GLP-1 Receptor Agonists Versus Empagliflozin or Dapagliflozin in the MASLD-Heart Failure Phenotype: An Exploratory Propensity-Matched Cohort Study

Background: Treatment selection may confound observational comparisons of glucagon-like peptide-1 receptor agonists (GLP-1 RAs) and sodium-glucose cotransporter-2 inhibitors (SGLT2i) in adults with metabolic dysfunction-associated steatotic liver disease and heart failure (HF). Aims: To compare recorded outcomes between treatment groups and examine whether the associations persisted after restricting the analysis by agent and diabetes status and separating early and later outcome windows. Study Design: Retrospective, initiation-based, active-comparator cohort study (new-user design not verified). Methods: We used the TriNetX US Collaborative Network and a common index eligibility period from January 1, 2018, through June 2, 2026. The revised analyses were performed on September 12, 2026. An additional sensitivity analysis of acute HF excluded patients with this outcome before the start of the day 1-1825 window. The primary analysis compared semaglutide, liraglutide, or dulaglutide with empagliflozin or dapagliflozin using 1:1 propensity score matching and assessed outcomes during days 1-1825. Separate analyses of the broader GLP-1 RA group examined outcomes during days 1-90 and 91-1825 and among patients with type 2 diabetes and no recorded type 1 diabetes. We report endpoint-specific event proportions, risk differences, hazard ratios (HRs), 95% confidence intervals (CIs), and proportionality test results. Results: The primary analysis included 8,550 patients per group after matching. Mortality occurred in 599/8,514 (7.0%) versus 861/8,509 (10.1%); the HR was 0.607 (95% CI, 0.546-0.673). Coded acute HF diagnoses occurred in 1,217/8,550 (14.2%) versus 2,525/8,550 (29.5%); the HR was 0.398 (95% CI, 0.372-0.427), with evidence of non-proportional hazards. In separately matched descriptive interval analyses (8,489 patients per group), the HR for coded acute HF was 0.262 (95% CI, 0.236-0.291) during days 1-90 and 0.507 (95% CI, 0.469-0.548) during days 91-1825. The HR for acute kidney injury during the later interval was 0.740 (95% CI, 0.650-0.842). Among 5,709 matched pairs with type 2 diabetes and no recorded type 1 diabetes, the HR for coded acute HF was 0.431 (95% CI, 0.397-0.468), whereas the HR for ketoacidosis was 0.683 (95% CI, 0.455-1.025). No advantage was observed for acute pancreatitis in any analysis. In a separate analysis excluding patients with prior acute HF, the HR was 0.421 (95% CI, 0.362-0.489), with evidence of non-proportional hazards. Albumin and bilirubin remained imbalanced in a direction favoring the GLP-1 RA group; body mass index also remained imbalanced. Conclusion: GLP-1 RA exposure was associated with lower observed hazards for several outcomes, and the association with coded acute HF was weaker after day 90. The early concentration and magnitude of this association make treatment channeling a plausible explanation, although this analysis cannot establish that explanation. These exploratory findings do not support causal inferences or treatment-preference conclusions and do not challenge randomized evidence supporting SGLT2i use in HF.