Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2024 Jun 6.
Published in final edited form as: Nat Cardiovasc Res. 2024 Apr 3;3(4):431–440. doi: 10.1038/s44161-024-00453-9

Effectiveness of glucose-lowering medications on cardiovascular outcomes in patients with type 2 diabetes at moderate cardiovascular risk

Rozalina G McCoy 1,2,3,4, Jeph Herrin 5, Kavya Sindhu Swarna 4,6, Yihong Deng 4,6, David M Kent 7, Joseph S Ross 8,9, Guillermo E Umpierrez 10, Rodolfo J Galindo 11, William H Crown 12, Bijan J Borah 6, Victor M Montori 13,14, Juan P Brito 13,14, Joshua J Neumiller 15,16, Mindy M Mickelson 6, Eric C Polley 17
PMCID: PMC11156225  NIHMSID: NIHMS1993206  PMID: 38846711

Abstract

Cardiovascular disease (CVD) is the leading cause of death among people with type 2 diabetes1–5, most of whom are at moderate CVD risk6, yet there is limited evidence on the preferred choice of glucose-lowering medication for CVD risk reduction in this population. Here, we report the results of a retrospective cohort study where data for US adults with type 2 diabetes and moderate risk for CVD are used to compare the risks of experiencing a major adverse cardiovascular event with initiation of glucagon-like peptide-1 receptor agonists (GLP-1RA; n = 44,188), sodium-glucose cotransporter 2 inhibitors (SGLT2i; n = 47,094), dipeptidyl peptidase-4 inhibitors (DPP4i; n = 84,315) and sulfonylureas (n = 210,679). Compared to DPP4i, GLP-1RA (hazard ratio (HR) 0.87; 95% confidence interval (CI) 0.82–0.93) and SGLT2i (HR 0.85; 95% CI 0.81–0.90) were associated with a lower risk of a major adverse cardiovascular event, whereas sulfonylureas were associated with a higher risk (HR 1.19; 95% CI 1.16–1.22). Thus, GLP-1RA and SGLT2i may be the preferred glucose-lowering agents for cardiovascular risk reduction in patients at moderate baseline risk for CVD. ClinicalTrials.gov registration: NCT05214573.

CVD is the leading cause of death among people living with type 2 diabetes1–5. Randomized controlled trials (RCTs) of glucose-lowering medications that focused on cardiovascular outcomes have revealed the benefits of GLP-1RA and SGLT2i compared to placebo or sulfonylureas, with respect to cardiovascular (both classes), heart failure (SGLT2i only) and mortality (both classes) outcomes in patients with or at high risk for CVD6,7. While these RCTs generated great interest in using GLP-1RA and SGLT2i for patients with established CVD8,9, they have important limitations that left optimal medication management unclear for most patients with type 2 diabetes. Specifically, these trials focused exclusively on patients with established or high risk for CVD, not patients at moderate risk who make up the majority of people with type 2 diabetes10, and their comparisons were limited to placebo or sulfonylureas, with no data on the comparative effectiveness between GLP-1RAs, SGLT2is and DPP4i. Although multiple meta-analyses11–19 have attempted to extend the available evidence from RCTs to pairwise comparisons across drug classes, these findings are still of limited utility due to limitations of the underlying trial data (that is, focus on high-risk patients only). The recently completed Glycemia Reduction Approaches in Type 2 Diabetes: A Comparative Effectiveness (GRADE) trial, which focused on glycemic control rather than cardiovascular risk reduction, did not demonstrate conclusive evidence of differences in cardiovascular outcomes (examined as secondary endpoints) in patients with low baseline CVD risk treated with the GLP-1RA liraglutide, DPP4i sitagliptin, sulfonylurea glimepiride and insulin glargine (there was no SGLT2i arm). Whether there are cardiovascular benefits with GLP-1RA or SGLT2i use among patients at moderate risk for CVD, as well as whether there are differences between GLP-1RA and SGLT2i in this patient population, is unknown20.

As a result, people living with diabetes and their clinicians lack the evidence necessary to make treatment decisions in the absence of high CVD risk. Given the operational complexity and cost required to conduct an RCT that could address these knowledge gaps, rendering such a trial infeasible, real-world data offer the opportunity to emulate an idealized comparative effectiveness target trial across diverse settings and populations at a fraction of the time and cost. Accordingly, we leveraged linked administrative claims data of enrollees in commercial, Medicare Advantage and Medicare fee-for-service health plans across the United States to emulate a target RCT comparing the effectiveness of GLP-1RA, SGLT2i, DPP4i and sulfonylureas in adults with type 2 diabetes at moderate risk of CVD, either as first-line or second-line therapy, with regard to major adverse cardiovascular events (MACE). To minimize the risks of confounding and bias by indication that are inherent to observational studies using real-world data, we used propensity score inverse probability of treatment weighting to balance the study arms on measured confounders and followed an a priori specified analytic plan in accordance with the target trial framework21.

We identified 386,276 adults with type 2 diabetes who first started a DPP4i (n = 84,315), GLP-1RA (n = 44,188), SGLT2i (n = 47,094) or sulfonylurea (n = 210,679) drug (Supplementary Fig. 1). Their baseline characteristics are shown in Supplementary Table 1. Before inverse probability weighting, there were differences between the treatment groups (maximum standardized mean differences (s.m.d.) ≥ 0.10) in most variables (Supplementary Table 2). After inverse probability weighting, the final study population was balanced on all covariates (Table 1 and Supplementary Table 3). The final (weighted) study population included 82,438 patients in the DPP4i group (mean age 65.8 (s.d. 8.4), 75.0% non-Hispanic white, 50.9% male, 79.0% on metformin), 44,255 patients in the GLP-1RA group (mean age 65.5 (s.d. 8.3), 75.9% non-Hispanic white, 51.1% male, 79.3% on metformin), 46,473 patients in the SGLT2i group (mean age 65.5 (s.d. 8.4), 75.6% non-Hispanic white, 50.6% male, 79.0% on metformin) and 207,186 patients in the sulfonylurea group (mean age 65.7 (s.d. 8.4), 74.8% non-Hispanic white, 51.0% male, 78.9% on metformin). The mean follow-up times were 1,262 days (interquartile range (IQR) 690–1,318) for the DPP4i group, 674 days (IQR 259–803) for the GLP-1RA group, 830 days (IQR 343–945) for the SGLT2i group and 1,221 days (IQR 665–1,285) for the sulfonylurea group.

Table 1 |.

Baseline patient characteristics, by treatment arm, after inverse probability of treatment weighting

