Comparative Outcomes of Empagliflozin to Dapagliflozin in Patients With Heart Failure

Key Points

Question  What are the comparative outcomes of empagliflozin and dapagliflozin in reducing all-cause mortality and hospitalizations in patients with heart failure?

Findings  In this cohort study of 28 075 patients with heart failure naive to sodium-glucose cotransporter-2 inhibitor therapy, patients who started using empagliflozin were less likely to experience the primary outcome of composite of all-cause mortality or hospitalization compared with those who started using dapagliflozin.

Meaning  These findings suggest that further studies are needed to clarify mechanisms that could explain outcomes of empagliflozin and dapagliflozin in patients with heart failure.

Importance  Sodium-glucose cotransporter-2 (SGLT2) inhibitors have been shown to have benefits when used in patients with heart failure. The comparative outcomes of SGLT2 inhibitors relative to each other has not been well defined and may impact medication selection.

Objective  To determine the comparative outcomes of empagliflozin and dapagliflozin on reducing the composite of all-cause mortality and hospitalizations in patients with heart failure.

Design, Setting, and Participants  This multicenter retrospective cohort study included patients with heart failure from August 18, 2021, and December 6, 2022, in the TriNetX Research Collaborative, a centralized database of deidentified electronic medical record data from a network of 81 health care organizations. Eligible patients had a diagnosis of heart failure, had never received an SGLT2 inhibitor previously, and were newly started on empagliflozin or dapagliflozin. Patients were followed up for 1 year.

Exposure  Initiation of dapagliflozin or empagliflozin.

Main Outcomes and Measures  The primary outcome was the time to the composite of all-cause mortality or hospitalization between study days 1 to 365. Kaplan-Meier analyses, hazard ratios (HRs), and 95% CIs were used to assess the primary outcome.

Results  Among 744 914 eligible patients, 28 075 began empagliflozin (15 976 [56.9%]) or dapagliflozin (12 099 [43.1%]). After nearest-neighbor matching for demographics, diagnoses, and medication use, there were 11 077 patients in each group. Of patients who received empagliflozin, 9247 (57.9%) were male, 3130 (19.6%) were Black individuals, and 9576 (59.9%) were White individuals. Similarly, of those who received dapagliflozin, 7439 (61.5%) were male, 2445 (20.2%) were Black individuals, and 7131 (58.9%) were White individuals. Patients receiving empagliflozin were less likely to experience the composite of all-cause mortality or hospitalization compared with those initiated on dapagliflozin (3545 [32.2%] vs 3828 [34.8%] events; HR, 0.90 [95% CI, 0.86-0.94]) in the year following SGLT2 inhibitor initiation and less likely to be hospitalized (HR, 0.90 [95% CI, 0.86-0.94]). All-cause mortality did not differ between exposure groups (HR, 0.91 [95% CI, 0.82-1.00]). There was no difference in mean hemoglobin A_1c or adverse events between groups.

Conclusions and Relevance  In this cohort study, patients who initiated empagliflozin were less likely to experience the composite of all-cause mortality or hospitalization compared with patients who started dapagliflozin. Additional studies are needed to confirm these finding.

The sodium-glucose cotransporter-2 (SGLT2) inhibitors empagliflozin and dapagliflozin reduce cardiovascular death and heart failure hospitalizations in patients with heart failure.¹ However, cardiac medications within the same class may not all have the same benefit. For example, carvedilol reduces mortality by 16% relative to metoprolol in patients with heart failure,² and chlorthalidone is more potent than hydrochlorothiazide in the treatment of essential hypertension.³ In patients with diabetes, empagliflozin may be associated with greater weight loss, reduction of blood pressure, and reduction of cholesterol compared with dapagliflozin.⁴ In patients with heart failure, a single center retrospective study suggested that empagliflozin may be associated with improvements in left ventricular ejection fraction and functional status compared with dapagliflozin.⁵ However, the outcomes of empagliflozin vs dapagliflozin on clinically important patient-centered outcomes for patients with heart failure is unclear. In this multicenter retrospective cohort study, we sought to compare the composite outcome of all-cause mortality and hospitalization between those initiated on empagliflozin vs dapagliflozin in patients with heart failure.

