Use of SGLT2 Inhibitors Versus DPP-4 Inhibitors and Age-Related Macular Degeneration in Patients WithType 2 Diabetes: A Multinational Cohort Study

Abstract

Purpose

To compare the impact of sodium–glucose cotransporter 2 (SGLT2) and dipeptidyl peptidase 4 (DPP-4) inhibitors on age-related macular degeneration (AMD) risk among patients with type 2 diabetes mellitus (T2DM).

Methods

This multinational, retrospective cohort study used electronic medical records from healthcare institutions across 21 countries. Adults 50 years or older with T2DM who had a prior prescription of metformin and initiated SGLT2 or DPP-4 inhibitors from 2013 to 2023 were included. The SGLT2 and DPP-4 inhibitor groups were propensity score matched in a 1:1 ratio to balance baseline characteristics and were followed for up to 5 years to observe the occurrence of AMD. Statistical analysis was performed using the Cox proportional hazards model and Kaplan–Meier analysis.

Results

Our final analysis included 20,966 T2DM patients prescribed SGLT2 inhibitors and 20,966 prescribed DPP-4 inhibitors. Compared to the DPP-4 inhibitor group, the SGLT2 inhibitor group was associated with significantly lower risks of AMD (hazard ratio [HR], 0.71; 95% confidence interval [CI], 0.58–0.85) and dry AMD (HR, 0.61; 95% CI, 0.46–0.80) but not wet AMD (HR, 0.74; 95% CI, 0.48–1.16). SGLT2 inhibitors compared with DPP-4 inhibitors were linked to a reduced risk of AMD in the White population, patients prescribed empagliflozin or dapagliflozin, and individuals with glycated hemoglobin < 8.5%, estimated glomerular filtration rate ≥ 60 mL/min/1.73 m², hypertension, or dyslipidemia, regardless of body mass index level.

Conclusions

In patients with T2DM, those prescribed SGLT2 inhibitors may experience lower risks of AMD and dry AMD compared to those prescribed DPP-4 inhibitors.

Type 2 diabetes mellitus (T2DM) is a prevalent chronic disease projected to affect 7079 individuals per 100,000 by 2030.¹ T2DM is associated with an increased risk of ocular complications, including age-related macular degeneration (AMD), a leading cause of irreversible blindness around the world.^2–7 Studies have consistently shown that patients with T2DM face a higher risk of developing AMD,^8–12 likely due to the shared pathophysiology between T2DM and AMD such as accumulation of glycation end products, increased inflammatory responses, excessive oxidative stress, abnormal lipid profiles, adverse microvascular complications, and retinal tissue hypoxia.^11–16

Among AMD subtypes, dry AMD presents limited treatment options with suboptimal efficacy.^5,17 Recent therapeutic advancements have focused on targeting the key mechanisms of dry AMD. Novel agents such as pegcetacoplan and avacincaptad pegol, recently approved by the U.S. Food and Drug Administration (FDA), aim to inhibit the complement cascade and thereby reduce inflammation—a primary driver of retinal pigmented epithelium (RPE) dysfunction and atrophy.^6,18–20 In addition, antioxidative treatments such as lutein and astaxanthin have been employed to mitigate oxidative stress, another critical factor in AMD progression.^21,22 Emerging evidence also suggests that obstruction and reduced perfusion of choriocapillaris, induced by aging and cardiovascular risk factors, contribute to an increased risk of AMD.¹⁴ These multifaceted mechanisms underline the importance of comprehensive treatment approaches addressing both inflammatory and vascular pathways of AMD.

Sodium–glucose cotransporter 2 (SGLT2) inhibitors, initially introduced as glucose-lowering medications, have shown potential benefits beyond their primary use.²³ With recognized cardiovascular and renal benefits, they also exhibit complement inactivation and anti-inflammatory and antioxidative properties, suggesting a possible role in reducing AMD risk in T2DM patients.^24–28 Among second-line glucose-lowering medications, including SGLT2 inhibitors, glucagon-like peptide 1 receptor agonists (GLP-1RAs), and dipeptidyl peptidase 4 (DPP-4) inhibitors, SGLT2 inhibitors and DPP-4 inhibitors are the most commonly prescribed, and their usage is increasing.^29,30 This may be due to their convenient oral administration, unlike GLP-1RAs, which require injection. However, evidence comparing the effects of SGLT2 inhibitors and DPP-4 inhibitors on AMD risk in patients with T2DM remains limited.

