Heart failure with preserved ejection fraction (HFpEF) is increasingly recognized as a cardiometabolic syndrome driven by obesity and related metabolic risk factors, yet strategies to prevent progression to clinically overt heart failure in at-risk populations remain limited. We evaluated the association between initiation of glucagon-like peptide-1 receptor agonist (GLP-1RA) therapy and subsequent heart failure hospitalization among adults with cardiometabolic risk and no prior heart failure using the TriNetX US Collaborative Network (2010 to 2024). In a new-user, active-comparator design, adults initiating a GLP-1RA were compared with those initiating sitagliptin, with 1:1 propensity score matching to balance demographics, cardiometabolic comorbidities, medications, and laboratory measures. The primary outcome was heart failure hospitalization, assessed from 1 day through 5 years after treatment initiation. Secondary outcomes included clinically diagnosed HFpEF, defined by International Classification of Diseases, Tenth Revision codes, and all-cause mortality. After matching, 219,189 patients were included in each group with well-balanced baseline characteristics. During follow-up, GLP-1RA initiation was associated with a lower cumulative incidence of heart failure hospitalization compared with sitagliptin (5.4% vs 8.1%; risk ratio 0.66; 95% confidence interval 0.65 to 0.68), with consistent findings across landmark, era-restricted, and agent-specific sensitivity analyses. Associations were also observed for lower rates of clinically diagnosed HFpEF (risk ratio 0.72; 95% confidence interval 0.70 to 0.74) and all-cause mortality. In conclusion, among adults with cardiometabolic risk and no prior heart failure, initiation of GLP-1RA therapy was associated with lower rates of heart failure hospitalization compared with sitagliptin, suggesting potential modification of clinical trajectories leading to overt heart failure that warrants confirmation in prospective studies.
Heart failure with preserved ejection fraction (HFpEF) now accounts for approximately half of all heart failure cases and continues to increase in prevalence in parallel with rising rates of obesity, type 2 diabetes mellitus, chronic kidney disease, and other cardiometabolic conditions. ,, Contemporary models conceptualize HFpEF as a systemic cardiometabolic syndrome characterized by chronic inflammation, endothelial dysfunction, coronary microvascular disease, myocardial fibrosis, and impaired ventricular–vascular coupling, rather than an isolated myocardial disorder. ,, Despite its growing burden, effective strategies to prevent HFpEF or delay its clinical manifestation among individuals at high cardiometabolic risk remain limited. Glucagon-like peptide-1 receptor agonists (GLP-1RAs), originally developed for glycemic control, have demonstrated robust cardiovascular benefits beyond glucose lowering, including sustained weight reduction, blood pressure lowering, improvement in insulin sensitivity, and attenuation of systemic inflammation. ,, Large cardiovascular outcome trials have shown that GLP-1RAs reduce major adverse cardiovascular events across diverse populations with and without diabetes, although early studies in heart failure with reduced ejection fraction raised concerns about neutral or adverse effects, underscoring the importance of phenotype-specific evaluation. In contrast, recent randomized trials in patients with established HFpEF and obesity have demonstrated meaningful improvements in symptoms, functional capacity, and quality of life with semaglutide and tirzepatide, providing proof of concept for benefit in this phenotype, though these studies were not designed to evaluate incident heart failure. , Emerging evidence from broader cardiometabolic populations further suggests favorable effects of GLP-1RAs on heart failure–related outcomes, including reductions in heart failure events and cardiovascular mortality, but these analyses have largely included patients with prevalent heart failure or have not distinguished HFpEF from other phenotypes. ,,, Whether initiation of GLP-1 receptor agonist therapy is associated with lower rates of clinically diagnosed HFpEF among individuals without prior heart failure therefore remains uncertain. Given that HFpEF is frequently identified in routine practice through administrative diagnoses rather than systematic imaging, understanding whether cardiometabolic therapies are associated with modification of clinical trajectories leading to HFpEF recognition is clinically relevant. Accordingly, we evaluated the association between initiation of GLP-1RA therapy and subsequent rates of clinically diagnosed HFpEF in a large real-world cohort enriched for cardiometabolic risk and explicitly excluding individuals with prior heart failure.