This study follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. Informed consent and review were not required because this study was designated to not be human participant research by the institutional review board of Boston University.

We used the TriNetX Research Collaborative Network,^6,7 a network of 81 health care organizations primarily in North America that contribute deidentified electronic medical record data to a central database. We included patients with heart failure (International Statistical Classification of Diseases and Related Health Problems, Tenth Revision [ICD-10] code: I50x), who had never received SGLT2 inhibitors previously, and were newly started on empagliflozin or dapagliflozin. Patients were included if they met criteria between August 18, 2021 (after publication of the Dapagliflozin Evaluation to Improve the Lives of Patients with Preserved Ejection Fraction Heart Failure, or DELIVER, Trial⁸ and after both SGLT2 inhibitors were approved for heart failure by the US Food and Drug Administration),^9,10 and December 6, 2022 (to allow all patients to have 1-year of follow-up time). Study day 0 was defined as the day of SGLT2 inhibitor initiation.

The intervention group was defined as initiation of empagliflozin, and the comparator group was defined as initiation of dapagliflozin. Consistent with an intention-to-treat clinical trial, all patients were analyzed according to the SGLT2 inhibitor initially received, regardless of changes in medication use after initiation.

The primary outcome was the time to the composite of all-cause mortality or hospitalization between study days 1 to 365. Due to limitations in the online TriNetX Query Builder and Analytics platforms, we were unable to determine cause-specific mortality or hospitalizations. Secondary outcomes were all-cause mortality, hospitalization, and last measured hemoglobin A_1c. Adverse effects were defined as occurrence of urinary tract infection (ICD-10 code: N39.0) or diabetic ketoacidosis (ICD-10 codes: E10.1x or E11.1x).

From each patient, we extracted covariables from study days −365 to 0 for inclusion in propensity score models that we thought were likely to confound the association between SGLT2 inhibitor selection and the composite outcome. Covariables included demographics, cardiac and diabetes-related diagnoses and medication use, glomerular filtration rate, hemoglobin A_1c, natriuretic peptides, left ventricular ejection fraction, and hospitalizations. A complete list of covariables and covariable definitions are included in eTable 1 in Supplement 1.

Prior to matching, covariables were summarized using mean (SD) and No. (%) as appropriate. We generated propensity scores for empagliflozin vs dapagliflozin initiation using logistic regression and included all covariables in the model. Covariables, except for age, were included in the model as categorical variables. For other variables that were originally continuous (glomerular filtration rate, hemoglobin A_1c, natriuretic peptides, and left ventricular ejection fraction), values were assigned into clinically relevant categories. This categorization allows for the inclusion of all covariables and patients in the models, even in the setting of multiple measurements or missing data. For example, a patient with no hemoglobin A_1c measured in the year prior to study day 0 would be assigned a 0 for all hemoglobin A_1c categories, and a patient with multiple hemoglobin A_1c measurements in the year prior to study day 0 could potentially be assigned a 1 for multiple hemoglobin A_1c categories. After propensity score generation, we used 1:1 greedy nearest neighbor propensity score matching without replacement to create balanced cohorts. Balance was visualized using density curves of the propensity score and formally assessed using absolute standardized mean differences (SMDs) with a SMD of less than 0.1 defining similarity. We then used the Kaplan-Meier method¹¹ and log-rank test to compare outcomes between matched groups; hazard ratios (HRs), 95% CIs, and absolute risk differences (95% CI) were also reported. Patients were censored from the Kaplan-Meier analysis on the day after the last fact in their TriNetX record. E-values¹² were calculated using the observed HRs and a freely available online calculator.^13-15 We repeated analyses in patients with heart failure with reduced ejection fraction (HFrEF) (ICD-10 codes: I50.2x or I50.4x) and heart failure with preserved ejection fraction (HFpEF) (ICD-10 codes: I50.3x or I50.4x). For the secondary outcome of last measured hemoglobin A_1c, patients with no values were excluded from the analysis and the t statistic was used to compare mean values between patients who received empagliflozin and dapagliflozin.

Cohort identification and statistical analyses were conducted using the TriNetX Platform¹⁶ Query Builder and Analytics Functions, respectively, on December 6, 2023. This point-and-click platform allows users to conduct observational studies using curated variable definitions and a limited set of analytic approaches. Kaplan-Meier curves were recreated using R version 4.2.1 (R Project for Statistical Computing). α was set at .05, and all hypothesis tests were 2-sided. We did not account for multiple testing. Thus, secondary outcomes and analyses should be considered hypothesis generating only.

