Assessing the risk of diabetic retinopathy progression with GLP-1 receptor agonists: a systematic review and meta-analysis Authors: Qian Yang, Waseem Hassan, Hammad Ahmed, Haiyan Zheng Published: 2026 ABSTRACT BackgroundDiabetic retinopathy (DR) is a significant cause of vision impairment in patients with diabetes. The impact of glucagon-like peptide-1 receptor agonists (GLP-1RAs) on the progression of DR remains unclear. This systematic review and meta-analysis aimed to evaluate the association between GLP-1RA therapy and DR progression.MethodsA comprehensive search of electronic databases identified 161 records, of which 11 studies met the inclusion criteria. Included studies comprised randomized controlled trials (RCTs) and retrospective cohort studies conducted in North America, Europe, and East Asia, with sample sizes ranging from 137 to 9,463 participants. The primary GLP-1RAs analyzed were albiglutide, liraglutide, semaglutide, exenatide, and dulaglutide, compared to either placebo or SGLT-2 inhibitors. The primary outcome was DR progression, assessed through pooled relative risk (RR) estimates using a random-effects model. Subgroup analyses were conducted by GLP-1RA type and geographical location. Sensitivity analyses and assessment of publication bias (via funnel plot, Egger’s test, and Begg’s test) were performed. Risk of bias was evaluated using the Cochrane tool.ResultsThe meta-analysis indicated no statistically significant association between GLP-1RA use and DR progression (pooled RR = 1.07, 95% confidence interval [CI]: 0.90–1.28). Subgroup analyses by GLP-1RA type and geographical location also showed no significant differences. However, semaglutide demonstrated the highest RR for DR progression in two studies (RR = 1.76 [1.11–2.79] and RR = 6.41 [0.67–61.46]), a unique finding contrasting with the overall results. Sensitivity analyses confirmed the robustness of the findings, and no publication bias was detected (P > 0.05 for Egger’s and Begg’s tests). Risk of bias analysis revealed a low risk across most domains.ConclusionThis meta-analysis suggests that GLP-1RA therapy is not significantly associated with an increased risk of DR progression. However, heterogeneity in outcomes, particularly with semaglutide, warrants further investigation.Trial registrationPROSPERO CRD420251007882.Clinical trial numberNot applicable.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12886-026-05036-9. Introduction Diabetic retinopathy (DR) remains a leading cause of vision impairment and blindness among individuals with diabetes mellitus worldwide [1]. The pathophysiology of DR involves chronic hyperglycemia-induced microvascular damage, inflammation, and oxidative stress, which lead to progressive retinal deterioration [2]. With the increasing prevalence of diabetes, effective therapeutic strategies to manage both systemic glycemic control and microvascular complications have become a major area of research interest [3]. Among these therapeutic agents, glucagon-like peptide-1 receptor agonists (GLP-1 RAs) have gained attention for their cardiometabolic benefits, yet concerns have been raised regarding their potential impact on DR progression [1, 2]. GLP-1 RAs are widely prescribed for the management of type 2 diabetes (T2DM) due to their ability to enhance insulin secretion, suppress glucagon release, and promote weight loss [4]. In addition to their glycemic effects, emerging evidence suggests that GLP-1 RAs exhibit anti-inflammatory and neuroprotective properties, which may confer benefits in preventing microvascular complications, including DR [5]. However, contradictory findings from clinical trials and observational studies have led to ongoing debate about the safety of GLP-1 RAs concerning DR progression [1]. Previous meta-analyses have produced mixed findings regarding the association of GLP-1 receptor agonists with diabetic retinopathy outcomes. Some analyses reported an increased risk of rapidly worsening retinopathy in major cardiovascular outcome trials, while others found no significant association overall or when stratified by study design [6–10]. The landmark SUSTAIN-6 trial first raised concerns by reporting an increased risk of DR-related complications in patients receiving semaglutide, particularly among those with pre-existing retinopathy and rapid glycemic improvement [4]. Since then, multiple studies have investigated this association, with some confirming an increased risk [11], while others found no significant relationship or even potential protective effects [3]. Given these discrepancies, it is crucial to systematically assess the collective evidence to determine whether GLP-1 RAs influence DR progression. This systematic review and meta-analysis integrate findings from eleven studies examining the association between GLP-1 RAs and DR progression in patients with T2DM. While two studies reported a significant increased risk, the overall meta-analysis demonstrated no statistically significant relationship between GLP-1 RA use and DR progression. These findings contribute to addressing the existing research question regarding the safety of GLP-1 RAs in the context of diabetic microvascular complications. This review aims to provide a comprehensive synthesis of the current literature to aid clinicians and researchers in understanding the implications of GLP-1 RA therapy on diabetic retinopathy. By critically analyzing the available evidence, we seek to clarify whether the reported concerns regarding DR progression are supported by robust clinical data or if they stem from confounding factors such as glycemic variability and pre-existing retinal disease. The findings of this meta-analysis will help inform clinical decision-making regarding the use of GLP-1 RAs in patients at risk of DR, ensuring that therapeutic strategies balance optimal glycemic control with microvascular safety considerations. Methods Design The systematic review was conducted and reported in accordance with 2020 Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines (PRISMA2020); Supplementary Fig. 1 [12]. The study protocol has been registered in the International Prospective Register of Systematic Reviews (PROSPERO) under the registration number CRD420251007882. Search strategy We performed a literature search on Pubmed and the Cochrane Library from their inceptions to February 2026. The search was conducted using keywords, Medical Subject Headings (MeSH) terms, as well as related text words. The detailed search strategy for Pubmed and Cochrane Library is presented in Supplementary Fig. 2. Similar search combinations were used for particular databases. The reference lists of all relevant articles were reviewed to identify potential missed studies. Study selection and inclusion/exclusion criteria This meta-analysis was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. A systematic search of PubMed and the Cochrane Central Register of Controlled Trials (CENTRAL) was performed to identify relevant studies assessing the risk of