Introduction

A novel type of drug, sodium-glucose cotransporter-2 (SGLT2) inhibitors, has been approved for diabetes treatment by the U.S. Food and Drug Administration (FDA) since 2014 [1]. SGLT2 is primarily expressed in the proximal convoluted tubule and functions as a reabsorption for glucose [2, 3]. SGLT2 inhibitors block the reabsorption of glucose by inhibiting SGLT2, thereby promoting glucose renal excretion; this lowers the elevated glucose levels of patients with diabetes and produces an osmotic diuretic effect [4].

Compared to other antihyperglycemic agents, SGLT2 inhibitors have a lower risk of hypoglycemia because they only affect the reabsorption of glucose that is filtered rather than blood glucose [5,6,7]. Moreover, SGLT2 inhibitors have cardiovascular protective effects and renal benefits [8,9,10,11]. In addition to cardiovascular and kidney problems, however, macular degeneration (MD) is also a complication among patients with diabetes [12,13,14]. A former review showed that the global incidence of MD was 1.59 (95% CrI 1.18–2.11) per 100 person-years [15]. A recent study revealed that from 2005 to 2014 [16], 3.75 to 3.95% of Taiwanese patients with diabetes experienced diabetic eye diseases, and 0.29 to 0.35% encountered poor vision and even blindness. A meta-analysis also showed that SGLT2 inhibitors have greater protective effects against retinal diseases compared to other hypoglycemic agents [17]. However, related evidence about SGLT2 inhibitor usage and MD is limited. Thus, we explore the association between SGLT2 inhibitor usage and MD in Taiwanese patients with diabetes.

Methods

Data source

The National Health Insurance (NHI) program of Taiwan was established in 1995, and this program's claim data are released as the National Health Insurance Research Database (NHIRD). This database covers more than 99% of the residents in Taiwan, and diseases such as type 2 diabetes mellitus or MD can be defined by the International Classification of Diseases, 9th Revision, Clinical Modification, and the International Classification of Diseases, 10th Revision (ICD-9-CM and ICD-10-CM). Additionally, therapeutic uses such as drugs for type 2 diabetes mellitus can be defined by the Anatomical Therapeutic Chemical (ATC) code. All analyses were performed in the Health and Welfare Data Center, and privacy policies were implemented to protect the privacy of beneficiaries. The Institutional Review Board (IRB) of China Medical University (CMUH110-REC3-133) has approved this study.

Study population

We included data on patients who were newly diagnosed with type 2 diabetes mellitus (ICD-9-CM: 250, exclude 250.1x; ICD-10-CM: E11), with an age at diagnosis of over 20 years as our study population. Patients who received (sodium-glucose cotransporter 2 inhibitor) SGLT2i (ATC code: A10BK) over 90 days in 2016–2019 were defined as the SGLT2i cohort. Conversely, patients who did never received SGLT2i were defined as the non-SGLT2i cohort. The index date was defined as the first date of receiving SGLT2i in the SGLT2i cohort and as a random date between 2016 and 2019 in the non-SGLT2i cohort. The exclusion criteria were having MD before the index date, receiving SGLT2i within 1 to 89 days, and missing data on sex, age, or days of SGLT2i usage. Two cohorts were matched by 1:1 propensity score matching, which was based on age, sex, payroll bracket grade, urbanization, comorbidities, and medications.

Main outcome and covariates

All study participants were followed from the index date until the onset of MD (ICD-9-CM: 362.5, ICD-10-CM: H35.3), withdrawal from the NHI program, or the end of 2019. The covariates included sex, age group (20–39, 40–59, 60 + years), payroll bracket grade (low, medium, high), urbanization (high, medium, low), comorbidities, and medications. Payroll bracket grade levels were defined as monthly income < 700 USD, 700–1300 USD, and > 1300 USD, respectively.

Comorbidities diagnosed before the index date, including hypertension (ICD-9-CM: 401–405; ICD-10-CM: I10-I13, I15, and N26.2), dyslipidemia (ICD-9-CM: 272; ICD-10-CM: E71.30, E75.21, E75.22, E75.24, E75.3, E75.5, E75.6, E77, E78.0-E78.6, E78.70, E78.79, E78.8, and E78.9), coronary artery disease (CAD) (ICD-9-CM: 410–414; ICD-10-CM: I20-I25), alcoholism (ICD-9-CM: 291, 303, 305.0, 571.0–571.3, 790.3, V11.3, and V79.1; ICD-10-CM: F10, K70, R78.0, and Z65.8), chronic obstructive pulmonary disease (COPD) (ICD-9-CM: 490–496, and 504–506; ICD-10-CM: J40-J47, and J64-J68), diabetic retinopathy (DR) (ICD-9-CM: 362.0; ICD-10-CM: E11.31-E11.35), and macular edema (ICD-9-CM: 362.83; ICD-10-CM: H35.81), were considered.