Among 744 914 patients with heart failure and naive to SGLT2 inhibitor therapy, 28 075 began empagliflozin (15 976 [56.9%]) or dapagliflozin (12 099 [43.1%]) (Figure 1). Prior to matching, the mean (SD) age was 66.4 (13.4) years for participants who received empagliflozin and 63.8 (14.2) years for participants who received dapagliflozin. Of those who received empagliflozin, 9247 (57.9%) were male, 3130 (19.6%) were Black patients, and 9576 (59.9%) were White patients. Similarly, of those who received dapagliflozin, 7439 (61.5%) were male, 2445 (20.2%) were Black patients, and 7131 (58.9%) were White patients (Table). The use of β blockers (SMD, 0.043) and angiotensin converting enzyme inhibitors (SMD, 0.065) inhibitors was similar between prematched groups. The use of angiotensin II inhibitors (7731 [63.9%] vs 8852 [55.4%]; SMD, 0.174) and sacubitril (5471 [45.2%] vs 5258 [32.9%]; SMD, 0.254) was higher in those who received dapagliflozin compared with those who received empagliflozin (Table). Postmatching characteristics (11 007 patients per group) were all similar between groups (Table). The largest difference in covariables after matching was for the use of angiotensin converting enzyme inhibitors (empagliflozin: 2917 [26.5%] vs dapagliflozin: 2982 [27.1%]; SMD, 0.013). Propensity score density curves before and after matching are shown in eFigure 1 in Supplement 1.

In the 1 year after SGLT2 inhibitor initiation, 3545 patients (32.2%) who received empagliflozin experienced the composite outcome of death or hospitalization vs 3828 (34.8%) of those who received dapagliflozin (HR, 0.90 [95% CI, 0.86-0.94]; log-rank P < .001; E-value 1.36) (Figure 2). Similarly, patients who received empagliflozin were less likely to be hospitalized (3270 [29.7%] events vs 3537 [32.1%] events; HR, 0.90 [95% CI, 0.86-0.94]), but all-cause mortality did not differ between exposure groups (691 [6.3%] events vs 764 [6.9%] events; HR, 0.91 [95% CI, 0.82-1.00]) (eFigures 2 and 3 in Supplement 1). A total of 4188 (38.0%) of patients who received empagliflozin and 4422 (40.2%) who received dapagliflozin had at least 1 hemoglobin A_1c measurement in the 1 year after SGLT2 inhibitor initiation. Among patients with hemoglobin A_1c measurements the mean (SD) of last measured hemoglobin A_1c levels was 6.8% (1.6%) in both groups (difference in means P = .64) (to convert to proportion of total hemoglobin, multiply by 0.01). In those who received empagliflozin, 652 patients (5.9%) experienced at least 1 adverse event (diabetic ketoacidosis: 40 events [6.1%]; urinary tract infection: 620 events [95.1%]) vs 714 patients (6.5%) among those who received dapagliflozin (diabetic ketoacidosis: 44 events [6.2%]; urinary tract infection: 681 events [95.4%]) (difference in proportions, P = .08).

Among those with HFrEF (16 892 patients), characteristics were balanced after matching (eTable 2 in Supplement 1). In the 1-year following SGLT2 inhibitor initiation, patients who received empagliflozin had fewer composite outcome events compared with those who received dapagliflozin (2430 [33.2%] events vs 2598 [35.5%] events; HR, 0.92 [95% CI, 0.87-0.97]) (eFigure 4 in Supplement 1). Similarly, among those with HFpEF (10 911 individuals) (eTable 3 in Supplement 1), patients who received empagliflozin were less likely to experience all-cause mortality or hospitalization compared with those who received dapagliflozin (1332 [34.3%] events vs 1424 [36.7%] events; HR, 0.91 [95% CI, 0.84-0.98]) (eFigure 5 in Supplement 1).