DR progression in patients receiving GLP-1 RAs. Studies were included if they met the following criteria: they were randomized control trials (RCTs) or clinical trials, including retrospective cohort studies, evaluating the impact of GLP-1 RAs on DR progression. Only studies involving patients diagnosed with T2DM were considered. There were no restrictions on the type, dose, or duration of GLP-1 RA treatment. Eligible studies included those comparing GLP-1 RAs with either placebo or other anti-diabetic agents. Additionally, studies were required to report DR initiation or progression as an outcome at the end of the study duration. Articles published in English were included, and no restrictions were placed on the gender of participants. Studies were excluded if they involved patients with type 1 diabetes mellitus (T1DM), were review articles, animal studies, or in-vitro studies, or if they lacked quantifiable effect estimates (HR, OR, or RR) for DR progression. Furthermore, studies including participants less than 18 years of age were excluded. The study selection process was conducted independently by two reviewers, with any discrepancies resolved through discussion or consultation with a third reviewer. Eligible studies meeting the inclusion criteria were included in the final quantitative synthesis for meta-analysis. The PICO criteria are provided in Table 1. The study characteristics are detailed in Table 2. Table 1PICOCriteriaDescriptionParticipant(s)Type 2 Diabetes Mellitus, Diabetic RetinopathyIntervention(s)GLP-1 Receptor AgonistsComparison(s)Placebo, Other Antidiabetic DrugsOutcome(s)Progression of Diabetic Retinopathy, Vision-related OutcomesStudy designRCTs and clinical trials Table 1PICOCriteriaDescriptionParticipant(s)Type 2 Diabetes Mellitus, Diabetic RetinopathyIntervention(s)GLP-1 Receptor AgonistsComparison(s)Placebo, Other Antidiabetic DrugsOutcome(s)Progression of Diabetic Retinopathy, Vision-related OutcomesStudy designRCTs and clinical trials PICO Data extraction Relevant data were systematically extracted from each eligible study using a standardized data collection form. The extracted information included study identification details (Study ID, year of publication, and country), study design (RCTs or retrospective cohort study), sample size with group distribution, mean age of participants, and gender distribution. Information regarding the intervention was collected, including the specific GLP-1 receptor agonist used, its drug class, dosage, and the number of participants who discontinued treatment (drop-outs). Additionally, details on DR assessment were recorded, including the method of DR diagnosis, the definition of DR progression used in each study, the duration of T2DM in participants, and follow-up duration. Study characteristics such as trial duration, concomitant conditions, and adjusted risk estimates were also noted. For effect size estimation, hazard ratios (HRs), odds ratios (ORs), and risk ratios (RRs) were extracted from the studies. To maintain consistency across studies, HRs and ORs were converted to RRs using an established formula [13]. The final meta-analysis results were presented in terms of RR to facilitate a standardized interpretation of the association between GLP-1 receptor agonists and DR progression. A random effects model was used to account for potential heterogeneity among studies [14, 15]. When data were unavailable, authors were contacted, and studies were considered irretrievable if no answer was received after three attempts. Two reviewers (QY and WH) independently performed literature screening, data extraction, and risk of bias assessment using pre-established criteria. Extracted data included study characteristics, patient demographics, intervention details, and outcomes related to diabetic retinopathy progression. Any discrepancies between the two reviewers were resolved through discussion [16], and if consensus was not reached, a third reviewer (HA or HZ) was consulted to make the final decision. Statistical analysis and data interpretation were conducted collaboratively by HA and HZ. Risk of bias assessment Using the RevMan Web Risk of Bias Assessment tool, two independent reviewers evaluated each study for bias. Sequence generation, allocation concealment, blinding, outcome data, and reporting were among the assessment domains [17]. Trials were classified as "unclear risk" if insufficient information was provided to identify the danger, "low risk" if methodological problems were thought to be insignificant, and "high risk" if they had flaws that could have affected the results. Disputes were settled by agreement. This tool aligns with the Cochrane Risk of Bias 2 (RoB 2) framework for randomized controlled trials. Publication bias and sensitivity analysis Publication bias and sensitivity analysis were assessed using RevMan Web tool. P<0.05 was regarded as evidence of small-study effects. Publication bias was quantitatively tested using the Egger's and Begg's tests and visually examined using funnel plots. Sensitivity tests were carried out, whereby each trial was eliminated from the meta-analysis and the effect size was recalculated using the remaining trials, in order to ascertain whether any one trial had a particularly significant impact on the overall results. Additionally, sensitivity assessments were conducted using leave-one-out technique [18] in Open metanalyst (version released on May 26, 2014) software. Certainty of evidence assessment The certainty of evidence was evaluated using the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) approach, which systematically assesses the quality of evidence across key domains [19]. The GRADE methodology considers five domains for downgrading the certainty of evidence: risk of bias, inconsistency, indirectness, imprecision, and publication bias. Studies were critically appraised for methodological rigor, including randomization procedures, allocation concealment, blinding, and completeness of outcome data to assess the risk of bias [20]. Inconsistency was examined by assessing heterogeneity across studies using statistical measures such as the I² statistic and Cochran’s Q test. Indirectness was considered when study populations, interventions, or outcomes did not directly align with the research question. Imprecision was evaluated based on confidence intervals and optimal information size, ensuring adequate statistical power. Lastly, publication bias was assessed using visual inspection of funnel plots and statistical tests such as Egger’s regression and Begg’s rank correlation test. The certainty of evidence was categorized as high, moderate, low, or very low, following the standard GRADE framework. RCTs initially started as high-certainty evidence but could be downgraded based on the five aforementioned domains. Conversely, observational studies could be upgraded if they demonstrated a large effect size, a dose-response relationship, or if all plausible biases were likely to underestimate the effect [21]. The