Eight common antidiabetic drugs, including metformin, sulphonylurea, meglitinides, a-glucosidase inhibitors (AGI), thiazolidinedione, dipeptidyl peptidase 4 inhibitors (DPP4i), glucagon-like peptide-1 receptor agonists (GLP-1RA), and insulin, were considered.

Statistical method

Mean and standard error (SD) were used to perform characteristics of continuous variables such as age and follow-up time, and percentage was used to perform characteristics of category variables. The difference can be evaluated using the standardized mean difference (SMD). The incidence rate was performed per 1,000 person-years (PY) and was calculated by n/PY*1,000. Univariable and multivariable Cox regression models were used to calculate the hazard ratio and the 95% interval estimation of the hazard ratio. The multivariable Cox regression adjusted age, sex, payroll bracket grade, urbanization, comorbidities, and medications. All analyses were performed using SAS version 9.4 (SAS Institute, Inc., Cary, NC, USA), and p values of 0.05 or less were considered statistically significant.

Result

Table 1 presents the demographic characteristics of the two cohorts. There were 147,664 patients matched 1:1 with another 147,664 patients in both cohorts. The SMD shows that age, sex, payroll bracket grade, urbanization, comorbidities, and medications had no significant difference between non-SGLT2i and SGLT2i cohorts. There were approximately 58% male patients among patients with type 2 diabetes mellitus (DM). The major urbanization level was high, with approximately 52%, and the major payroll bracket grade level was medium, with approximately 54%. The top two comorbidities, including hypertension and dyslipidemia, were performed. The proportions of patients receiving medications with and without SGLT2i were as follows: metformin (97.34% versus 97.51%), sulphonylurea (83.63% versus 84.25%), meglitinides (18.06% versus 18.12%), AGI (36.71% versus 35.82%), thiazolidinedione (39.21% versus 37.94%), DPP4i (72.00% versus 71.69%), GLP-1RA (2.02% versus 1.89%), and insulin (40.60% versus 41.55%), respectively. The mean (SD) follow-up times were 2.24 (0.78) years in the study group and 2.21 (0.98) years in the control group.

Table 1 Comparison of demographic characteristics and comorbidities between T2DM patients with SGLT2i and controls
Full size table

Table 2 shows the risk of MD according to whether the patients have SGLT2i, comorbidities, or medications. Compared to patients who did not receive SGLT2i, patients who received SGLT2i had a significantly lower risk of MD (adjusted hazard ratio [aHR] = 0.70, 95% CI = 0.66–0.75). For different age groups, we found that patients over 60 had a higher risk of MD compared to those aged 20–39 (aHR = 4.06, 95% CI = 3.17–5.18). There was no significantly different risk of MD between female and male patients. Patients with a payroll bracket grade higher than 700 USD had lower risks of MD than those with a payroll bracket grade lower than 700 USD. Moreover, patients with comorbidities, including CAD (aHR = 1.11, 95% CI = 1.03–1.19), DR (aHR = 2.47, 95% CI = 2.30–2.65), macular edema (aHR = 1.65, 95% CI = 1.32–2.05), as well as those who received medications such as AGI (aHR = 1.16, 95% CI = 1.08–1.24), thiazolidinedione (aHR = 1.09, 95% CI = 1.01–1.16), DPP4i (aHR = 1.12, 95% CI = 1.02–1.22), and insulin (aHR = 1.13, 95% CI = 1.06–1.21) had higher risks of MD compared with the corresponding groups.

Table 2 Hazard ratios and 95% confidence intervals of macular degeneration development
Full size table

When patients were categorized according to covariates, the aHRs were statistically lower in the SGLT2i group than in the control group, except in alcoholism patients, macular edema patients, without metformin, without sulphonylurea, and with GLP-1RA; the aHR were 0.72 (95% CI = 0.43–1.23), 0.78 (95% CI = 0.50–1.24), 0.71 (95% CI =  0.42–1.18), 0.82 (95% CI = 0.67–1.00), and 0.63 (95% CI = 0.40, 1.01), respectively (Table 3).

