Abstract
Objective:
To investigate whether the choice of glucose-lowering agent for type 2 diabetes (T2D) impacts a patient’s risk of developing sight-threatening diabetic retinopathy complications.
Design:
Retrospective observational database study emulating an idealized target trial.
Subjects:
Adult (≥21 years) enrollees in United States commercial, Medicare Advantage, and Medicare fee-for-service plans from January 1, 2014, to December 31, 2021, with T2D and moderate cardiovascular disease (CVD) risk who had no baseline history of advanced diabetic retinal complications, initiating treatment with glucagon-like peptide-1 receptor agonists (GLP-1 RA), sodium-glucose cotransporter 2 inhibitors (SGLT2i), dipeptidyl peptidase-4 inhibitors (DPP-4i), and sulfonylureas.
Methods:
We used inverse propensity scoring weights in time-to-event Cox proportional hazards models.
Main Outcome Measures:
Treatment for either diabetic macular edema or proliferative diabetic retinopathy.
Results:
The final study population included 371 698 patients, of whom 42 265 initiated GLP-1 RA, 53 476 initiated SGLT2i, 78 444 initiated DPP-4i, and 197 513 initiated sulfonylurea agents. The probability of treatment for sight-threatening retinopathy within 2 and 5 years was 0.3% and 0.7% for patients initiating SGLT2i (median follow-up 830 [interquartile range (IQR), 343–1401] days), 0.4% and 1.0% for GLP-1 RA (669 [IQR, 256–1167] days), 0.4% and 0.9% for DPP-4i (1263 [IQR, 688–1938] days), and 0.5% and 1.2% for sulfonylurea (1223 [IQR, 662–1879] days). Sodium-glucose cotransporter 2 inhibitors use was associated with a lower risk of treatment for sight-threatening retinopathy compared with all other medication classes, including GLP-1 RA (hazard ratio [HR], 0.73; 95% confidence interval [CI], 0.55–0.97), DPP-4i (HR, 0.79; 95% CI, 0.64–0.97), and sulfonylurea (HR, 0.61; 95% CI, 0.50–0.74). Glucagon-like peptide-1 receptor agonists use was associated with a similar risk of sight-threatening retinopathy as DPP-4i (HR, 1.07; 95% CI, 0.85–1.35) and sulfonylurea (HR, 0.83; 95% CI, 0.67–1.03).
Conclusions:
Sodium-glucose cotransporter 2 inhibitors use was associated with a lower risk of sight-threatening diabetic retinopathy among adults with T2D and moderate CVD risk compared with other glucose-lowering therapies. Glucagon-like peptide-1 receptor agonists do not confer increased retinal risk, relative to DPP-4i and sulfonylurea medications.
Financial Disclosure(s):
Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.
Keywords: Diabetic macular edema, Diabetic retinopathy, GLP-1 RA, Proliferative diabetic retinopathy, SGLT2i
Diabetic retinopathy is a leading cause of severe vision loss among working-age adults in the United States.1 Globally, the prevalence of diabetes is projected to rise from an estimated 537 million adults aged 20 to 79 years in 2021, >90% of whom have type 2 diabetes (T2D), to 783 million worldwide by 2045.1,2 Blood glucose-lowering medications play an important role in helping people living with T2D maintain glycemic control and reduce their risk of sight-threatening retinal complications.3,4 In recent years, several new classes of antihyperglycemic medications for T2D have been approved by the United States Food and Drug Administration including glucagon-like peptide-1 receptor agonists (GLP-1 RA, first-in-class approval in 2004), dipeptidyl peptidase-4 inhibitors (DPP-4i, 2005), and sodium-glucose cotransporter 2 inhibitors (SGLT2i, 2013). Clinical guidelines and practice patterns have rapidly evolved to reflect the growing evidence that SGLT2i and GLP-1 RA medications offer beneficial effects in T2D beyond glycemic control, including weight loss and protection against cardiovascular and kidney events. However, the relative benefits and risks of these new medications with respect to vision-threatening retinal complications remain incompletely understood.
Preclinical investigations involving SGLT2i, GLP-1 RA, and DPP-4i agents have identified multiple class-specific retinal microvascular, inflammatory, and neuroprotective findings that are predominantly favorable and potentially independent of glucose-lowering effects.5–14 The combination of these distinct local retinal effects and the interclass differences in both glucose-lowering and broader pleiotropic systemic effects could plausibly lead to meaningful differences in risk of retinal complication rates among patients taking different classes over time. Pivotal randomized controlled trials (RCTs) of these medications have focused primarily on cardiovascular and kidney outcomes in patients with T2D and high-cardiovascular risk, but ophthalmic outcomes have been variably reported. In 2016, the SUSTAIN-6 trial evaluating semaglutide (GLP-1 RA) identified an increased risk of diabetic retinopathy (DR) complications (vitreous hemorrhage, DR-related blindness, intravitreal pharmacotherapy, and laser photocoagulation) compared with placebo over 2 years, which drew attention to the potential impact of GLP-1 RA pharmacotherapy on DR outcomes.15 A meta-analysis of RCTs involving glucose-lowering medications identified no increased risk of DR complications with GLP-1 RA, SGLT2i, or DPP-4i agents versus the placebo but did show an increased DR risk with sulfonylurea use compared with the placebo.16 Inherent limitations in the underlying clinical trial data, however, reduce the utility of these DR-related findings due to the relatively short RCT follow-up, narrow inclusion criteria based on nonophthalmic conditions, limited and variable prespecified ophthalmic outcomes, heterogeneous baseline DR status, and the lack of direct comparison between newer medication classes. Therefore, we sought to leverage linked administrative claims data for a geographically and demographically diverse cohort of United States adults enrolled in commercial, Medicare Advantage, and Medicare fee-for-service health plans to emulate an idealized trial investigating the longer-term comparative effectiveness of GLP-1 RA, SGLT2i, DPP-4i, and sulfonylurea medications with respect to sight-threatening retinal complications.
Methods
Study Design
This is a prespecified secondary analysis of a retrospective observational study emulating an idealized target trial examining head-to-head the comparative effectiveness of initiation of GLP-1 RA, SGLT2i, DPP-4i, and sulfonylureas by adults with T2D at moderate risk of cardiovascular disease (CVD, defined below) with regard to cardiovascular events, microvascular complications, severe hypoglycemia, and other adverse events. The study protocol was preregistered on ClinicalTrials.gov (NCT05214573). The study was exempt from the Mayo Clinic Institutional Review Board review and the requirement for informed consent was waived. The study adhered to the Declaration of Helsinki and is reported according to the RECORD and START-RWE reporting guidelines.17,18
Study Cohort
We used deidentified medical and pharmacy claims from OptumLabs Data Warehouse, which includes enrollees in commercial and Medicare Advantage health plans across the United States, linked using personal identifiers to a 100% sample of Medicare fee-for-service claims. The OptumLabs Data Warehouse represents a diverse population with a wide range of ages, racial and ethnic groups, income levels, geographic regions, health systems, and health plans across the United States.19,20 We used claims data from January 1, 2014, to December 31, 2021.