To bridge this literature gap, we conducted a multinational retrospective cohort study to evaluate the impact of SGLT2 inhibitors on AMD, including its dry and wet subtypes, compared to DPP-4 inhibitors. SGLT2 and DPP-4 inhibitors are both second-line glucose-lowering medications following metformin.^31,32 This study aimed to identify potential therapeutic agents that could mitigate AMD risk in patients with T2DM who had previously used metformin as a first-line glucose-lowering treatment, addressing the urgent need for more effective treatments in this population.

Methods

Data Source

We utilized the TriNetX Analytics Network as our data source. TriNetX is a multinational database that incorporates de-identified electronic health records from 160 healthcare organizations and over 200 million patients spanning 21 countries, including information on the participants’ age, race, diagnostic codes, medications prescribed, procedures, and laboratory data. We performed the analysis in March 2025 using the Global Collaborative Network, which includes the largest number of healthcare organizations across the globe participating in TriNetX. TriNetX ensures patient privacy by adhering to the U.S. Health Insurance Portability and Accountability Act (HIPAA), which safeguards healthcare data security by providing de-identified electronic health records and concealing the exact event numbers for outcomes involving 10 patients or fewer.

Study Design

Patients diagnosed with T2DM who initiated SGLT2 or DPP-4 inhibitors between 2013 and 2023 were included in our study. To prevent discrepancies in follow-up durations between patients using SGLT2 and DPP-4 inhibitors, enrollment was delayed until after U.S. FDA approval of SGLT2 inhibitors in 2013 (canagliflozin), following the approval of DPP-4 inhibitors in 2006 (sitagliptin). The date each patient was first prescribed SGLT2 or DPP-4 inhibitors was defined as the index date, and participants were followed for up to 5 years to monitor the occurrence of AMD, dry AMD, and wet AMD. Follow-up for each patient ended upon the occurrence of outcome or the end of the study period, whichever came earlier.

Study Population

Our inclusion criteria were as follows: (1) patients ≥ 50 years old; (2) patients coded with a diagnosis of T2DM prior to the index date; (3) patients who initiated SGLT2 inhibitors or DPP-4 inhibitors between 2013 and 2023; (4) participants who had a refill or an active prescription for the medication within 3 to 12 months following the initial prescription; (5) patients with a history of metformin prescription prior to the use of SGLT2 inhibitors or DPP-4 inhibitors, consistent with current guidelines for glucose-lowering drug use in most patients with T2DM³³; and (6) individuals with glycated hemoglobin (HbA1c), body mass index (BMI), and estimated glomerular filtration rate (eGFR) measurements recorded within 1 year before the index date. We limited inclusion to individuals with available HbA1c, BMI, and eGFR data to accurately reflect the updated conditions of each participant and ensure complete propensity score matching. The Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation was used to estimate eGFR.

Our exclusion criteria were as follows: (1) patients with a diagnosis of AMD or macular degeneration before the index date, (2) patients in the SGLT2 inhibitor group who concurrently used DPP-4 inhibitors during the study period, (3) patients in the DPP-4 inhibitor group who concurrently used SGLT2 inhibitors during the study period, and (4) patients prescribed GLP-1RAs during the study observation period.

