My vision has worsened since my diabetes came under better control: understanding early worsening of diabetic retinopathy

You’ve started a new treatment, your glycated haemoglobin has dropped markedly, your diabetes specialist is pleased with you — and yet your vision has become blurred. This is neither a coincidence nor anything you’ve done wrong. It’s a phenomenon that has been described for close to thirty years: early worsening of diabetic retinopathy. It deserves a calm explanation: it resolves in around one patient in two within eighteen months, it can be monitored, and it does not call into question the value of bringing diabetes under control.

UNDERSTANDING

The paradox: better control, yet worsening vision

This scene plays out again and again in consultations at the Diabet’ Paris 13 clinic. A patient arrives with their latest blood test results: their HbA1c has fallen from 10% to 7% in four months, sometimes less. They’ve lost weight, they feel better. And they’ve come to see me because their vision has become blurred, because lines of text seem to ripple slightly, or because they’re noticing new dark spots. The link with the new treatment seems obvious to them, and they’re not wrong — but not quite for the reason they imagine.

What is then observed has a name in ophthalmology: early worsening of diabetic retinopathy, known by the acronym EWDR. It refers to a progression of retinal damage occurring in the months following a rapid and substantial improvement in blood sugar control. The paradox is real and well documented: treating diabetes well can, transitorily, worsen retinopathy.

In concrete terms, “worsening” does not mean blindness. On fundus photographs, this means new small haemorrhages, cotton-wool spots, sometimes oedema of the macula that accounts for blurred or distorted vision, and, more rarely, the appearance of new vessels in a patient whose retinopathy was already advanced. These are signs that can be recognised, dated and treated, provided they are looked for at the right time.

Let’s say this straight away, because it’s the most common misunderstanding: this phenomenon has nothing to do with non-arteritic anterior ischaemic optic neuropathy (NAION), a vascular event affecting the optic nerve that is much rarer, which was the subject of a separate safety signal under semaglutide and which we cover in our article on GLP-1 and vision.

The second misunderstanding would be to conclude that one should give up on bringing diabetes under control: that would be the exact opposite of what the data show. Early worsening is a transitional phenomenon; chronic hyperglycaemia, by contrast, damages the retina lastingly and silently. The question, then, is how to get through this transition under proper monitoring.

What the DCCT trial established as early as 1998

Early worsening was not discovered with GLP-1 analogues. It was described and quantified in the Diabetes Control and Complications Trial (DCCT), the landmark trial on intensive treatment of type 1 diabetes, whose analysis specifically devoted to this topic was published in 1998 (PMID 9682700). Patients receiving intensive insulin therapy were compared with those receiving conventional treatment, with fundus photographs taken at regular intervals over several years.

The result is clear-cut. Early worsening of retinopathy was seen at six or twelve months in around 13% of patients in the intensive-treatment group, compared with about 8% in the conventional-treatment group. Expressed in absolute terms, that amounts to roughly five additional patients for every hundred treated: a real, measurable difference, but one that leaves more than eight patients in ten completely unaffected by the phenomenon.

The decisive point is what happened next. By the eighteen-month visit, the worsening had resolved in around one patient in two (51% in the intensive-treatment group, 55% in the conventional-treatment group). And, most importantly, over the long term, the intensive-treatment group had markedly less severe retinopathy than the conventional-treatment group: intensively treated patients who had experienced early worsening did just as well, if not better, than conventional-treatment patients who had not experienced it. The authors explicitly concluded that the long-term benefit of good blood sugar control far outweighs the drawback of this transitory worsening. That remains the reference position today.

Early worsening is a transitional episode. The long-term benefit of good blood sugar control for the retina remains far greater than it: the practical conclusion is not to treat diabetes less, it is to keep watching the retina while treating it.

Summary of DCCT data on early worsening, 1998

The DCCT also identified the two most important factors behind this worsening: a high HbA1c before treatment was intensified, and the extent of its fall over the first six months — without a more gradual fall having been shown to be protective. Severe forms following early worsening remained rare in the trial (two high-risk proliferative retinopathies, three cases of macular oedema), and all responded well to treatment. It is this framework, rather than the name of the drug prescribed, that remains the most useful one today.