DPP4i (n = 82,437.58) GLP-1RA (n = 44,254.83) SGLT2i (n = 46,472.99) Sulfonylurea (n = 207,186.18) Largest s.m.d.
Age, mean (s.d.) 65.76 (8.38) 65.47 (8.25) 65.46 (8.35) 65.69 (8.38) 0.04
Age group, n (%) 0.06
 <45 696.4 (0.8) 356.6 (0.8) 400.4 (0.9) 1,902.6 (0.9)
 45–49 3,239.8 (3.9) 1,782.3 (4.0) 1,848.7 (4.0) 8,014.5 (3.9)
 50–54 6,618.1 (8.0) 3,661.8 (8.3) 3,946.1 (8.5) 16,841.0 (8.1)
 55–59 8,679.1 (10.5) 4,740.4 (10.7) 5,094.5 (11.0) 21,847.4 (10.5)
 60–64 8,422.4 (10.2) 4,674.1 (10.6) 4,919.8 (10.6) 21,377.4 (10.3)
 65–69 23,394.6 (28.4) 13,185.8 (29.8) 13,412.9 (28.9) 59,435.0 (28.7)
 70–74 22,186.8 (26.9) 11,645.9 (26.3) 12,161.2 (26.2) 55,035.6 (26.6)
 ≥75 9,200.4 (11.2) 4,207.9 (9.5) 4,689.4 (10.1) 22,732.8 (11.0)
Sex, n (%) 0.01
 Female 40,506.38 (49.1) 21,653.53 (48.9) 22,971.49 (49.4) 101,537.38 (49.0)
 Male 41,931.2 (50.9) 22,601.3 (51.1) 23,501.5 (50.6) 105,648.8 (51.0)
Race/ethnicity, n (%) 0.04
 Asian 2,486.9 (3.0) 1,029.9 (2.3) 1,289.4 (2.8) 6,176.2 (3.0)
 Black 8,567.0 (10.4) 4,525.1 (10.2) 4,698.8 (10.1) 21,662.7 (10.5)
 Hispanic 7,016.2 (8.5) 3,693.6 (8.3) 3,901.3 (8.4) 17,839.4 (8.6)
 White 61,815.0 (75.0) 33,602.0 (75.9) 35,125.2 (75.6) 155,032.2 (74.8)
 Unknown 2,552.4 (3.1) 1,404.3 (3.2) 1,458.4 (3.1) 6,475.6 (3.1)
Region, n (%) 0.04
 Midwest 21,806.2 (26.5) 11,701.7 (26.4) 12,128.8 (26.1) 54,935.4 (26.5)
 Northeast 10,976.3 (13.3) 5,360.0 (12.1) 6,061.3 (13.0) 27,349.0 (13.2)
 South 38,555.7 (46.8) 20,999.2 (47.5) 21,918.5 (47.2) 96,976.0 (46.8)
 West 10,949.5 (13.3) 6,130.7 (13.9) 6,299.6 (13.6) 27,531.1 (13.3)
 Unknown 149.9 (0.2) 63.2 (0.1) 64.8 (0.1) 394.7 (0.2)
Prescriber specialty, n (%) 0.04
 Cardiology 84.6 (0.1) 54.4 (0.1) 93.7 (0.2) 175.5 (0.1)
 Endocrinology 1,086.7 (1.3) 672.3 (1.5) 706.0 (1.5) 2,704.0 (1.3)
 Nephrology 51.5 (0.1) 17.4 (0.0) 21.7 (0.0) 115.9 (0.1)
 Primary care 49,245.0 (59.7) 26,431.5 (59.7) 27,839.5 (59.9) 123,700.2 (59.7)
 Other 12,492.0 (15.2) 6,779.2 (15.3) 7,070.2 (15.2) 31,597.7 (15.3)
 Unknown 19,477.7 (23.6) 10,300.0 (23.3) 10,742.0 (23.1) 48,892.9 (23.6)
Index year, n (%) 0.07
 2014 10,756.0 (13.0) 5,480.4 (12.4) 5,069.1 (10.9) 26,956.8 (13.0)
 2015 11,369.2 (13.8) 5,724.1 (12.9) 6,269.9 (13.5) 28,492.6 (13.8)
 2016 11,779.3 (14.3) 6,086.5 (13.8) 6,657.3 (14.3) 29,466.5 (14.2)
 2017 12,720.7 (15.4) 6,720.9 (15.2) 7,354.6 (15.8) 31,863.5 (15.4)
 2018 12,468.2 (15.1) 6,867.1 (15.5) 7,203.3 (15.5) 31,221.4 (15.1)
 2019 12,164.9 (14.8) 6,819.3 (15.4) 7,158.7 (15.4) 30,438.3 (14.7)
 2020 5,389.9 (6.5) 3,096.8 (7.0) 3,216.0 (6.9) 13,786.5 (6.7)
 2021 5,789.4 (7.0) 3,459.7 (7.8) 3,544.1 (7.6) 14,960.5 (7.2)
Source data, n (%) 0.02
 OLDW 37,317 (45.3) 20,318 (45.9) 21,584 (46.4) 94,401 (45.6)
 Medicare fee-for-service 45,120.5 (54.7) 23,936.5 (54.1) 24,888.6 (53.6) 112,785.2 (54.4)
Comorbidities, n (%)
 Acute MI, months 1–9 144.2 (0.2) 72.7 (0.2) 83.5 (0.2) 393.8 (0.2) 0.01
 Acute MI, months 10–12 35.0 (0.0) 16.6 (0.0) 18.0 (0.0) 108.5 (0.1) 0.01
 Coronary artery disease (other than MI) 12,446.4 (15.1) 6,953.4 (15.7) 6,919.1 (14.9) 31,340.9 (15.1) 0.02
 Acute stroke, months 1–9 92.0 (0.1) 50.3 (0.1) 47.2 (0.1) 238.8 (0.1) 0.0
 Acute stroke, months 10–12 23.8 (0.0) 10.4 (0.0) 10.5 (0.0) 60.4 (0.0) 0.0
 Cerebrovascular disease (other than stroke) 4,755.8 (5.8) 2,515.6 (5.7) 2,594.5 (5.6) 11,979.1 (5.8) 0.01
 HHF, months 1–9 82.9 (0.1) 50.9 (0.1) 45.0 (0.1) 225.0 (0.1) 0.01
 HHF, months 10–12 68.8 (0.1) 27.1 (0.1) 30.7 (0.1) 181.0 (0.1) 0.01
 Heart failure (other than HHF) 1,511.2 (1.8) 836.4 (1.9) 889.3 (1.9) 3,834.0 (1.9) 0.01
 Revascularization, months 1–9 884.3 (1.1) 447.2 (1.0) 461.8 (1.0) 2,269.2 (1.1) 0.01
 Revascularization, months 10–12 423.1 (0.5) 212.9 (0.5) 223.7 (0.5) 1,128.2 (0.5) 0.01
 Atrial fibrillation/flutter 3,521.5 (4.3) 1,895.4 (4.3) 1,965.6 (4.2) 8,820.9 (4.3) 0.0
 Hypertension 67,381.5 (81.7) 36,506.5 (82.5) 37,741.6 (81.2) 169,182.4 (81.7) 0.03
 Nephropathy with CKD <stage 3 6,366.4 (7.7) 3,475.0 (7.9) 3,402.8 (7.3) 16,052.4 (7.7) 0.02
 CKD, stages 3–4 4,298.9 (5.2) 2,317.7 (5.2) 2,160.3 (4.6) 10,746.1 (5.2) 0.03
 CKD stage 5 or ESKD 86.7 (0.1) 44.3 (0.1) 38.1 (0.1) 224.3 (0.1) 0.01
 Renal replacement therapy 278.3 (0.3) 140.1 (0.3) 140.5 (0.3) 714.5 (0.3) 0.01
 Acute kidney injury 2,010.9 (2.4) 972.9 (2.2) 998.8 (2.1) 5,143.4 (2.5) 0.02
 Peripheral vascular disease 5,443.6 (6.6) 2,918.8 (6.6) 3,037.5 (6.5) 13,801.5 (6.7) 0.01
 Neuropathy 14,960.1 (18.1) 8,210.1 (18.6) 8,389.7 (18.1) 37,657.2 (18.2) 0.01
 Amputation 280.8 (0.3) 144.4 (0.3) 141.2 (0.3) 723.7 (0.3) 0.01
 Other lower extremity complications 1,269.2 (1.5) 659.7 (1.5) 687.5 (1.5) 3,274.9 (1.6) 0.01
 Retinopathy 7,392.3 (9.0) 3,975.4 (9.0) 4,025.4 (8.7) 18,634.9 (9.0) 0.01
 Retinopathy treatment 353.1 (0.4) 195.1 (0.4) 189.6 (0.4) 899.3 (0.4) 0.01
 Blindness 414.3 (0.5) 188.2 (0.4) 198.3 (0.4) 1,092.7 (0.5) 0.01
 Hyperglycemic crisis 74.3 (0.1) 48.8 (0.1) 38.3 (0.1) 228.5 (0.1) 0.01
 Hypoglycemic crisis 70.2 (0.1) 43.1 (0.1) 45.2 (0.1) 199.7 (0.1) 0.0
 Obesity 27,772.6 (33.7) 15,568.2 (35.2) 16,199.1 (34.9) 70,042.1 (33.8) 0.03
 Metabolic/bariatric surgery 1,158.9 (1.4) 596.5 (1.3) 654.0 (1.4) 2,902.6 (1.4) 0.01
 Pancreatitis 387.8 (0.5) 184.8 (0.4) 215.6 (0.5) 1,060.5 (0.5) 0.01
 Pancreatic cancer 104.3 (0.1) 64.6 (0.1) 56.3 (0.1) 280.1 (0.1) 0.01
 Cancer 7,297.8 (8.9) 3,765.4 (8.5) 4,039.5 (8.7) 18,263.5 (8.8) 0.01
 Smoking 7,159.4 (8.7) 3,821.7 (8.6) 3,973.2 (8.5) 18,129.2 (8.8) 0.01
 Thyroid cancer 307.9 (0.4) 186.2 (0.4) 170.5 (0.4) 757.5 (0.4) 0.01
 Genitourinary tract infection 3,931.9 (4.8) 2,011.5 (4.5) 2,082.7 (4.5) 9,943.4 (4.8) 0.02
 Cirrhosis 734.9 (0.9) 384.0 (0.9) 396.2 (0.9) 1,838.5 (0.9) 0.0
 Dementia 938.8 (1.1) 462.3 (1.0) 441.5 (0.9) 2,369.7 (1.1) 0.02
 Falls 3,406.1 (4.1) 1,785.1 (4.0) 2,001.0 (4.3) 8,495.7 (4.1) 0.01
 Urinary incontinence 3,197.6 (3.9) 1,676.3 (3.8) 1,765.1 (3.8) 8,033.0 (3.9) 0.0
 Unplanned hospitalizations 6,607.9 (8.0) 3,308.6 (7.5) 3,548.5 (7.6) 16,785.2 (8.1) 0.02
Medications, n (%)
 Antiplatelet drug 3,679.0 (4.5) 2,096.9 (4.7) 2,017.8 (4.3) 9,386.5 (4.5) 0.02
 Diuretics 32,495.1 (39.4) 17,580.5 (39.7) 18,155.6 (39.1) 81,575.0 (39.4) 0.01
 RAAS inhibitors 55,638.6 (67.5) 30,107.8 (68.0) 31,059.4 (66.8) 139,745.3 (67.4) 0.03
 Sacubitril/valsartan 73.1 (0.1) 55.5 (0.1) 64.6 (0.1) 185.5 (0.1) 0.01
 MRA 1,882.8 (2.3) 1,043.9 (2.4) 1,058.2 (2.3) 4,722.0 (2.3) 0.01
 Beta blockers 27,425.3 (33.3) 14,987.5 (33.9) 15,215.6 (32.7) 68,909.0 (33.3) 0.02
 Calcium channel blocker 21,075.9 (25.6) 11,306.7 (25.5) 11,648.7 (25.1) 53,036 (25.6) 0.01
 Other antihypertensives 3,928.9 (4.8) 2,142.0 (4.8) 2,131.8 (4.6) 9,946.7 (4.8) 0.01
 Lipid- lowering meds 56,316.6 (68.3) 3,0481.3 (68.9) 31,588.2 (68.0) 141,341.0 (68.2) 0.02
 Anti coagulants 4,060.4 (4.9) 2,176.2 (4.9) 2,262.4 (4.9) 10,190.5 (4.9) 0.0
 Metformin 65,126.6 (79.0) 35,102.3 (79.3) 3,6691.3 (79.0) 163,472.0 (78.9) 0.01
 Thiazolidinedione 3,720.5 (4.5) 2,213.3 (5.0) 2,235.0 (4.8) 9,488.3 (4.6) 0.02
Open in a new tab

CKD, chronic kidney disease; ESKD, end-stage kidney disease; MRA, mineralocorticoid receptor antagonist; RAAS, renin angiotensin aldosterone system.

