Two separate eye findings attach to GLP-1 medications, and they get discussed as though they were one. They are not related, they involve different parts of the eye, and confusing them makes both harder to think about clearly.
The first is diabetic retinopathy, which lives in the retina and appeared as a numerical signal in a 2016 cardiovascular outcomes trial. It is on the semaglutide label. The second is a sudden loss of vision in one eye called NAION, which involves the optic nerve rather than the retina, surfaced in a 2024 clinic study, and was added to European product information in 2025 as a very rare effect.
The retinopathy signal has a mechanism that was described in the 1990s and has nothing to do with these drugs specifically. The NAION question is younger, smaller in absolute terms, and much less settled. Both are worth understanding before starting anything, and neither is a reason to interpret a change in vision without an ophthalmologist.
Key Takeaways
- SUSTAIN-6, published in the New England Journal of Medicine in 2016, reported diabetic retinopathy complications in about 3.0 percent of the semaglutide arm against about 1.8 percent on placebo over roughly two years, a hazard ratio near 1.76. This is the finding behind the diabetic retinopathy warning on semaglutide labelling.
- The excess in SUSTAIN-6 was concentrated almost entirely in participants who entered with existing retinopathy, were using insulin, and had a large HbA1c fall in the first sixteen weeks. Participants without retinopathy at baseline showed very few events in either arm.
- Early worsening of retinopathy after rapid glycaemic improvement was documented in the Diabetes Control and Complications Trial in 1993, where about 13.1 percent of the intensive arm worsened at twelve months against 7.6 percent conventional. The effect reversed, and the intensive arm ended with substantially less retinopathy years later.
- Non-arteritic anterior ischaemic optic neuropathy is an optic nerve event, not a retinal one, and it is a separate question from retinopathy. A JAMA Ophthalmology study in July 2024 from a Boston neuro-ophthalmology practice reported a large relative increase among patients prescribed semaglutide, and subsequent database studies were mixed.
- In June 2025 the European Medicines Agency safety committee concluded NAION is a very rare side effect of semaglutide, with a frequency of roughly one in ten thousand treated people, and required it to be listed in the product information.
- Blurred vision during the first weeks of a large glycaemic change is usually an osmotic refractive shift in the lens rather than retinal damage, which is why spectacle prescriptions are not normally updated during that window.
- No trial has tested whether a reduced dose or a slower escalation protects the retina. The mechanistic argument rests on the rate of HbA1c fall, which is an inference from the glucose control literature rather than a finding about GLP-1 dosing.
Who Actually Had the Retinopathy Events in SUSTAIN-6?
Approximate percent of participants with diabetic retinopathy complications over roughly two years, semaglutide arm versus placebo arm, split by status at enrolment
- Semaglutide arm
- Placebo arm
Approximate values reflecting the whole-trial result and the direction of the subgroup reporting rather than exact published subgroup figures, several of which involved small event counts. The first pair of bars is the number that reached the label. The next three explain it. A participant who entered without retinopathy sat in a risk band roughly ten times lower than one who entered with it, in both arms.
What did the diabetic retinopathy signal actually show?
One trial, one subgroup, one mechanism
SUSTAIN-6 randomised 3,297 people with type 2 diabetes and high cardiovascular risk to semaglutide or placebo, and it reported more diabetic retinopathy complications in the semaglutide arm, with a hazard ratio near 1.76. The complications counted were serious ones, meaning vitreous haemorrhage, blindness, or a need for intravitreal injection or retinal photocoagulation. Across roughly two years the rates were about 3.0 percent and 1.8 percent.
The trial population matters more than the headline. These were people with long-standing diabetes, a mean HbA1c around 8.7 percent at entry, and a high rate of insulin use. Participants who already had retinopathy when they enrolled accounted for the overwhelming majority of events, and those who did not had very few events in either arm. The signal was never a statement about people with healthy retinas.
What the trial investigators proposed, and what the label reflects, is that the excess was tied to the magnitude and speed of the glycaemic improvement rather than to a toxic effect of the molecule on retinal tissue. Semaglutide produced a larger and faster HbA1c fall than placebo, and the retinopathy events clustered in people whose fall in the first sixteen weeks was largest. That explanation has a long history behind it.
Why Does Retinopathy Get Worse Before It Gets Better?
Approximate percent showing retinopathy worsening by month after a large improvement in glycaemic control, drawn from the intensive glucose control literature
- Rapid glycaemic improvement
- Gradual or no change
Illustrative curves anchored to the twelve month early worsening figures reported in the DCCT and the direction of its longer follow up, rather than digitized values from one publication. The lines cross somewhere between eighteen and thirty-six months. That crossing is the whole point of the graph, and it is why early worsening is treated as something to watch through rather than something to prevent by leaving glucose high.