In these data, we identified adults aged ≥21 years who had filled their first prescription for a GLP-1 RA, SGLT2i, DPP-4i, or sulfonylurea between January 1, 2014, and December 31, 2021 (index date was defined as the first date they filled the study medication), had second fill of the same class to confirm use with no more than a 30-day gap, and had 12 months of enrollment before the index date to ascertain baseline covariates and establish incident use of these drugs. We excluded patients with (1) missing sex, year of birth, or region; (2) previous fill for any study drug class, glinide, or insulin; (3) fill for any second study drug class during days 0 to 30 after index date; (4) type 1 diabetes (defined by the presence of any type 1 diabetes International Classification of Diseases-9/10 codes); (5) pregnancy; and (6) metastatic cancer. To address the aims of the original study we restricted the study population to patients at moderate CVD risk, defined using the Annualized Claims-based Major Adverse Cardiovascular Event Estimator as 1% to 5% annualized predicted risk of experiencing a major adverse cardiovascular event.21
As the last step, we restricted this cohort to patients without previous proliferative diabetic retinopathy (PDR) diagnosis, treatment, or surgery; without previous diabetic macular edema (DME) diagnosis or treatment; and without previous diagnosis of blindness or low vision (Table S1, available at www.ophthalmologyretina.org). Included patients were followed from the time of index date of 1 of the 4 study medications during the study period until either the end of the study (December 31, 2021), disenrollment from insurance plans in our database, or death.
Outcomes
The primary outcome was time to treatment for sight-threatening diabetic retinopathy, defined as a composite of treatment with intravitreal pharmacotherapy, laser photocoagulation, or vitreoretinal surgery for either DME or PDR, whichever occurred first (Table S1). Secondary outcomes were time to PDR treatment, DME treatment, and diagnosis of blindness or low vision, separately.
Covariates
Covariates included baseline demographic characteristics (age, sex, race/ethnicity, and United States region) ascertained from enrollment files and comorbidities and nonstudy medications ascertained from the baseline 12-month period (Tables S2 and S3, available at www.ophthalmologyretina.org).
Statistical Analyses
Our primary analyses were time-to-event models with the intention-to-treat framework. We first used a diverse set of binomial prediction algorithms included in a super learner ensemble22 to estimate the probability of treatment assignment for each medication class.23 The super learner framework allows for flexible estimation of an ensemble predictor with the theoretical properties of crossvalidation for model selection to control for overfitting.24 We included all baseline covariates (Tables S2 and S3) in the super learner ensemble, estimating a separate propensity model for each treatment (vs. the other 3). The predicted propensity scores were evaluated for evidence of violations of the positivity assumption. Because the inverse propensity score weights had extreme values when examining their distributions, we used the stabilized version proposed by Yoshida et al25 for the analysis.
We evaluated the balance across treatment groups using the weights to calculate standardized mean differences of baseline covariates for each pairwise comparison.23 The final stabilized weights were then incorporated into cause specific Cox proportional hazards models, one for each outcome, with all treatment groups and any covariates that were not balanced after weighting (standardized mean difference: > 0.2) were also included in the models; the cause specific models accounted for the competing risk of death.26 We calculated all pairwise treatment effects, reporting hazard ratios (HRs), and used simulation to estimate 95% confidence intervals (CIs). For each model, we tested the proportional hazards assumption for the treatment group and overall using the Grambsch and Therneau test.27,28 A single overall omnibus test for the treatment hazard rates all being equal to each other versus at least one different was conducted at the 0.05 significance level. Cumulative incidence curves by medication class were estimated using the Kaplan–Meier method. Finally, to better interpret the magnitude of the effects, we used these models to estimate adjusted event rates at 1, 2, 3, and 5 years (Table S4, available at www.ophthalmologyretina.org).
Sensitivity Analyses
Two sensitivity analyses were conducted using an as-treated approach, whereby patients were censored upon the following: (1) discontinuation of the assigned treatment (no fills after 30 days beyond the dispensed pill count), death, or disenrollment from the health plan, whichever came first; and (2) discontinuation of the assigned treatment, the addition of another glucose-lowering drug or insulin, death, or disenrollment from the health plan, whichever came first (Tables S5, available at www.ophthalmologyretina.org).
Data management activities were conducted using SAS 9.04 (SAS Institute Inc.), whereas analyses were conducted using R version 4.1 (R Foundation for Statistical Computing).
Results
We identified 368 810 adult patients with T2D who met the inclusion criteria, of whom 42 468 initially received GLP-1 RA, 45 109 received SGLT2i, 80 284 received DPP-4i, and 200 949 received sulfonylurea agents. Baseline characteristics before weighting are shown in Table S6 (available at www.ophthalmologyretina.org). After inverse probability weighting, the final study population (Table 7) was well-balanced across all characteristics, including demographics, comorbidities, and index years (all standardized mean difference ≤ 0.10). Of the 371 698 final weighted study patients, 42 265 initiated treatment with GLP-1 RA (65.4 ± 8.3 years, 51.4% male), 53 476 began SGLT2i (65.3 ± 8.4 years, 50.7% male), 78 444 began DPP-4i (65.6 ± 8.4 years, 51.0% male), and 197 513 began sulfonylureas (65.6 ± 8.4 years, 51.1% male). The median follow-up by initial exposure was 669 days for GLP-1 RA (interquartile range [IQR] 256–1167), 830 days for SGLT2i (IQR, 343–1401), 1263 days for DPP-4i (IQR, 688–1938), and 1223 days for sulfonylurea (IQR, 662–1879).
Table 7.