Results of the intention-to-treat (ITT) analyses in the inverse probability weighted cohort for MACE, expanded MACE and their components are summarized in Table 2 and Fig. 1. Compared to DPP4i, GLP-1RA were associated with significantly lower risks of MACE (HR 0.87; 95% CI 0.82–0.93) and all-cause mortality (HR 0.86; 95% CI 0.78–0.94); there was no difference between them in the other endpoints. SGLT2i were associated with significantly lower risk of MACE (HR 0.85; 95% CI 0.81–0.90), expanded MACE (HR 0.93; 95% CI 0.81–0.96), all-cause mortality (HR 0.79; 95% CI 0.73–0.85) and hospitalization for heart failure (HHF) (HR 0.74; 95% CI 0.66–0.83) compared to DPP4i. SGLT2i were associated with significantly lower risk of HHF (HR 0.79; 95% CI 0.67–0.92) compared to GLP-1RA, with no statistically significant difference in any other outcome. Sulfonylureas were associated with significantly higher risk of all outcomes (MACE, expanded MACE, all-cause mortality, stroke, myocardial infarction (MI), HHF and revascularization procedure) compared to DPP4i, GLP-1RA and SGLT2i. Comparisons of outcome risks in the unweighted study cohort are shown in Extended Data Table 1 and are greater in magnitude compared to the main study findings.

Table 2 |.

Association between glucose-lowering treatment and cardiovascular outcomes: ITT analysis

MACE Expanded MACE All-cause mortality Acute stroke Acute MI HHF Revascularization
HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI)
GLP-1RA versus DPP4i 0.87 (0.82–0.93) 0.95 (0.91–1.00) 0.86 (0.78–0.94) 0.89 (0.78–1.02) 0.89 (0.78–1.01) 0.94 (0.83–1.07) 1.02 (0.96–1.08)
SGLT2i versus DPP4i 0.85 (0.81–0.90) 0.93 (0.89–0.96) 0.79 (0.73–0.85) 0.89 (0.81–0.98) 0.93 (0.85–1.02) 0.74 (0.66–0.83) 1.00 (0.96–1.05)
SU versus DPP4i 1.19 (1.16–1.22) 1.14 (1.12–1.17) 1.22 (1.18–1.26) 1.18 (1.13–1.24) 1.21 (1.15–1.26) 1.27 (1.21–1.33) 1.11 (1.09–1.14)
SGLT2i versus GLP-1RA 0.97 (0.90–1.05) 0.97 (0.92–1.02) 0.92 (0.82–1.03) 1.00 (0.86–1.16) 1.05 (0.91–1.21) 0.79 (0.67–0.92) 0.98 (0.92–1.05)
SU versus GLP-1RA 1.36 (1.28–1.46) 1.20 (1.15–1.26) 1.42 (1.30–1.56) 1.33 (1.17–1.50) 1.35 (1.20–1.53) 1.35 (1.20–1.52) 1.09 (1.04–1.15)
SU versus SGLT2i 1.40 (1.33–1.47) 1.24 (1.20–1.28) 1.55 (1.44–1.66) 1.33 (1.22–1.45) 1.30 (1.19–1.41) 1.71 (1.54–1.90) 1.11 (1.07–1.15)
Wald P value <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001
Open in a new tab

SU, sulfonylurea.

Fig. 1 |. Cumulative incidence of study outcomes.

Fig. 1 |

Open in a new tab

Cumulative hazards for MACE (composite of myocardial infarction, stroke, all-cause mortality), expanded MACE (composite of MACE, hospitalizations for heart failure and revascularization procedure endpoints), all-cause mortality, stroke, myocardial infarction, HHF and revascularization procedure.

Because there was evidence of nonproportional hazards for nearly all outcomes (all except stroke and revascularization), we calculated the pairwise HRs (Extended Data Table 2) and predicted probabilities of outcome (Table 3) separately for follow-up of up to 1 year, 1–2 years and 2 years or more from index. The greatest difference was seen in the sulfonylurea-based comparisons, with sulfonylureas becoming progressively more inferior to all other drug classes with prolonged treatment for expanded MACE (driven mostly by changes in the stroke outcome), but progressively less inferior for HHF. Number needed to treat (NNT) for the nonsulfonylurea medications relative to sulfonylurea are reported in Extended Data Table 3. By year 3 of treatment, NNT for DPP4i ranged from 49 (expanded MACE) to 251 (stroke); for GLP-1RA, from 37 (expanded MACE) to 158 (stroke, HHF) and for SGLT2i, from 32 (expanded MACE) to 158 (acute MI). The lowest NNTs were identified for SGLT2i (versus DPP4i) for MACE, expanded MACE, all-cause mortality, acute stroke and HHF outcomes; for GLP-1RA for acute MI, and for DPP4i for revascularization procedure.

Table 3 |.

Predicted probabilities of experiencing MACE at 1, 2 and 3 years after treatment initiation

MACE Expanded MACE All-cause mortality Acute stroke Acute MI HHF Revascularization
1 year
DPP4i 1.9 6.7 0.8 0.7 0.7 0.5 5.7
GLP-1RA 1.7 6.4 0.7 0.6 0.7 0.5 5.8
SGLT2i 1.6 6.3 0.6 0.6 0.7 0.4 5.7
Sulfonylurea 2.3 7.7 1.0 0.8 0.9 0.7 6.3
2 years
DPP4i 4.4 11.4 2.1 1.4 1.5 1.2 8.5
GLP-1RA 3.8 10.9 1.8 1.3 1.3 1.1 8.7
SGLT2i 3.7 10.6 1.7 1.3 1.4 0.9 8.5
Sulfonylurea 5.2 12.9 2.6 1.7 1.8 1.5 9.4
3 years
DPP4i 7.1 15.6 3.7 2.2 2.3 2.0 10.6
GLP-1RA 6.2 14.9 3.2 2.0 2.1 1.8 10.8
SGLT2i 6.1 14.5 2.9 2.0 2.2 1.5 10.6
Sulfonylurea 8.4 17.7 4.5 2.6 2.8 2.5 11.7
Open in a new tab

Sensitivity analyses conducted according to the as-treated approach, censored when patients discontinue the assigned treatment (Extended Data Table 4), found the magnitude of associations between treatments and the examined outcomes to have increased compared to the ITT analyses, but remained in the same direction. Results were similar in the second as-treated sensitivity analysis, which censored patients when they discontinued the assigned treatment or added another drug class, whichever came first (Extended Data Table 5). Results were also unchanged after adjusting for baseline CVD risk (Extended Data Table 6). Finally, heterogeneity of treatment effect analyses by patient age revealed no heterogeneity in any of the examined endpoints with the exception of HHF, where the SGLT2i had greater magnitude of risk reduction as compared to DPP4i, GLP-1RA and sulfonylureas among patients under 65 years compared to those 65 years and older (Extended Data Table 7).

In the analysis of falsification endpoints (Extended Data Table 8), there was no significant difference in any of the pairwise comparisons with respect to appendicitis (P = 0.52). However, there were differences between the treatment arms for the pneumonia and colonoscopy endpoints, mostly of lesser magnitude but in the same direction as the main treatment effects.

In an emulated target trial approach using data from a nationwide cohort of over 380,000 adults with type 2 diabetes at moderate baseline risk of CVD, SGLT2i and GLP-1RA were associated with similar risks for MACE, expanded MACE, stroke, acute MI and need for revascularization procedure, whereas SGLT2i use was associated with lower risk of HHF than GLP-1RA (21% risk reduction). Compared to DPP4i, GLP-1RA were associated with lower risks of MACE (13% risk reduction) and all-cause mortality (14% risk reduction). Similarly, compared to DPP4i, SGLT2i were associated with lower risks of MACE (15% risk reduction), expanded MACE (7% risk reduction), all-cause mortality (21% risk reduction), stroke (11% decrease) and HHF (26% risk reduction). Finally, sulfonylureas were associated with significantly higher risks of all outcomes compared to DPP4i, GLP-1RA and SGLT2i. Overall, SGLT2i appeared to be the preferred treatment option, with lowest NNTs (for example, just 32 for expanded MACE and 64 for death with 3 years of treatment) compared to the least costly sulfonylurea for nearly all examined outcomes. These NNTs are comparable to those observed for moderate intensity statins when used for primary prevention of MACE22 as well as ezetimibe and PCSK9 inhibitors when used in high-risk patients23.

