Section Structure
1. Highlights
2. Clinical Background
3. Study Design and Methods
4. Main Results
5. Interpretation for Clinical Practice
6. Strengths and Limitations
7. Conclusion
8. Funding and Trial Registration
9. Citation and Selected References
Highlights
In a multinational, retrospective, new-user active-comparator cohort study including 1,245,211 adults with type 2 diabetes mellitus receiving metformin, individual GLP-1 receptor agonists and SGLT2 inhibitors showed broadly similar cardiovascular effectiveness.
Semaglutide and empagliflozin, two of the most commonly used agents, had comparable risks of both 3-point major adverse cardiovascular events and 4-point major adverse cardiovascular events in meta-analytic comparisons.
The overall pattern of similarity extended across within-class and between-class comparisons, and remained consistent in a secondary subgroup of 316,242 patients with established cardiovascular diseases.
These findings support current guideline approaches that treat both drug classes as cardioprotective options, while emphasizing that agent selection in practice may depend more on safety, tolerability, adherence, route of administration, renal and heart failure considerations, cost, and patient preference than on major differences in atherosclerotic cardiovascular effectiveness.
Clinical Background
Type 2 diabetes mellitus is a major driver of cardiovascular morbidity and mortality worldwide. Beyond glycemic control, contemporary diabetes management increasingly prioritizes prevention of myocardial infarction, stroke, heart failure, chronic kidney disease progression, and premature cardiovascular death. Two therapeutic classes have transformed this landscape: glucagon-like peptide 1 receptor agonists, or GLP-1RAs, and sodium-glucose cotransporter 2 inhibitors, or SGLT2Is.
Randomized cardiovascular outcome trials have established that several agents from both classes reduce clinically important cardiovascular events in high-risk populations. However, the evidence base has largely been built from placebo-controlled trials designed to show class or agent benefit against standard care rather than direct head-to-head comparisons between active drugs. As a result, clinicians often face a practical question that trial programs have not fully answered: when choosing among specific GLP-1RAs and SGLT2Is for a patient already receiving metformin, do individual agents meaningfully differ in cardiovascular effectiveness?
This question matters because the agents are not identical. They differ in route of administration, frequency of dosing, effects on weight, gastrointestinal tolerability, heart failure and renal data, adverse effect profiles, and cost. Some agents also have more robust randomized trial evidence than others. Real-world comparative effectiveness studies can therefore provide useful complementary evidence, especially when they apply rigorous methods to reduce confounding and assess consistency across healthcare systems.
The study by Bu and colleagues addresses this gap through a large-scale, multinational, real-world comparison of six GLP-1RAs and four SGLT2Is among adults with type 2 diabetes who initiated second-line therapy after metformin.
Study Design and Methods
This investigation was a retrospective, new-user active-comparator cohort study conducted across 10 U.S. and non-U.S. administrative claims and electronic health record databases. The design is notable for several reasons. First, the new-user approach reduces bias introduced by including prevalent users, whose treatment experience may already reflect survivorship or tolerability. Second, active-comparator analyses are generally more clinically meaningful than comparisons against nonuse, because treatment groups are more likely to have similar indications for therapy. Third, the use of multiple databases improves scale and heterogeneity, while allowing cross-database diagnostics and meta-analysis.
The cohort included 1,245,211 adults with type 2 diabetes mellitus who were receiving metformin and initiated one of the following as second-line treatment: six GLP-1RAs, namely albiglutide, dulaglutide, exenatide, liraglutide, lixisenatide, and semaglutide; or four SGLT2Is, namely canagliflozin, dapagliflozin, empagliflozin, and ertugliflozin. The most commonly used agents were empagliflozin in 393,499 patients, semaglutide in 235,585, dapagliflozin in 208,666, and dulaglutide in 207,348. A secondary subgroup analysis focused on 316,242 patients with established cardiovascular diseases.
The primary endpoints were 3-point major adverse cardiovascular events, defined as acute myocardial infarction, stroke, or sudden cardiac death, and 4-point major adverse cardiovascular events, which added hospitalization or emergency room visit with heart failure. Secondary outcomes included the individual components of these composite endpoints.
Hazard ratios were estimated for pairwise comparisons between agents using Cox proportional hazards models. To strengthen causal inference in the setting of observational data, the investigators used propensity score adjustment, negative control calibration, and prespecified diagnostics intended to assess exchangeability, covariate balance, and residual bias. Random-effects meta-analysis was then used to summarize results across those data sources that passed diagnostic thresholds.
The authors also evaluated whether the study had sufficient discriminatory capacity to identify clinically meaningful similarity. Importantly, diagnostics indicated adequate power to detect similarity in hazard ratios between 0.8 and 1.2 for the commonly used agents. This point matters because a null result in underpowered real-world comparisons can be uninformative; here, the investigators attempted to show that many of the key comparisons were adequately powered and methodologically defensible.