Baseline Patient Characteristics after Inverse Probability of Treatment Weighting by Treatment Arm
| DPP-4i (N = 78 444) |
GLP-1 RA (N = 42 265) |
SGLT2i (N = 53 476) |
Sulfonylurea (N = 197 513) |
Total (N = 371 698) |
Largest SMD |
|
|---|---|---|---|---|---|---|
|
| ||||||
| Age, mean (SD) | 65.7 (8.4) | 65.4 (8.3) | 65.3 (8.4) | 65.6 (8.4) | 65.5 (8.4) | 0.04 |
| Age group (y), N (%) | 0.06 | |||||
| <45 | 672 (0.9) | 346 (0.8) | 452 (0.8) | 1845 (0.9) | 3314 (0.9) | |
| 45–49 | 3146 (4.0) | 1740 (4.1) | 2163 (4.0) | 7812 (4.0) | 14 862 (4.0) | |
| 50–54 | 6429 (8.2) | 3580 (8.5) | 4638 (8.7) | 16 381 (8.3) | 31 028 (8.3) | |
| 55–59 | 8423 (10.7) | 4617 (10.9) | 5987 (11.2) | 21 199 (10.7) | 40 225 (10.8) | |
| 60–64 | 8084 (10.3) | 4507 (10.7) | 5732 (10.7) | 20 548 (10.4) | 38 871 (10.5) | |
| 65–69 | 22 220 (28.3) | 12 542 (29.7) | 15 403 (28.8) | 56 517 (28.6) | 106 682 (28.7) | |
| 70–74 | 20 897 (26.6) | 11 026 (26.1) | 13 826 (25.9) | 51 967 (26.3) | 97 715 (26.3) | |
| 75+ | 8572 (11.0) | 3910 (9.2) | 5276 (9.8) | 21 244 (10.8) | 39 001 (10.5) | |
| Sex, N (%) | ||||||
| Male | 40 033 (51.0) | 21 714 (51.4) | 27 120 (50.7) | 101 027 (51.1) | 189 895 (51.1) | 0.01 |
| Female | 38 411 (49.0) | 20 551 (48.6) | 26 356 (49.3) | 96 486 (48.9) | 181 803 (48.9) | |
| Race/ethnicity, N (%) | 0.04 | |||||
| Asian | 2344 (3.0) | 975 (2.3) | 1477 (2.8) | 5844 (3.0) | 10 640 (2.9) | |
| Black | 8071 (10.3) | 4263 (10.1) | 5358 (10.0) | 20 457 (10.4) | 38 149 (10.3) | |
| Hispanic | 6657 (8.5) | 3542 (8.4) | 4480 (8.4) | 16 945 (8.6) | 31 624 (8.5) | |
| White | 58 936 (75.1) | 32 155 (76.1) | 40 493 (75.7) | 148 105 (75.0) | 279 689 (75.2) | |
| Unknown | 2435 (3.1) | 1330 (3.1) | 1668 (3.1) | 6162 (3.1) | 11 596 (3.1) | |
| Region, N (%) | 0.04 | |||||
| Midwest | 20 840 (26.6) | 11 193 (26.5) | 14 028 (26.2) | 52 629 (26.6) | 98 689 (26.6) | |
| Northeast | 10 308 (13.1) | 5049 (11.9) | 6895 (12.9) | 25 725 (13.0) | 47 978 (12.9) | |
| South | 36 695 (46.8) | 20 061 (47.5) | 25 198 (47.1) | 92 469 (46.8) | 174 423 (46.9) | |
| West | 10 460 (13.3) | 5900 (14.0) | 7286 (13.6) | 26 315 (13.3) | 49 961 (13.4) | |
| Unknown | 141 (0.2) | 62 (0.1) | 69 (0.1) | 375 (0.2) | 647 (0.2) | |
| Prescriber specialty, N (%) | 0.03 | |||||
| Cardiology | 82 (0.1) | 58 (0.1) | 103 (0.2) | 177 (0.1) | 419 (0.1) | |
| Endocrinology | 1040 (1.3) | 653 (1.5) | 803 (1.5) | 2622 (1.3) | 5118 (1.4) | |
| Primary care | 46 929 (59.8) | 25 287 (59.8) | 32 110 (60.0) | 118 052 (59.8) | 222 379 (59.8) | |
| Others | 11 987 (15.2) | 6536 (15.4) | 8209 (15.3) | 30 360 (15.3) | 57 091 (15.3) | |
| Unknown | 18 407 (23.5) | 9732 (23.0) | 12 251 (22.9) | 46 302 (23.4) | 86 692 (23.3) | |
| Index yr, N (%) | 0.07 | |||||
| 2014 | 10 343 (13.2) | 5302 (12.5) | 5875 (11.0) | 25 945 (13.1) | 47 464 (12.8) | |
| 2015 | 10 917 (13.9) | 5519 (13.1) | 7289 (13.6) | 27 386 (13.9) | 51 111 (13.8) | |
| 2016 | 11 218 (14.3) | 5815 (13.8) | 7666 (14.3) | 28 081 (14.2) | 52 780 (14.2) | |
| 2017 | 12 039 (15.3) | 6372 (15.1) | 8435 (15.8) | 30 222 (15.3) | 57 068 (15.4) | |
| 2018 | 11 807 (15.1) | 6526 (15.4) | 8249 (15.4) | 29 574 (15.0) | 56 155 (15.1) | |
| 2019 | 11 469 (14.6) | 6469 (15.3) | 8186 (15.3) | 28 823 (14.6) | 54 948 (14.8) | |
| 2020 | 5142 (6.6) | 2955 (7.0) | 3708 (6.9) | 13 157 (6.7) | 24 962 (6.7) | |
| 2021 | 5509 (7.0) | 3307 (7.8) | 4069 (7.6) | 14 325 (7.3) | 27 210 (7.3) | |
| Data source, N (%) | 0.02 | |||||
| Medicare fee-for-service | 42 733 (54.5) | 22 779 (53.9) | 28 489 (53.3) | 106 999 (54.2) | 201 000 (54.1) | |
| OLDW | 35 711 (45.5) | 19 486 (46.1) | 24 987 (46.7) | 90 514 (45.8) | 170 | |
| 698 (45.9) | ||||||
| Comorbidities, N (%) | ||||||
| Acute MI | 170 (0.2) | 87 (0.2) | 110 (0.2) | 477 (0.2) | 845 (0.3) | 0.00 |
| Coronary artery disease (other than acute MI) | 11 770 (15.0) | 6591 (15.6) | 7849 (14.7) | 29 688 (15.0) | 55 898 (15.0) | 0.03 |
| Acute stroke | 103 (0.1) | 57 (0.1) | 59 (0.1) | 268 (0.1) | 487 (0.1) | 0.01 |
| Cerebrovascular disease (other than acute stroke) | 4403 (5.6) | 2341 (5.5) | 2909 (5.4) | 11 119 (5.6) | 20 773 (5.6) | 0.01 |
| HHF | 143 (0.2) | 75 (0.2) | 77 (0.2) | 396 (0.2) | 691 (0.2) | 0.01 |