This is the first direct comparison, to our knowledge, of the four commonly used second-line glucose-lowering medications conducted among patients with type 2 diabetes at moderate baseline risk of CVD, addressing a critical knowledge gap left by previously conducted RCTs that focused exclusively on patients with established CVD or at high risk for CVD events11–19. CVD outcomes were also examined in the GRADE trial for patients with low baseline CVD risk, though as secondary outcomes and without a SGLT2i arm (finding a small benefit to the GLP-1RA liraglutide when compared to a pooled comparator group of patients treated with insulin glargine, the sulfonylurea glimepiride and the DPP4i sitagliptin)24. Importantly, our findings are consistent with RCTs and network meta-analyses conducted in high-risk populations11–19 and make the case for preferential use of GLP-1RA and SGLT2i to reduce the risk of CVD events and of SGLT2i to reduce the risk of HHF in moderate-risk patients in addition to those at high risk7,8. Indeed, current clinical practice guidelines only advise on CVD risk reduction independent of glycemic control in patients with high baseline CVD risk, as those were the only data available, and our results suggest that these recommendations should be extended to moderate-risk subgroups. These findings are particularly important as an RCT of all four second-line drugs in moderate-risk populations, among whom event rates are expected to be lower than in high-risk populations, is unlikely to ever be conducted given the large sample size, long duration and high cost that would be necessary to sufficiently power and enable such a trial.

We found no difference between GLP-1RA and SGLT2i in reducing the risk of stroke in this moderate-risk population, though SGLT2i were superior to DPP4i (HR 0.89; 95% CI 0.81–0.98) and GLP-1RA had a similarly favorable point estimate but did not reach statistical significance (HR 0.89; 95% CI 0.78–1.02). These findings add to the emerging and inconclusive literature demonstrating the potential benefit of GLP-1RA on stroke risk reduction in high-risk patients in some cardiovascular outcomes trials18,19,25–27, but suggest that SGLT2i may be similarly efficacious in the moderate-risk, real-world population.

Sulfonylurea was the only medication class consistently associated with greater risk of all MACE and expanded MACE outcomes, calling for caution with their use in this patient population. However, avoidance of sulfonylurea drugs may not be feasible for many patients due to the high cost of brand-name diabetes drugs. This is particularly important for older patients with Medicare Advantage or Medicare fee-for-service health plans who are not eligible for copay reduction with the use of savings cards28. Medicare beneficiaries, as well as low-income individuals and racial and ethnic minoritized populations, face gaps in access to these medications29,30 and it will be important to ensure equitable access to and utilization of these agents if their use becomes recommended or preferred for broader patient populations.

In the absence of RCTs directly examining the comparative effectiveness of second-line glucose-lowering medications in adults with type 2 diabetes at moderate risk for CVD, causal inference analytic methods applied to observational data can provide the evidence needed to inform clinical practice and shared decision-making. In contrast to other observational studies examining second-line diabetes medications with respect to CVD outcomes, which also did not compare all four drug classes head-to-head31–33, our analyses adhered to the ITT principle to fully align with the target trial framework. Our study is further strengthened by the size, heterogeneity and diversity of the included patient population. Though we were not able to include Medicaid beneficiaries, in the absence of a database for all Americans, the linked OptumLabs Data Warehouse (OLDW) and Medicare fee-for-service data are the closest alternative to a population-based sample of longitudinal clinical data that includes individuals across the entire United States, from different health systems, regions and socioeconomic backgrounds.

Nevertheless, even with robust analytic methods and adherence to the target trial framework34, observational studies remain prone to residual unmeasured confounding35,36. Cognizant of these limitations, we conducted analyses examining the effect of treatment on several falsification endpoints. Whereas we found no association with acute appendicitis, we did observe unanticipated associations of treatment with pneumonia events and receipt of screening colonoscopy, raising concerns about the presence of unmeasured confounding. Including social determinants of health as covariates in the propensity score models may have mitigated this bias if these factors influence both access to more expensive medications and the risk of the falsification endpoints. However, social determinants of health data were not available in our dataset, which was prioritized over more granular but much smaller and nonrepresentative databases due to its size, diversity, heterogeneity and complete capture of exposures and outcomes. Although these falsification results do not nullify our primary results, they underscore the importance of conducting falsification endpoint analyses in all observational studies and even in prospective trials. All studies, including prospective RCTs, are potentially subject to confounding and bias and while rigorous randomization reduces this risk in RCTs, it does not eliminate it entirely (indeed, RCTs often report heightened risks of clinically unrelated adverse events). Yet, such analyses are rarely, if ever, included in the published literature, including in previous observational studies comparing second-line glucose-lowering medications25,26,31–33 and the cardiovascular outcomes trials.

Additionally, we were not able to specifically examine cardiovascular mortality, as cause of death is not available in the data, and instead assessed all-cause mortality as part of the MACE and expanded MACE endpoints. We also could not assess the impact of clinical variables such as hemoglobin A1c and obesity on drug effectiveness and study outcomes as these data were not available within claims.

This emulation of a target trial comparing the effectiveness of second-line GLP-1RA, SGLT2i, DPP4i and sulfonylureas with respect to reducing cardiovascular events and death in adults with type 2 diabetes at moderate baseline risk for CVD confirmed the superiority of GLP-1RA and SGLT2i compared to DPP4i and sulfonylureas, and the superiority of DPP4i over sulfonylureas, on all outcomes. In the absence of RCTs comparing these second-line agents head-to-head, especially in the moderate CVD risk subgroup, robustly designed observational studies provide the best evidence for comparative effectiveness in real-world care settings and diverse patient populations. Our findings suggest that GLP-1RA and SGLT2i (and especially SGLT2i with the lowest NNT for nearly all study outcomes) may be the preferred glucose-lowering agents for cardiovascular risk reduction in patients at moderate risk for CVD, building on and consistent with existing RCT evidence in the high-risk population.

Methods

Ethical review

The study was exempt from Mayo Clinic Institutional Review Board review and is reported according to the RECORD and START-RWE reporting guidelines37,38. Informed consent requirements are not applicable as the study used deidentified administrative claims data.

Study design and data source

This is a retrospective observational study emulating an idealized target trial examining head-to-head the comparative effectiveness of first- or second-line initiation of GLP-1RA, SGLT2i, DPP4i and sulfonylureas by adults with type 2 diabetes at moderate risk of CVD with regard to MACE (Extended Data Fig. 1). The study protocol was preregistered on ClinicalTrials.gov (registration: NCT05214573).

For the analyses, we linked medical and pharmacy claims from OLDW (which includes enrollees in commercial and Medicare Advantage health plans across the United States) to a 100% sample of Medicare fee-for-service claims, thereby including a diverse cohort with a wide range of ages, racial and ethnic groups, income levels, geographic regions, health systems and health plans across the United States39,40. OLDW and Medicare fee-for-service claims were linked on personal identifiers and deidentified by OptumLabs before being made available to researchers41, allowing us to analyze the data as a single cohort and follow patients as they switched among commercial, Medicare Advantage and Medicare fee-for-service plans.

Study population

We identified adults aged 21 years or older who had filled a new prescription for a GLP-1RA, SGLT2i, DPP4i or sulfonylurea between 1 January 2014 and 31 December 2021. The date of the first fill was used as the index date for each patient, but patients were required to have a second fill of the study drug class to confirm its use and 12 months of enrollment before the index date to ascertain baseline covariates.

We allowed a 30-day gap between the last day covered by a preceding prescription and the new prescription to count as continued use. Patients could enter the cohort only once, the first time they met the eligibility criteria for a medication class. Patients with missing sex, year of birth or region were excluded (<1%).

Exclusion criteria (applied to the baseline 12-month period) were the following: fill for any study drug or glinide (excluded due to its similarity to sulfonylurea) during the baseline period; any second study drug during days 0–30 after index date; insulin use; type 1 diabetes (defined by presence of any type 1 diabetes ICD9 or 10 codes during the baseline period); pregnancy and metastatic cancer. We then restricted the study population to patients at moderate CVD risk, defined as an estimated 1–5% annualized predicted risk of experiencing a MACE event. CVD risk was estimated using the annualized claims-based MACE estimator, developed and internally validated among patients with type 2 diabetes included in OLDW and Medicare fee-for-service claims6. In the derivation cohort, annualized claims-based MACE estimator had concordance index 0.74 (s.e. = 0.0002), with a Monte Carlo cross-validation mean index 0.740 (range 0.739–0.741).

Included patients were followed from the time of first-observed use of one of the four study medications during the study period until either the end of the study (31 December 2021), disenrollment from insurance plans in our database or death.

Outcomes

The primary outcome was time to MACE, defined as the composite of hospitalization for nonfatal acute MI or nonfatal stroke or all-cause mortality, whichever occurred first. Secondary outcomes were time to expanded MACE, defined as the composite of MACE, HHF and revascularization procedure, and the time to individual components of expanded MACE. Diagnosis codes used to ascertain all baseline covariates and outcomes are detailed in Supplementary Table 4.

Independent variables and covariates

Our independent variable was the index treatment (GLP-1RA, SGLT2i, DPP4i or sulfonylurea). Covariates included baseline demographic characteristics (age, sex, race or ethnicity and US region) ascertained from enrollment files and comorbidities (Supplementary Table 5) and nonindex medications (Supplementary Table 6) ascertained from medical and pharmacy claims, respectively, during the baseline 12-month period.