Statistical Analysis

Propensity score matching was performed by assigning a propensity score to each patient and then matching between the two groups. Cohorts were randomly shuffled before matching. Logistic regression was used to generate propensity scores. Matching was conducted in a 1:1 ratio using the greedy nearest-neighbor matching algorithm without replacement. Caliper width was specified as 0.1 of pooled standard deviations (SDs). A standardized mean difference (SMD) of <0.1 indicated a well-balanced comparison, reflecting minimal differences between groups. Matching variables included age, sex (female), race (White, Black or African American, and Asian), smoking history, clinical measurements (HbA1c, BMI, and eGFR), medications (metformin, sulfonylureas, thiazolidinediones, and insulin), comorbidities (hypertension, ischemic heart disease, cerebrovascular diseases, heart failure, atrial fibrillation and flutter, dyslipidemia, end-stage renal disease, and hypermetropia), different stages of diabetic retinopathy (mild, moderate, severe nonproliferative diabetic retinopathy and proliferative diabetic retinopathy), and the prescription of antioxidants (lutein, zeaxanthin, and astaxanthin). The codes used to determine medication prescriptions and disease diagnoses, including those from the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) and the Anatomical Therapeutic Chemical (ATC) classification system, are listed in Supplementary Table S1.

Our primary analysis examined the risks of AMD, dry AMD, and wet AMD between patients prescribed SGLT2 inhibitors and DPP-4 inhibitors. We also performed stratified analyses based on sex (female and male), race (White, Black or African American, and Asian), type of SGLT2 inhibitors used (empagliflozin, canagliflozin, and dapagliflozin), HbA1c (<8.5% and ≥8.5%), BMI (<30 and ≥30 kg/m²), eGFR (<60 and ≥60 mL/min/1.73 m²), history of hypertension, and history of dyslipidemia. The cut-off values for HbA1c, BMI, and eGFR were determined according to criteria established in previous randomized controlled trials investigating the effectiveness of SGLT2 inhibitors.^34–36 In each stratification, patients were separately identified from the TriNetX platform based on predefined inclusion and exclusion criteria, rather than being directly extracted from the participants in the main analysis. This approach facilitates complete propensity score matching, ensuring balanced characteristics between SGLT2 inhibitor and DPP-4 inhibitor users in each stratification. Participants were followed for up to 5 years in the stratification analysis.

The Cox proportional hazards model was used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). We applied Kaplan–Meier analyses to calculate cumulative incidence based on survival probabilities over 5 years. Participants with outcomes coded prior to the index date were excluded from the final analysis. The TriNetX platform offers built-in functions for data analysis through its online website. Statistical outputs are generated using various software languages and packages, including Java 11.0.16 (Apache Commons Math 3.6.1), R 4.0.2 (Hmisc 1.1, and Survival 3.2-3; R Foundation for Statistical Computing), and Python 3.7 (lifelines 0.22.4, matplotlib 3.5.1, numpy 1.21.5, pandas 1.3.5, scipy 1.7.3, and statsmodels 0.13.2). Further details about the underlying technology of the platform are proprietary and safeguarded as TriNetX trade secrets.

In Kaplan–Meier analyses, participants eligible for study inclusion were censored after their last recorded clinical event within the study time frame. The TriNetX platform presents data as recorded in medical records without imputing or estimating missing clinical values, except for eGFR, which is derived from each reported serum, plasma, or blood creatinine value if the patient's sex and age are available. TriNetX does not have a built-in function to determine the proportion of participants lost to follow-up during the observation period.

The study protocol for retrospective data collection from the TriNetX platform was approved by the Institutional Review Board (IRB) of Taichung Veterans General Hospital under the registration number CE24430C. Because TriNetX handles secondary analysis of de-identified patient data and does not directly intervene with human subjects, the IRB granted a waiver for informed consent.

Results

The final analysis included 20,966 participants treated with SGLT2 inhibitors and 20,966 participants treated with DPP-4 inhibitors. The mean age ± SD was 64.6 ± 9.8 years for the SGLT2 inhibitor group, and it was 64.7 ± 10.9 years for the DPP-4 inhibitor group. The study flowchart is depicted in Figure 1, and the study timeline is shown in Supplementary Figure S1. Baseline characteristics were well balanced between the SGLT2 inhibitor and DPP-4 inhibitor groups, with SMD < 0.1 (Table).

Figure 1.

Study flowchart depicting the inclusion and exclusion criteria for the study population. PSM, propensity score matching.