What is happening in the retina during a rapid drop in glucose?
Four proposed mechanisms, none of them fully settled
Early worsening is a well-documented observation attached to several competing explanations, and no single mechanism has been established as the cause. The candidates below are the ones the ophthalmology literature returns to, and they are not mutually exclusive.
Retinal blood flow autoregulation resetting
Chronically high blood glucose raises retinal blood flow, and the vessels adapt to operating in that state over years. A rapid correction reduces flow toward normal faster than the adapted vasculature adjusts, producing a period of relative retinal ischaemia. Ischaemic retina upregulates vascular endothelial growth factor, which is the same pathway that drives proliferative disease. This is the explanation that connects most cleanly to why the effect is transient and why it is proportional to the size of the glycaemic change.
Insulin-like growth factor and VEGF signalling
Rapid glycaemic improvement alters circulating IGF-1 and its binding proteins, and IGF-1 acts on the retinal vasculature in ways that overlap with VEGF signalling. Some investigators have proposed that GLP-1 receptors present in retinal tissue contribute an additional direct effect on this axis. Evidence for a direct retinal action of GLP-1 receptor agonists in humans is thin, and animal work has pointed in both protective and harmful directions. This mechanism remains a hypothesis rather than a demonstrated pathway.
Loss of the osmotic and metabolic status quo
Retinal tissue that has been metabolically adapted to high glucose for a decade responds to sudden normalisation as a stress rather than a relief, which is the general principle behind early worsening in every setting where it has been observed. The same pattern appears after intensive insulin therapy, after pancreas transplantation, and during pregnancy in women with pre-existing diabetes. None of those involve GLP-1 receptor agonists. That consistency across unrelated interventions is the strongest argument that the mechanism is about the change in glycaemia rather than about any drug.
A shorter time horizon than the trials that reassure
The long-term retinal benefit of better glycaemic control is one of the most firmly established results in diabetes medicine, and it accrues over years. A two year cardiovascular outcomes trial captures the early worsening window and only the beginning of the payoff period. FOCUS, the dedicated five year semaglutide retinopathy trial reported in 2025, was designed specifically to look past that window, and its results are the most direct evidence available on the longer horizon. Interpreting either trial requires knowing which part of the curve it was measuring.
Does the Size of the HbA1c Fall Matter?
Approximate relative odds of early retinopathy worsening by the size of the HbA1c fall achieved in roughly the first four months, with the smallest fall set to one
Constructed gradient illustrating the dose-response relationship described across the intensive glucose control literature rather than odds ratios from a single study. The relationship is directional and consistent, and the exact numbers vary by population and by how worsening was defined. What is not in dispute is that a larger and faster fall carries more early risk, in every intervention that has been examined.
What is NAION and how strong is the evidence linking it to semaglutide?
A different structure, a different mechanism, a much smaller number
Non-arteritic anterior ischaemic optic neuropathy is a sudden, usually painless loss of vision in one eye caused by compromised blood supply to the head of the optic nerve. It has nothing to do with the retina, which is why it is a genuinely separate question from the retinopathy signal. It typically affects people over fifty and is more common in those with a small, crowded optic disc, and the vision loss is often permanent.
The signal emerged from a matched retrospective cohort study conducted at a Boston neuro-ophthalmology practice and published in JAMA Ophthalmology in July 2024, which reported a hazard ratio above four in patients with type 2 diabetes prescribed semaglutide and higher still in those prescribed it for overweight or obesity. That study drew on a population already referred for optic nerve problems, which inflates absolute rates dramatically and is the main reason its numbers cannot be read as population risk.
Larger database studies published afterwards produced inconsistent results, with some finding a modest association and others none. In June 2025 the European Medicines Agency safety committee completed its review and concluded that NAION is a very rare side effect occurring in approximately one in ten thousand treated people, and required it to be added to the product information for semaglutide-containing medicines. That is a regulatory determination that the association is real and that its absolute magnitude is small.
How Do the NAION Numbers Compare With Each Other?
Approximate cases per 100,000 people per year across three framings of the same condition, on a shared scale
Approximate figures converted to a common annual rate for comparison, which the original sources did not report in identical units, so treat the scale as indicative. The bottom two bars come from a clinic population already referred for optic nerve disease and are hundreds of times higher than any population estimate for that reason alone. The highlighted bar is the regulator's frequency classification, and it is the one closest to what a general population figure would look like.