Statistical analyses

Our primary analyses were time-to-event models with the ITT framework. We first estimated propensity score models using baseline covariates (Supplementary Tables 5 and 6) to account for the multiple treatments and estimate the probability of treatment assignment for each medication class42. Potential confounders were identified by the study team and a patient and stakeholder advisory group, which was formed to provide feedback on study design, implementation and dissemination. We estimated a separate model for each medication class versus the pool of the other three, which allowed us to use more flexible models to estimate the probability of treatment. Specifically, we used a diverse set of individual binomial prediction algorithms included in the super learner ensemble43 to estimate the propensity score models. The super learner framework allows for flexible estimation of an ensemble predictor with the theoretical properties of cross-validation for model selection to control for overfitting44. The predicted propensity scores were evaluated for evidence of violations of the positivity assumption. Since the inverse propensity score weights had extreme values when examining their distributions (Supplementary Fig. 2), we used the stabilized version proposed by Yoshida for the analysis45.

We evaluated the balance across treatment groups using the weights to calculate s.m.d. of baseline covariates42. The s.m.d. values were calculated for each pairwise comparison. The final stabilized inverse propensity scores were then incorporated as weights into a Cox proportional hazards model with all treatment groups. Any baseline covariates with maximum s.m.d. ≥ 0.10 (across all pairwise comparisons) and hence indicative of imbalance in that variable, were included as adjustment variables in the outcome model. A separate model was estimated for each outcome, reporting HRs and 95% CI for each treatment relative to DPP4i. We then calculated all pairwise treatment effects, using simulation to estimate 95% CIs. To aid interpretation of results, we additionally calculated the NNT with DPP4i, GLP-1RA and SGLT2i versus sulfonylurea to avoid one event at 1, 2 and 3 years. For each model we tested the proportional hazards assumption for treatment group using the Grambsch and Therneau test46,47. A single overall omnibus test for the HRs 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 within 1, 2 and 3 years. All statistical tests were two-sided.

Data management activities were conducted using SAS 9.04 (SAS Institute Inc.), whereas analyses were conducted using R v.4.1 (R Foundation for Statistical Computing).

Sensitivity analyses

To evaluate the sensitivity of the treatment effects to potential unmeasured confounders, we examined falsification endpoints48,49 of pneumonia (during the overall study period and truncated at 2019 to eliminate the potential impact of the COVID-19 pandemic); hospitalizations for appendicitis and completion of screening colonoscopy. We also assessed for robustness of study findings to baseline CVD risk by adding the estimated CVD risk score to each of our main models as a continuous variable. To examine the degree to which inverse probability of treatment weighting impacted comparisons, we repeated all analyses using the unweighted cohort (that is, setting the weights to one).

Two sensitivity analyses were conducted using an as-treated approach (Extended Data Fig. 1), 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, addition of another glucose-lowering drug or insulin, death or disenrollment from the health plan, whichever came first. We further assessed for heterogeneity of treatment effects of age by testing the interaction between age and treatment groups and by conducting subgroup analyses among patients under 65 and 65 years of age and older.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Extended Data

Extended Data Fig. 1 |. Study Design.

Extended Data Fig. 1 |

Open in a new tab

CVD, cardiovascular disease. MACE, major adverse cardiovascular event.

Extended Data Table 1 |.

Association between glucose-lowering treatment and cardiovascular outcomes in the unweighted study cohort

MACE Expanded MACE All-cause mortality Acute stroke Acute MI HHF Revascularization
HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI)
GLP-1RA vs. DPP4i 0.72 (0.68, 0.75) 0.76 (0.73, 0.79) 0.68 (0.63, 0.73) 0.69 (0.63, 0.76) 0.70 (0.64, 0.77) 0.76 (0.69, 0.84) 0.74 (0.71, 0.77)
SGLT2i vs. DPP4i 0.76 (0.73, 0.79) 0.98 (0.95, 1.01) 0.65 (0.61, 0.70) 0.78 (0.72, 0.84) 0.86 (0.80, 0.93) 0.64 (0.59, 0.71) 1.12 (1.08, 1.16)
SU vs. DPP4i 1.22 (1.19, 1.25) 1.17 (1.15, 1.19) 1.24 (1.20, 1.28) 1.19 (1.14, 1.25) 1.24 (1.18, 1.29) 1.29 (1.23, 1.35) 1.13 (1.11, 1.16)
SGLT2i vs. GLP-1RA 1.06 (1.00, 1.13) 1.29 (1.24, 1.34) 0.97 (0.89, 1.05) 1.12 (1.01, 1.26) 1.23 (1.10, 1.36) 0.85 (0.75, 0.96) 1.51 (1.44, 1.58)
SU vs. GLP-1RA 1.70 (1.62, 1.78) 1.54 (1.49, 1.59) 1.84 (1.72, 1.97) 1.72 (1.57, 1.88) 1.76 (1.61, 1.92) 1.70 (1.55, 1.86) 1.53 (1.47, 1.59)
SU vs. SGLT2i 1.60 (1.54, 1.67) 1.19 (1.16, 1.22) 1.90 (1.79, 2.02) 1.53 (1.41, 1.64) 1.43 (1.34, 1.54) 2.00 (1.84, 2.19) 1.01 (0.98, 1.05)
Wald p-value <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001
Open in a new tab

DPP4i, dipeptidyl peptidase-4 inhibitors. GLP-1RA, glucagon-like peptide-1 receptor agonists. HHF, hospitalization for heart failure. MACE, major adverse cardiovascular events. MI, myocardial infarction. SGLT2i, sodium-glucose cotransporter 2 inhibitors. SU, sulfonylurea.

Extended Data Table 2 |.

Association between glucose-lowering treatment and cardiovascular outcomes by duration of treatment: intention-to-treat analysis

MACE Expanded MACE All-cause mortality Acute stroke Acute MI HHF Revascularization
HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI)
≤ 1 year of treatment
GLP-1RA vs. DPP4i 0.79 (0.69, 0.91) 1.00 (0.94, 1.07) 0.79 (0.64, 0.98) 0.78 (0.62, 0.99) 0.75 (0.59, 0.95) 0.89 (0.68, 1.15) 1.04 (0.97, 1.12)
SGLT2i vs. DPP4i 0.92 (0.83, 1.03) 1.01 (0.96, 1.06) 0.82 (0.68, 0.99) 0.99 (0.83, 1.18) 0.85 (0.71, 1.01) 0.59 (0.47, 0.75) 1.03 (0.98, 1.09)
SU vs. DPP4i 1.15 (1.08, 1.22) 1.04 (1.00, 1.08) 1.23 (1.11, 1.35) 1.09 (0.98, 1.20) 1.15 (1.04, 1.27) 1.22 (1.09, 1.37) 1.03 (0.99, 1.07)
SGLT2i vs. GLP-1RA 1.16 (0.99, 1.36) 1.01 (0.94, 1.08) 1.03 (0.80, 1.35) 1.27 (0.97, 1.65) 1.14 (0.87, 1.50) 0.67 (0.48, 0.93) 0.99 (0.91, 1.07)
SU vs. GLP-1RA 1.44 (1.28, 1.64) 1.04 (0.98, 1.10) 1.55 (1.27, 1.90) 1.39 (1.11, 1.74) 1.54 (1.22, 1.94) 1.38 (1.07, 1.77) 0.98 (0.92, 1.05)
SU vs. SGLT2i 1.24 (1.13, 1.37) 1.03 (0.99, 1.08) 1.50 (1.25, 1.79) 1.10 (0.94, 1.29) 1.35 (1.15, 1.58) 2.06 (1.65, 2.56) 1.00 (0.95, 1.05)
1–2 years of treatment
GLP-1RA vs. DPP4i 0.89 (0.77, 1.02) 0.93 (0.85, 1.02) 0.80 (0.65, 0.98) 1.10 (0.84, 1.43) 0.78 (0.60, 1.02) 0.83 (0.64, 1.08) 0.96 (0.85, 1.08)
SGLT2i vs. DPP4i 0.79 (0.71, 0.88) 0.87 (0.81, 0.93) 0.70 (0.60, 0.83) 0.92 (0.76, 1.11) 0.83 (0.69, 1.00) 0.73 (0.57, 0.92) 0.92 (0.84, 1.01)
SU vs. DPP4i 1.14 (1.08, 1.21) 1.10 (1.05, 1.15) 1.19 (1.10, 1.28) 1.16 (1.05, 1.30) 1.13 (1.02, 1.25) 1.25 (1.12, 1.40) 1.06 (1.00, 1.12)
SGLT2i vs. GLP-1RA 0.89 (0.76, 1.04) 0.93 (0.84, 1.03) 0.88 (0.69, 1.12) 0.83 (0.61, 1.12) 1.06 (0.79, 1.44) 0.88 (0.63, 1.22) 0.96 (0.84, 1.10)
SU vs. GLP-1RA 1.29 (1.12, 1.47) 1.18 (1.08, 1.28) 1.49 (1.22, 1.82) 1.06 (0.82, 1.37) 1.45 (1.12, 1.88) 1.51 (1.17, 1.95) 1.11 (0.99, 1.24)
SU vs. SGLT2i 1.45 (1.32, 1.60) 1.27 (1.18, 1.35) 1.69 (1.45, 1.96) 1.27 (1.07, 1.52) 1.36 (1.14, 1.62) 1.72 (1.38, 2.15) 1.15 (1.06, 1.25)
≥2 years of treatment
GLP-1RA vs. DPP4i 0.90 (0.82, 0.98) 0.93 (0.86, 1.00) 0.89 (0.79, 1.00) 0.86 (0.72, 1.04) 1.01 (0.85, 1.20) 1.00 (0.85, 1.18) 1.03 (0.92, 1.16)
SGLT2i vs. DPP4i 0.85 (0.80, 0.92) 0.90 (0.85, 0.95) 0.81 (0.74, 0.88) 0.84 (0.73, 0.97) 1.02 (0.89, 1.16) 0.80 (0.69, 0.93) 1.04 (0.96, 1.14)
SU vs. DPP4i 1.18 (1.14, 1.21) 1.16 (1.12, 1.19) 1.19 (1.15, 1.24) 1.18 (1.11, 1.25) 1.20 (1.13, 1.27) 1.23 (1.16, 1.31) 1.11 (1.06, 1.16)
SGLT2i vs GLP-1RA 0.95 (0.86, 1.06) 0.96 (0.88, 1.05) 0.91 (0.79, 1.05) 0.98 (0.78, 1.21) 1.00 (0.82, 1.23) 0.80 (0.65, 0.99) 1.01 (0.89, 1.16)
SU vs. GLP-1RA 1.31 (1.20, 1.43) 1.24 (1.15, 1.34) 1.34 (1.20, 1.51) 1.37 (1.13, 1.63) 1.18 (1.01, 1.40) 1.23 (1.05, 1.45) 1.08 (0.96, 1.21)
SU vs. SGLT2i 1.38 (1.29, 1.47) 1.29 (1.22, 1.36) 1.48 (1.35, 1.61) 1.40 (1.23, 1.60) 1.18 (1.04, 1.34) 1.54 (1.34, 1.77) 1.06 (0.98, 1.15)
Open in a new tab