| Heart failure (other) | 1436 (1.8) | 801 (1.9) | 979 (1.8) | 3653 (1.8) | 6869 (1.8) | 0.00 |
| Revascularization procedure | 1222 (1.5) | 608 (1.5) | 763 (1.5) | 3192 (1.6) | 5785 (1.5) | 0.01 |
| Atrial fibrillation/flutter | 3330 (4.2) | 1792 (4.2) | 2237 (4.2) | 8380 (4.2) | 15 739 (4.2) | 0.00 |
| Hypertension | 63 933 (81.5) | 34 776 (82.3) | 43 295 (81.0) | 160 808 (81.4) | 302 812 (81.5) | 0.03 |
| Nephropathy with CKD less than stage 3 | 5977 (7.6) | 3284 (7.8) | 3842 (7.2) | 15 114 (7.7) | 28 218 (7.6) | 0.02 |
| Acute kidney injury | 1880 (2.4) | 903 (2.1) | 1110 (2.1) | 4827 (2.4) | 8720 (2.3) | 0.02 |
| CKD stage 3–4 | 3988 (5.1) | 2147 (5.1) | 2436 (4.6) | 10 023 (5.1) | 18 594 (5.0) | 0.02 |
| CKD stage 5, ESRD | 78 (0.1) | 41 (0.1) | 39 (0.1) | 203 (0.1) | 360 (0.1) | 0.01 |
| RRT | 267 (0.3) | 120 (0.3) | 160 (0.3) | 678 (0.3) | 1226 (0.3) | 0.01 |
| Peripheral vascular disease | 5073 (6.5) | 2746 (6.5) | 3406 (6.4) | 12 879 (6.5) | 24 103 (6.5) | 0.01 |
| Neuropathy | 13 979 (17.8) | 7733 (18.3) | 9473 (17.7) | 35 276 (17.9) | 66 461 (17.9) | 0.02 |
| Amputation | 248 (0.3) | 131 (0.3) | 146 (0.3) | 644 (0.3) | 1168 (0.3) | 0.01 |
| Other lower extremity complication | 1170 (1.5) | 618 (1.5) | 751 (1.4) | 3010 (1.5) | 5549 (1.5) | 0.01 |
| Diabetic retinopathy | 4122 (5.3) | 2250 (5.3) | 2758 (5.2) | 10 374 (5.3) | 19 504 (5.2) | 0.01 |
| Obesity | 26 390 (33.6) | 14 839 (35.1) | 18 628 (34.8) | 66 666 (33.8) | 126 522 (34.0) | 0.03 |
| Metabolic/bariatric surgery | 1086 (1.4) | 560 (1.3) | 745 (1.4) | 2711 (1.4) | 5102 (1.4) | 0.01 |
| Pancreatitis | 370 (0.5) | 179 (0.4) | 247 (0.5) | 1003 (0.5) | 1799 (0.5) | 0.01 |
| Cancer | 6874 (8.8) | 3562 (8.4) | 4560 (8.5) | 17 241 (8.7) | 32 237 (8.7) | 0.01 |
| Smoking (current) | 6866 (8.8) | 3682 (8.7) | 4584 (8.6) | 17 419 (8.8) | 32 551 (8.8) | 0.01 |
| Thyroid cancer | 284 (0.4) | 177 (0.4) | 186 (0.3) | 704 (0.4) | 1351 (0.4) | 0.01 |
| Genitourinary tract infection | 3718 (4.7) | 1893 (4.5) | 2395 (4.5) | 9434 (4.8) | 17 441 (4.7) | 0.01 |
| Cirrhosis | 696 (0.9) | 371 (0.9) | 451 (0.8) | 1738 (0.9) | 3257 (0.9) | 0.00 |
| Dementia | 871 (1.1) | 420 (1.0) | 487 (0.9) | 2204 (1.1) | 3982 (1.1) | 0.02 |
| Falls | 3204 (4.1) | 1709 (4.0) | 2259 (4.2) | 7988 (4.0) | 15 159 (4.1) | 0.01 |
| Urinary incontinence | 3007 (3.8) | 1572 (3.7) | 2024 (3.8) | 7605 (3.9) | 14 208 (3.8) | 0.01 |
| Unplanned hospitalizations | 6221 (7.9) | 3149 (7.4) | 4021 (7.5) | 15 885 (8.0) | 29 276 (7.9) | 0.02 |
| Medications, N (%) | ||||||
| Antiplatelet drug | 3469 (4.4) | 1974 (4.7) | 2289 (4.3) | 8878 (4.5) | 16 610 (4.5) | 0.02 |
| Diuretic | 30 828 (39.3) | 16 711 (39.5) | 20 830 (39.0) | 77 505 (39.2) | 145 874 (39.2) | 0.01 |
| RAAS inhibitor | 52 745 (67.2) | 28 665 (67.8) | 35 611 (66.6) | 132 696 (67.2) | 249 717 (67.2) | 0.03 |
| MRA | 1713 (2.2) | 1248 (3.0) | 1296 (2.4) | 4223 (2.1) | 8480 (2.3) | 0.05 |
| Beta-blocker | 25 887 (33.0) | 14 214 (33.6) | 17 329 (32.4) | 65 173 (33.0) | 122 602 (33.0) | 0.03 |
| Calcium channel blocker | 19 846 (25.3) | 10 692 (25.3) | 13 282 (24.8) | 50 034 (25.3) | 93 854 (25.2) | 0.01 |
| Other antihypertensive | 3697 (4.7) | 2019 (4.8) | 2435 (4.6) | 9386 (4.8) | 17 537 (4.7) | 0.01 |
| Lipid-lowering medication | 53 460 (68.2) | 29 032 (68.7) | 36 247 (67.8) | 134 411 (68.1) | 253 150 (68.1) | 0.02 |
| Oral anticoagulant | 3845 (4.9) | 2073 (4.9) | 2584 (4.8) | 9674 (4.9) | 18 175 (4.9) | 0.00 |
| Metformin | 61 924 (78.9) | 33 518 (79.3) | 42 209 (78.9) | 155 758 (78.9) | 293 409 (78.9) | 0.01 |
| Thiazolidinedione | 3491 (4.5) | 2086 (4.9) | 2527 (4.7) | 8945 (4.5) | 17 049 (4.6) | 0.02 |
| Sacubitril/valsartan | 65 (0.1) | 36 (0.1) | 133 (0.2) | 92 (0.0) | 325 (0.1) | 0.05 |
CKD = chronic kidney disease; DPP-4i = dipeptidyl peptidase-4 inhibitors; ESRD = end stage renal disease; GLP-1 RA = glucagon-like peptide-1 receptor agonists; HHF = hospitalization for heart failure; MI = myocardial infarction; MRA = mineralocorticoid receptor antagonist; OLDW = OptumLabs Data Warehouse; RAAS = renin angiotensin aldosterone system; RRT = renal replacement therapy; SD = standard deviation; SGLT2i = sodium-glucose cotransporter 2 inhibitors; SMD = standardized mean difference; SU = sulfonylurea.
