026 and runs to December 31, 2027, covering selected GLP-1 drugs for obesity for eligible Part D beneficiaries at a $50 copay, administered through a single central processor that handles prior authorization, claims adjudication and pharmacy payment. Part D sponsors do not carry the risk for eligible drugs furnished under the Bridge, so the decision of who qualifies sits with one federal gate rather than dozens of plan formularies.

That is the mechanism by which a labeling change becomes an economic event without touching a single prescription. A label that names NAION, describes it as sudden and painless, tells prescribers to refer suspected cases, and identifies the risk factors a claims database can see — prior NAION, and the crowded optic disc that only an eye examination can reveal — supplies everything a prior-authorization criterion needs and nothing a prescription forecast can capture. Every additional member of the roughly 11% of American adults now using a GLP-1 drug for weight loss becomes a candidate for a coverage edit at renewal.

The differentiation is statistical rather than rhetorical. Every head-to-head signal in the record is semaglutide-based: the hazard ratio of 2.27 against empagliflozin in an OHDSI analysis of 37.1 million adults with type 2 diabetes, the self-controlled case series incidence rate ratios of 1.32 and 1.50, the pooled Danish and Norwegian registry hazard ratio of 2.81, and TriNetX hazard ratios of 2.39, 2.44 and 2.05 at two, three and four years of treatment. No equivalent tirzepatide-specific estimate appears in the FDA review, the MHRA update or the ophthalmology statement, because none has been produced.

That asymmetry is what makes this a relative call rather than a class call, and it arrives at an awkward moment for one of the two companies. Novo Nordisk is cutting US list prices roughly 50% on Wegovy and 35% on Ozempic from January 1, which the company says will hit 2027 cash flow, while the molecule carrying the eye-risk signal is the one whose injectable diabetes base is already shrinking. Semaglutide's relative evidence weakens exactly as its pricing resets, whereas tirzepatide gains a clean comparative profile it never had to argue for.

The obvious second-order idea from a vision-loss warning is that eye diagnostics and ophthalmology volumes inflect. The evidence says no. Neither the FDA communication nor the AAO and NANOS statement recommends any routine or baseline eye examination before or during GLP-1 therapy; referrals are triggered by symptoms, which means the diagnostic load is bounded by the case count, not by the 11% of American adults now taking a GLP-1 drug for weight loss. There is also no approved treatment to monitor, since management of established NAION is observation, so a treated population generates no recurring screening revenue.

The listed diagnostics vehicle is thinner still. Topcon, the Japanese maker of the Maestro2 optical coherence tomography and fundus camera that images the optic nerve head and retinal nerve fibre layer used in NAION workup, was taken private in a KKR and JIC Capital buyout and its shares were delisted on December 2, 2025, so the sector's most direct listed proxy for the theme no longer trades. The remaining exposure is diversified eye-health names whose growth has nothing to do with NAION. There is no volume behind the eye-care re-rating, and the honest conclusion is that this is a coverage-and-litigation event with a thin, largely private diagnostics tail.

Note what the map shows about scale. The two upstream peptide suppliers are expanding capacity on the assumption that GLP-1 volumes keep growing, and the label is far too small to change that: a 2.6 per 100,000 patient-year excess risk does not dent peptide orders, while the eight-fold-plus growth in molecules served at WuXi AppTec says nothing about safety-adjusted persistence either way. The downstream substitution effect is real but narrow, because it applies to a small, identifiable pool of type 2 diabetes patients with a crowded optic disc or prior NAION rather than to the whole obese population, and because a coverage criterion can do the work of switching a prescriber's mind without a patient ever asking for a refund. The label's cash-flow footprint is narrow, documented and monitorable, which is the opposite of what a boxed warning would have delivered.

Synthesising the layers: the facts are that the FDA required Warnings and Precautions language on the class on November 9, that the agency's own excess-risk estimate is 2.6 events per 100,000 patient-years, that the label prescribes neither stopping nor screening, and that the litigation docket is concentrated on one manufacturer. The inference is that the economic consequences run through coverage criteria and case valuation rather than through prescription volume, because a rare complication with a named referral pathway is a better prior-authorization input than a prescription forecast. The speculation, clearly labelled, is that the 2027 Medicare renewal cycle converts the label into explicit ocular monitoring, and that European script data reveal the discontinuation response the US wording invites. What is not disclosed, and should not be assumed, is any evidence that US prescribers have begun discontinuing on the basis of the label; no such data exist yet, which is precisely why the trade is a monitoring exercise today rather than a positioning one.