In-Hospital Dapagliflozin Misses Primary Endpoint but Mortality Signal Persists

Confirmatory Impact 3/5

The formally neutral primary endpoint (HR 0.86; 95% CI 0.68–1.08; p=0.20) does not change practice on its own, but the consistent mortality signal and pooled data reinforce existing guideline support for safe in-hospital SGLT2i initiation.

Evidence Grade: Moderate

The individual trial was double-blind with central adjudication and low loss to follow-up, but the primary result was imprecise and underpowered owing to lower-than-expected event rates, and the supporting meta-analysis includes post hoc subgroup data, lowering certainty below High.

Study Design

Investigator-initiated, multinational, randomized, double-blind, placebo-controlled cardiovascular outcomes trial with a prespecified two-month follow-up, plus a prespecified random-effects meta-analysis of trials evaluating in-hospital initiation of sodium-glucose cotransporter-2 inhibitors (SGLT2i). Enrollment ran from September 2020 through March 2025 across 210 sites in 5 countries. Registered as NCT04363697. Study drug and funding were provided by AstraZeneca.

Population

2401 adults hospitalized with a primary diagnosis of heart failure (HF) with signs of fluid overload, elevated natriuretic peptides, and clinical stabilization (no intravenous diuretic intensification within 12 hours and no intravenous vasodilators or inotropes within 24 hours before randomization). Median age 69 years (interquartile range [IQR] 58–77), 33.9% women, 18.7% Black race, 35.5% with type 2 diabetes, 71.5% with left ventricular ejection fraction (LVEF) ≤40%, and 44.7% with newly diagnosed HF. Median time from admission to randomization was 3.6 days. Background therapy at randomization included beta-blockers (82.5%), renin-angiotensin system inhibitors (70.1%, of whom 27.0% on an angiotensin receptor-neprilysin inhibitor [ARNI]), and mineralocorticoid receptor antagonists (MRA) (48.6%).

Baseline Characteristics

Primary Outcome Not Met

Composite of time to cardiovascular death or first worsening heart failure event (worsening HF defined as in-hospital deterioration requiring inotropes, mechanical circulatory support, invasive ventilation, or transplantation; unplanned HF rehospitalization; or urgent ambulatory HF visit requiring intravenous diuretics) (2 months)

Statistical Methods

The primary time-to-event outcome was analyzed by stratified log-rank test with stratification by type 2 diabetes and prior HF history; cumulative incidence was estimated by Kaplan-Meier, and treatment effects by stratified Cox proportional-hazards models with the proportional-hazards assumption verified via Schoenfeld residuals. A supportive Bayesian survival model was prespecified. The hierarchical composite was analyzed with an unmatched stratified win ratio, and the KCCQ-12 change with analysis of covariance adjusting for baseline. Safety events were compared with chi-square or Fisher's exact tests. The prespecified meta-analysis used random-effects models with inverse-variance weighting, with heterogeneity assessed by Cochran's Q.

Secondary Outcomes

Safety & Tolerability

⚠ = serious adverse event

Subgroup Analyses

Number Needed to Treat / Harm

Limitations

The trial was underpowered relative to plan because observed primary event rates fell well below the assumed 16% placebo rate, driven by a high proportion of newly diagnosed HF patients. The deliberately short two-month follow-up may have truncated benefit that emerges later, as suggested by EMPULSE data showing the largest worsening HF effect between days 60 and 90. The meta-analysis leans on prior trials with modest samples and, for SOLOIST-WHF, a post hoc subgroup analysis without central adjudication, which may inflate pooled estimates.

Clinical Interpretation

This was a well-conducted, adequately blinded trial with central adjudication that produced a formally neutral primary result, so a large early benefit of in-hospital dapagliflozin on cardiovascular death or worsening HF cannot be claimed. The consistent, statistically significant all-cause mortality reduction and the pooled meta-analytic signal support initiating SGLT2i during hospitalization, but the true effect size is likely closer to the modest DAPA ACT HF-TIMI 68 estimate than to the more optimistic earlier trials. Safety was reassuring and consistent with the known class profile.

Clinical Pearls

Pharmacist Implications

• In-hospital initiation of dapagliflozin 10 mg daily remains appropriate and safe once patients are hemodynamically stabilized, defined here as no intravenous diuretic intensification within 12 hours and no intravenous vasodilators or inotropes within 24 hours.
• Counsel the team that the strongest and most consistent signal across all three trials is early all-cause mortality reduction, which supports starting SGLT2i before discharge rather than deferring to outpatient follow-up.
• Anticipate a modestly higher rate of symptomatic hypotension with dapagliflozin and mitigate by avoiding concurrent overdiuresis and by pacing up-titration of neurohormonal antagonists.
• Reassure prescribers that major hypoglycemia was rare and no diabetic ketoacidosis occurred, so the metabolic safety of inpatient initiation is favorable across diabetic and non-diabetic patients.
• Expect and communicate the known early dip in estimated glomerular filtration rate (eGFR); a modest rise in worsening kidney function events should not prompt reflexive discontinuation absent a true doubling of serum creatinine or other defined criteria.

Guideline Context

The 2022 American Heart Association/American College of Cardiology/Heart Failure Society of America (AHA/ACC/HFSA) guideline and the 2023 European Society of Cardiology (ESC) focused update already give SGLT2i a Class 1 recommendation across the ejection fraction spectrum. These results add nuance rather than contradiction: they reinforce the safety of inpatient initiation and the class benefit but temper expectations of a large early reduction in worsening HF specifically within the first two months.

Monitoring Guidance

Monitor blood pressure and volume status for symptomatic hypotension, particularly when co-titrating renin-angiotensin system inhibitors, ARNIs, and MRAs. Check baseline and follow-up renal function and serum potassium; expect an initial eGFR decline that typically stabilizes, and reserve concern for a doubling of serum creatinine or clinically significant deterioration. Assess volume status to avoid additive diuresis. For diabetic patients, screen for hypoglycemia and diabetic ketoacidosis, though rates here were very low. For any future oncology-context toxicity grading, the current standard is Common Terminology Criteria for Adverse Events (CTCAE) v6.0, though this trial used HF-specific clinical event definitions rather than CTCAE.

Nuanced Considerations

No heterogeneity of treatment effect was observed across prespecified subgroups, including LVEF above versus at or below 40%, diabetes status, and newly diagnosed versus worsening chronic HF. The high fraction of newly diagnosed HF patients (44.7%) lowered event rates and generalizes to a contemporary hospitalized population but reduced statistical power. Enrollment was limited to North America and Europe, which may limit generalizability to other regions.

Controversies & Gaps

The central tension is a neutral primary endpoint in the largest, most rigorous trial versus a significant pooled benefit driven partly by smaller trials with methodologic limitations. Whether the early mortality signal reflects a true class effect or residual optimism from earlier trials remains unresolved. The optimal timing window within hospitalization and whether benefit accrues mainly after two months (as EMPULSE suggested) are open questions that no further dedicated in-hospital initiation trial is planned to answer.

Cost & Logistics

Not discussed. Dapagliflozin is widely available; formulary access and copay considerations for inpatient initiation and discharge continuity are relevant but not addressed in the source.

Full text source: PubMed Central (Open Access)