Because all baseline covariates were balanced across treatment groups, our final Cox models included only medication group as the independent variable. The intention-to-treat analyses for composite sight-threatening diabetic retinopathy (primary outcome; Fig 1), DME treatment, PDR treatment, and clinical coding for blindness outcomes are summarized in Table 8. The probability of treatment for composite sight-threatening retinopathy within 2 and 5 years of index date was 0.3% and 0.7%, respectively, for patients initiating SGLT2i, 0.4% and 1.0%, respectively, for GLP-1 RA, 0.4% and 0.9%, respectively, for DPP-4i, and 0.5% and 1.2%, respectively, for sulfonylurea (Table S4). Treatment for DME and PDR, respectively, within 5 years of index date was 0.6% and 0.3%, respectively, for patients initiating SGLT2i, 0.8% and 0.4%, respectively, for GLP-1 RA, 0.7% and 0.4%, respectively, for DPP-4i, and 0.9% and 0.5%, respectively, for sulfonylurea (Table S4). Sodium-glucose cotransporter 2 inhibitors use was associated with a lower risk of treatment for sight-threatening retinopathy compared with all other medication classes, including GLP-1 RA (HR, 0.73; 95% CI, 0.55–0.97), DPP-4i (HR, 0.79; 95% CI, 0.64–0.97), and sulfonylurea (HR, 0.61; 95% CI, 0.50–0.74). Glucagon-like peptide-1 receptor agonists use was associated with a similar risk of sight-threatening retinopathy as DPP-4i (HR, 1.07; 95% CI, 0.85–1.35) and sulfonylurea (HR, 0.83; 95% CI, 0.67–1.03). When analyzing DME and PDR outcomes individually, the pairwise HRs for each retinopathy complication were similar to those in the primary composite outcome analyses, with the exception that PDR treatment risk with SGLT2i versus GLP-1 RA (HR, 0.73; 95% CI, 0.48–1.13) was not significant. There was a lower risk of diagnostic coding for blindness outcomes with SGLT2i compared with both GLP-1 RA (HR, 0.69; 95% CI, 0.56–0.85) and sulfonylurea (HR, 0.81; 95% CI, 0.71–0.93) and an increased risk of blindness outcomes with GLP-1 RA versus DPP-4i (HR, 1.27; 95% CI, 1.07–1.51).
Figure 1.
Cumulative incidence of treatment for sight-threatening retinopathy (diabetic macular edema and/or proliferative diabetic retinopathy). DPP-4i = dipeptidyl peptidase-4 inhibitors; GLP-1 RA = glucagon-like peptide-1 receptor agonists; SGLT2i = sodium-glucose cotransporter 2 inhibitors.
Table 8.
Association between Glucose-Lowering Treatment and Sight-Threatening Retinopathy Outcomes: Intention-to-Treat Analysis
| DME and/or PDR HR (95% CI) |
DME HR (95% CI) |
PDR HR (95% CI) |
Blindness HR (95% CI) |
|
|---|---|---|---|---|
|
| ||||
| SGLT2i vs. GLP-1 RA | 0.73 (0.55-0.97) | 0.72 (0.52-0.99) | 0.73 (0.48-1.13) | 0.69 (0.56-0.85) |
| SGLT2i vs. DPP-4i | 0.79 (0.64-0.97) | 0.76 (0.60-0.96) | 0.73 (0.53-0.99) | 0.88 (0.76-1.01) |
| SGLT2i vs. SU | 0.61 (0.50-0.74) | 0.58 (0.47-0.73) | 0.56 (0.42-0.75) | 0.81 (0.71-0.93) |
| GLP-1 RA vs. DPP-4i | 1.07 (0.85-1.35) | 1.06 (0.82-1.37) | 0.99 (0.69-1.40) | 1.27 (1.07-1.51) |
| GLP-1 RA vs. SU | 0.83 (0.67-1.03) | 0.81 (0.64-1.04) | 0.76 (0.55-1.06) | 1.19 (1.00-1.39) |
| SU vs. DPP-4i | 1.29 (1.17-1.42) | 1.30 (1.16-1.45) | 1.30 (1.11-1.51) | 1.07 (1.00-1.15) |
| Wald P value | < 0.001 | < 0.001 | < 0.001 | 0.001 |
CI = confidence interval; DME = diabetic macular edema; DPP-4i = dipeptidyl peptidase-4 inhibitors; GLP-1 RA = glucagon-like peptide-1 receptor agonists; HR = hazard ratio; PDR = proliferative diabetic retinopathy; SGLT2i = sodium-glucose cotransporter 2 inhibitors; SU = sulfonylurea.
Because of evidence of nonproportional hazards in the intention-to-treat analyses, we replicated the main models including only patients with follow-up of ≤1 year, >1 year but ≤2 years, and >2 years (Fig 2 and Table S9, available at www.ophthalmologyretina.org). The relative benefit of SGLT2i for all retinopathy outcomes, as well as a relative increased retinopathy risk with sulfonylurea agents, was consistent across time periods.
Figure 2.
Association between glucose-lowering treatment and sight-threatening retinopathy outcomes by duration of treatment: intention-to-treat analysis. DME = diabetic macular edema; DPP-4i = dipeptidyl peptidase-4 inhibitors; GLP-1 RA = glucagon-like peptide-1 receptor agonists; PDR = proliferative diabetic retinopathy; SGLT2i = sodium-glucose cotransporter 2 inhibitors; SU = sulfonylurea.
The 2 planned sensitivity analyses performed under an as-treated approach showed relatively consistent results (Table S5). Sulfonylurea use continued to be associated with an increased retinopathy risk across pairwise comparisons in both sensitivity analyses. Sodium-glucose cotransporter 2 inhibitors and GLP-1 RA were similar with respect to retinopathy treatment risk when censoring on discontinuation of the assigned treatment or the addition of another glucose-lowering drug, whichever came first.
Discussion
In this cohort of approximately 370 000 patients with T2D and without advanced diabetic eye disease beginning treatment with SGLT2i, GLP-1 RA, DPP-4i, and sulfonylureas, SGLT2i use was associated with a lower risk of treatment for composite sight-threatening diabetic retinopathy (DME and PDR) compared with all other classes, including a 39% risk reduction versus sulfonylurea, 27% risk reduction versus GLP-1 RA, and 21% risk reduction versus DPP-4i. Sodium-glucose cotransporter 2 inhibitors use was associated with a lower risk of treatment for DME relative to all other classes, and the risk of treatment for PDR was lower among initial SGLT2i users than in patients who received DPP-4i and sulfonylureas. Patients receiving GLP-1 RA were treated for composite sight-threatening retinopathy, DME, and PDR at similar rates to those starting DPP-4i and sulfonylurea medications. Thus, in this first head-to-head comparison of 4 classes of glucose-lowering drugs, GLP-1 RA use was not associated with heighted risk of advanced diabetic eye disease compared with other drugs, whereas SGLT2i was associated with a lower risk of these complications.