Why a rapid fall in blood sugar can put the retina at risk

What is understood about the mechanism

No single explanation has reached consensus, but several converging mechanisms have been put forward, notably in a recent review on antihyperglycaemic treatments and the retina (PMID 42193290), covering the retina’s metabolic adaptation and the role of pre-existing ischaemic tissue; the hormonal hypothesis, which is older, comes from other work.

• A retina exposed to high blood sugar over a long period has adapted to an abnormal metabolic environment; rapid normalisation of glucose abruptly changes the energy supply to tissue that is already weakened.
• Retinal blood flow and the self-regulation of capillaries readjust, which can encourage the appearance or worsening of macular oedema in the first few months.
• Rapid correction of blood sugar is accompanied by hormonal changes, notably in the growth factor IGF-1, which may stimulate the proliferative component of pre-existing retinopathy.
• Areas of ischaemia that have already formed do not disappear: they remain the ground on which all of this plays out, which explains why the phenomenon largely spares retinas that are otherwise unaffected.

It’s not about the drug

This is the most important point in this article, and the one that press headlines systematically miss. Early worsening has been described with intensive insulin therapy; it is reported after bariatric surgery, where blood sugar sometimes normalises within a few weeks; and it is routinely monitored during pregnancy in a woman with diabetes, a situation in which control is tightened rapidly. GLP-1 analogues are added to this list because they lower HbA1c effectively, not because they are toxic to the retina.

In other words: the factor responsible is not the drug, it is the speed of the fall in someone who already has damage. If you’d like to understand what this HbA1c figure means for your eyes, we’ve devoted an entire article to the link between blood sugar and vision.

Semaglutide: what the SUSTAIN-6 trial actually shows

The SUSTAIN-6 trial assessed the cardiovascular safety of semaglutide in high-risk patients with type 2 diabetes, followed for about two years (PMID 27633186). Its cardiovascular results were favourable. But a secondary signal attracted attention, and it is this signal that fuelled the press coverage: diabetic retinopathy complications — vitreous haemorrhage, blindness, the need for laser treatment or an intravitreal injection — were more frequent in the semaglutide group.

These figures deserve to be given in absolute terms, because they are far more meaningful than a risk ratio: about 3 patients in 100 in the semaglutide group, versus just under 2 in 100 on placebo, over two years. The difference amounts to roughly one additional patient for every hundred treated. That is not negligible, but nor does it justify alarmist headlines.

The dedicated analysis published afterwards (PMID 29178519) is the real key to interpreting this. It shows that this signal is concentrated in patients already on insulin, who had retinopathy and poor blood sugar control at the start of the trial, and who had the strongest and fastest fall in HbA1c during the first sixteen weeks. In the SUSTAIN 1 to 5 trials and the Japanese trial, whose participants far less often had retinopathy at the outset, no imbalance in retinal events was observed — an indirect signal, rather than proof, in favour of patients with unaffected retinas. This is exactly the same pattern as the DCCT, twenty years later, with a different class of drug.

Review articles on incretins and the microvascular complications of diabetes (PMID 37597048), as well as more recent overviews of new blood-sugar-control drugs and retinopathy progression (PMID 40586870), reach the same cautious conclusion: the link has more to do with the speed of blood sugar correction than with any specific effect of the drug on the retina, and long-term data are still lacking.

A trial specifically dedicated to the retina under semaglutide, the FOCUS study, was set up to settle the question; its results have not been published to date and are still awaited. Let’s be frank about it: on this particular point, the definitive answer does not yet exist, and any categorical statement, in either direction, goes beyond what the data allow.

Tirzepatide (Mounjaro): a recent and contradictory body of literature

Tirzepatide, a more recent drug, acts on two receptors (GIP and GLP-1) and often lowers HbA1c even more markedly. The retinal question arises for it in the same terms — except that the literature does not say the same thing from one publication to the next.