Table.

Baseline Characteristics of the Study Population

	Before Propensity Score Matching			After Propensity Score Matching
Characteristic	SGLT2 Inhibitor	DPP-4 Inhibitor	SMD	SGLT2 Inhibitor	DPP-4 Inhibitor	SMD
Total number	28,067	26,114	—	20,966	20,966	—
Age (y), mean ± SD	64.9 ± 9.9	65.0 ± 11.0	0.013	64.6 ± 9.8	64.7 ± 10.9	0.006
Female, n (%)	10,746 (38.3)	12,935 (49.5)	0.228	9194 (43.9)	9319 (44.4)	0.012
Race, n (%)*
White	16,854 (60.0)	14,545 (55.7)	0.088	12,259 (58.5)	12,255 (58.5)	<0.001
Black	5830 (20.8)	5364 (20.5)	0.006	4359 (20.8)	4326 (20.6)	0.004
Asian	3470 (12.4)	4978 (19.1)	0.185	3,210 (15.3)	3249 (15.5)	0.005
Smoking, n (%)	1049 (3.7)	684 (2.6)	0.064	621 (3.0)	602 (2.9)	0.005
Laboratory values, mean ± SD
HbA1c (%)	8.1 ± 1.8	8.1 ± 1.8	0.004	8.1 ± 1.8	8.1 ± 1.8	0.012
BMI (kg/m2)	32.0 ± 7.0	31.3 ± 7.2	0.103	31.8 ± 7.0	31.6 ± 7.1	0.023
eGFR (mL/min/1.73m2)	75.9 ± 22.6	76.2 ± 25.2	0.011	77.0 ± 22.9	76.6 ± 24.4	0.017
Medications, n (%)
Metformin	21,298 (75.9)	21,017 (80.5)	0.112	16,651 (79.4)	16,616 (79.3)	0.004
Sulfonylureas	7271 (25.9)	9817 (37.6)	0.253	6551 (31.2)	6713 (32.0)	0.017
Thiazolidinediones	1351 (4.8)	1478 (5.7)	0.038	1128 (5.4)	1146 (5.5)	0.004
Insulin	10,915 (38.9)	8267 (31.7)	0.152	6821 (32.5)	6851 (32.7)	0.003
Comorbidities, n (%)
Hypertension	21,145 (75.3)	19,612 (75.1)	0.005	15,693 (74.8)	15,677 (74.8)	0.002
Ischemic heart disease	8976 (32.0)	5209 (19.9)	0.277	4673 (22.3)	4748 (22.6)	0.009
Cerebrovascular diseases	2481 (8.8)	2014 (7.7)	0.041	1587 (7.6)	1626 (7.8)	0.007
Heart failure	6308 (22.5)	2495 (9.6)	0.358	2259 (10.8)	2420 (11.5)	0.024
Atrial fibrillation	3975 (14.2)	2161 (8.3)	0.187	1892 (9.0)	1962 (9.4)	0.012
Dyslipidemia	22,030 (78.5)	19,720 (75.5)	0.071	16,058 (76.6)	16,115 (76.9)	0.006
End-stage renal disease	177 (0.6)	285 (1.1)	0.050	166 (0.8)	173 (0.8)	0.004
Hypermetropia	297 (1.1)	231 (0.9)	0.018	205 (1.0)	204 (1.0)	<0.001
Diabetic retinopathy, n (%)
Mild NPDR	563 (2.0)	367 (1.4)	0.046	336 (1.6)	330 (1.6)	0.002
Moderate NPDR	168 (0.6)	78 (0.3)	0.045	67 (0.3)	76 (0.4)	0.007
Severe NPDR	60 (0.2)	24 (0.1)	0.031	18 (0.1)	23 (0.1)	0.008
PDR	170 (0.6)	116 (0.4)	0.022	97 (0.5)	99 (0.5)	0.001
Antioxidants, n (%)
Lutein	88 (0.3)	95 (0.4)	0.009	68 (0.3)	72 (0.3)	0.003
Zeaxanthin	20 (0.1)	16 (0.1)	0.004	12 (0.1)	13 (0.1)	0.002
Astaxanthin	≤10 (<0.1)	≤10 (<0.1)	0.001	≤10 (<0.1)	≤10 (<0.1)	<0.001

NPDR, nonproliferative diabetic retinopathy; PDR, proliferative diabetic retinopathy.