GRADE assessments were performed independently by two reviewers, with discrepancies resolved through discussion or consultation with a third reviewer. Statistical analysis We assessed heterogeneity among the included studies using the I² statistic, with I² > 50% indicating substantial heterogeneity. A random-effects model was used for primary analysis to account for between-study variability, while a fixed-effects model was considered when heterogeneity was low (I² < 50%). Pooled effect sizes were reported as RRs with 95% Confidence Intervals (CIs). To evaluate result stability, we conducted a sensitivity analysis, sequentially omitting each study to assess its impact on the overall estimate. Publication bias was assessed using Begg’s and Egger’s tests, provided that at least10 studies were available. AP-value < 0.05 was considered statistically significant. Statistical analyses were conducted using RevMan Web [22]. Results Study selection The initial literature search across databases identified 98 records, with an additional 63 records retrieved from the Cochrane Library, resulting in a total of 161 records. After removing duplicates (n = 14), 147 unique records remained for screening. Titles and abstracts of the 147 unique records were independently screened for relevance by two reviewers. A total of 77 records were excluded because they were observational studies, non-human studies, reviews, meta-analyses, systematic reviews, conference abstracts, non-English publications, or incomplete trials. Following abstract screening, 70 full-text articles were assessed for eligibility. Of these, 16 articles were excluded as they were commentaries, Cochrane reviews, or study protocols. Additionally, 3 articles were excluded because full texts could not be retrieved, while 10 studies were excluded because their methodologies did not meet the predefined eligibility criteria. Furthermore, 11 articles were excluded as they did not evaluate retinopathy as a direct outcome of GLP-1 receptor agonist therapy, and another 11 articles were removed because they did not include a placebo-controlled comparison. Ultimately, 11 studies fulfilled all eligibility criteria and were included in quantitative meta-analysis. The selection process is detailed in the PRISMA flow diagram (Fig. 1). Characteristics of the included studies A total of eleven studies were included in this systematic review and meta-analysis, encompassing a range of study designs, sample sizes, and durations. These studies varied in geographical locations, with trials conducted in North America, Europe, and East Asia, including multi-country trials. The included studies primarily investigated GLP-1 receptor agonists in patients with T2DM and assessed their impact on DR progression. The included studies comprised RCTs and retrospective cohort studies. The sample sizes ranged from 137 to 9,463 participants. The largest study involved 9,463 participants [23] across multiple countries; whereas the smallest study [24] included 137 participants. The mean age of participants across studies ranged from 55 to 69.7 years. Most studies had a higher proportion of male participants than female participants. The gender distribution varied significantly, with one study [25] not specifying gender breakdowns, while others [4, 23, 24, 26–32] had a predominant male representation. The studies primarily focused on GLP-1 receptor agonists, including albiglutide, liraglutide, semaglutide, exenatide, and dulaglutide comparing results with placebo and SGLT-2 inhibitors such as empagliflozin, canagliflozin, and dapagliflozin. The included studies investigated various GLP-1 receptor agonists, with semaglutide administered in doses ranging from 0.1 mg to 14 mg (weekly or daily) [4, 25, 27, 29, 31, 32], liraglutide at 0.9–1.8 mg daily, and exenatide at 5–10 µg [26, 28–30]. Albiglutide (30–50 mg weekly) and dulaglutide (0.75 mg) [23, 25, 31]. The dropout rates varied across studies, with some studies reporting high discontinuation rates. Hernandez et al. reported the highest number of discontinuations, with 1,161 participants dropping out from the albiglutide group and 1,318 from the placebo group [23]. Conversely, Yabe et al. reported a lower dropout rate, with only 10 participants (2%) discontinuing treatment [31].The duration of T2DM among study participants ranged from 2.6 years [29] to over 20 years [23]. Follow-up durations varied, with some studies [25] spanning up to 11 years, while others [29] had shorter follow-up periods of 17 weeks. Table 2 summarizes the main characteristics of the included studies [4, 23–33]. Table 2Summarizes the main characteristics of the included studies [4, 23–33]Study ID, yearCountryStudy typeSample size (Group distribution)Mean age (years)Gender DistributionDrug class/Drug nameDose of the drugDrop-out numberDR diagnosisDefinition of DR progressionT2DM durationFollow up Duration of TrialConcommitant stateAdjusted Risk Ratio (RR)Adrian F Hernandez et al. 2018Multi-countryA double-blind, randomised placebo-controlledtrial (HARMONY))Total: 9463; 4731 albiglutideand 4732 patients to receive placeboAlbiglutide: 64·1 (8·7); Placebo: 64·2 (8·7)Females: 2894; males: 6569Albiglutide (GLP-1 receptor agonist)Subcutaneous injection of albiglutide (30–50 mg, based on glycaemic response and tolerability) weekly1161 discontinued in albiglutide; 1318 discontinued in placeboNot specefiedNot specefied< 10 years: 3355 ≥10 years to <20 years: 3925 ≥20 years: 2167 Medianfollow-up of at least 1·5 years 4.7 years.T2DM0·88 (0·65–1·18)Bénédicte Gaborit et al. 2020FranceRandomized open label(NCT02671864) (AngioSafe-1)315462 (Range: 55–69)Males: 1819 (57.8); Females: 1335 (42.2)GLP-1RA (Liraglutide)Liraglutide (1.2 mg) dailysubcutaneous injection194Fundus photographyTransition from non-severe/no DR to severe DRNot specified3 years2 yearsAdults diagnosed with T2D1.10 (0.91–1.35)Bernard Zinman et al. 2019CanadaRandomised,placebo-controlled trial (SUSTAIN 9)302 57.0 yearsMales: 176; Females: 126Semaglutide (GLP-1 receptor agonists)Semalglutide: 1mg16 participantsFundus photography or dilatedfundoscopyNot specefied9.7 yearsFollow-up for 5 weeks30 weeksT2DM0.38 (0.10 to 1.45)Daisuke Yabe et al. 2019JapanRandomized, double-blind, placebo-controlled trial (PIONEER-10)Total: 458; oralsemaglutide 3 mg (n=131), 7 mg (n=132), or 14 mg (n=130),Mean age: 58341 males; 117 femalesSemaglutide and Dulaglutide (GLP-1 receptor agonists)Semaglutide 3 mg, 7 mg and 14 mg; Placebo: Dulaglutide 0.75 mg10 patients (2%).Not specefiedNot specefiedMean duration ofdiabetes was 9·4 (6·3) years52 weeksTreatment duration was 52 weeksT2DM 1.17 (0.29, 4.67)Julia H. Joo et al. 2024USARetrospective cohort studyTotal: 981; GLP-1RA group: 692, SGLT-2I 289 groupGLP-1RA group: 66SGLT-2 group: 65Not specifiedGLP-1RA (semaglutide, dulaglutide, and exenatide) or SGLT-2I(empagliflozin, canagliflozin, and dapagliflozin) from January2012 through February 2023 were included in theNot specefiedNot specefiedDiabetic retinopathy (DR)Worsening of DR