• An assessment of the effect of tirzepatide on the risk of retinopathy in type 2 diabetes has been published (PMID 38456523).
• A real-world cohort subsequently reported cases of early worsening in patients treated with tirzepatide (PMID 40637847).
• This work immediately sparked a methodological controversy: two separate letters challenged the robustness of these findings, one arguing that there is no evidence that tirzepatide worsens retinopathy (PMID 41051409), the other proposing that the risk estimate be reassessed entirely (PMID 41051411), and the original authors responded to both in a single article (PMID 41051410).
• A comparison of ocular outcomes between tirzepatide and GLP-1 analogues provides further evidence (PMID 41655764), as does a study on the occurrence of retinopathy after one year of tirzepatide as adjunctive treatment for type 1 diabetes (PMID 42452840).

We are not going to settle this on the literature’s behalf. When competent teams disagree, the role of a doctor writing for their patients is to say so, not to pick whichever version sounds most reassuring or most dramatic. This disagreement does not change the practical approach: it is the same whichever drug is involved, because it is based on the state of the retina and the speed of the fall in HbA1c.

Who is affected, and who is much less so

The situations that call for particular vigilance overlap from one study to the next:

• Diabetic retinopathy that is already present, particularly if it is moderate to severe, or proliferative.
• A high HbA1c before the change in treatment, typically above 9% or 10%.
• A fall in HbA1c of more than two percentage points in under six months.
• Long-standing diabetes, poorly controlled high blood pressure, associated kidney disease, or a current pregnancy.

The window for vigilance is also well defined: it is the three to twelve months following the rapid fall in HbA1c. Beyond that, the balance tips clearly towards benefit.

Conversely, a patient with a normal fundus examination, whose HbA1c was already close to target and whose fall is gradual, falls into a category in which this phenomenon is very poorly documented. And a person without diabetes being treated for obesity cannot, by definition, develop diabetic retinopathy: the question simply does not arise for them in these terms, even though normal ophthalmological follow-up remains worthwhile.

In practice: fundus examination first, close monitoring after

Here is what all this means in concrete practice.

• Before, or right at the start: a fundus examination, ideally with retinal photographs and macular optical coherence tomography (OCT), for any patient with diabetes who is about to start a treatment likely to bring HbA1c down quickly.
• This examination serves as a starting point: without a baseline image, any worsening seen six months later is impossible to interpret honestly.
• If retinopathy is already present: close follow-up, in practice at around three months and then around six months, during the first year.
• If the fundus examination is normal and HbA1c is moderately raised: the usual screening schedule is most often sufficient.
• If proliferative retinopathy is already known, the retinal treatment required (panretinal photocoagulation, intravitreal injections) is carried out without delay, independently of whatever is decided regarding the diabetes.

One point must be stated without the slightest ambiguity: none of this is a reason to stop, delay or change a diabetes treatment. These decisions are for your diabetes specialist or your GP to make, not your ophthalmologist — prescribing for diabetes is not within an ophthalmologist’s remit. The ophthalmologist’s role is to look at the retina, to establish its current state and to monitor it at the right pace. If this situation applies to you, the right course of action can be summed up in one sentence: talk to your diabetes specialist or your GP about it, and have your fundus examined.

Cachan practice · Tel. +33 1 45 47 08 11

Organised screening in Île-de-France: the OPHDIAT network

Patients followed up in Île-de-France have access to a tool that many are unaware of: OPHDIAT, a telemedicine-based screening network for diabetic retinopathy. The principle: retinal photographs taken at a local centre by trained staff, then interpreted remotely by an ophthalmologist. A review of the first five years was published as early as 2012 (PMID 22765981).

Two more recent publications are directly useful here. The first looks at the ten-year incidence of retinopathy and at the screening intervals that this data makes acceptable (PMID 32522790): not all patients need the same frequency, and it is possible to reasonably space out monitoring for patients with unaffected retinas and good control, provided that monitoring is tighter for others. The second focuses on patients over 70 (PMID 41547437), a growing population that has long been understudied.