^*The racial and ethnic classifications (e.g., White, Black and African American, Asian) are directly taken from the original terminology provided in the TriNetX database. Specifically, the term “Black and African American” has been shortened to “Black” due to space constraints.

Risk of AMD and Its Subtypes

Compared to patients prescribed DPP-4 inhibitors, patients prescribed SGLT2 inhibitors were found to have a significantly lower risk of developing AMD (HR, 0.71; 95% CI, 0.58–0.85) and dry AMD (HR, 0.61; 95% CI, 0.46–0.80) but not wet AMD (HR, 0.74; 95% CI, 0.48–1.16). Results are shown in Figure 2. Patients prescribed SGLT2 inhibitors had a lower cumulative incidence of AMD and dry AMD over 5 years of follow-up compared to those on DPP-4 inhibitors. At the end of the 5-year follow-up, the cumulative incidence of AMD was 1.43% (95% CI, 1.19%–1.71%) for the SGLT2 inhibitor group, and it was 2.02% (95% CI, 1.81%–2.25%) for the DPP-4 inhibitor group (log-rank P = 0.003). For dry AMD, after 5 years of follow-up the cumulative incidence was 0.62% (95% CI, 0.47%–0.81%) for SGLT2 inhibitor users, and it was 1.14% (95% CI, 0.98%–1.32%) for DPP-4 inhibitor users (log-rank P = 0.003). The cumulative incidence curves are presented in Figure 3.

Figure 2.

Associations between SGLT2 and DPP-4 inhibitors for AMD, dry AMD, and wet AMD. The graph illustrates the comparative associations of SGLT2 inhibitors and DPP-4 inhibitors with the risk of incident AMD, dry AMD, and wet AMD. A lower hazard ratio indicates a reduced risk of the outcome with SGLT2 inhibitors compared to DPP-4 inhibitors.

Figure 3.

Curves illustrate the 5-year cumulative incidence of (A) AMD and (B) dry AMD for patients prescribed SGLT2 and DPP-4 inhibitors. The light blue line represents patients prescribed SGLT2 inhibitors, and the dark blue line represents patients prescribed DPP-4 inhibitors.

AMD and Dry AMD Risk Stratified by Different Characteristics

Study participants were further divided into subpopulations based on sex, race, type of SGLT2 inhibitors prescribed, HbA1c, BMI, eGFR, history of hypertension, and history of dyslipidemia. The baseline characteristics were well matched in each group with SMD < 0.1 (Supplementary Table S2). SGLT2 inhibitors compared to DPP-4 inhibitors were associated with a lower risk of AMD in the White participants (HR, 0.70; 95% CI, 0.54–0.89); patients prescribed empagliflozin (HR, 0.72; 95% CI, 0.58–0.89) or dapagliflozin (HR, 0.61; 95% CI, 0.44–0.85); and patients with HbA1c < 8.5% (HR, 0.80; 95% CI, 0.65–0.98), BMI < 30 kg/m² (HR, 0.73; 95% CI, 0.57–0.95), BMI ≥ 30 kg/m² (HR, 0.74; 95% CI, 0.56–0.98), eGFR ≥ 60 mL/min/1.73 m² (HR, 0.77; 95% CI, 0.62–0.97), history of hypertension (HR, 0.72; 95% CI, 0.59–0.89), or history of dyslipidemia (HR, 0.80; 95% CI, 0.65–0.98). These findings are presented in Figure 4.

Figure 4.

