DAPA-HF Subgroup Analyses: Who Responded Most and Least
At a glance
| Parameter | Detail |
|---|---|
| N | 4,744 |
| Intervention | Dapagliflozin 10 mg once daily |
| Comparator | Placebo |
| Duration | Median 18.2 months |
| Primary endpoint | Composite of worsening heart failure (hospitalization or urgent IV therapy) or cardiovascular death |
| Key result | HR 0.74 (95% CI 0.65 to 0.85; P <0.001), a 26% relative risk reduction |
Why Subgroup Analyses Matter Here
The top-line DAPA-HF result, a 26% reduction in the primary composite, changed practice overnight. But a single hazard ratio tells prescribers nothing about the 67-year-old woman with an LVEF of 28% and no diabetes sitting in their clinic. Pre-specified subgroup analyses test whether the overall effect holds across clinically meaningful strata. When interaction P-values are non-significant, the message is: the drug works broadly. When point estimates shift, even without crossing unity, that pattern can sharpen clinical intuition.
The primary DAPA-HF publication reported 14 pre-specified subgroups. Subsequent post-hoc analyses, published as companion papers and presented at ACC and ESC, extended the lens to biomarker tertiles, geographic region, and background therapy.
Pre-Specified Subgroup Results
The framework below organizes every pre-specified subgroup by its point estimate, confidence interval, and interaction P-value. We flag subgroups where the 95% CI crossed 1.0 (suggesting a possibility of no benefit in that stratum) and those where point estimates were notably lower than the overall HR 0.74.
Diabetes Status: The Defining Question
| Subgroup | n | HR (95% CI) | P-interaction |
|---|---|---|---|
| Type 2 diabetes | 2,139 | 0.75 (0.63 to 0.90) | 0.80 |
| No diabetes | 2,605 | 0.73 (0.60 to 0.88) | , |
This was the subgroup the field watched most closely. Prior SGLT2 inhibitor cardiovascular outcome trials (EMPA-REG OUTCOME, CANVAS, DECLARE-TIMI 58) enrolled only patients with type 2 diabetes. DAPA-HF was the first major trial to include patients without diabetes. The virtually identical hazard ratios (0.75 vs. 0.73) and a non-significant interaction P-value of 0.80 confirmed that the mechanism of benefit in heart failure is independent of glucose lowering. This finding directly informed the FDA label expansion for dapagliflozin to include HFrEF irrespective of diabetes.
A subsequent analysis by Petrie et al. (Lancet Diabetes Endocrinol, 2020) further stratified by HbA1c and found that even patients with HbA1c <5.7% (definitively non-diabetic, non-prediabetic) derived full benefit. This closed the door on residual glycemic confounding.
Age
| Subgroup | n | HR (95% CI) | P-interaction |
|---|---|---|---|
| <55 years | 900 | 0.70 (0.51 to 0.96) | 0.76 |
| 55 to <65 years | 1,242 | 0.76 (0.59 to 0.97) | , |
| 65 to <75 years | 1,717 | 0.77 (0.63 to 0.95) | , |
| ≥75 years | 885 | 0.68 (0.53 to 0.88) | , |
Point estimates favored dapagliflozin across all age quartiles. The oldest patients (≥75 years) showed, if anything, a slightly larger effect (HR 0.68). This is clinically relevant because older patients are often excluded from new therapies due to safety concerns. Martinez et al. (JACC Heart Fail, 2020) published a dedicated age-stratified analysis confirming tolerability in elderly DAPA-HF participants, with no excess in falls, fractures, or volume depletion events requiring hospitalization.
Sex
| Subgroup | n | HR (95% CI) | P-interaction |
|---|---|---|---|
| Male | 3,563 | 0.73 (0.63 to 0.85) | 0.60 |
| Female | 1,181 | 0.79 (0.59 to 1.06) | , |
Women comprised only 23.4% of the trial population, a recognized limitation. The point estimate for women (HR 0.79) favored dapagliflozin, but the confidence interval crossed 1.0. The interaction P-value (0.60) was non-significant, meaning this overlap with unity reflects underpowering in the female subgroup rather than a true differential effect. The 2022 AHA/ACC/HFSA guideline update does not differentiate SGLT2 inhibitor recommendations by sex.
LVEF Strata
| Subgroup | n | HR (95% CI) | P-interaction |
|---|---|---|---|
| LVEF ≤30% | 2,756 | 0.72 (0.61 to 0.84) | 0.49 |
| LVEF >30% | 1,988 | 0.79 (0.63 to 0.99) | , |
Both strata benefited. The slightly lower HR in the ≤30% group aligns with the general principle that sicker patients have more events to prevent, yielding larger absolute risk reductions.
Race and Geographic Region
| Subgroup | n | HR (95% CI) | P-interaction |
|---|---|---|---|
| White | 3,292 | 0.76 (0.65 to 0.89) | 0.74 |
| Black | 225 | 0.73 (0.38 to 1.38) | , |
| Asian | 1,095 | 0.69 (0.52 to 0.91) | , |
The Asian subgroup showed a numerically strong effect (HR 0.69). The Black subgroup had wide confidence intervals (n = 225), limiting interpretability. This underrepresentation is a known gap in SGLT2 inhibitor trial data broadly.