In this multicenter retrospective cohort study that used clinical data, patients who initiated empagliflozin were less likely to experience the composite of all-cause mortality or hospitalization compared with patients who initiated dapagliflozin. These results were determined by differences in rates of hospitalization. Our results suggest that there were possible differences in medication outcomes between specific medications within the SGLT2 inhibitor class. Future studies are needed to clarify potential mechanisms that could explain the observed differences.

Our results should be considered in the context of prior work. A small, single center retrospective study found that empagliflozin (vs dapagliflozin) was associated with greater increases in left ventricular ejection fraction and New York Heart Association Class.⁵ The results of Hao et al¹⁷ and our own may suggest that there are differences in the degree of cardiac remodeling between different SGLT2 inhibitors. Unlike our results, a recent meta-analysis¹ of clinical trials showed similar improvements in cardiovascular outcomes between empagliflozin and dapagliflozin compared with placebo. We speculate that the discrepant results between the association and observational outcomes data could be explained by differences in adherence (that were unable to be determined in our study) or differences in comorbidity burden (eg, diabetes) and treatments^4,19 that could lead to synergistic outcomes of empagliflozin relative to dapagliflozin⁴ or differences in all-cause outcomes associated with heart failure-specific outcomes. Future studies should explore potential mechanisms by which the association and outcomes may differ and directly compare outcomes of empagliflozin to dapagliflozin in a pragmatic comparative effectiveness randomized controlled trial.

This study has limitations. Observational studies are at risk for unmeasured confounding (eg, New York Heart Association functional class); however, the E-value of 1.36 suggests that our findings would only be altered in the presence of an unmeasured confounder with a 30% or greater association with treatment assignment and outcome. Current guidelines¹⁸ recommend treating patients with heart failure with quadruple therapy, including angiotensin receptor-neprilysin inhibitors, SGLT2 inhibitors, β blockers, and mineralocorticoid receptor antagonists. In our study, use of each component of quadruple therapy ranged from approximately 30% to 80% of patients. The reasons that patients did not receive all components of quadruple therapy is unclear. Our results should not be used as evidence to include empagliflozin over dapagliflozin in quadruple therapy regimens. We were unable to calculate cause-specific mortality or hospitalization due to limitations in the TriNetX platform. TriNetX platform limitations did not allow us to calculate cluster-robust standard errors to account for pair membership or patient practice clustering. Finally, due to the platform limitations, we were unable to determine the specific onset of heart failure in included patients, thus unable to quantify potential risks for immortal time bias.²⁰ However, both exposures in this study have the same indications, and inclusion was limited to time after both treatments were available and approved for use in patients with heart failure, all of which likely minimize risks of immortal time.

In this cohort study using propensity score matching, empagliflozin was associated with lower rates of hospitalization at 1 year compared with dapagliflozin. Future studies are needed confirm these findings and to understand why outcomes may differ from those of efficacy trial meta-analyses.

Accepted for Publication: March 4, 2024.

Published: May 2, 2024. doi:10.1001/jamanetworkopen.2024.9305

Correction: This article was corrected on July 24, 2024, to fix a numerical error in the results section.

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2024 Modzelewski KL et al. JAMA Network Open.

Corresponding Author: Nicholas A. Bosch, MD, The Pulmonary Center, Boston University Chobanian and Avedisian School of Medicine, 72 E Concord St, R-304, Boston, MA 02118 (nabosch@bu.edu).

Author Contributions: Dr Bosch had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Drs Modzelewski and Pipilas are cofirst authors.

Concept and design: Modzelewski, Bosch.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Modzelewski, Bosch.

Critical review of the manuscript for important intellectual content: All authors.

Statistical analysis: Bosch.

Obtained funding: Bosch.

Administrative, technical, or material support: Modzelewski.

Supervision: Bosch.

Conflict of Interest Disclosures: Dr Bosch reported receiving grants from the US Department of Defense and the National Heart, Lung, and Blood Institute outside the submitted work. No other disclosures were reported.

Funding/Support: This study was supported by grants 1KL2TR001411 and 1UL1TR001430 from the National Institute of Health National Center for Advancing Translational Sciences and the Boston University Chobanian and Avedisian School of Medicine Department of Medicine Career Investment Award.

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: This study’s contents are solely the responsibility of the authors and do not necessarily represent the official views of the National Institutes of Health or Boston University.

Data Sharing Statement: See Supplement 2.