Improving our understanding of the relative impacts of SGLT2i and GLP-1 RA agents on retinopathy outcomes in T2D is critically important because their use has rapidly increased in the wake of strong evidence for cardiorenal protection and weight loss.29–31 In 2019, the American Diabetes Association updated its Standards of Medical Care in Diabetes to recommend use of either an SGLT2i or GLP-1 RA agent for patients with T2D and established atherosclerotic CVD independent of glycemic control status (level A), a recommendation that was endorsed by the American College of Cardiology.32 By 2020, an estimated 9.4% of Americans with T2D were treated with SGLT2i agents, a 21-fold increase from 2013.33 American Diabetes Association clinical guidelines as of 2024 recommend use of SGLT2i or GLP-1 RA for T2D and established or multiple risk factors for established atherosclerotic CVD (independent of hemoglobin A1c [HbA1c] or metformin use), use of SGLT2i for T2D and established heart failure (with reduced or preserved ejection fraction), use of SGLT2i (if estimated glomerular filtration rate of 20–60 ml/min/1.73 m2 or albuminuria) or GLP-1 RA (if estimated glomerular filtration rate <30 ml/min/1.73 m2) for established chronic kidney disease, considered use of either GLP-1 RA or SGLT2i as a part of antihyperglycemic management for those with T2D and overweight, and preferential use of GLP-1 RA agents before insulin for glycemic control in T2D.34,35 These rapidly evolving practice patterns offer substantial systemic health benefits for those living with T2D. Elucidating the relative impact of these medications on retinopathy risk will help inform the continued optimization of pharmacotherapeutic strategies in T2D and will potentially aid in improving targeted retinal surveillance for complications requiring treatment.
When assessing the impact of antihyperglycemic therapies on retinal complications, one must consider that more rapid and effective glycemic control can be associated with early worsening of retinopathy, while ultimately lowering retinopathy risk over a longer-term. This paradoxical effect was observed among patients with type 1 diabetes in the Diabetes Control and Complications Trial when early retinopathy worsening was seen in patients who were randomized to intensive insulin treatment compared with those receiving less strict conventional treatment (Diabetes Control and Complications Trial). After 18 months, however, retinopathy outcomes became favorable for the cohort receiving intensive insulin treatment, and tighter glycemic control during the study period was shown to offer durable retinal protection years later. Furthermore, those with intensive treatment and early worsening of retinopathy ultimately had favorable retinal outcomes to those with less strict glycemic control and no early worsening.36
Early retinopathy worsening has been less well-documented in T2D, in part because pivotal T2D trials have mostly lacked interim retinopathy severity assessments within the first 2 years of treatment when early worsening would occur.37 In the SUSTAIN-6 trial, however, early retinopathy worsening because of rapid glycemic control has been proposed as a potential explanation for the increased rate of retinal complications with semaglutide versus placebo (3.0% vs. 1.8%, HR, 1.76; 95% CI, 1.1–2.8).38 Participants receiving semaglutide experienced rapid lowering of HbA1c within 16 weeks and had a 1.1% to 1.4% HbA1c reduction over 2 years compared with 0.4% reduction in placebo groups. Because of the relatively shorter trial duration, it was unclear whether the favorable glycemic control would ultimately benefit patients beyond the 2-year end point. In our emulated target trial, secondary analyses of retinal outcomes at either shorter (≤ 1 or ≤ 2 years) or longer (> 2 years) follow-up (Fig 2) were mostly consistent with the primary analyses. The risk of retinopathy outcomes with GLP-1 RA use relative to other glucose-lowering medications was similar at shorter and longer durations of use. Findings of relatively lower retinopathy risk with SGLT2i and relatively increased retinopathy risk with sulfonylurea were consistent across time comparisons.
Our findings expand on previous investigations of retinopathy outcomes with glucose-lowering therapies in routine clinical practice. A Korean administrative database study found patients receiving SGLT2i had a lower hazard for treatment with laser photocoagulation or vitrectomy for PDR compared with DPP-4i over a 2-year period (HR, 0.89; 95% CI, 0.83–0.97).39 Another retrospective cohort study of 23 378 Taiwanese patients receiving SGLT2i or GLP-1 RA identified a lower hazard with SGLT2i use for diagnostic coding of PDR (HR, 0.53; 95% CI, 0.42–0.68), intravitreal injection (HR, 0.65; 95% CI, 0.47–0.91), and laser photocoagulation (HR, 0.59; 95% CI, 0.47–0.74).40 In this context, our findings provide compelling evidence that SGLT2i pharmacotherapy offers a relatively favorable diabetic retinopathy risk profile compared with other glucose-lowering therapies.
As with all observational studies, these analyses have the inherent limitation that unobserved confounding may have impacted results. Another important limitation is the lack of clinical variables such as HbA1c and body weight in the source claims data, which prevents more robust analysis of associations between systemic and ocular outcomes. Additionally, drug class-based analyses must be viewed with an understanding that intraclass differences between medications may limit extrapolation of composite class outcomes to any specific drug.41
Despite these limitations, this analysis of a large geographically and demographically diverse population comparing head-to-head 4 classes of glucose-lowering medications contribute valuable evidence from routine clinical practice on the relative impact of these T2D therapeutics on retinal complications. Existing evidence from clinical trials lacks head-to-head comparisons of these classes and has relatively short follow-up and suboptimal methodologies for evaluating ocular outcomes. In addition, the majority of individuals living with T2D in the United States have a moderate CVD risk profile,21 which limits the generalizability of outcomes from clinical trials consisting primarily of individuals at high CVD risk. Widespread replication of RCT investigations for moderate CVD risk populations may not be forthcoming as significantly lower event rates would require much larger study populations, but understanding the relative retinal risk in this moderate CVD risk population is particularly important because these patients are less likely than those with T2D and high CVD risk to have an overriding indication for either SGLT2i or GLP-1 RA. The emulated target trial design offers an opportunity to perform robust analyses on individuals at moderate CVD risk in a more resource-efficient manner.
Given the effectiveness of many newer glucose-lowering agents at improving glycemic control, patients with elevated HbA1c initiating any antihyperglycemic medication anticipated to rapidly improve control may benefit from temporarily shortening their retinal screening interval to monitor for early retinopathy worsening that requires treatment. In this study, although the specific mechanisms contributing to observed interclass differences remain unclear, the relative protective effect of SGLT2i medications for sight-threatening retinopathy was significant, as was the lack of increased retinal risk with GLP-1 RA medications compared with DPP-4i and sulfonylureas. As we continue to improve our understanding of the comparative effectiveness and cost-effectiveness of these medications, we will need to integrate new evidence on meaningful outcomes, including sight-threatening diabetic retinopathy, into clinical guidelines for patients with T2D, and work toward ensuring equitable access to higher cost medications when beneficial.
