Tirzepatide is a potent glucose-lowering agent and, at higher doses, provides greater HbA1c lowering than GLP-1RAs including dulaglutide and semaglutide [20, 21]. Our observation that tirzepatide use was associated with PDR but not with diabetic macular oedema is consistent with recent large post-marketing studies of GLP-1RA use [22]. Our finding of increased incidence of PDR in tirzepatide-exposed individuals contrasts with the low rates of PDR reported in the SURPASS 1–5 clinical trials, which reported a total of 21 adverse retinal events in tirzepatide-exposed individuals and 19 events in tirzepatide-unexposed individuals [13]. Although diabetes duration was not an inclusion criterion in the SURPASS trials, individuals enrolled in SURPASS-1 were previously untreated, those in SURPASS-2 were treated with metformin alone, and those in SURPASS-3 were treated with metformin with or without an SGLT2i, while only SURPASS-5 and SURPASS-6 included participants using insulin at baseline, suggesting that the average diabetes duration at inclusion in these trials may have been relatively short. The average incidence rate of any retinal adverse event in these studies was approximately 5.6 per 1000 person-years or 0.6%, lower than the present study and also lower than previously reported in type 2 diabetes cohorts [23, 24]. Importantly, individuals with moderate NPDR or higher did not meet the inclusion criteria for the SURPASS trials and consequently our study is the first to explore retinal outcomes after tirzepatide exposure in individuals at higher risk of retinopathy progression. The difference in retinopathy incidence between the present study and SURPASS also highlights the value of using regular retinal screening, rather than event reporting, in future trials of potent glucose-lowering agents.

Although it has been reported that the risk of EWDR is increased by faster or more effective lowering of HbA1c in individuals treated with insulin or a sulfonylurea [14, 25], in our study reduction in HbA1c was not independently associated with increased odds of the PDR endpoint. A post hoc analysis of the SUSTAIN 6 study found that semaglutide exposure and the degree of reduction in HbA1c at week 16 were independently associated with increased risk of new diabetic retinopathy, with the majority of events contributing to this finding being mild or moderate NPDR [26]. Exposure to exenatide is associated with increased risk of diabetic retinopathy, with the majority of events occurring in individuals in whom HbA1c improved on treatment [9]. In contrast, in the LEADER trial, liraglutide exposure was not associated with a significant increase in retinopathy hazard [11]. A potentially important distinction between these trials is that the mean reduction in HbA1c in LEADER was −0.40%, compared with −1.9% at the 0.5 mg/week dose, and −2.5% at the 1.0 mg/week dose in SUSTAIN 6. The mean HbA1c at baseline in SUSTAIN 6 was 8.7% [11]. In the present real-world study, the mean ± SD change in HbA1c in the tirzepatide-exposed group was −4.6 ± 11.2 mmol/mol (−0.4 ± 1.0%), the mean ± SD HbA1c at baseline was 55.2 ± 15.5 mmol/mol (7.2 ± 1.4%), and a large proportion of tirzepatide-exposed individuals were previously treated with semaglutide. The apparent disparity in our findings is therefore potentially explained by the significantly better baseline glycaemic control of our participants as well as previous exposure to semaglutide [13]. Direct beneficial effects of GLP-1 agonism on retinal vasculature have also previously been reported in animal models [27], which could potentially offer another mechanism by which risks of EWDR associated with rapid HbA1c reduction might be mitigated.

Our study endpoints were ascertained from annual digital retinal photography images and were not synchronised with the date of tirzepatide initiation. Consequently, the CIs for time to progression were large, making it difficult to infer from our data how soon, or how often, individuals at higher risk of progression to PDR should be followed up after initiation of tirzepatide. It should be noted that in the SUSTAIN-6 trial, new diabetic retinopathy events among individuals with pre-existing retinopathy were more frequent among semaglutide-treated individuals from week 0 of follow-up, suggesting that the increase in hazard of EWDR associated with GLP-1RAs may begin at initiation [26]. In the present study, new PDR events occurred in 17 of 113 (15%) of tirzepatide-treated individuals with R2M0 or R2M1 at baseline, which by the ETDRS referral criteria and the 2024 NICE guidelines would justify referral to specialist ophthalmology follow-up. The NHS Diabetic Eye Screening Programme has also recently published criteria for high-risk features of R2, based on ETDRS criteria, which would guide Eye Service referral [10, 17, 26, 28].

Study participants had a relatively long mean diabetes duration (<10 years) and the majority of individuals in the treatment group were switched to tirzepatide from a GLP-1RA, potentially both limiting the number of endpoints and reducing the applicability of our findings to GLP-1RA naive patients being initiated on tirzepatide. Although this study is based on real-world data in a representative population of people with type 2 diabetes, enhancing its external validity, the improvements in HbA1c seen after introduction of tirzepatide were modest in comparison with those seen in clinical trials and may not be applicable in a population with less-optimised baseline glycaemic control. The use of 1:1 matching, rather than a higher ratio of tirzepatide-exposed to tirzepatide-unexposed individuals, resulted in a loss of statistical power but conversely would tend to reduce the potential for type 1 error. Due to matching for sex and the use of conditional logistic regression for analysis, it was not possible to evaluate the influence of sex on study outcomes.

Strengths of this study include a large study population, enabling good quality of matching and providing sufficient endpoints for multivariate analysis. Other important strengths include the comprehensive consultant-led in-house retinal screening programme at ICLDC, and a relatively consistent 3 monthly cycle of clinic attendance, which enables frequent clinical and laboratory measures and enhances uptake of screening for complications of diabetes.

Tirzepatide exposure was not associated with progression of retinopathy in individuals with mild NPDR at baseline (R1M0, R1M1) and appeared to reduce the odds of new retinopathy in individuals without retinopathy (R0M0). However, tirzepatide was significantly and independently associated with new onset of PDR, with the majority of events occurring in individuals with mild NPDR with maculopathy (R1M1) or moderate-to-severe NPDR with maculopathy (R1M1 or R2M1). Our findings, in a group of relatively well-controlled individuals with diabetes, not only reinforce the need for additional retinal photography in those already at high risk of PDR who commence treatment with tirzepatide but also provide reassurance that EWDR is rare in individuals with lower pre-treatment risk.