This forest plot illustrates the stratified associations of SGLT2 inhibitors and DPP-4 inhibitors with the risk of AMD. To ensure complete propensity score matching, patients were separately identified from the TriNetX platform in each stratification according to the predefined inclusion and exclusion criteria, rather than being directly extracted from the participants in the main analysis. Therefore, slight variations exist between the sum of events in each subpopulation and the overall population. To protect patient privacy, the TriNetX platform conceals the precise event count for groups with ≤10 events.

Patients prescribed SGLT2 inhibitors, compared to those prescribed DPP-4 inhibitors, were associated with a reduced risk of dry AMD in males (HR, 0.61; 95% CI, 0.42–0.90); the White population (HR, 0.57; 95% CI, 0.41–0.80); those prescribed empagliflozin (HR, 0.60; 95% CI, 0.44–0.83) or dapagliflozin (HR, 0.49; 95% CI, 0.29–0.83); and those with HbA1c < 8.5% (HR, 0.71; 95% CI, 0.53–0.97), BMI < 30 kg/m² (HR, 0.62; 95% CI, 0.40–0.96), BMI ≥ 30 kg/m² (HR, 0.58; 95% CI, 0.39–0.87), eGFR ≥ 60 mL/min/1.73 m² (HR, 0.65; 95% CI, 0.46–0.92), or history of hypertension (HR, 0.67; 95% CI, 0.50–0.90). These findings are presented in Supplementary Figure S2.

Discussion

This multinational cohort study demonstrated that, among patients with T2DM, those prescribed SGLT2 inhibitors experience a lower risk of AMD and dry AMD, but not wet AMD, compared to those prescribed DPP-4 inhibitors. In our study, SGLT2 inhibitors compared with DPP-4 inhibitors were associated with a lower risk of AMD in the White population, patients prescribed empagliflozin or dapagliflozin, and patients with HbA1c < 8.5%, eGFR ≥ 60 mL/min/1.73 m², hypertension, or dyslipidemia, regardless of BMI level. These findings provide insights into the potential of SGLT2 inhibitors in mitigating AMD and dry AMD risk, compared with DPP-4 inhibitors.

Limited research has examined the impact of novel glucose-lowering medications on AMD. Although previous studies have highlighted the potential of metformin to reduce AMD risk by modulating the aging process,^37,38 few have explored the role of SGLT2 inhibitors in AMD. Only a recent study using the Taiwan National Health Insurance Research Database had provided partial insights into the potential protective effects of SGLT2 inhibitors against macular degeneration, a broader spectrum of disease than AMD, in patients with T2DM.³⁹ However, there are several limitations of this Taiwan National Health Insurance program study. First, they investigated the risk of macular degeneration (H35.3) instead of specifically AMD. Second, the study population was drawn from a localized database in Taiwan, which may limit its generalizability to other populations. Third, the study did not adjust for HbA1c, BMI, or eGFR, increasing the risk of imbalanced diabetes severity and differing risk profiles between SGLT2 inhibitor users and non-users. Our study addressed these gaps by focusing on more specific AMD diagnoses (H35.30, H35.31, and H35.32), leveraging data from a multinational database, including HbA1c, BMI, and eGFR into the propensity score matching process, and further exploring the risk of dry and wet AMD.

Pathophysiology of SGLT2 Inhibitors in Reducing Risk of AMD

The reduced risk of AMD associated with SGLT2 inhibitors may be attributed to their ability to counteract key pathological processes of AMD. First, lipid dysfunction is linked to dry AMD pathogenesis, with drusen (lipid-rich deposits under the RPE) contributing to disease progression through impaired cholesterol homeostasis, RPE dysfunction, and complement activation.⁴⁰ Animal studies by Sene et al.⁴¹ suggested that impaired cholesterol efflux in aging macrophages promotes pathological vascular changes, further supporting a lipid-driven mechanism in AMD. This mechanism is reflected in clinical studies showing that cholesterol-lowering medications may reduce AMD risk.⁴² Given the ability of SGLT2 inhibitors to improve lipid profiles by lowering total cholesterol, low-density lipoprotein, and triglycerides while increasing high-density lipoprotein,^43–45 their association with a reduced AMD risk compared to DPP-4 inhibitors may be attributed to these lipid-lowering effects.