Does a Slower Escalation Change the Shape of the Glycaemic Fall?
Illustrative HbA1c trajectory under a standard escalation schedule versus a slower ramp arriving at a broadly similar endpoint
- Standard escalation
- Slower ramp
Constructed trajectories illustrating an argument that no trial has tested, not measured data from any cohort. The steep early segment of the grey line is the window in which early worsening was observed in the intensive control literature. Whether flattening that segment reduces retinal risk in practice is unknown, and a slower ramp also means longer exposure to higher glucose, which carries its own retinal cost.
Why does vision blur in the first few weeks?
Usually the lens, occasionally something else
Fluctuating blurred vision during a period of rapidly falling blood glucose is most often a refractive effect rather than a retinal one. The lens of the eye takes up and releases water in response to changes in blood osmolarity, and a change in its water content changes its focusing power. Vision that drifts in and out of focus over days, affects both eyes, and is worse for reading or for distance depending on the direction of the shift fits this pattern.
This is the reason optometrists generally decline to write a new spectacle prescription during a window of large glycaemic change. A prescription measured mid-shift describes an eye that will not have the same refraction in six weeks. Waiting until glucose has been stable for a period, usually several weeks to a couple of months, produces a prescription that lasts.
A different set of symptoms points away from this benign explanation entirely. Sudden loss of vision in one eye, a dark or missing area in part of the visual field, a curtain descending, a sharp increase in floaters, or flashes of light are not refractive symptoms and are not something to monitor at home. Each of those warrants same-day ophthalmological assessment regardless of what medication anyone is taking.
How Does Baseline Retinal Status Change What Gets Monitored?
Illustrative number of retinal examinations across a first year of treatment, by retinal status documented before starting
Illustrative pattern showing how surveillance intensity scales with baseline findings, not a schedule for anyone and not drawn from a specific guideline document. Actual intervals are set by an ophthalmologist based on the individual retina, the glycaemic trajectory and local guidance. The point of the graph is the slope, which is that the value of a documented baseline examination rises steeply with how much there already is to see.
What is worth documenting and raising with a prescriber?
Six observations, none of which are a monitoring plan
Everything below describes information worth having and questions worth asking. None of it sets a screening interval, and no visual symptom should be interpreted at home. Decisions about whether to start, how quickly to escalate, and how closely to watch the retina belong with an ophthalmologist and a prescriber who can see the actual eye.
A baseline examination is what makes any later finding interpretable, because retinal changes discovered at month six cannot be dated without one. This matters most for anyone with type 2 diabetes of several years' standing, a high HbA1c, or existing retinopathy, which is precisely the group in which the SUSTAIN-6 events occurred. Retinal photography creates a record that a subsequent clinician can compare against directly. Twenty minutes before starting removes an ambiguity that is impossible to resolve afterward.
Diabetic retinopathy is graded, and the difference between mild non-proliferative changes and severe non-proliferative or proliferative disease is the difference between routine surveillance and a situation where the pace of glycaemic improvement becomes a specific clinical decision. Asking for the grade and writing it down gives a prescriber something concrete to work with. A patient who can say what stage was documented and when is a much easier person to advise. The grade also determines how urgently any later symptom needs assessment.
The combination of long-standing poor control, existing retinopathy, and a rapid fall in HbA1c is the exact profile in which early worsening has been observed across every intervention that produces it. Whether that changes the escalation schedule, adds an ophthalmology review before a dose step, or changes nothing at all is a clinical judgement. Raising it explicitly makes it a considered decision rather than an unexamined one. This is a conversation for the visit before starting, not the one after a symptom.
Painless loss of vision in one eye, a dark area in part of the visual field, or a curtain effect require immediate assessment because several of the possible causes are time-sensitive and some are treatable only within a narrow window. NAION is one item on that list and is not the most common one. Nothing about being on a GLP-1 medication changes the urgency in either direction. Same-day emergency ophthalmology is the correct response, not a message to a prescriber.
The lens changes its refraction as blood osmolarity shifts, so a prescription measured during a period of rapid glycaemic change describes an eye that will not have those optics in two months. Optometrists routinely ask about recent changes in diabetes control for exactly this reason. Waiting until glucose has been steady for several weeks produces glasses that remain correct. This is a practical annoyance rather than a safety issue, but it is a common and avoidable expense.
A previous NAION event in either eye and an anatomically small, crowded optic disc are both recognised risk factors independent of any medication, and a prescriber cannot weigh a very rare risk without knowing whether that background is present. Most people have no idea what their optic disc looks like, which is another reason a baseline eye examination is useful. This information changes very little for most people and is genuinely relevant for a small number. It is worth mentioning once rather than assuming it is on file.