Methods
Study design and data source
We conducted a retrospective cohort study using the TriNetX US Collaborative Network, a federated electronic health record platform comprising 71 health care organizations across the United States. The database contains deidentified longitudinal patient-level data on demographics, diagnoses coded using the International Classification of Diseases, Tenth Revision (ICD-10), medications mapped to RxNorm, procedures, laboratory values, and vital signs. Because all data are deidentified, this study was exempt from institutional review board oversight.
Study population
Adults aged ≥18 years with cardiometabolic risk who received care between January 1, 2010, and December 31, 2024, were eligible for inclusion. Cardiometabolic risk was defined by the presence of obesity, type 2 diabetes mellitus, or related metabolic comorbidities, identified using diagnosis codes and available anthropometric data. To evaluate incident heart failure, individuals with any prior diagnosis of heart failure (ICD-10 I50.x) before cohort entry were excluded. Additional exclusions included pregnancy, prior bariatric surgery, end-stage kidney disease or dialysis dependence, and active or prior malignancy.
Exposure and comparator
The exposure group consisted of patients initiating a GLP-1RA, including semaglutide, liraglutide, dulaglutide, or tirzepatide. The comparator group consisted of patients initiating sitagliptin, selected a priori as a weight-neutral dipeptidyl peptidase-4 inhibitor with established cardiovascular safety and no known benefit for heart failure outcomes. A new-user, active-comparator design was employed, requiring no prior exposure to GLP-1RAs or sitagliptin. The index date was defined as the date of first prescription for the qualifying agent.
Outcome definitions
The primary outcome was heart failure hospitalization, defined as inpatient admission with a primary or secondary diagnosis of heart failure (ICD-10 I50.x) occurring after the index date. Secondary outcomes included incident clinically diagnosed HFpEF, defined by new ICD-10 codes for diastolic heart failure (I50.3x), and all-cause mortality. Patients with any heart failure diagnoses prior to cohort entry were excluded from all analyses. Outcomes were assessed beginning 1 day after treatment initiation and continued for up to 5 years.
Because systematic echocardiographic adjudication was not available within the federated electronic health record platform, HFpEF ascertainment relied on administrative diagnosis codes, reflecting routine clinical recognition rather than standardized imaging-based phenotyping.
Covariates
Baseline covariates assessed prior to the index date included age, sex, race and ethnicity, cardiometabolic comorbidities (including diabetes, hypertension, dyslipidemia, ischemic heart disease, chronic kidney disease, liver disease, and pulmonary disease), medication use, and available laboratory values such as body mass index and hemoglobin A1c ( Supplementary Table S2 ).
Propensity score matching
To mitigate confounding by indication, propensity score matching was performed using a 1:1 nearest-neighbor approach without replacement and a caliper width of 0.1 pooled standard deviations of the logit of the propensity score. The propensity model included demographic, clinical, medication, and laboratory variables ( Supplementary Table S2 ). Covariate balance was assessed using standardized mean differences, with values <0.10 indicating adequate balance.
Follow-up and statistical analysis
Follow-up began 1 day after the index date and continued until the occurrence of the outcome of interest, death, last recorded clinical encounter, or 5 years, whichever occurred first. Cumulative incidence was estimated using risk proportions, and associations were summarized as risk ratios with corresponding 95% confidence intervals.
Cox proportional hazards models were explored; however, because follow-up duration differed substantially between treatment groups due to treatment persistence and calendar-time effects, resulting in violation of the proportional hazards assumption and potential informative censoring, hazard ratio estimates were not used for primary inference.
A formal competing risk analysis accounting for death as a competing event was not performed because the current TriNetX analytics environment does not support competing risk regression or cumulative incidence function modeling. Accordingly, mortality was treated as a censoring event in time-to-event analyses, and cumulative risk estimates should be interpreted as associations with clinically diagnosed HFpEF rather than cause-specific or subdistribution hazard effects.