defined by ICD-10 codes or indication for procedures; Development of macular edema usingICD-10 codes; Development of vitreous hemorrhage or tractional retinal detachment using ICD-10 codesNot specefied1.54 years (GLP-1RA), 1.38 years (SGLT-2I)2012–2023 (11 years, Retrospective Study)92% of the included patients were having T2DM0.30 (0.11–0.97)Mansoor Husain et al. 2019CanadaRandomized, double-blind, placebo-controlled trial (PIONEER-6)Total: 3183; Oral Semaglutide(N = 1591)Placebo(N = 1592)Mean age: 66 yearsFemales: 1007; Males: 2176Semaglutide (GLP-1 receptor agonists)Semaglutide 14 mg288Scheduled ophthalmic examinations, including fundus photography or dilated fundoscopy.Not specefiedMean duration of diabetes 14.9±8.5 yearsFollow-up appointments occurred every 6 to7 weeks in person or by telephone. (median follow-up period of 15.9 months)Till 122 adjudication-confirmed first major adverse cardiovascular events (MACEs) occurredT2DM1.11 (0.92–1.34)Michael A. Nauck et al. 2015Multi-countryRandomized, double-blind phase 2 trial (NCT00691273) (DIRECT)Total: 415The mean age of patientswas 55 years270 males; females: 145Semaglutide; Liraglutide (GLP-1 receptor agonists)Semaglutide once weeklywithout dose escalation (0.1–0.8 mg); liraglutide once daily (1.2 or 1.8 mg)115 patientsNot specefiedNot specefiedThe mean duration of diabetes was2.6 6 3.1 years, with marked variation,ranging from 0.2 to 30.7 years.The 12-week treatment period was followed by a5-week follow-up period and a follow-upvisit (week 17).17 weeksT2DM6.41 (0.67 to 61.59)Nobuya Inagak et al. 2011JapanDouble-blind, randomized controlledtrial; open-label extension study (NCT00577824)Total: 137; Exenatide group: 109; Placebo: 28Average age: 57.75 years.Men: 121; Women: 21Exenatide (GLP-1 receptor agonist)Exenatide: 5 µg and 10µg150 out of 24 weektrial enrolled in the 28 weeks extension phase of the trialNot specefiedNoted as side effect of Exenatide 10 µg12.05±6.65 yearsFollow-up assessments were made every 4 weeks.28-weekT2DM patients who had suboptimal glycemic control0.98(0.042-23.01)Søren Gullaksen et al. 2023DenmarkA double blind, placebo-controlled, randomised clinical trial EudraCT 2019- 000781-38)Total: 120; Placebo: 30 Semaglutide: 30 Empagliflozin: 30 Combination: 30Placebo: 66.7 ±6.8 Semaglutide: 69.7 ±6.6 Empagliflozin: 70.4 ±6.6 Combination: 69.5 ±6.8Males (Placebo group): 24 (80%), females 06 (20%); Males (Semaglutide group): 26 (87%), females: 04 (13%), Males (Empagliflozin group) 22 (73%), females 08 (27%); Combination group: Males 25 (83%), females 05 (17%)Empagliflozin (SGLT2 inhibitors); Semaglutide (GLP-1 receptor agonists)Empagliflozin 10 mg once-daily and once-weekly subcutaneous SemaglutideNot mentionedFundus photographyCentral retinal thicknessAvergae of all four groups: 10.125 yearsNot specified32 weeksAdults diagnosed with T2DM.Steven P. Marso et al. 2016Multi-countryNCT01720446 (SUSTAIN-6)Total 3297; (Semaglutide group: 1648; Placebo group: 1649)Semaglutide: 64.65±7.2; Placebo: 64.6±7.5Males (Semaglutide group): 1013 (61.45%); Females (Semaglutide group): 635 (38.5%); Males (Placebo group): 989 (60%); Females: 660 (40%)Semaglutide (GLP-1 receptor agonists)Semaglutide: 0.5mg and 1mgNot specifiedScheduled Ophthalmic Assessments; Adverse Event ReportingVitreous hemorrhage, blindness,or conditions requiring treatment with an intravitreal agent or photocoagulationMean duration: 13.9 years5-week follow-up104 weeks (2 years)T2DM who were at high risk for cardiovascular events.1.76 (1.11–2.78)Steven P. Marso et al. 2016USARandomized double-blind trial (LEADER)Total: 9340 (Liraglutide group = 4668); (Placebo group = 4672) 64.3 (Range: 57.1 and 71.5)Females: 3337; Males: 6003Liraglutide (GLP-1 receptor agonists)Maximum tolerated dose) of Liraglutide (1.8mg)784Scheduled Ophthalmic Assessments; Adverse Event ReportingVitreous hemorrhage, blindness,or conditions requiring treatment with an intravitreal agent or photocoagulationNot specefiedMedian follow-up was 3.8 years3.8 yearsT2DM + High cardiovascular risk1.15 (0.87–1.52)Yutaka Seino et al. 2016JapanRandomized, double-blind, parallel-group trial (LIRA-ADD2INSULIN JAPAN)Total 257: (n = 127) or placebo (n = 130)Mean age 60.5 years143 males and 114 femalesLiraglutide (GLP-1 receptor agonist)Liraglutide 0.9 mg/day22Not specefiedNot specefiedMean diabetes duration 14.5 years 36 weeks36 weeksT2DM0.54 (0.26–1.13) Table 2Summarizes the main characteristics of the included studies [4, 23–33]Study ID, yearCountryStudy typeSample size (Group distribution)Mean age (years)Gender DistributionDrug class/Drug nameDose of the drugDrop-out numberDR diagnosisDefinition of DR progressionT2DM durationFollow up Duration of TrialConcommitant stateAdjusted Risk Ratio (RR)Adrian F Hernandez et al. 2018Multi-countryA double-blind, randomised placebo-controlledtrial (HARMONY))Total: 9463; 4731 albiglutideand 4732 patients to receive placeboAlbiglutide: 64·1 (8·7); Placebo: 64·2 (8·7)Females: 2894; males: 6569Albiglutide (GLP-1 receptor agonist)Subcutaneous injection of albiglutide (30–50 mg, based on glycaemic response and tolerability) weekly1161 discontinued in albiglutide; 1318 discontinued in placeboNot specefiedNot specefied< 10 years: 3355 ≥10 years to <20 years: 3925 ≥20 years: 2167 Medianfollow-up of at least 1·5 years 4.7 years.T2DM0·88 (0·65–1·18)Bénédicte Gaborit et al. 2020FranceRandomized open label(NCT02671864) (AngioSafe-1)315462 (Range: 55–69)Males: 1819 (57.8); Females: 1335 (42.2)GLP-1RA (Liraglutide)Liraglutide (1.2 mg) dailysubcutaneous injection194Fundus photographyTransition from non-severe/no DR to severe DRNot specified3 years2 yearsAdults diagnosed with T2D1.10 (0.91–1.35)Bernard Zinman et al. 2019CanadaRandomised,placebo-controlled trial (SUSTAIN 9)302 57.0 yearsMales: 176; Females: 126Semaglutide (GLP-1 receptor agonists)Semalglutide: 1mg16 participantsFundus photography or dilatedfundoscopyNot specefied9.7 yearsFollow-up for 5 weeks30 weeksT2DM0.38 (0.10 to 1.45)Daisuke Yabe et al. 2019JapanRandomized, double-blind, placebo-controlled trial (PIONEER-10)Total: 458; oralsemaglutide 3 mg (n=131), 7 mg (n=132), or 14 mg (n=130),Mean age: 58341 males; 117 femalesSemaglutide and Dulaglutide (GLP-1 receptor agonists)Semaglutide 3 mg, 7 mg and 14 mg; Placebo: Dulaglutide 0.75 mg10 patients (2%).Not specefiedNot specefiedMean duration ofdiabetes was 9·4 (6·3) years52 weeksTreatment duration was 52 weeksT2DM 1.17 (0.29, 4.67)Julia H. Joo et al. 2024USARetrospective cohort studyTotal: 981; GLP-1RA group: 692, SGLT-2I 289 groupGLP-1RA group: 66SGLT-2 group: 65Not specifiedGLP-1RA (semaglutide, dulaglutide, and exenatide) or SGLT-2I(empagliflozin, canagliflozin, and dapagliflozin) from January2012 through February 2023 were included in theNot specefiedNot specefiedDiabetic retinopathy (DR)Worsening of DR defined by ICD-10 codes or indication for procedures; Development of macular edema usingICD-10 codes; Development of vitreous hemorrhage or tractional retinal detachment using ICD-10 codesNot specefied1.54 years (GLP-1RA), 1.38 years (SGLT-2I)2012–2023 (11 years, Retrospective Study)92% of the included patients were having T2DM0.30 (0.11–0.97)Mansoor Husain et al. 2019CanadaRandomized, double-blind, placebo-controlled trial (PIONEER-6)Total: 3183; Oral Semaglutide(N = 1591)Placebo(N = 1592)Mean age: 66 yearsFemales: 1007; Males: 2176Semaglutide (GLP-1 receptor agonists)Semaglutide 14 mg288Scheduled ophthalmic examinations, including fundus photography or dilated fundoscopy.Not specefiedMean duration of diabetes 14.9±8.5 yearsFollow-up appointments occurred every 6 to7 weeks in person or by telephone. (median follow-up period of 15.9 months)Till 122 adjudication-confirmed first major adverse cardiovascular events (MACEs) occurredT2DM1.11 (0.92–1.34)Michael A. Nauck et al. 2015Multi-countryRandomized, double-blind phase 2 trial (NCT00691273) (DIRECT)Total: 415The mean age of patientswas 55 years270 males; females: 145Semaglutide; Liraglutide (GLP-1 receptor agonists)Semaglutide once weeklywithout dose escalation (0.1–0.8 mg); liraglutide once daily (1.2 or 1.8 mg)115 patientsNot specefiedNot specefiedThe mean duration of diabetes was2.6 6 3.1 years, with marked variation,ranging from 0.2 to 30.7 years.The 12-week treatment period was followed by a5-week follow-up period and a follow-upvisit (week 17).17 weeksT2DM6.41 (0.67 to 61.59)Nobuya Inagak et al. 2011JapanDouble-blind, randomized controlledtrial; open-label extension study (NCT00577824)Total: 137; Exenatide group: 109; Placebo: 28Average age: 57.75 years.Men: 121; Women: 21Exenatide (GLP-1 receptor agonist)Exenatide: 5 µg and 10µg150 out of 24 weektrial enrolled in the 28 weeks extension phase of the trialNot specefiedNoted as side effect of Exenatide 10 µg12.05±6.65 yearsFollow-up assessments were made every 4 weeks.28-weekT2DM patients who had suboptimal glycemic control0.98(0.042-23.01)Søren Gullaksen et al. 2023DenmarkA double blind, placebo-controlled, randomised clinical trial EudraCT 2019- 000781-38)Total: 120; Placebo: 30 Semaglutide: 30 Empagliflozin: 30 Combination: 30Placebo: 66.7 ±6.8 Semaglutide: 69.7 ±6.6 Empagliflozin: 70.4 ±6.6 Combination: 69.5 ±6.8Males (Placebo group): 24 (80%), females 06 (20%); Males (Semaglutide group): 26 (87%), females: 04 (13%), Males (Empagliflozin group) 22 (73%), females 08 (27%); Combination group: Males 25 (83%), females 05 (17%)Empagliflozin (SGLT2 inhibitors); Semaglutide (GLP-1 receptor agonists)Empagliflozin 10 mg once-daily and once-weekly subcutaneous SemaglutideNot mentionedFundus photographyCentral retinal thicknessAvergae of all four groups: 10.125 yearsNot specified32 weeksAdults diagnosed with T2DM.Steven P. Marso et al. 2016Multi-countryNCT01720446 (SUSTAIN-6)Total 3297; (Semaglutide group: 1648; Placebo group: 1649)Semaglutide: 64.65±7.2; Placebo: 64.6±7.5Males (Semaglutide group): 1013 (61.45%); Females (Semaglutide group): 635 (38.5%); Males (Placebo group): 989 (60%); Females: 660 (40%)Semaglutide (GLP-1 receptor agonists)Semaglutide: 0.5mg and 1mgNot specifiedScheduled Ophthalmic Assessments; Adverse Event ReportingVitreous hemorrhage, blindness,or conditions requiring treatment with an intravitreal agent or photocoagulationMean duration: 13.9 years5-week follow-up104 weeks (2 years)T2DM who were at high risk for cardiovascular events.1.76 (1.11–2.78)Steven P. Marso et al. 2016USARandomized double-blind trial (LEADER)Total: 9340 (Liraglutide group = 4668); (Placebo group = 4672) 64.3 (Range: 57.1 and 71.5)Females: 3337; Males: 6003Liraglutide (GLP-1 receptor agonists)Maximum tolerated dose) of Liraglutide (1.8mg)784Scheduled Ophthalmic Assessments; Adverse Event ReportingVitreous hemorrhage, blindness,or conditions requiring treatment with an intravitreal agent or photocoagulationNot specefiedMedian follow-up was 3.8 years3.8 yearsT2DM + High cardiovascular risk1.15 (0.87–1.52)Yutaka Seino et al. 2016JapanRandomized, double-blind, parallel-group trial (LIRA-ADD2INSULIN JAPAN)Total 257: (n = 127) or placebo (n = 130)Mean age 60.5 years143 males and 114 femalesLiraglutide (GLP-1 receptor agonist)Liraglutide 0.9 mg/day22Not specefiedNot specefiedMean diabetes duration 14.5 years 36 weeks36 weeksT2DM0.54 (0.26–1.13) Summarizes the main characteristics of the included studies [4, 23–33] Fig. 1PRISMA flow chart. From: Moher D, Liberati A, Tetzlaff J, Altman DG, The PRISMA Group (2009). Preferred Reporting Items for Systematic Reviews and Meta-Analyses: The PRISMA Statement. PLoS Med 6(6): e1000097. doi:10.1371/journal.pmed1000097. For more information, visit HYPERLINK "http://www.consort-statement.org/" www.prisma-statement.org PRISMA flow chart. From: Moher D, Liberati A, Tetzlaff J, Altman DG, The PRISMA Group (2009). Preferred Reporting Items for Systematic Reviews and Meta-Analyses: The PRISMA Statement. PLoS Med 6(6): e1000097. doi:10.1371/journal.pmed1000097. For more information, visit HYPERLINK "http://www.consort-statement.org/" www.prisma-statement.org Association between GLP-1 receptor agonists and diabetic retinopathy A meta-analysis was conducted to evaluate the association between GLP-1 receptor agonists and the progression of DR across 11 eligible clinical studies. The overall pooled RR was estimated at 1.07 (95% CI: 0.90–1.28), suggesting no statistically significant association between GLP-1 receptor agonist use and DR progression. Individual studies reported varying results, with some indicating a potential protective effect while others suggested an increased risk. Notably, the study by Hernandez et al. (HARMONY) [23] reported an RR of 0.88 (95% CI: 0.65–1.18), indicating a potential reduction in DR risk with albiglutide use. Similarly, Joo et al. [25] observed a lower RR of 0.30 (95% CI: 0.11–0.97), suggesting a significant protective effect of GLP-1 receptor agonists against DR progression. Conversely, studies such as Marso et al. (SUSTAIN-6) [4] reported an RR of 1.76 (95% CI: 1.11–2.78), indicating a higher risk of DR progression associated with semaglutide. Similarly, Husain et al. (PIONEER-6) observed an RR of 1.11 (95% CI: 0.92–1.34), while Gaborit et al. (2020) (AngioSafe-1) found an RR of 1.10 (95% CI: 0.91–1.35), suggesting a possible but non-significant increased risk. Although statistical heterogeneity was low (I² = 0%; Tau² = 0), clinical heterogeneity remained possible because of differences in study populations, treatment duration, comparator groups, and DR assessment methods. Studies such as Zinman et al. (2019) (SUSTAIN-9) and Seino et al. (2016) (LIRA-ADD2INSULIN JAPAN) reported RRs of 0.38 (95% CI: 0.10–1.45) and 0.54 (95% CI: 0.26–1.13), respectively, indicating a trend towards a lower risk. However, Nauck et al. (2015) (DIRECT) reported an RR of 6.41 (95% CI: 0.67–61.59), suggesting extreme variability likely due to sample size limitations and methodological differences. Overall, the pooled estimate suggests no significant association between GLP-1 receptor agonist use and an increased risk of DR (Fig. 2). Three recent large real-world cohort studies evaluated incretin-based therapies in