DPP4i, dipeptidyl peptidase-4 inhibitors. GLP-1RA, glucagon-like peptide-1 receptor agonists. HHF, hospitalization for heart failure. MACE, major adverse cardiovascular event. MI, myocardial infarction. SGLT2i, sodium-glucose cotransporter 2 inhibitors. SU, sulfonylurea.

Extended Data Table 3 |.

Number needed to treat (NNT) to experience one fewer adverse health outcome compared to treatment with sulfonylurea

MACE Expanded MACE All-cause mortality Acute stroke Acute MI HHF Revascularization
1-year
DPP4i 277 107 585 787 667 690 161
GLP-1RA 166 80 351 496 435 564 195
SGLT2i 155 70 293 494 498 351 165
2-year
DPP4i 122 65 220 382 329 310 110
GLP-1RA 73 49 131 241 215 254 133
SGLT2i 68 43 110 240 246 158 112
3-year
DPP4i 77 49 128 251 212 194 89
GLP-1RA 46 37 76 158 138 158 108
SGLT2i 43 32 64 157 158 98 91
Open in a new tab

DPP4i, dipeptidyl peptidase-4 inhibitors. GLP-1RA, glucagon-like peptide-1 receptor agonists. HHF, hospitalization for heart failure. MACE, major adverse cardiovascular event. MI, myocardial infarction. SGLT2i, sodium-glucose cotransporter 2 inhibitors.

Extended Data Table 4 |.

Association between glucose-lowering treatment and cardiovascular outcomes: as-treated analysis, censored upon discontinuation of the study drug

MACE Expanded MACE All-cause mortality Acute stroke Acute MI HHF Revascularization
HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI)
GLP-1RA vs. DPP4i 0.74 (0.63, 0.87) 0.96 (0.89, 1.04) 0.60 (0.42, 0.86) 0.67 (0.52, 0.86) 0.75 (0.58, 0.97) 0.71 (0.52, 0.97) 0.99 (0.91, 1.08)
SGLT2i vs. DPP4i 0.90 (0.79, 1.02) 0.99 (0.94, 1.05) 0.73 (0.54, 0.98) 0.88 (0.73, 1.05) 0.87 (0.73, 1.05) 0.46 (0.35, 0.60) 1.01 (0.96, 1.07)
SU vs. DPP4i 1.29 (1.21, 1.38) 1.18 (1.15, 1.22) 1.46 (1.29, 1.64) 1.21 (1.10, 1.34) 1.34 (1.21, 1.47) 1.40 (1.26, 1.56) 1.17 (1.13, 1.22)
SGLT2i vs. GLP-1RA 1.21 (1.00, 1.47) 1.03 (0.95, 1.13) 1.22 (0.79, 1.87) 1.31 (0.99, 1.77) 1.16 (0.88, 1.55) 0.65 (0.44, 0.96) 1.02 (0.93, 1.12)
SU vs. GLP-1RA 1.75 (1.49, 2.05) 1.23 (1.14, 1.33) 2.43 (1.74, 3.42) 1.82 (1.43, 2.32) 1.78 (1.40, 2.26) 1.97 (1.48, 2.68) 1.19 (1.10, 1.29)
SU vs. SGLT2i 1.44 (1.28, 1.62) 1.19 (1.14, 1.25) 2.00 (1.51, 2.64) 1.38 (1.17, 1.63) 1.53 (1.30, 1.80) 3.06 (2.35, 3.99) 1.16 (1.10, 1.22)
Wald p-value <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001
Open in a new tab

DPP4i, dipeptidyl peptidase-4 inhibitors. GLP-1RA, glucagon-like peptide-1 receptor agonists. HHF, hospitalization for heart failure. MACE, major adverse cardiovascular event. MI, myocardial infarction. SGLT2i, sodium-glucose cotransporter 2 inhibitors. SU, sulfonylurea.

Extended Data Table 5 |.

Association between glucose-lowering treatment and cardiovascular outcomes: as-treated analysis, censored upon discontinuation of the study drug or addition of another drug class, whichever came first

MACE Expanded MACE All-cause mortality Acute stroke Acute MI HHF Revascularization
HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI)
GLP-1RA vs. DPP4i 0.74 (0.62, 0.89) 0.92 (0.85, 1.00) 0.56 (0.38, 0.83) 0.67 (0.50, 0.90) 0.71 (0.54, 0.93) 0.74 (0.52, 1.02) 0.94 (0.86, 1.03)
SGLT2i vs. DPP4i 0.88 (0.77, 1.01) 1.00 (0.94, 1.06) 0.73 (0.53, 1.02) 0.83 (0.69, 1.01) 0.83 (0.68, 1.00) 0.44 (0.33, 0.59) 1.02 (0.96, 1.08)
SU vs. DPP4i 1.30 (1.21, 1.39) 1.19 (1.15, 1.23) 1.42 (1.24, 1.63) 1.16 (1.05, 1.29) 1.36 (1.23, 1.51) 1.44 (1.28, 1.62) 1.18 (1.14, 1.23)
SGLT2i vs GLP-1RA 1.19 (0.97, 1.46) 1.08 (0.99, 1.19) 1.29 (0.81, 2.09) 1.24 (0.89, 1.72) 1.17 (0.85, 1.59) 0.61 (0.40, 0.92) 1.09 (0.99, 1.20)
SU vs. GLP-1RA 1.76 (1.48, 2.09) 1.29 (1.19, 1.40) 2.53 (1.75, 3.65) 1.74 (1.32, 2.30) 1.92 (1.48, 2.49) 1.96 (1.42, 2.72) 1.26 (1.16, 1.38)
SU vs. SGLT2i 1.48 (1.30, 1.67) 1.19 (1.13, 1.25) 1.95 (1.43, 2.63) 1.40 (1.17, 1.67) 1.65 (1.38, 1.96) 3.25 (2.45, 4.31) 1.16 (1.10, 1.22)
Wald p-value <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001
Open in a new tab

DPP4i, dipeptidyl peptidase-4 inhibitors. GLP-1RA, glucagon-like peptide-1 receptor agonists. HHF, hospitalization for heart failure. MACE, major adverse cardiovascular event. MI, myocardial infarction. SGLT2i, sodium-glucose cotransporter 2 inhibitors. SU, sulfonylurea.

Extended Data Table 6 |.

Association between glucose-lowering treatment and cardiovascular outcomes: intention-to-treat analysis adjusted for baseline cardiovascular disease risk

MACE Expanded MACE All-cause mortality Acute stroke Acute MI HHF Revascularization
HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI)
GLP-1RA vs. DPP4i 0.88 (0.82, 0.94) 0.97 (0.92, 1.01) 0.86 (0.79, 0.95) 0.90 (0.79, 1.02) 0.90 (0.79, 1.02) 0.95 (0.84, 1.08) 1.04 (0.99, 1.11)
SGLT2i vs. DPP4i 0.86 (0.82, 0.91) 0.95 (0.91, 0.98) 0.80 (0.74, 0.87) 0.91 (0.82, 1.00) 0.94 (0.86, 1.04) 0.76 (0.68, 0.85) 1.03 (0.99, 1.07)
SU vs. DPP4i 1.19 (1.16, 1.22) 1.14 (1.12, 1.16) 1.21 (1.17, 1.25) 1.18 (1.13, 1.24) 1.20 (1.15, 1.26) 1.26 (1.20, 1.32) 1.11 (1.08, 1.14)
SGLT2i vs GLP-1RA 0.98 (0.91, 1.06) 0.98 (0.93, 1.03) 0.93 (0.83, 1.04) 1.01 (0.87, 1.17) 1.05 (0.91, 1.22) 0.80 (0.68, 0.93) 0.99 (0.93, 1.05)
SU vs. GLP-1RA 1.35 (1.26, 1.44) 1.18 (1.13, 1.23) 1.40 (1.28, 1.53) 1.31 (1.16, 1.48) 1.34 (1.18, 1.51) 1.32 (1.17, 1.49) 1.06 (1.01, 1.12)
SU vs. SGLT2i 1.37 (1.30, 1.44) 1.20 (1.16, 1.25) 1.51 (1.40, 1.62) 1.30 (1.19, 1.42) 1.27 (1.17, 1.39) 1.66 (1.50, 1.85) 1.08 (1.04, 1.12)
Wald p-value <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001
Open in a new tab

DPP4i, dipeptidyl peptidase-4 inhibitors. GLP-1RA, glucagon-like peptide-1 receptor agonists. HHF, hospitalization for heart failure. MACE, major adverse cardiovascular event. MI, myocardial infarction. SGLT2i, sodium-glucose cotransporter 2 inhibitors. SU, sulfonylurea.