In this large emulated target trial comparing the relative impact of GLP-1 RA, SGLT2i, DPP-4i, and sulfonylurea medications on the risk sight-threatening diabetic retinopathy in patients with T2D with moderate CVD risk, SGLT2i use was protective against advanced diabetic retinal complications compared with other glucose-lowering therapies. Furthermore, GLP-1 RA did not increase the risk of sight-threatening retinopathy compared with DPP-4i and sulfonylurea medications in either the short-term or long-term. In the context of rapidly increasing use of SGLT2i and GLP-1 RA agents, and, given the inherent limitations of existing clinical trial data, these robust findings from routine clinical practice will be reassuring for clinicians and people living with T2D seeking the well-established cardiovascular, renal, and weight management benefits of these therapies.
Supplementary Material
Supplemental material available at www.ophthalmologyretina.org.
Disclosure(s):
All authors have completed and submitted the ICMJE disclosures form.
The authors have made the following disclosures:
J.H.: Grant support – Centers for Medicare and Medicaid Services, NIH, Agency for Healthcare Research and Quality, Patient-Centered Outcomes Research Institute, American Heart Association, Pfizer.
G.E.U.: Grant support – Dexcom, Abbott, Bayer; Advisor – Dexcom, GlyCare.
R.J.G.: Grants support – Boehringer Ingelheim, Dexcom, Novo Nordisk, Eli Lilly; Consultant – Abbott, Bayer, Boehringer Ingelheim, Dexcom, Novo Nordisk, Eli Lilly, Medtronic, AstraZeneca; Honoraria – Eli Lilly.
J.S.R.: Grant support –Johnson & Johnson, Medical Device Innovation Consortium, Food and Drug Administration, Agency for Healthcare Research and Quality, National Heart, Lung and Blood Institute, Arnold Ventures; Expert witness – Greene Law Firm.
R.G.M.: Grants support – National Institute of Diabetes and Digestive and Kidney Diseases, National Institute on Aging, National Center for Advancing Translational Sciences, AARP, American Diabetes Association; Consultant – EmmiEducate, Yale New Haven Health System; Received payment or honoraria – American Diabetes Association; Meetings/travel support – American Diabetes Association.
The other authors have no proprietary or commercial interest in any materials discussed in this article.
Research reported in this work was funded through a Patient-Centered Outcomes Research Institute (PCORI) Award (DB-2020C2–20306). R.J.G. was supported in part by the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) of the National Institutes of Health (NIH) under Award Numbers P30DK111024, K23DK123384, and R03DK138255.
Abbreviations and Acronyms:
- CI
confidence interval
- CVD
cardiovascular disease
- DME
diabetic macular edema
- DPP-4i
dipeptidyl peptidase-4 inhibitor
- DR
diabetic retinopathy
- GLP-1 RA
glucagon-like peptide-1 receptor agonist
- HbA1c
hemoglobin A1c
- HR
hazard ratio
- IQR
interquartile range
- PDR
proliferative diabetic retinopathy
- RCT
randomized controlled trial
- SGLT2i
sodium-glucose cotransporter 2 inhibitor
- T2D
type 2 diabetes
Footnotes
The statements in this report are solely the responsibility of the authors and do not necessarily represent the views of the Patient-Centered Outcomes Research Institute (PCORI), its Board of Governors or Methodology Committee.
HUMAN SUBJECTS: Human subjects were included in this study. The study was exempt from the Mayo Clinic Institutional Review Board review and the requirement for informed consent was waived. The study adhered to the Declaration of Helsinki and is reported according to the RECORD and START-RWE reporting guidelines.
No animal subjects were used in this study.
This study was conducted using deidentified data from OptumLabs Data Warehouse and linked 100% sample of Medicare fee-for-service claims. These data are third party data owned by OptumLabs and contain sensitive patient information; therefore, the data are only available upon request. Interested researchers engaged in HIPAA compliant research may contact connected@optum.com for data access requests. The data use requires researchers to pay for rights to use and access the data. These data are subject to restrictions on sharing as a condition of access.
References
- 1.Flaxel CJ, Adelman RA, Bailey ST, et al. Diabetic retinopathy preferred practice pattern®. Ophthalmology. 2020;127:P66–P145. Published correction appears in Ophthalmology. 2020;127:1279. [DOI] [PubMed] [Google Scholar]
- 2.Sun H, Saeedi P, Karuranga S, et al. IDF Diabetes Atlas: global, regional and country-level diabetes prevalence estimates for 2021 and projections for 2045. Diabetes Res Clin Pract. 2022;183:109119. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Buehler AM, Cavalcanti AB, Berwanger O, et al. Effect of tight blood glucose control versus conventional control in patients with type 2 diabetes mellitus: a systematic review with meta-analysis of randomized controlled trials. Cardiovasc Ther. 2013;31:147–160. [DOI] [PubMed] [Google Scholar]
- 4.UK Prospective Diabetes Study (UKPDS) Group. Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33). Lancet. 1998;352:837–853. [PubMed] [Google Scholar]
- 5.Maeda S, Yamagishi S, Matsui T, et al. Beneficial effects of vildagliptin on retinal injury in obese type 2 diabetic rats. Ophthalmic Res. 2013;50:221–226. [DOI] [PubMed] [Google Scholar]
- 6.Gonçalves A, Marques C, Leal E, et al. Dipeptidyl peptidase-IV inhibition prevents blood-retinal barrier breakdown, inflammation and neuronal cell death in the retina of type 1 diabetic rats. Biochim Biophys Acta. 2014;1842:1454–1463. [DOI] [PubMed] [Google Scholar]
- 7.Lee CS, Kim YG, Cho HJ, et al. Dipeptidyl peptidase-4 inhibitor increases vascular leakage in retina through VE-cadherin phosphorylation. Sci Rep. 2016;6:29393. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Dietrich N, Kolibabka M, Busch S, et al. The DPP4 inhibitor linagliptin protects from experimental diabetic retinopathy. PLoS One. 2016;11:e0167853. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Hernández C, Bogdanov P, Corraliza L, et al. Topical administration of GLP-1 receptor agonists prevents retinal neurodegeneration in experimental diabetes. Diabetes. 2016;65:172–187. [DOI] [PubMed] [Google Scholar]
- 10.Takakura S, Toyoshi T, Hayashizaki Y, Takasu T. Effect of ipragliflozin, an SGLT2 inhibitor, on progression of diabetic microvascular complications in spontaneously diabetic Torii fatty rats. Life Sci. 2016;147:125–131. [DOI] [PubMed] [Google Scholar]
- 11.Kang YM, Jung CH. Effects of incretin-based therapies on diabetic microvascular complications. Endocrinol Metab (Seoul). 2017;32:316–325. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Khat DZ, Husain M. Molecular mechanisms underlying the cardiovascular benefits of SGLT2i and GLP-1RA. Curr Diab Rep. 2018;18:45. [DOI] [PubMed] [Google Scholar]