This logic of adjusted intervals links directly to the question of early worsening: starting a treatment that is going to bring HbA1c down is one of the reasons to temporarily tighten the monitoring schedule for a patient who would otherwise be due for more widely spaced check-ups. Details of how we organise this are given on the page devoted to diabetic retinopathy and on that of the Paris 13 ophthalmology practice.

Signs that call for an appointment without delay

• A rapid loss of vision, in one eye or both.
• The sudden appearance of numerous floaters, a sensation of soot, or a shower of black dots, which may indicate a vitreous haemorrhage.
• Distorted straight lines or a spot at the centre of the visual field, suggestive of macular oedema.
• A veil, a shadow to the side, or a curtain coming down, which calls for an urgent examination.

Slightly fluctuating vision for a few weeks after a substantial change in blood sugar is, by contrast, common: it is most often due to changes in the hydration of the lens, which alters optical correction. It does not call for the same urgency, but it is worth mentioning at your check-up.

Frequently asked questions

Cachan practice · Tel. +33 1 45 47 08 11

Scientific sources

• DCCT Research Group. Early worsening of diabetic retinopathy in the Diabetes Control and Complications Trial. Arch Ophthalmol, 1998. PMID 9682700
• Marso SP, et al. Semaglutide and Cardiovascular Outcomes in Patients with Type 2 Diabetes (SUSTAIN-6). N Engl J Med, 2016. PMID 27633186
• Vilsbøll T, et al. Semaglutide, reduction in glycated haemoglobin and the risk of diabetic retinopathy. Diabetes Obes Metab, 2018. PMID 29178519
• Wang T, et al. Retinal Outcomes in Diabetes: Antihyperglycemic Therapy, EWDR, and Perioperative Considerations. Biomedicines 2026. PMID 42193290
• Vujosevic S, et al. New generation agents for glycemic control and diabetic retinopathy progression: what we need to know ?. Acta Diabetol 2025. PMID 40586870
• Goldney J, et al. Incretins and microvascular complications of diabetes: neuropathy, nephropathy, retinopathy and microangiopathy. Diabetologia 2023. PMID 37597048
• Popovic DS, et al. Effect of tirzepatide on the risk of diabetic retinopathy in type 2 diabetes. Diabetes Obes Metab 2024. PMID 38456523
• Buckley AJ, et al. Early worsening of diabetic retinopathy in individuals with type 2 diabetes treated with tirzepatide: a real-world cohort study. Diabetologia 2025. PMID 40637847
• Fadini GP. No evidence that tirzepatide can worsen diabetic retinopathy. Diabetologia 2025. PMID 41051409
• Buckley AJ, et al. No evidence that tirzepatide can worsen diabetic retinopathy [réponse conjointe aux lettres de Fadini GP et de Zhang N, Wang W, Wan G]. Diabetologia 2025. PMID 41051410
• Zhang N, Wang W, Wan G. Reassessing retinopathy risk with tirzepatide therapy. Diabetologia 2025. PMID 41051411
• Hong AT, et al. Ocular Outcomes with Tirzepatide versus Glucagon-like Peptide-1 Receptor Agonists in Type 2 Diabetes. Ophthalmol Retina 2026. PMID 41655764
• Polsky S, et al. Incident Diabetic Retinopathy after Adjunctive Tirzepatide Treatment for 1 Year in Adults with Type 1 Diabetes. Diabetes Technol Ther 2026. PMID 42452840
• Torres-Villaros H, et al. Dépistage de la rétinopathie diabétique chez les plus de 70 ans, réseau OPHDIAT. Diabetes Metab, 2026. PMID 41547437
• Chamard C, et al. Incidence à dix ans de la rétinopathie diabétique et intervalles de dépistage sûrs, réseau OPHDIAT. Br J Ophthalmol, 2021. PMID 32522790
• Schulze-Döbold C, et al. Ophdiat: cinq ans d’expérience d’un réseau de dépistage par télémédecine en Île-de-France. Diabetes Metab, 2012. PMID 22765981

This article is intended for information purposes and does not replace a medical consultation. Indications, expected benefits and risks depend on each individual situation. For any question concerning your vision, consult an ophthalmologist.