Extended Data Table 7 |.

Age-based heterogeneity of treatment effects analysis

MACE Expanded MACE All-cause mortality Acute stroke Acute MI HHF Revascularization
HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI)
Patients <65 years old
GLP-1RA vs. DPP4i 0.83 (0.72, 0.95) 0.95 (0.88, 1.04) 0.81 (0.65, 1.00) 0.96 (0.75, 1.23) 0.81 (0.65, 1.01) 0.90 (0.70, 1.16) 1.00 (0.90, 1.11)
SGLT2i vs. DPP4i 0.81 (0.73, 0.91) 0.92 (0.85, 0.98) 0.68 (0.56, 0.82) 0.95 (0.78, 1.17) 0.88 (0.74, 1.05) 0.61 (0.49, 0.76) 1.00 (0.92, 1.09)
SU vs. DPP4i 1.23 (1.15, 1.32) 1.16 (1.11, 1.22) 1.20 (1.09, 1.32) 1.33 (1.17, 1.52) 1.29 (1.15, 1.44) 1.43 (1.26, 1.62) 1.13 (1.06, 1.20)
SGLT2i vs GLP-1RA 0.98 (0.84, 1.15) 0.96 (0.87, 1.05) 0.84 (0.65, 1.09) 0.99 (0.75, 1.32) 1.08 (0.85, 1.37) 0.68 (0.51, 0.91) 1.00 (0.90, 1.12)
SU vs. GLP-1RA 1.49 (1.31, 1.69) 1.22 (1.13, 1.32) 1.49 (1.22, 1.83) 1.39 (1.10, 1.75) 1.58 (1.29, 1.93) 1.58 (1.26, 1.98) 1.13 (1.03, 1.24)
SU vs. SGLT2i 1.51 (1.36, 1.68) 1.27 (1.20, 1.35) 1.77 (1.48, 2.11) 1.40 (1.16, 1.68) 1.47 (1.26, 1.71) 2.33 (1.91, 2.85) 1.13 (1.05, 1.21)
Patients ≥65 years old
GLP-1RA vs. DPP4i 0.88 (0.82, 0.95) 0.95 (0.91, 1.01) 0.87 (0.78, 0.96) 0.88 (0.76, 1.03) 0.91 (0.79, 1.05) 0.95 (0.82, 1.10) 1.03 (0.96, 1.10)
SGLT2i vs. DPP4i 0.86 (0.81, 0.92) 0.93 (0.90, 0.97) 0.81 (0.74, 0.88) 0.88 (0.80, 0.98) 0.95 (0.85, 1.05) 0.77 (0.68, 0.88) 1.01 (0.96, 1.06)
SU vs. DPP4i 1.18 (1.15, 1.22) 1.14 (1.12, 1.16) 1.22 (1.18, 1.26) 1.16 (1.10, 1.22) 1.19 (1.13, 1.25) 1.24 (1.17, 1.30) 1.11 (1.08, 1.14)
SGLT2i vs GLP-1RA 0.98 (0.89, 1.07) 0.98 (0.92, 1.04) 0.93 (0.82, 1.05) 1.00 (0.85, 1.19) 1.04 (0.88, 1.23) 0.81 (0.68, 0.97) 0.98 (0.91, 1.06)
SU vs. GLP-1RA 1.34 (1.24, 1.45) 1.20 (1.14, 1.26) 1.41 (1.28, 1.56) 1.31 (1.14, 1.52) 1.31 (1.14, 1.50) 1.30 (1.14, 1.50) 1.08 (1.02, 1.15)
SU vs. SGLT2i 1.37 (1.30, 1.45) 1.22 (1.18, 1.27) 1.51 (1.40, 1.63) 1.31 (1.19, 1.45) 1.25 (1.13, 1.39) 1.60 (1.42, 1.81) 1.10 (1.05, 1.15)
Interaction p-value 0.19 0.66 0.38 0.29 0.14 0.003 0.86
Open in a new tab

DPP4i, dipeptidyl peptidase-4 inhibitors. GLP-1RA, glucagon-like peptide-1 receptor agonists. HHF, hospitalization for heart failure. MACE, major adverse cardiovascular event. MI, myocardial infarction. SGLT2i, sodium-glucose cotransporter 2 inhibitors. SU, sulfonylurea.

Extended Data Table 8 |.

Association between glucose-lowering treatment and falsification endpoints of pneumonia, appendicitis hospitalizations, and screening colonoscopy

Pneumonia (2014–2021) Pneumonia (2014–2019) Appendicitis Screening Colonoscopy
HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI)
GLP-1RA vs. DPP4i 1.00 (0.95, 1.05) 1.01 (0.96, 1.07) 0.86 (0.55, 1.34) 1.01 (0.98, 1.04)
SGLT2i vs. DPP4i 0.86 (0.83, 0.90) 0.89 (0.85, 0.93) 0.78 (0.55, 1.10) 1.02 (0.99, 1.04)
SU vs. DPP4i 1.09 (1.07, 1.12) 1.08 (1.06, 1.11) 0.95 (0.80, 1.13) 0.93 (0.92, 0.95)
SGLT2i vs. GLP-1RA 0.86 (0.81, 0.92) 0.88 (0.82, 0.94) 0.90 (0.53, 1.53) 1.01 (0.97, 1.05)
SU vs. GLP-1RA 1.09 (1.04, 1.15) 1.07 (1.01, 1.13) 1.10 (0.72, 1.70) 0.93 (0.90, 0.96)
SU vs. SGLT2i 1.27 (1.22, 1.32) 1.21 (1.16, 1.27) 1.22 (0.88, 1.69) 0.92 (0.90, 0.94)
Wald p-value <0.001 <0.001 0.52 <0.001
Open in a new tab

To check whether the falsification endpoint of pneumonia was influenced by the COVID-19 pandemic, we assessed it for both the overall study period and for 2014–2019. DPP4i, dipeptidyl peptidase-4 inhibitors. GLP-1RA, glucagon-like peptide-1 receptor agonists. SGLT2i, sodium-glucose cotransporter 2 inhibitors. SU, sulfonylurea.

Supplementary Material

Supplementary Material

Acknowledgements

We thank the Patient and Stakeholder Advisory Group convened in support of this work for their insight and feedback on model covariates and study design. Members of the Patient and Stakeholder Advisory Group include: J. P. W. Bynum (University of Michigan School of Medicine); J. K. Cuddeback (American Medical Group Association); W. B. DeHart (OptumLabs); R. A. Gabbay (American Diabetes Association); J. Gockerman (Grand Rapids, MI); E. H. Golembiewski (Mayo Clinic); J. Haag (Mayo Clinic); B. Labatte (Rochester, MN); R. J. Stroebel (Mayo Clinic); M. Tesulov (Rochester, MN) and S. Violette (UnitedHealth Group). Funding: research reported in this work was funded through a Patient-Centered Outcomes Research Institute Award PCS-1409–24099 (R.G.M.). The funder had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript. Disclaimer: 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, its Board of Governors or Methodology Committee.

Footnotes

Competing interests

G.E.U. reports unrestricted support for research studies to Emory University from Dexcom, Abbott and Bayer, and serves on the advisory board of Directors for GlyCare. R.J.G. has received unrestricted research support (to Emory University) from Novo Nordisk, Eli Lilly and Dexcom, and consulting fees from Sanofi, Novo Nordisk, Eli Lilly, Pfizer, Boehringer, Bayer and Weight Watchers. W.H.C. has received unrestricted research consulting support from Janssen Scientific Affairs LLC, Viatris, Merck and Optum. J.J.N. reports serving as a consultant to Sanofi, Bayer, Eli Lilly and Boehringer Ingelheim. The other authors declare no competing interests.

Additional information

Extended data is available for this paper at https://doi.org/10.1038/s44161-024-00453-9.

Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s44161-024-00453-9.

