Background: Sodium-glucose cotransporter 2 (SGLT2) inhibitors, originally developed as glucose-lowering agents for type 2 diabetes mellitus, have been shown in large, randomized trials to improve clinical outcomes in heart failure (HF) across the ejection fraction (EF) spectrum. The imaging-derived and hemodynamic changes that accompany these benefits remain incompletely characterized, and it is not known whether the patterns of cardiac remodeling observed differ between HF with reduced ejection fraction (HFrEF, left ventricular ejection fraction [LVEF] ≤ 40%) and HF with preserved ejection fraction (HFpEF, LVEF ≥ 50%). This systematic review synthesized randomized evidence on the effects of SGLT2 inhibitors on cardiac remodeling parameters and described the patterns observed in each phenotype.
Methods: PubMed/MEDLINE, Scopus and Embase were searched for randomized controlled trials (RCTs) published between February 2016 and March 2026. Eligible studies enrolled adults with chronic HF receiving an SGLT2 inhibitor, included a non-SGLT2-inhibitor comparator and assessed cardiac structure, function or loading by echocardiography, cardiac magnetic resonance (CMR) or cardiac catheterization over at least three months. Risk of bias was assessed with Cochrane RoB 2, separately for each outcome domain and targeting the effect of assignment to intervention, and certainty of evidence with GRADE, separately by phenotype and imaging modality. A structured narrative synthesis following Synthesis Without Meta-analysis (SWiM) guidance that prioritized the direction, magnitude and precision of randomized between-group treatment effects was performed.
Results: Fourteen publications from thirteen unique RCTs (1232 participants) were included: four trials enrolled HFrEF (LVEF ≤ 40%), five enrolled mixed reduced and mildly reduced EF cohorts (LVEF < 50% or 35–49%), four enrolled HFpEF (LVEF ≥ 50%) and one enrolled a mixed ejection-fraction population. No trial enrolled an exclusively HFmrEF population. In reduced EF, the direction of effect on LV volumes favored SGLT2 inhibition in four of eight trials reporting volumes, but the estimates were not compatible in magnitude, and the one adequately powered trial at low risk of bias (EFFORT) was neutral; the largest volumetric estimates came from EMPA-TROPISM, which published no adjusted between-group estimate. The only prespecified, adjusted and confidence-interval-bearing volumetric estimates were those of SUGAR-DM-HF (left ventricular end-systolic volume index −6.0 mL/m
2, 95% CI −10.8 to −1.2; end-diastolic volume index −8.2 mL/m
2, 95% CI −13.7 to −2.6). The global longitudinal strain improved in the trial with blinded core-laboratory reading (EFFORT −1.44%, 95% CI −2.42 to −0.46), but effect sizes in the open-label trials without stated reader blinding were two to three times larger, and the CMR-derived strain was unchanged. In HFpEF, volumes, EF, mass and strain were unchanged throughout. The most secure finding of the review was invasive: in CAMEO-DAPA, dapagliflozin reduced pulmonary capillary wedge pressure at rest by 3.5 mmHg (95% CI −6.6 to −0.4) and at peak exercise by 5.7 mmHg (95% CI −10.8 to −0.7). NT-proBNP results were directionally inconsistent, and the trials in which NT-proBNP was a prespecified primary endpoint (Empire HF, CANDLE) were neutral. Certainty of evidence was very low for 18 of 20 evidence profiles and low for the remaining two.
Conclusions: Randomized imaging and hemodynamic studies suggest phenotype-related differences in the cardiac remodeling associated with SGLT2 inhibitor therapy. In studies enrolling patients with reduced EF, treatment was associated predominantly with reverse remodeling of left ventricular volumes and modest improvements in myocardial function, whereas in preserved EF, the observed changes were mainly confined to filling pressures and diastolic function, with no consistent evidence of left ventricular volumetric remodeling. Given the limited number of studies, substantial heterogeneity, and the overall low certainty of evidence, these findings should be considered hypothesis-generating and warrant confirmation in adequately powered phenotype-specific mechanistic trials.
Full article