relation to diabetic retinopathy outcomes. Buckley et al. [34] investigated tirzepatide, a dual GLP-1/GIP receptor agonist, and reported increased odds of new-onset proliferative diabetic retinopathy (RR = 2.15, 95% CI 1.24–3.74) among patients with type 2 diabetes mellitus, while observing decreased odds in participants without baseline retinopathy (R0M0: RR = 0.73, 95% CI 0.62–0.86). Similarly, Shah et al. [35] found a lower risk of diabetic retinopathy progression over 12 months among tirzepatide users (NPDR: RR = 0.864, 95% CI 0.758–0.985; PDR: RR = 0.705, 95% CI 0.564–0.882). In contrast, Ramsey et al. [36] evaluated GLP-1 receptor agonists as a class in a large propensity-matched cohort and reported a modestly increased risk of incident diabetic retinopathy (HR = 1.07, 95% CI 1.03–1.11), with no significant association for progression to proliferative diabetic retinopathy (HR = 1.06, 95% CI 0.97–1.15) or development of diabetic macular edema (HR = 0.98, 95% CI 0.95–1.01). These studies were not included in the quantitative meta-analysis because tirzepatide is a dual GLP-1/GIP receptor agonist rather than a GLP-1 receptor agonist alone and because of differences in outcome ascertainment methods (administrative coding versus standardized ophthalmologic assessment). Therefore, their findings were evaluated qualitatively and summarized in Table 2. Fig. 2Forest plot illustrating the pooled risk estimates for diabetic retinopathy progression associated with GLP-1 receptor agonists Forest plot illustrating the pooled risk estimates for diabetic retinopathy progression associated with GLP-1 receptor agonists Subgroup analysis Subgroup analysis by GLP-1 receptor agonist type A subgroup analysis based on specific GLP-1 receptor agonists (Albiglutide, Semaglutide, Exenatide, and Liraglutide) was performed to explore their differential effects on DR progression. The pooled RR and CI for each agent were as follows: Albiglutide, RR = 0.88 [0.65–1.19]; Semaglutide, RR = 0.88 [0.65–1.19]; Exenatide, RR = 0.99 [0.39–2.51]; and Liraglutide, RR = 0.98 [0.04–22.94]. The overall RR across all GLP-1 receptor agonists was 1.07 [0.90–1.28], suggesting no statistically significant increased or decreased risk of DR (Fig. 3). Notably, two individual studies within the Semaglutide subgroup reported conflicting results. One study showed a markedly elevated RR of 6.41 [0.67–61.46], while another reported an RR of 1.76 [1.11–2.79]. These findings contrast with the overall pooled estimate and suggest the need for further investigation to clarify the potential variation in DR risk associated with Semaglutide treatment. Fig. 3Forest plot of subgroup analysis according to GLP-1 receptor agonist type Forest plot of subgroup analysis according to GLP-1 receptor agonist type Subgroup analysis by geographical location A geographical subgroup analysis was conducted to assess whether location-specific factors influence the relationship between GLP-1 receptor agonist use and DR progression. The RR estimates were as follows: Europe, RR = 1.10 [0.90–1.34]; North America, RR = 0.78 [0.28–2.20]; multi-country studies, RR = 1.41 [0.25–7.92]; and East Asia, RR = 0.65 [0.24–1.74]. The overall RR remained 1.07 [0.90–1.28] (Fig. 4). While no statistically significant association was observed in any region, there was considerable variability between geographic locations. Notably, studies conducted in North America and East Asia suggested a potential protective effect (RR < 1), whereas studies from Europe and multi-country trials showed a tendency toward increased risk (RR > 1). This heterogeneity underscores the importance of accounting for regional differences in future research to better understand the effects of GLP-1 receptor agonists on diabetic retinopathy. Fig. 4Forest plot of subgroup analysis according to geographical region Forest plot of subgroup analysis according to geographical region Publication bias A visual examination of the funnel plots (Supplementary Fig. 3) indicated no apparent asymmetry. This observation was further supported by Egger’s and Begg’s tests, which showed no significant small-study effects for any of the outcomes (P > 0.05). Risk of bias The Cochrane Risk of Bias Tool (Supplementary Figs. 4 and 5) was used to assess individual trials, categorizing them as having a high, low, or unclear risk of bias across different domains. The evaluation considered (A) blinding of outcome assessment (detection bias), (B) incomplete outcome data (attrition bias), and (C) selective reporting (reporting bias). Most studies demonstrated a low risk of bias in all domains, except for three trials, which were classified as having an “unclear risk” in blinding of outcome assessment (detection bias). The data analysis was conducted independently by two researchers, and any discrepancies were resolved through mutual discussion. Sensitivity analyses The systematic exclusion of individual trials did not impact the overall results. Sensitivity analyses using leave-one-out forest plots confirmed that no single study disproportionately influenced the findings. The overall association between GLP-1 receptor agonists and DR remained consistent (Supplementary Fig. 6). GRADE analysis A total of 11 included studies were analyzed to assess the risk of DR progression in patients treated with GLP-1RAs compared to placebo or other treatments. The certainty of evidence ranged from low to moderate due to clinical and methodological variability across studies, subgroup inconsistency (particularly within semaglutide trials) and imprecision. The risk of bias was moderate, as some studies lacked specific DR diagnosis criteria and had high dropout rates. Inconsistencies arose due to variations in DR progression definitions, while imprecision was noted due to wide confidence intervals. Indirectness was low, as studies directly addressed DR progression in T2DM patients treated with GLP-1RA. Possible publication bias was identified, indicating the need for larger, long-term studies. The relative effect varied across studies. While 9 of 11 trials showed no significant effect [23–28, 30–32], two studies—SUSTAIN-6 (RR = 1.76) and DIRECT (RR = 6.41) [4, 29] suggested a potential increased risk, particularly with semaglutide. The absolute effect was mixed, with most studies reporting no significant impact, except for a subset indicating possible risk. Given the low to moderate certainty of evidence, further well-controlled trials are needed to confirm these findings. While the majority of studies do not indicate a significant risk of DR progression with GLP-1RA use, semaglutide may warrant further evaluation because a limited number of studies reported a potential increased risk of DR progression. GRADE analysis is shown in Supplementary Table 1. Discussion The findings of this systematic review and meta-analysis provide a comprehensive evaluation of the association between GLP-1 receptor agonist use and DR progression. Our pooled analysis of 11 eligible studies, comprising RCTs and a retrospective