Peer review information Nature Cardiovascular Research thanks Hans-Peter Brunner–La Rocca, Koos Zwinderman and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Data availability

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 is 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.Baena-Díez JM et al. Risk of cause-specific death in individuals with diabetes: a competing risks analysis. Diabetes Care 39, 1987–1995 (2016). [DOI] [PubMed] [Google Scholar]
  • 2.Rao Kondapally Seshasai S. et al. Diabetes mellitus, fasting glucose, and risk of cause-specific death. N. Engl. J. Med. 364, 829–841 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Tancredi M. et al. Excess mortality among persons with type 2 diabetes. New Engl. J. Med. 373, 1720–1732 (2015). [DOI] [PubMed] [Google Scholar]
  • 4.American Diabetes Association. Economic costs of diabetes in the US in 2017. Diabetes Care 41, 917–928 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Einarson TR, Acs A, Ludwig C. & Panton UH Prevalence of cardiovascular disease in type 2 diabetes: a systematic literature review of scientific evidence from across the world in 2007–2017. Cardiovasc. Diabetol. 17, 83 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Chilton RJ, Dungan KM, Shubrook JH & Umpierrez GE Cardiovascular risk and the implications for clinical practice of cardiovascular outcome trials in type 2 diabetes. Prim. Care Diabetes 14, 193–212 (2020). [DOI] [PubMed] [Google Scholar]
  • 7.McCoy RG et al. Derivation of an annualized claims-based major adverse cardiovascular event estimator in type 2 diabetes. JACC: Advances 3, 100852 (2024). [Google Scholar]
  • 8.ElSayed NA et al. Standards of Care in Diabetes—2023. Chapter 9. Pharmacologic Approaches to Glycemic Treatment. Diabetes Care 46, S140–S157 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.ElSayed NA et al. Standards of Care in Diabetes—2023. Chapter 10. Cardiovascular Disease and Risk Management. Diabetes Care 46, S158–S190 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Ismail-Beigi F, Moghissi E, Kosiborod M. & Inzucchi SE Shifting paradigms in the medical management of type 2 diabetes: reflections on recent cardiovascular outcome trials. J. Gen. Intern. Med. 32, 1044–1051 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Tsapas A. et al. Comparative effectiveness of glucose-lowering drugs for type 2 diabetes. Ann. Intern. Med. 173, 278–286 (2020). [DOI] [PubMed] [Google Scholar]
  • 12.Zheng SL et al. Association between use of sodium-glucose cotransporter 2 inhibitors, glucagon-like peptide 1 agonists, and dipeptidyl peptidase 4 inhibitors with all-cause mortality in patients with type 2 diabetes: a stematic review and meta-analysis. JAMA 319, 1580–1591 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Zelniker TA et al. Comparison of the effects of glucagon-like peptide receptor agonists and sodium-glucose cotransporter 2 inhibitors for prevention of major adverse cardiovascular and renal outcomes in type 2 diabetes mellitus. Circulation 139, 2022–2031 (2019). [DOI] [PubMed] [Google Scholar]
  • 14.Zelniker TA et al. SGLT2 inhibitors for primary and secondary prevention of cardiovascular and renal outcomes in type 2 diabetes: a systematic review and meta-analysis of cardiovascular outcome trials. Lancet 393, 31–39 (2019). [DOI] [PubMed] [Google Scholar]
  • 15.Selvin E. et al. Cardiovascular outcomes in trials of oral diabetes medications: a systematic review. Arch. Intern. Med. 168, 2070–2080 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Varvaki Rados D, Catani Pinto L, Reck Remonti L, Bauermann Leitao C. & Gross JL The association between sulfonylurea use and all-cause and cardiovascular mortality: a meta-analysis with trial sequential analysis of randomized clinical trials. PLoS Med. 13, e1001992 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Bain S. et al. Cardiovascular events and all-cause mortality associated with sulphonylureas compared with other antihyperglycaemic drugs: a Bayesian meta-analysis of survival data. Diabetes Obes. Metab. 19, 329–335 (2017). [DOI] [PubMed] [Google Scholar]
  • 18.Giugliano D. et al. The effect of DPP-4 inhibitors, GLP-1 receptor agonists and SGLT-2 inhibitors on cardiorenal outcomes: a network meta-analysis of 23 CVOTs. Cardiovasc. Diabetol. 21, 42 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Sim R. et al. Comparative effectiveness of cardiovascular, renal and safety outcomes of second-line antidiabetic drugs use in people with type 2 diabetes: a systematic review and network meta-analysis of randomised controlled trials. Diabet. Med. 39, e14780 (2022). [DOI] [PubMed] [Google Scholar]
  • 20.Downing NS et al. Participation of the elderly, women and minorities in pivotal trials supporting 2011–2013 US Food and Drug Administration approvals. Trials 17, 199 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Hernán MA, Wang W. & Leaf DE Target trial emulation: a framework for causal inference from observational Data. JAMA 328, 2446–2447 (2022). [DOI] [PubMed] [Google Scholar]
  • 22.Mortensen MB & Nordestgaard BG Statin use in primary prevention of atherosclerotic cardiovascular disease according to 5 major guidelines for sensitivity, specificity, and number needed to treat. JAMA Cardiology 4, 1131–1138 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Robinson JG et al. Determining when to add nonstatin therapy: a quantitative approach. J. Am. Coll. Cardiol. 68, 2412–2421 (2016). [DOI] [PubMed] [Google Scholar]
  • 24.Glycemia Reduction in Type 2 Diabetes — Microvascular and cardiovascular outcomes. New Engl. J. Med. 387, 1075–1088 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Fu EL et al. Comparative effectiveness of SGLT2i versus GLP1-RA on cardiovascular outcomes in routine clinical practice. Int. J. Cardiol. 352, 172–179 (2022). [DOI] [PubMed] [Google Scholar]
  • 26.Ueda P. et al. The comparative cardiovascular and renal effectiveness of sodium-glucose co-transporter-2 inhibitors and glucagon-like peptide-1 receptor agonists: a Scandinavian cohort study. Diabetes Obes. Metab. 24, 473–485 (2022). [DOI] [PubMed] [Google Scholar]
  • 27.Kristensen SL 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. 7, 776–785 (2019). [DOI] [PubMed] [Google Scholar]
  • 28.Herges JR, Neumiller JJ & McCoy RG Easing the financial burden of diabetes management: a guide for patients and primary care clinicians. Clin. Diabetes 39, 427–436 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.McCoy RG et al. Second-line therapy for type 2 diabetes management: the treatment/benefit paradox of cardiovascular and kidney comorbidities. Diabetes Care 44, 2302–2311 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.McCoy RG et al. Race and sex differences in the initiation of diabetes drugs by privately insured US adults. Endocrine 73, 480–484 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.D’Andrea E. et al. Comparing effectiveness and safety of SGLT2 inhibitors versus DPP-4 inhibitors in patients with type 2 diabetes and varying baseline HbA1c Levels. JAMA Intern. Med. 183, 242–254 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Shin H, Schneeweiss S, Glynn RJ & Patorno E. Cardiovascular outcomes in patients initiating first-line treatment of type 2 diabetes with sodium-glucose cotransporter-2 inhibitors versus Metformin. Ann. Intern. Med. 175, 927–937 (2022). [DOI] [PubMed] [Google Scholar]
  • 33.Htoo PT et al. Comparative effectiveness of Empagliflozin versus Iraglutide or Sitagliptin in older adults with diverse patient characteristics. JAMA Netw. Open 5, e2237606 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Hernan MA & Robins JM Using big data to emulate a target trial when a randomized trial is not available. Am. J. Epidemiol. 183, 758–764 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Fewell Z, Davey Smith G. & Sterne JA The impact of residual and unmeasured confounding in epidemiologic studies: a simulation study. Am. J. Epidemiol. 166, 646–655 (2007). [DOI] [PubMed] [Google Scholar]
  • 36.Hernan MA & Robins JM Estimating causal effects from epidemiological data. J. Epidemiol. Community Health 60, 578–586 (2006). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Benchimol EI et al. The Reporting of Studies Conducted using Observational Routinely-collected Health Data (RECORD) statement. PLoS Med. 12, e1001885 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Wang SV et al. STaRT-RWE: structured template for planning and reporting on the implementation of real-world evidence studies. Brit. Med. J. 372, m4856 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Wallace PJ, Shah ND, Dennen T, Bleicher PA & Crown WH Optum Labs: building a novel node in the learning health care system. Health Aff. 33, 1187–1194 (2014). [DOI] [PubMed] [Google Scholar]
  • 40.OptumLabs. OptumLabs and OptumLabs Data Warehouse (OLDW) Descriptions and Citation. n.p., March 2023, used with permission from OptumLabs. [Google Scholar]
  • 41.von Elm E. et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Ann. Intern. Med. 147, 573–577 (2007). [DOI] [PubMed] [Google Scholar]
  • 42.McCaffrey DF et al. A tutorial on propensity score estimation for multiple treatments using generalized boosted models. Stat. Med. 32, 3388–3414 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Polley EC, LeDell E, Kennedy C, Lendle S. & van der Laan MJ SuperLearner: Super Learner Prediction. R package version 2.0–28 (2021). [Google Scholar]
  • 44.van der Laan MJ, Polley EC & Hubbard AE Super learner. Stat. Appl. Genet. Mol. Biol. 6, 25 (2007). [DOI] [PubMed] [Google Scholar]
  • 45.Yoshida K. et al. Matching weights to simultaneously compare three treatment groups: comparison to three-way matching. Epidemiology 28, 387–395 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Moore KL & Van der Laan MJ in Design and Analysis of Clinical Trials with Time-to-Event Endpoints (ed. Peace KE) 455–482 (Chapman and Hall/CRC, 2009). [Google Scholar]
  • 47.Grambsch PM & Therneau TM Proportional hazards tests and diagnostics based on weighted residuals. Biometrika 81, 515–526 (1994). [Google Scholar]
  • 48.Prasad V. & Jena AB Prespecified falsification end points: can they validate true observational associations? JAMA 309, 241–242 (2013). [DOI] [PubMed] [Google Scholar]
  • 49.Lipsitch M, Tchetgen Tchetgen E. & Cohen T. Negative controls: a tool for detecting confounding and bias in observational studies. Epidemiology 21, 383–388 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Material
NIHMS1993206-supplement-Supplementary_Material.pdf (1.1MB, pdf)

Data Availability Statement

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 is 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.

RESOURCES