Main Results
The principal finding was that individual GLP-1RAs and SGLT2Is demonstrated broadly similar cardiovascular effectiveness, both within each class and across classes. In practical terms, no single commonly used agent emerged as clearly superior to another for major cardiovascular event prevention in this metformin-treated population.
The most clinically intuitive example is the comparison between semaglutide and empagliflozin. For 3-point major adverse cardiovascular events, the meta-analytic hazard ratio was 1.05 with a 95% confidence interval of 0.79 to 1.39. For 4-point major adverse cardiovascular events, the meta-analytic hazard ratio was 0.95 with a 95% confidence interval of 0.81 to 1.12. Both estimates are centered near unity, and the confidence intervals do not suggest a compelling advantage of either drug for these broad cardiovascular outcomes.
The same general pattern extended across other pairwise comparisons among the more commonly prescribed agents. The article’s overall conclusion is not merely that semaglutide and empagliflozin were similar, but that the comparative landscape among the evaluated GLP-1RAs and SGLT2Is was characterized by substantial overlap in cardiovascular effectiveness estimates. That held for comparisons within the GLP-1RA class, within the SGLT2I class, and across the two classes.
In the subgroup with established cardiovascular diseases, findings were consistent with the primary analysis. This is particularly relevant because guideline-directed use of these therapies is often strongest in patients with known atherosclerotic cardiovascular disease, heart failure, or chronic kidney disease. The persistence of broadly comparable effects in a higher-risk subgroup increases the clinical credibility of the study’s main message.
Another important observation is that the analysis incorporated heart failure into the 4-point composite endpoint. SGLT2 inhibitors are often favored clinically when heart failure risk is prominent, given randomized trial evidence for heart failure hospitalization reduction across diabetic and non-diabetic populations. Yet, in this study’s broad comparison of 4-point major adverse cardiovascular events, no dramatic separation was seen between commonly used SGLT2Is and GLP-1RAs. This should not be interpreted as negating known heart failure benefits of SGLT2Is, but rather as indicating that in a real-world second-line diabetes treatment setting, overall composite cardiovascular performance may be more similar than many clinicians assume when focusing only on class narratives.
The authors report that study diagnostics confirmed adequate equipoise, covariate balance, and sufficient power for common-agent comparisons. These are not trivial methodological footnotes. In observational comparative effectiveness research, the central concern is confounding by indication: patients selected for one agent may differ systematically from those selected for another in ways that affect cardiovascular risk. By demonstrating diagnostic adequacy before pooling results, the investigators aimed to avoid overinterpreting biased or poorly calibrated comparisons.
At the same time, the confidence intervals for some pairwise estimates remain wide, particularly for less commonly used medications or outcomes with fewer events. Therefore, the most robust inferences apply to the more frequently prescribed agents and to the overarching conclusion of broad similarity, rather than to very fine distinctions between less common drugs.
Interpretation for Clinical Practice
For clinicians, this study reinforces an increasingly practical view of modern diabetes therapeutics: both GLP-1RAs and SGLT2Is are evidence-based cardioprotective classes, and the choice among individual agents may often rest less on differences in atherosclerotic cardiovascular outcomes and more on other dimensions of care.
Those dimensions include heart failure and chronic kidney disease status, route of administration, expected weight reduction, adverse effects, injection burden, gastrointestinal tolerability, risk of genital mycotic infection, volume depletion, access, and affordability. For example, if minimizing heart failure hospitalization or slowing kidney disease progression is the dominant goal, many clinicians may still preferentially select an SGLT2 inhibitor because that preference is anchored in robust randomized outcome trial data and guideline recommendations. Conversely, if a patient prioritizes weight loss and is willing to use an injectable or oral semaglutide where available, a GLP-1RA may be favored. This study suggests that such choices do not appear to impose a major tradeoff in broad cardiovascular protection for most patients.
The findings also align with network meta-analyses of cardiovascular outcome trials, which generally show that both classes reduce major cardiovascular events, though with somewhat different strengths across specific cardiorenal endpoints. Current diabetes and cardiology guidelines already recommend using either a GLP-1RA or an SGLT2I in appropriately selected patients with type 2 diabetes and cardiovascular or renal risk, independent of baseline glycated hemoglobin in many cases. Bu and colleagues provide real-world evidence that supports this class-level equivalence framework while refining it to the level of individual agents.
One valuable implication is therapeutic flexibility. When insurance coverage changes, drug shortages occur, or tolerability limits the use of one agent, switching to another agent within or across these classes may preserve cardiovascular benefit better than clinicians might fear. In systems-level care, that could support formulary strategies that prioritize access and adherence without necessarily compromising major cardiovascular outcomes.
