Risk of Heart Failure Hospitalization for GLP-1 Receptor Agonists Versus DPP-4 Inhibitors or SGLT-2 Inhibitors in Patients With Type 2 Diabetes: A Target Trial Emulation
Xu Y, Huang T, Zhang Y, Ji D, Tuttle KR, Carrero JJ, Fu EL
Circulation (2026) · n=63,083
In a Stockholm 2010–2021 target-trial emulation of 63,083 adults with type 2 diabetes, GLP-1 receptor agonists reduced 3-year heart-failure hospitalization versus DPP-4 inhibitors (HR 0.77; 95% CI 0.66–0.91) but were not different from SGLT-2 inhibitors (HR 1.02; 95% CI 0.85–1.18) — the cleanest comparative-effectiveness signal to date that the GLP-1RA cardioprotective effect is real against DPP-4i but does not match SGLT-2i for the heart-failure endpoint.
Xu and colleagues at the Karolinska Institutet (Carrero), Leiden University Medical Center (Fu), and the University of Washington (Tuttle) report a target-trial-emulation analysis in Circulation of heart-failure hospitalization risk for GLP-1 receptor agonists relative to the two other modern type-2 diabetes drug classes — DPP-4 inhibitors and SGLT-2 inhibitors. The target-trial-emulation design is the current methodological gold standard for using routinely-collected health-care data to approximate a randomized comparison: it specifies eligibility, treatment strategies, time-zero, and outcome ascertainment as if running an RCT, and uses propensity-score and inverse-probability weighting to balance baseline characteristics across treatment arms.
The Stockholm 2010–2021 cohort yielded two emulated trials. Trial 1 compared GLP-1RA versus DPP-4 inhibitors in 32,979 patients (42% GLP-1RA initiators, 58% DPP-4i initiators). Trial 2 compared GLP-1RA versus SGLT-2 inhibitors in 30,104 patients (49% GLP-1RA, 51% SGLT-2i). The primary outcome was hospitalization for heart failure (HHF) over a 3-year follow-up. GLP-1RA produced lower 3-year HHF incidence than DPP-4i (3.4% vs 4.3%, HR 0.77, 95% CI 0.66–0.91). GLP-1RA was effectively equivalent to SGLT-2i (3.6% vs 3.3%, HR 1.02, 95% CI 0.85–1.18). Results were consistent across individual agents within the class — relevant for the semaglutide and liraglutide pages on this site, since both are represented in the GLP-1RA arm — and across subgroups defined by age, sex, baseline cardiovascular and kidney status.
The findings sit alongside the existing trial-grade evidence anchored by LEADER (liraglutide), SUSTAIN-6 (semaglutide), and SELECT (semaglutide CV) — those trials established the cardiovascular benefit of GLP-1RAs versus placebo. The Xu 2026 paper extends the picture by anchoring GLP-1RAs against active comparators in the same modern T2D pharmacopoeia: GLP-1RAs are clearly better than DPP-4i for heart-failure prevention, and clearly indistinguishable from SGLT-2i for that endpoint.
This is a target-trial-emulation cohort study using routinely-collected health-care data from Stockholm, Sweden — not a randomized trial. The methodology is among the most rigorous available for non-randomized comparative effectiveness, but residual confounding by indication is the always-present limit. Patients initiating GLP-1RAs in Stockholm during this period skewed slightly different from SGLT-2i initiators in ways the propensity weighting attempts to balance but cannot eliminate. The 3-year follow-up is sufficient for an HHF event-rate readout but short relative to the lifelong chronic-therapy framing of T2D management; longer-term differential effects are not addressed.
The cohort is Stockholm-resident, predominantly White European, and embedded in a single-payer health system with broadly accessible prescribing of all three drug classes; generalizability to U.S. real-world prescribing patterns, where access and adherence are heterogeneous and shaped by insurance coverage, is a different question. The paper does not break out individual agents within the GLP-1RA class by frequency of events; the per-agent consistency claim is reported in subgroup analyses but the per-agent absolute event counts are small. There is no declared commercial sponsorship; the authors have institutional grant support from European research councils. Read this entry as the strongest comparative-effectiveness signal for GLP-1RAs in T2D heart-failure prevention — but as one that confirms equivalence with, rather than superiority over, SGLT-2 inhibitors.
07·Member discussion
No member discussion yet.
Member-only conversation lives here — cycle notes, practitioner commentary, pattern-matching. Be the first paying member to start the thread.