Second, SGLT2 inhibitors have been shown to inhibit complement activation, resulting from reduced HIF-1α accumulation and increased expression of complement receptor type 1–related protein y (Crry) in renal proximal tubular cells from mice with diabetic kidney disease.²⁴ These findings suggest that similar mechanisms may help mitigate complement-driven retinal inflammation, thereby reducing risks associated with AMD. Further studies are necessary to confirm the potential effect of complement systems in AMD.

Third, persistent hyperglycemia in diabetes leads to retinal damage through increased oxidative stress and inflammation.²⁵ In diabetes, reduced glucose utilization for energy production due to insulin resistance results in increased free fatty acids and fat oxidation. This heightened fat oxidation demands more oxygen and produces greater oxidative stress, which exacerbates retinal damage by causing further oxidative harm to the mitochondria.²⁵ Alternatively, studies have shown that SGLT2 inhibitors can reduce markers of oxidative stress and systemic inflammation, such as retinal H₂O₂ and serum IL-6 levels, as demonstrated in mice models with diabetes.²⁶

Fourth, SGLT2 inhibitors may also reduce the risk of AMD by protecting the cardiovascular system and improving retinal microcirculation.^27,28 Studies have shown that cardiovascular risk factors such as aging, smoking, and hypertension can cause microvascular damage, leading to reduced perfusion and increased obstruction in the choriocapillaris, which may subsequently contribute to the development of AMD.¹⁴ In brief, by improving lipid profiles and mitigating complement activation, oxidative stress, inflammatory response, and retinal vascular dysfunction, SGLT2 inhibitors may alleviate risk associated with AMD.

SGLT2 Inhibitors May Primarily Reduce Dry AMD Risk Rather Than Wet AMD

Our findings suggest that SGLT2 inhibitors are associated with a reduced risk of dry AMD but have no significant association with the risk of wet AMD compared to DPP-4 inhibitors. This distinction highlights the different pathophysiological mechanisms between dry and wet AMD. Dry AMD accounts for 85% to 90% of AMD and is linked to complement activation, increased oxidative stress, and microvascular damage.^6,14,17 These pathological processes contribute to RPE dysfunction and drusen formation, which are hallmarks of dry AMD. In contrast, wet AMD affects 10% to 15% of AMD patients and typically develops on the basis of pre-existing dry AMD. The pathogenesis of wet AMD involves the accumulation of vascular endothelial growth factor (VEGF), which induces choroidal neovascularization and leads to the leakage of blood or plasma, ultimately resulting in severe vision loss.¹⁹ A key factor differentiating wet from dry AMD is the increased production of VEGF-A. This angiogenic factor, triggered by altered cytokine and immune cell activity, induces the formation of abnormal new blood vessels in wet AMD.⁴⁶ Although SGLT2 inhibitors may target the complement, oxidative, and inflammatory mechanisms relevant to dry AMD, SGLT2 inhibitors do not appear to directly impact VEGF-A–driven neovascularization, which may explain their limited effect on wet AMD.

Risk of AMD in Participants Stratified by Different Factors

Our stratified analysis reveals that the association between SGLT2 inhibitors and a reduced risk of AMD, compared to DPP-4 inhibitors, is consistent across the White population, patients prescribed empagliflozin or dapagliflozin, and individuals with HbA1c < 8.5%, eGFR ≥ 60 mL/min/1.73 m², history of hypertension, or history of dyslipidemia, regardless of BMI group. The higher prevalence of AMD in the White population and among individuals with hypertension or dyslipidemia may lead to a more pronounced reduction in AMD risk with SGLT2 inhibitors compared to DPP-4 inhibitors in these groups. Racial distributions suggest that AMD is most common in Whites, with studies reporting a prevalence of 5.4% in Whites, 4.6% in Asians, and 2.4% in Blacks.^47–49 Hypertension and dyslipidemia are also risk factors for AMD.^50,51 Moreover, our stratified analysis indicates that the reduced AMD risk associated with SGLT2 inhibitors compared to DPP-4 inhibitors is observed only in patients with HbA1c < 8.5%, not in those with HbA1c ≥ 8.5%. This suggests that patients with mild to moderate T2DM may experience greater benefits in this context.