cohort study, revealed no statistically significant association between GLP-1RA use and an increased risk of DR progression (RR = 1.07, 95% CI: 0.90–1.28). This result aligns with recent large-scale cardiovascular outcomes trials (CVOTs) that have assessed GLP-1RA and their potential ocular impacts [37]. Our results show a nuanced relationship between GLP-1RA and DR progression. While the overall pooled estimate suggests no significant risk, individual studies reported divergent outcomes. Notably, the SUSTAIN-6 trial [4] reported a significantly increased risk of DR with semaglutide (RR = 1.76, 95% CI: 1.11–2.78), which contrasts with the HARMONY trial [23] that observed a potential protective effect with albiglutide (RR = 0.88, 95% CI: 0.65–1.18). This variability in individual study results may stem from differences in study design, baseline characteristics, and treatment durations. Although the overall meta-analysis demonstrated low statistical heterogeneity (I² = 0%; Tau² = 0), variability in clinical characteristics and study methodologies across trials remains important for interpretation. Notably, subgroup analysis of semaglutide trials revealed substantial heterogeneity (I² = 77%), suggesting that treatment effects may not be uniform across all GLP-1 receptor agonists. Differences in baseline glycemic control (HbA1c), duration of diabetes, and pre-existing severity of diabetic retinopathy likely contribute to variability in outcomes. Rapid reductions in HbA1c may exacerbate DR in susceptible individuals, particularly those with advanced baseline retinopathy. Additionally, variations in DR assessment and grading systems across studies, including differences in ophthalmologic examination methods or definitions of progression, could further account for variability in results. Understanding these sources is critical for interpreting study findings and highlights the need for standardized DR assessment in future trials. One possible explanation for the elevated DR risk observed in some trials is the “early worsening phenomenon,” a transient exacerbation of DR following rapid glycemic control. This effect is well-documented in the Diabetes Control and Complications Trial (DCCT) [38, 39] and has been suggested to occur with GLP-1RA due to their potent glucose-lowering effects [40]. In contrast, studies with more gradual glycemic improvements, such as HARMONY [23], did not observe this adverse effect, suggesting that the rate of glucose reduction may be a critical factor in DR outcomes. An additional mechanistic consideration is the magnitude and rapidity of HbA1c reduction. The phenomenon of early worsening of diabetic retinopathy following rapid glycemic improvement is well recognized. In SUSTAIN-6, semaglutide was associated with substantial HbA1c reductions (approximately 1.5–1.8%), which may partially explain the increased risk of DR progression observed in that trial. In contrast, studies reporting more modest HbA1c reductions generally did not demonstrate significant DR worsening. These findings suggest that the magnitude of glycemic reduction may act as an effect modifier rather than a direct drug-specific toxicity. However, a formal meta-regression assessing the association between HbA1c reduction and DR risk was not feasible due to inconsistent reporting of HbA1c changes and lack of patient-level data across studies. Future analyses incorporating standardized HbA1c reporting or individual participant data would help clarify this relationship. Recent large cohort studies provide additional context to the findings of our quantitative synthesis. Ramsey et al. [36] reported a modestly increased risk of incident diabetic retinopathy among GLP-1 receptor agonist users but found no significant association with progression to proliferative diabetic retinopathy or diabetic macular edema. In contrast, studies specifically evaluating tirzepatide, a dual GLP-1/GIP receptor agonist, have shown heterogeneous results. Buckley et al. [34] observed increased risk of new-onset proliferative diabetic retinopathy in patients with established disease, while noting reduced incident retinopathy in individuals without baseline retinal involvement. Similarly, Shah et al. [35] demonstrated a lower risk of non-proliferative and proliferative diabetic retinopathy over 12 months of follow-up. Subgroup analyses by GLP-1RA type revealed no significant differences among albiglutide (RR = 0.88, 95% CI: 0.65–1.19), semaglutide (RR = 1.76, 95% CI: 1.11–2.78), exenatide (RR = 0.99, 95% CI: 0.39–2.51), and liraglutide (RR = 0.98, 95% CI: 0.04–22.94). However, the heterogeneity within the semaglutide subgroup warrants further investigation. As demonstrated in prior research, the pharmacokinetic profiles of these agents vary, which may influence retinal vascular outcomes [41]. Geographical subgroup analysis indicated regional variability in the association between GLP-1RA and DR. While studies conducted in North America and East Asia suggested a potential protective effect, European and multi-country trials indicated a non-significant increased risk. This finding highlights the need to consider region-specific factors such as healthcare access, patient monitoring, and genetic predispositions [42]. Emerging molecular evidence highlights that diabetic DR is not merely a microvascular complication of chronic hyperglycemia, but also a disorder with significant inflammatory and immune components. Transcriptomic and proteomic analyses have identified multiple genes and proteins associated with DR that are involved in adaptive immune responses, complement activation, and immune system regulation, underscoring the role of systemic and local inflammation in retinal pathology. Moreover, pathways related to the formation of advanced glycation end products and mitochondrial dysfunction have been linked to the progression of DR, further emphasizing oxidative stress as a key driver of retinal cellular injury [43]. GLP-1RAs have been shown in other contexts to exert anti-inflammatory and antioxidative effects, including modulation of endothelial function and cytokine signaling. Incorporating this mechanistic perspective helps contextualize clinical findings, suggesting that the pleiotropic actions of GLP-1RAs could influence microvascular health and intersect with intrinsic inflammatory pathways implicated in DR pathogenesis. In addition to randomized trial data, emerging real-world evidence has provided further insight into the relationship between GLP-1RA therapy and diabetic retinopathy outcomes. Several large observational cohort studies have reported either neutral associations or only transient increases in DR risk following GLP-1RA initiation [44–47]. For example, a 10-year retrospective cohort study reported lower adjusted hazard ratios for DR and diabetic macular edema among GLP-1RA users compared with non-users, even after accounting for longitudinal glycemic control [44]. Similarly, a nationwide cohort from Sweden showed a significantly lower DR incidence