Picture a man in his fifties, a year into a weekly injection that has taken two stone off him and steadied his blood pressure. He wakes, covers his right eye to read the clock and finds a grey smudge where the numerals should be. There is no pain. He assumes it is sleep, makes tea and by lunchtime has decided it will pass. The condition he may have has no established treatment. What follows is a scramble to establish what has happened and whether the drug should continue.

The US Food and Drug Administration brought that scenario into regulatory language on 9 November. All twelve marketed GLP-1 receptor agonists must now carry a warning of a potential increased risk of non-arteritic anterior ischaemic optic neuropathy, known as NAION. By the standards of drug safety the evidence is thin. Across 126 randomised trials the agency counted 15 probable cases, ten among patients on GLP-1 drugs and five on placebo, which works out at 6.4 cases per 100,000 person-years against 3.8 for comparators. Its own analysis of insured American patients with type 2 diabetes found a higher risk against one older class of diabetes drug and no statistically significant difference against another. Animal studies offered no mechanism. Anyone claiming certainty in either direction is ahead of the data.

Britain is no newcomer to the question. In February the Medicines and Healthcare products Regulatory Agency updated semaglutide product information to list NAION as a very rare effect, affecting up to one in 10,000 patients, following the European regulator. That advice covers semaglutide alone. Washington has now extended the warning to the whole class, which takes in tirzepatide, the drug at the centre of the NHS's phased obesity rollout. The MHRA will need to say whether the evidence supports the same step. The wording of a label is the simplest part of what follows and probably the least consequential.

The harder problem is detection. The MHRA's own notice recorded three Yellow Card reports of suspected NAION with semaglutide between 2018 and August 2025, against roughly 10.2 million packs dispensed. That figure can be read as confirmation of extreme rarity. It can equally be read as a measure of how seldom a clinician or patient thinks to report sudden loss of vision in one eye to a voluntary scheme. The data cannot tell the two apart. NAION occurs mostly in people over 50, and its recognised risk factors include diabetes, hypertension, sleep apnoea and obesity, which describes the people being prescribed these drugs. A background rate therefore exists, and any drug signal has to be separated from it. Passive reporting is poorly suited to that task in a population this large and this varied.

Access to care follows. The MHRA tells patients with sudden visual change to attend eye casualty "if available in your area". That conditional clause is the system describing itself with unusual candour. Ophthalmology is among the busiest outpatient specialties in the NHS, and urgent eye provision varies between trusts and between community optometry arrangements. A patient who waits a day or two for a first assessment loses time that matters for establishing the cause and for deciding about the medication. Primary care will carry most of the prescribing, yet it has no routine pathway for recording a patient's visual history before a GLP-1 is started, or for flagging someone who has already lost vision in one eye, a group for whom caution is plainly warranted.

A further complication sits outside the NHS. A substantial private market supplies these drugs through online pharmacies whose records do not reach a patient's GP. Someone who loses vision after buying a pen online may present to an eye clinic where nobody knows the drug is involved, and the case will probably never be counted. The MHRA identified falsified tirzepatide pens supplied by one online pharmacy this year, which shows how porous that channel can be.

The practical implications are specific. Integrated care boards and trusts scaling up prescribing should build a short, recorded conversation into initiation that tells patients to treat sudden painless loss of vision as a same-day eye emergency. They should also agree an urgent route that does not depend on postcode. The MHRA should state promptly whether UK advice on tirzepatide will change and should press clinicians to report. Novo Nordisk and Eli Lilly, whose UK growth depends on public confidence, have a clear interest in supporting active surveillance. The NHS holds linked primary care and hospital records that could produce a far better estimate of the true rate than spontaneous reports ever will, and health-data teams in the service and in industry should treat that as a live project.

None of this justifies patients abandoning treatment on the strength of a label. The absolute risk is low, the mechanism is unexplained and the benefits of these drugs for people with diabetes, cardiovascular disease and severe obesity are well documented. The reasonable response is proportionate awareness. Rare harms in a drug taken by millions will occur as a matter of arithmetic. How early they are noticed, how accurately they are counted and how quickly an eye service can see the patient depend on decisions made well before anyone wakes with a grey smudge in one eye.