Sensitivity analyses
A series of prespecified sensitivity analyses were performed to evaluate robustness of the primary findings and assess potential sources of bias. To reduce influence of early events and exposure misclassification, landmark analyses were conducted at 30 and 90 days after treatment initiation. To account for secular trends in prescribing patterns and cardiometabolic care, era-restricted analyses were performed, including comparisons restricted to 2017 to 2020 (predominantly semaglutide use) and 2022 to 2024 (predominantly tirzepatide use). Agent-specific analyses evaluated semaglutide-based and tirzepatide-based regimens separately compared with sitagliptin, and direct comparisons between tirzepatide and semaglutide were explored to assess class consistency rather than comparative effectiveness.
To assess potential effect modification, additional sensitivity analyses were stratified by age (<65 vs ≥65 years), sex, and severe obesity (body mass index ≥40 kg/m²). Within each stratum, cumulative risks of heart failure hospitalization and clinically diagnosed HFpEF were estimated separately for each treatment group, with associations summarized using risk ratios and 95% confidence intervals. All sensitivity analyses were designed to evaluate consistency of associations across analytic assumptions, time periods, agents, and key demographic and cardiometabolic subgroups rather than to establish causal or subgroup-specific effects.
Statistical software
All analyses were performed within the TriNetX analytics environment using validated R-based statistical procedures. Two-sided p values <0.05 were considered statistically significant.
Results
Study population
After application of inclusion and exclusion criteria, the prematched cohorts included 557,329 adults initiating a GLP-1RA and 238,679 adults initiating sitagliptin, all without a prior diagnosis of heart failure ( Supplementary Table S1 ). Following 1:1 propensity score matching across demographic, clinical, medication, and laboratory variables, the final analytic cohort comprised 219,189 patients in each treatment group, with excellent balance across all covariates (standardized mean differences <0.10 for all variables) ( Table 1 ).
Table 1
Baseline characteristics of the propensity-score–matched cohort
| Characteristic | GLP-1 RA ( n = 219,189) | Sitagliptin ( n = 219,189) | p Value | SMD |
|---|---|---|---|---|
| Demographics | ||||
| Age, years (mean ± SD) | 61.0 ± 11.7 | 61.0 ± 12.3 | 0.082 | 0.005 |
| Female sex, n (%) | 103,962 (47.4) | 108,704 (49.6) | <0.001 | 0.043 |
| Male sex, n (%) | 115,227 (52.6) | 110,485 (50.4) | <0.001 | 0.043 |
| White race, n (%) | 134,366 (61.3) | 133,184 (60.8) | <0.001 | 0.011 |
| Black or African American, n (%) | 42,466 (19.4) | 41,943 (19.1) | 0.045 | 0.006 |
| Hispanic or Latino ethnicity, n (%) | 21,173 (9.7) | 20,541 (9.4) | 0.001 | 0.010 |
| Asian race, n (%) | 9,081 (4.1) | 10,762 (4.9) | <0.001 | 0.037 |
| Cardiometabolic comorbidities | ||||
| Type 2 diabetes mellitus | 101,161 (46.2) | 97,080 (44.3) | <0.001 | 0.037 |
| Hypertension | 78,999 (36.0) | 79,625 (36.3) | 0.049 | 0.006 |
| Dyslipidemia | 67,817 (30.9) | 68,048 (31.0) | 0.451 | 0.002 |
| Chronic kidney disease | 14,058 (6.4) | 14,371 (6.6) | 0.055 | 0.006 |
| Ischemic heart disease | 21,545 (9.8) | 22,533 (10.3) | <0.001 | 0.015 |
| Cerebrovascular disease | 7,638 (3.5) | 8,410 (3.8) | <0.001 | 0.019 |
| Anthropometric and laboratory measures | ||||
| Body mass index, kg/m² (mean ± SD) | 33.9 ± 7.4 | 33.0 ± 7.5 | <0.001 | 0.123 |
| Body weight, lb (mean ± SD) | 216.4 ± 54.5 | 208.4 ± 54.2 | <0.001 | 0.147 |
| HbA1c, % (mean ± SD) | 8.2 ± 2.1 | 8.3 ± 2.0 | <0.001 | 0.027 |
| LDL cholesterol, mg/dL | 92.2 ± 39.1 | 93.0 ± 39.5 | <0.001 | 0.020 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