A significantly reduced risk of AMD is observed in patients prescribed empagliflozin and dapagliflozin rather than canagliflozin. Given that empagliflozin and dapagliflozin have a higher relative affinity for SGLT2 over SGLT1 compared to canagliflozin, our findings suggest that SGLT1 and SGLT2 may be expressed differently in the retina.^52–54 However, the relative expression proportion of SGLT1 and SGLT2 in the retina remains poorly understood, warranting further investigation.⁵⁵ Moreover, our analysis suggests that patients with an eGFR ≥ 60 mL/min/1.73 m² are associated with a significantly reduced risk of AMD and dry AMD, whereas those with eGFR < 60 mL/min/1.73 m² are not. This finding may result from the better glucose eradication effect from glycosuria in those with better renal function.^56,57 Another reason for the lack of a significant association between SGLT2 inhibitors and AMD among patients with eGFR < 60 mL/min/1.73 m² may be attributed to the smaller number of participants involved. SGLT2 inhibitors are mainly indicated in patients with eGFR ≥ 60 mL/min/1.73 m², as reflected in our analysis, where participants with eGFR < 60 mL/min/1.73 m² were fewer than those with an eGFR ≥ 60 mL/min/1.73 m².⁵⁶ Findings from the stratified analysis could assist clinicians in identifying T2DM patients who are most likely to benefit from SGLT2 inhibitors compared to DPP-4 inhibitors.

Strengths and Limitations

Our study has several strengths. First, by leveraging a large multinational database, our findings offer robust statistical power and enhanced generalizability. Second, we focused on specific outcomes such as AMD, dry AMD, and wet AMD, offering a more nuanced evaluation compared to broader classifications of macular degeneration used in previous research. Third, we assessed the risk of outcomes across 5 years, providing a more thorough evaluation of how these medications affect AMD over time. Fourth, we conducted stratified analyses across different sexes, races, types of SGLT2 inhibitors, and patients with different HbA1c, BMI, and eGFR levels and histories of comorbidities. These stratified analyses allowed for more precise AMD risk predictions based on population characteristics, enabling personalized risk assessments for individual patients in the clinical setting.

However, our study has several limitations. First, the retrospective cohort design is prone to biases related to differing prescription preferences or indications between SGLT2 and DPP-4 inhibitors. Second, given that TriNetX is a de-identified database and individual medical records cannot be accessed to verify AMD diagnoses, there is potential for disease misclassification or coding errors. Third, we could not confirm the medical compliance of each individual, including whether prescriptions were filled or administered according to medical instructions. Fourth, as SGLT2 and DPP-4 inhibitors are often used in combination with other glucose-lowering medications rather than as monotherapy, potential drug–drug interactions may have affected our results. To minimize this bias, we excluded patients concurrently using GLP-1RAs and performed propensity score matching with other glucose-lowering medications at baseline. Fifth, although our findings suggest a statistically significant lower hazard for developing AMD and dry AMD with SGLT2 inhibitors compared to DPP-4 inhibitors, the incidence of AMD in our population remained relatively small. This raises concerns about whether the results are clinically meaningful or could be statistical artifacts due to low event rates in a large but low-risk cohort. Nonetheless, given the limited treatment options for dry AMD and its risk of progression to vision loss, our findings still provide insights into potential prevention strategies for dry AMD that may benefit patients at high risk.⁵⁸ Future studies should validate our findings in populations at higher risk of developing AMD and assess whether the benefits of SGLT2 inhibitors over DPP-4 inhibitors translate into meaningful advantages for patients with T2DM in the clinical setting.

Conclusions

Our findings suggest that SGLT2 inhibitors are associated with a reduced risk of AMD, particularly dry AMD, in patients with T2DM compared to DPP-4 inhibitors. These results address the challenges of managing dry AMD, where treatment options remain limited and suboptimal.