Table 3 Hazard ratios and 95% confidence intervals of macular degeneration development with and without SGLT2i stratified by age, comorbidities, and medication
Full size table

Discussions

The study investigates the use of SGLT2 inhibitors. SGLT2 inhibitors block the reabsorption of glucose by inhibiting SGLT2, promoting glucose renal excretion, lowering the elevated glucose levels of patients with diabetes, and producing an osmotic diuretic effect [18,19,20]. This study explored the association between SGLT2 inhibitors in patients with DM and the development of MD. The study used data from the Longitudinal Generation Tracking Database (LGTD), which was derived from a large database of a single-payer NHI program initiated in Taiwan since 1995, called the NHIRD. We aimed to explore the risk of MD in patients with diabetes treated with SGLT2 inhibitors. We found that the incidence of MD was lower in diabetic patients treated with SGLT2 inhibitors than in those treated without SGLT2 inhibitors. This may imply some positive correlation between SGLT2 inhibitors and the prevention of MD.

The mechanisms underlying the protective effect of SGLT2 inhibitors that reduce the risk of MD remain understudied. During the early stage of diabetic retinopathy, pericyte swelling and pericyte loss occur, which results in microaneurysm formation in the retina. The swollen pericytes lose their contractile ability and lead to retinal hyperperfusion [21, 22]. These processes result in a similar clinical observation as the early MD disease sign, edematous macula. The SGLT2 inhibitor has been proven to decrease inflammatory responses and oxidative processes, such as changing the phenotype of astrocytes and microglia into pro-inflammatory ones [22], which also plays a crucial role in MD disease. SGLT2 inhibitors elevate the ratios of adenosine diphosphate/adenosine triphosphate (ATP) and AMP/ATP, leading to a decrease in ATP production and oxidative phosphorylation. Consequently, cellular ATP levels are reduced, which triggers AMPK activation. AMPK activation hinders protein synthesis and cell proliferation. A previous study has shown that metformin exerts its anti-inflammatory and antioxidative effects primarily through the activation of AMP-activated protein kinase (AMPK). This activation leads to the inhibition of nuclear factor-kappa B, a reduction in reactive oxygen species, and the inhibition of the mammalian target of rapamycin. AMPK, a metabolic-sensing Ser/Thr kinase expressed in all cell types, plays a central role in maintaining energy homeostasis and regulating metabolic stress. It also contributes to the dysfunction of the metabolic ecosystem, potentially influencing the pathogenesis of age-related macular degeneration (AMD).

Metformin's protective effects against oxidative damage and mitochondrial dysfunction are considered additional mechanisms for its potential efficacy in AMD treatment [23]. These effects have been confirmed in retinal pigment epithelial (RPE) cells, where metformin stimulates autophagy through the activation of the AMPK pathway, effectively shielding RPE cells from oxidative damage. Furthermore, metformin safeguards photoreceptors and RPE cells from acute injury and delays inherited retinal degeneration by reducing oxidative stress and increasing mitochondrial energy production. Additionally, metformin demonstrates anti-inflammatory properties on the RPE by attenuating pro-inflammatory and adhesion molecule genes. This is particularly relevant in early AMD, characterized by drusen formation, as it results from an inflammatory reaction triggered by RPE damage. Thus, the anti-inflammatory and antioxidant effects of metformin help protect the RPE from early AMD lesions [13, 24]. Thus, in our study, SGLT2 inhibitors showed a similar effect to metformin and resulted in a decreased relative risk of MD in patients with DM.

Subretinal fibrosis and geographic atrophy are challenging clinical situations. Type IV collagen synthesis is increased under high-glucose conditions in bovine pericytes. SGLT2 inhibitors may inhibit mesangial expansion in the glomerulus and microvessel occlusion in the retina (22). Thus, the collagen synthesis modulation could be a positive impact on MD disease prevention.

The study has several strengths, such as the use of a large database, adjustment of potential confounders, and multivariate Cox proportional hazards models. However, the study has some limitations: It is a retrospective cohort study that could not determine the causal relationship between SGLT2 inhibitors and MD. Moreover, the study was conducted in Taiwan; therefore, the results may not be generalizable to other populations. Finally, the Image evaluation and results of hematology test are not available from NHIRD. However, all the patients who need anti-VEFG drugs to treat macular degeneration disease will turn in optical coherence tomography (OCT) image for audited. Similarly, all the patients who need anti-VEFG drugs to treating diabetic macular edema will turn in HbA1C data for audited. The data of HbA1C for first diagnosis should below 10%. Overall, the study suggests that SGLT2 inhibitors may have a protective effect on MD in patients with diabetes.

In conclusion, we found that SGLT2is has a stronger protective effect against MD in patients with diabetes than diuretics. SGLT2is may have benefits beyond glycemic control in patients with DR. Compared with other anti-diabetic drugs, SGLT2 inhibitors show strong ability of anti-inflammatory and antioxidant effects by regulating adenosine diphosphate/adenosine triphosphate (ATP) and AMP/ATP pathway. SGLT2 inhibitor may have a better chance to propone macular degeneration disease process and lead to the protecting effect. However, additional clinical and experimental studies are required.