- 13.Zhou HR, Ma XF, Lin WJ, et al. Neuroprotective role of GLP-1 analog for retinal ganglion cells via PINK1/Parkin-mediated mitophagy in diabetic retinopathy. Front Pharmacol. 2020;11:589114. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Hu Y, Xu Q, Li H, et al. Dapagliflozin reduces apoptosis of diabetic retina and human retinal microvascular endothelial cells through ERK1/2/cPLA2/AA/ROS pathway independent of hypoglycemic. Front Pharmacol. 2022;13:827896. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Marso SP, Bain SC, Consoli A, et al. Semaglutide and cardiovascular outcomes in patients with type 2 diabetes. N Engl J Med. 2016;375:1834–1844. [DOI] [PubMed] [Google Scholar]
- 16.Tang H, Li G, Zhao Y, et al. Comparisons of diabetic retinopathy events associated with glucose-lowering drugs in patients with type 2 diabetes mellitus: a network meta-analysis. Diabetes Obes Metab. 2018;20:1262–1279. [DOI] [PubMed] [Google Scholar]
- 17.Benchimol EI, Smeeth L, Guttmann A, et al. The reporting of studies conducted using observational routinely-collected health data (RECORD) statement. PLoS Med. 2015;12: e1001885. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Wang SV, Pinheiro S, Hua W, et al. STaRT-RWE: structured template for planning and reporting on the implementation of real world evidence studies. BMJ. 2021;372:m4856. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Wallace PJ, Shah ND, Dennen T, et al. Optum Labs: building a novel node in the learning health care system. Health Aff (Millwood). 2014;33:1187–1194. Published correction appears in Health Aff (Millwood). 2014;33:1703. [DOI] [PubMed] [Google Scholar]
- 20.OptumLabs. OptumLabs and OptumLabs Data Warehouse (OLDW) Descriptionsand Citation. n.p.; 2023. Reproducedwith permission from OptumLabs. https://www.optumlabs.com/. [Google Scholar]
- 21.McCoy RG, Swarna KS, Deng Y, et al. Derivation of an annualized claims-based major adverse cardiovascular event estimator in type 2 diabetes. JACC Adv. 2024;3:100852. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Polley EC, LeDell E, Kennedy C, et al. SuperLearner: superlearner prediction. https://CRAN.R-project.org/package=SuperLearner; 2024. Accessed October 4, 2023. [Google Scholar]
- 23.McCaffrey DF, Griffin BA, Almirall D, et al. A tutorial on propensity score estimation for multiple treatments using generalized boosted models. Stat Med. 2013;32:3388–3414. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.van der Laan MJ, Polley EC, Hubbard AE. Super learner. Stat Appl Genet Mol Biol. 2007;6:Article25. [DOI] [PubMed] [Google Scholar]
- 25.Yoshida K, Hernández-Díaz S, Solomon DH, et al. Matching weights to simultaneously compare three treatment groups: comparisontothree-waymatching. Epidemiology.2017;28:387–395. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Austin PC, Lee DS, Fine JP. Introduction to the analysis of survival data in the presence of competing risks. Circulation. 2016;133:601–609. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Moore KL, Van der Laan MJ. Application of time-to-event methods in the assessment of safety in clinical trials. In: Peace KE, ed. Design and Analysis of Clinical Trials with Time-to-Event Endpoints. Biostatistics Series. Chapman and Hall/CRC; 2009:455–482. [Google Scholar]
- 28.Grambsch PM, Therneau TM. Proportional hazards tests and diagnostics based on weighted residuals. Biometrika. 1994;81:515–526. [Google Scholar]
- 29.Galindo RJ, Trujillo JM, Low Wang CC, McCoy RG. Advances in the management of type 2 diabetes in adults. BMJ Med. 2023;2:e000372. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.McGuire DK, Shih WJ, Cosentino F, et al. Association of SGLT2 inhibitors with cardiovascular and kidney outcomes in patients with type 2 diabetes: a meta-analysis. JAMA Cardiol. 2021;6:148–158. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Kristensen SL, Rørth R, Jhund PS, et al. Cardiovascular, mortality, and kidney outcomes with GLP-1 receptor agonists in patients with type 2 diabetes: a systematic review and meta-analysis of cardiovascular outcome trials. Lancet Diabetes Endocrinol. 2019;7:776–785. [DOI] [PubMed] [Google Scholar]
- 32.American Diabetes Association. 10. Cardiovascular disease and risk management: Standards of Medical Care in Diabetes-2019. Diabetes Care. 2019;42(suppl 1):S103–S123. [DOI] [PubMed] [Google Scholar]
- 33.Shen TH, Farley JF. Utilization and impact of SLGT2 inhibitors among diabetes patients in a nationally representative survey: findings from NHANES 2013–2020. J Diabetes Complications. 2023;37:108625. [DOI] [PubMed] [Google Scholar]
- 34.American Diabetes Association Professional Practice Committee. 9. Pharmacologic approaches to glycemic treatment: standards of care in diabetes-2024. Diabetes Care. 2024;47(suppl 1):S158–S178. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.American Diabetes Association Professional Practice Committee. 10. Cardiovascular disease and risk management: standards of care in diabetes-2024. Diabetes Care. 2024;47(suppl 1):S179–S218. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Diabetes Control and Complications Trial (DCCT)/Epidemiology of Diabetes Interventions and Complications (EDIC) Research Group, Lachin JM, White NH, et al. Effect of intensive diabetes therapy on the progression of diabetic retinopathy in patients with type 1 diabetes: 18 years of follow-up in the DCCT/EDIC. Diabetes. 2015;64:631–642. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Bain SC, Klufas MA, Ho A, Matthews DR. Worsening of diabetic retinopathy with rapid improvement in systemic glucose control: a review. Diabetes Obes Metab. 2019;21:454–466. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Bethel MA, Diaz R, Castellana N, et al. HbA1c change and diabetic retinopathy during GLP-1 receptor agonist cardiovascular outcome trials: a meta-analysis and meta-regression. Diabetes Care. 2021;44:290–296. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Chung YR, Ha KH, Lee K, Kim DJ. Effects of sodium-glucose cotransporter-2 inhibitors and dipeptidyl peptidase-4 inhibitors on diabetic retinopathy and its progression: a real-world Korean study. PLoS One. 2019;14:e0224549. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Lin TY, Kang EY, Shao SC, et al. Risk of diabetic retinopathy between sodium-glucose cotransporter-2 inhibitors and glucagon-like peptide-1 receptor agonists. Diabetes Metab J. 2023;47:394–404. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Kapoor I, Sarvepalli SM, D’Alessio D, et al. GLP-1 receptor agonists and diabetic retinopathy: a meta-analysis of randomized clinical trials. Surv Ophthalmol. 2023;68:1071–1083. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.