What is genuinely unresolved here?
Three gaps worth naming
Almost none of this evidence comes from people without diabetes. SUSTAIN-6 enrolled people with type 2 diabetes, high cardiovascular risk, and a mean HbA1c near 8.7 percent, and the retinopathy events happened in the subset of that population with existing retinal disease. Someone taking a GLP-1 medication for weight with normal glycaemia has no meaningful HbA1c fall to drive early worsening and no retinopathy to worsen. Extending the retinopathy discussion to that population is an extrapolation across a very wide gap, and the honest answer is that the trials do not speak to it.
The NAION evidence base is young and inconsistent. A retrospective study in a referral clinic produced a striking hazard ratio, several larger database analyses that followed disagreed with each other, and a regulator reviewing all of it concluded the association is real but very rare. No prospective trial has been designed to answer the question, and none is likely to be, because an event affecting roughly one in ten thousand people requires an impractical sample size. The mechanism is also unexplained, which means there is no biological story to weigh against the epidemiology.
Reduced dose protocols have not been examined for any retinal endpoint. The argument that a slower glycaemic fall should reduce early worsening is a reasonable inference from a consistent literature, and it remains an inference. It also cuts both ways, since a slower ramp means a longer period at higher glucose, and chronic hyperglycaemia is the established driver of retinopathy over years. No trial has weighed those two against each other, and nobody should assume a smaller dose is the safer choice for their retina on the strength of the reasoning alone.
Frequently Asked Questions
Selected References
- Marso SP et al. Semaglutide and cardiovascular outcomes in patients with type 2 diabetes. New England Journal of Medicine, 2016. SUSTAIN-6 primary results including the diabetic retinopathy complications endpoint.
- Vilsboll T et al. Semaglutide, reduction in glycated haemoglobin and the risk of diabetic retinopathy. Diabetes, Obesity and Metabolism, 2018. Mediation analysis of the SUSTAIN-6 retinopathy finding.
- The Diabetes Control and Complications Trial Research Group. The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus. New England Journal of Medicine, 1993.
- Early Treatment Diabetic Retinopathy Study Research Group and related DCCT reporting on early worsening of retinopathy after intensive glycaemic control and its subsequent reversal.
- Hathaway JT et al. Risk of nonarteritic anterior ischemic optic neuropathy in patients prescribed semaglutide. JAMA Ophthalmology, July 2024. Matched retrospective cohort from a neuro-ophthalmology practice.
- European Medicines Agency, Pharmacovigilance Risk Assessment Committee. Recommendation on non-arteritic anterior ischaemic optic neuropathy as a very rare side effect of semaglutide-containing medicines, June 2025.
- Novo Nordisk. Prescribing information for semaglutide injection, section on diabetic retinopathy complications in patients with type 2 diabetes.
- Novo Nordisk. FOCUS trial, a long-term study of the effect of semaglutide on diabetic retinopathy in people with type 2 diabetes, results reported 2025.
- Bain SC et al. Worsening of diabetic retinopathy with rapid improvement in systemic glucose control, a review. Diabetes, Obesity and Metabolism, 2019.
- Hayreh SS and related work on non-arteritic anterior ischaemic optic neuropathy, including optic disc morphology as a structural risk factor.
This article is for general educational purposes only and does not constitute medical advice, diagnosis, treatment, or an eye monitoring plan. Nothing here should be used to start, stop or adjust any medication, to set a retinal screening interval, or to interpret a change in vision. GLP-1 medications are prescription drugs, and dose, escalation pace, eye surveillance and any low dose or microdosing approach should be decided and monitored by a qualified healthcare provider and, where the retina is involved, an ophthalmologist who can examine the eye directly. Nothing here is intended to diagnose, prevent, treat, or cure diabetes, diabetic retinopathy, optic neuropathy, obesity, or any other condition, and no low dose or microdosing approach has been shown in trials to be safe or effective or to protect the retina or optic nerve. Several charts here are constructed, approximate or illustrative rather than measured outcomes, and the trial and regulatory figures cited are approximations of published reporting rather than exact values. Sudden loss of vision in one eye, a dark or missing area in the visual field, a curtain effect, new flashes of light or a sudden increase in floaters require immediate emergency eye assessment. Individual results vary.
Smaller Doses. Sharper Protocols.
Metabolic health you can measure, built around a dose designed to work with your body rather than against it.
Find Your Protocol