with GLP-1RA treatment [45]. In many of these settings, the observed retinopathy worsening appeared closely linked to baseline DR severity and the magnitude of HbA1c reduction rather than a direct drug-specific retinal toxicity. Compared with RCT populations, real-world cohorts often include patients with broader comorbidity profiles and varying adherence patterns, which may influence retinal outcomes. The consistency between our pooled analysis and most real-world data supports the conclusion that GLP-1RA therapy does not confer a sustained independent risk of DR progression, though careful monitoring remains advisable in high-risk individuals. The clinical implications of these findings are particularly relevant for patients with pre-existing diabetic retinopathy who require intensification of glycemic management. Given the well-established cardiovascular and renal benefits of GLP-1 RA [48], our results support their continued use without routine contraindication based solely on DR concerns. However, individualized risk stratification remains essential. Patients with advanced baseline DR, long-standing diabetes or markedly elevated HbA1c levels may be at higher risk of transient retinopathy worsening, particularly if glycemic control improves rapidly. Although the overall pooled analysis did not demonstrate a statistically significant increase in DR progression, certain trials, particularly those evaluating semaglutide, reported higher point estimates of risk. These findings do not establish causality but suggest that enhanced vigilance may be prudent when initiating specific GLP-1 RA in high-risk individuals. In clinical practice, a baseline ophthalmologic evaluation prior to initiation of GLP-1RA therapy may be considered in high-risk patients. Additionally, closer retinal monitoring during the first 3–6 months of treatment may help identify early worsening phenomena [49]. Where feasible, gradual titration strategies and avoidance of abrupt reductions in HbA1c may further mitigate potential short-term retinal instability. Such an approach allows clinicians to balance the substantial cardiometabolic benefits of GLP-1RA therapy with appropriate microvascular vigilance. The certainty of evidence for DR progression with GLP-1RA therapy was rated as low-to-moderate according to GRADE, reflecting several limitations in the included studies. Imprecision arose from wide confidence intervals in studies such as SUSTAIN-6 (RR = 1.76) and DIRECT (RR = 6.41), which limits the ability to draw definitive conclusions. Inconsistency was driven by variation in DR progression definitions across studies and conflicting results for semaglutide, introducing heterogeneity. Moderate risk of bias due to incomplete DR assessment and high dropout rates further reduces confidence. Clinically, this means that while the majority of studies suggest no significant overall risk, the evidence is not strong enough to exclude the possibility of DR progression in high-risk subgroups, particularly patients with advanced baseline DR or those experiencing rapid HbA1c reductions. Therefore, careful monitoring is recommended, and further adequately powered trials with standardized DR assessment are needed to provide more definitive guidance. The heterogeneity observed emphasizes the importance of individualized risk assessment when prescribing GLP-1RA. Patients with advanced DR or those at high risk of rapid glycemic improvement should be monitored more closely. Further, the transient increase in DR risk underscores the need to adopt a gradual approach to glycemic targets in susceptible individuals [50]. Limitations Despite the strengths of our meta-analysis, several limitations should be acknowledged. First, the included studies varied widely in their design, sample size, and follow-up duration, contributing to clinical and methodological variability across studies. Second, the lack of uniform DR assessment criteria across studies may introduce outcome misclassification. Additionally, several trials had incomplete blinding of outcome assessments, raising concerns regarding detection bias. Finally, our analysis focused on published studies, and the potential for publication bias remains despite negative results indicated by funnel plot analysis. An additional limitation of this meta-analysis is the inability to perform subgroup analyses based on baseline DR severity or rate of HbA1c reduction. Although these factors are biologically plausible effect modifiers particularly given the phenomenon of early worsening associated with rapid glycemic improvement, most included studies did not provide stratified retinal outcomes according to baseline DR stage, nor did they consistently report the magnitude or rate of HbA1c reduction in relation to DR progression. Furthermore, variations in DR classification systems and outcome reporting precluded harmonization across studies. As a result, we were unable to determine whether any potential risk is confined to individuals with advanced baseline disease or those experiencing rapid glycemic lowering. Future prospective studies should incorporate standardized retinal grading and report outcomes stratified by baseline severity and glycemic trajectory to better clarify high-risk subgroups. Future directions Further research is needed to clarify the mechanisms underlying the early worsening phenomenon and to identify patient subgroups most at risk. Large-scale, long-term observational studies with standardized DR assessment and real-world evidence are essential. Additionally, future trials should explore whether specific GLP-1RA differ in their ocular safety profiles. Conclusion In conclusion, the available evidence does not show a statistically significant association between GLP-1 receptor agonists and progression of diabetic retinopathy. Overall heterogeneity across the pooled analyses was low, although some variability was observed in certain drug-specific subgroups, particularly with semaglutide, which should be interpreted with caution. Differences in study design, baseline retinopathy status, and methods of outcome assessment may partly explain these inconsistencies. While the overall findings are generally reassuring, careful monitoring remains appropriate, especially in patients with advanced baseline retinopathy or those undergoing rapid glycemic improvement. Further well-designed prospective studies with standardized retinal assessment and stratified analyses are needed to provide greater clarity. These findings are consistent with previous meta-analyses that have not demonstrated a clear increased risk of diabetic retinopathy with GLP-1RA use but highlight ongoing uncertainties due to heterogeneity in study designs and outcome measures [51]. Electronic Supplementary Material Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2 Supplementary Material 3 Supplementary Material 3 Supplementary Material 4 Supplementary Material 4 Supplementary Material 5 Supplementary Material 5 Supplementary Material 6 Supplementary Material 6