Strengths and Limitations
The study has several strengths. Its scale is exceptional, with more than 1.2 million patients across multiple healthcare systems. The new-user active-comparator framework is well suited to comparative effectiveness questions. The inclusion of both U.S. and non-U.S. data enhances external relevance. The use of prespecified diagnostics, propensity score methods, negative control calibration, and random-effects meta-analysis reflects a sophisticated analytic strategy that is stronger than many routine observational studies.
Nevertheless, important limitations remain. First, residual confounding cannot be eliminated in nonrandomized data, even with advanced statistical adjustment. Factors such as body mass index, diabetes duration, frailty, socioeconomic status, smoking, clinician preference, and unmeasured disease severity may not be perfectly captured across all databases. Second, claims and electronic health record data are dependent on coding accuracy; misclassification of exposure, outcomes, and comorbidities is possible. Third, not all agents were equally represented. Estimates for less commonly used drugs are likely less precise and should be interpreted cautiously.
Fourth, the on-treatment, or per-protocol, approach can better estimate effects while patients remain exposed but may also be influenced by informative discontinuation if patients stop therapy because of early adverse effects, cost, or loss of effectiveness. Total follow-up analyses help complement this but address a slightly different clinical question. Fifth, the composite endpoints may conceal heterogeneity in specific outcomes. It remains plausible that some agents differ modestly in stroke, myocardial infarction, or heart failure effects even if their overall composite performance appears similar.
Finally, this study primarily addresses cardiovascular effectiveness, not comparative safety, tolerability, persistence, weight change, glycemic durability, renal outcomes, or cost-effectiveness. Those domains are highly relevant in real-world decision-making and may ultimately matter more than small differences in major adverse cardiovascular event rates.
Conclusion
This large multinational real-world study suggests that individual GLP-1 receptor agonists and SGLT2 inhibitors confer broadly comparable cardiovascular benefit when used as second-line therapy after metformin in adults with type 2 diabetes mellitus. The pattern was consistent across the overall cohort and among patients with established cardiovascular diseases, and the findings are directionally concordant with network meta-analytic interpretations of cardiovascular outcome trials.
The practical message is not that all drugs are interchangeable in every respect. Rather, for major cardiovascular outcomes, no commonly used individual agent clearly separated itself from the others in a clinically decisive way. This supports current guideline strategies that view both GLP-1RAs and SGLT2Is as cardioprotective pillars of diabetes management. In day-to-day practice, treatment selection should therefore be individualized around patient phenotype, comorbid heart failure or kidney disease, safety profile, tolerability, route of administration, adherence, access, and patient preference.
Future work should move beyond major adverse cardiovascular events alone and compare these agents on kidney outcomes, heart failure-specific endpoints, safety, persistence, quality of life, and cost. Head-to-head randomized trials would remain the most definitive way to resolve lingering uncertainty, but until then, carefully conducted large-scale real-world analyses such as this one offer clinically useful guidance.
Funding and Trial Registration
The abstract provided does not report funding details or a ClinicalTrials.gov registration number. Readers should consult the full Journal of the American College of Cardiology article for complete disclosures, funding sources, and any study governance details.
Citation and Selected References
Primary study: Bu F, Wu R, Ostropolets A, Aminorroaya A, Chen HY, Chai Y, Dhingra LS, Falconer T, Hsu JC, Kim C, Lau WCY, Man KKC, Minty E, Morales DR, Nishimura A, Thangraraj P, Van Zandt M, Yin C, Khera R, Hripcsak G, Suchard MA. Comparative Cardiovascular Effectiveness of Glucagon-Like Peptide 1 Receptor Agonists and Sodium-Glucose Cotransporter 2 Inhibitors in Diabetes Mellitus. Journal of the American College of Cardiology. 2026-04-15;87(21):2963-2977. PMID: 41984016.
Selected background references: Marso SP, Daniels GH, Brown-Frandsen K, et al. Liraglutide and cardiovascular outcomes in type 2 diabetes. N Engl J Med. 2016;375:311-322. Zinman B, Wanner C, Lachin JM, et al. Empagliflozin, cardiovascular outcomes, and mortality in type 2 diabetes. N Engl J Med. 2015;373:2117-2128. Gerstein HC, Colhoun HM, Dagenais GR, et al. Dulaglutide and cardiovascular outcomes in type 2 diabetes (REWIND). Lancet. 2019;394:121-130. Hernandez AF, Green JB, Janmohamed S, et al. Albiglutide and cardiovascular outcomes in patients with type 2 diabetes and cardiovascular disease (Harmony Outcomes). Lancet. 2018;392:1519-1529. American Diabetes Association. Standards of Care in Diabetes—2025. Diabetes Care. 2025. McDonagh TA, Metra M, Adamo M, et al. 2023 focused update of the 2021 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure. Eur Heart J. 2023;44:3627-3639.