BMI Categories
| Subgroup | n | HR (95% CI) | P-interaction |
|---|---|---|---|
| BMI <30 kg/m² | 2,994 | 0.73 (0.62 to 0.86) | 0.78 |
| BMI ≥30 kg/m² | 1,679 | 0.76 (0.61 to 0.95) | , |
No meaningful differential. The caloric loss from glycosuria (~200 to 300 kcal/day) does not appear to explain the cardiovascular benefit, consistent with the diabetes-status finding above.
Post-Hoc and Extended Subgroup Analyses
NT-proBNP Tertiles
A dedicated biomarker analysis from DAPA-HF examined response by baseline NT-proBNP. Patients in the highest tertile (NT-proBNP >2,500 pg/mL) had the most events but still showed a consistent relative risk reduction. The absolute benefit was largest in this group: approximately 7.5 fewer primary endpoint events per 100 patient-years, compared to roughly 3 events per 100 patient-years in the lowest tertile.
Background Therapy
Dapagliflozin's benefit was additive to guideline-directed medical therapy. Subgroups defined by use of ACE inhibitors/ARBs/ARNI, beta-blockers, and mineralocorticoid receptor antagonists (MRAs) all showed consistent hazard ratios. Patients on sacubitril-valsartan (ARNI), the most contemporary background regimen, still derived a 26% relative risk reduction (HR 0.75, 95% CI 0.50 to 1.13; wide CI due to the smaller ARNI subgroup of ~500 patients).
Renal Function
| Subgroup | eGFR range | HR (95% CI) |
|---|---|---|
| ≥60 mL/min/1.73 m² | n ≈ 2,800 | 0.76 (0.63 to 0.91) |
| <60 mL/min/1.73 m² | n ≈ 1,900 | 0.71 (0.58 to 0.87) |
Patients with CKD stage 3 or worse responded at least as well, with a numerically lower HR. This is consistent with findings from the later DAPA-CKD trial, which demonstrated dapagliflozin's renal protection benefit.
What the Subgroup Pattern Tells Prescribers
Three clinical takeaways emerge from the aggregate subgroup data:
1. There is no identifiable subgroup that does not benefit. All 14 pre-specified subgroup analyses showed point estimates below 1.0. All interaction P-values were non-significant. This is unusual, since most HF trials show at least one subgroup where the CI crosses unity widely enough to raise doubt.
2. Diabetes status is irrelevant to the prescribing decision. The near-identical HRs (0.75 vs. 0.73) eliminated the concern that SGLT2 inhibitors were "diabetes drugs borrowed for cardiology." The 2022 AHA/ACC/HFSA guidelines now give SGLT2 inhibitors a Class I recommendation for HFrEF without any diabetes qualifier.
3. Older and sicker patients should not be excluded. Point estimates in patients ≥75 years, with LVEF ≤30%, and with eGFR <60 were all at or below the overall HR. The absolute benefit in high-risk strata was larger because event rates were higher.
Limitations of These Subgroup Data
Subgroup analyses, even pre-specified ones, are hypothesis-generating. The trial was powered for the overall comparison, not for individual strata. Several important caveats apply:
- Women were underrepresented (23.4%). The CI crossing 1.0 in the female subgroup does not mean women don't benefit, but it does mean we cannot confirm the magnitude with the same confidence.
- Black patients comprised only 4.7% of the cohort. Subgroup estimates in this stratum are unreliable due to wide confidence intervals.
- Multiple comparisons were not adjusted for. With 14 subgroups, the probability of a false-positive interaction signal by chance is non-trivial, though in this case the opposite concern applies: the uniformity itself is the finding.
- Post-hoc analyses (biomarker tertiles, background therapy) carry additional risk of data-driven bias. They should inform clinical reasoning, not dictate it.
- Geographic heterogeneity in background care standards (e.g., ARNI use ranged from ~10% in some regions to ~15% in others) could confound region-specific estimates.
Placing DAPA-HF Subgroups in Context
The EMPEROR-Reduced trial (empagliflozin in HFrEF, published 2020) showed a remarkably similar subgroup pattern: consistent benefit across diabetes status, age, sex, and renal function. A meta-analysis by Zannad et al. (Lancet, 2020) pooling DAPA-HF and EMPEROR-Reduced confirmed consistent effects across all major subgroups, strengthening the class-level recommendation for SGLT2 inhibitors in HFrEF.
Frequently asked questions
Did patients without diabetes benefit equally from dapagliflozin in DAPA-HF?
Yes. The hazard ratio was 0.73 (95% CI 0.60 to 0.88) in patients without diabetes versus 0.75 (0.63 to 0.90) in those with type 2 diabetes, with an interaction P-value of 0.80, indicating no differential effect by diabetes status.
Were older patients (≥75 years) safe to treat with dapagliflozin in DAPA-HF?
Patients aged 75 and older had a hazard ratio of 0.68 (0.53 to 0.88), numerically better than the overall trial result. Safety data showed no excess volume depletion, falls, or fractures in this group.
Did women respond differently to dapagliflozin in DAPA-HF?
The point estimate in women (HR 0.79) favored dapagliflozin, but the confidence interval crossed unity (0.59 to 1.06). The interaction P-value was 0.60, suggesting the overlap reflects underpowering (women were 23.4% of the cohort) rather than a true lack of effect.
Does baseline kidney function affect dapagliflozin's benefit in heart failure?
No. Patients with eGFR <60 mL/min/1.73 m² had a hazard ratio of 0.71, numerically better than those with preserved renal function (HR 0.76). The interaction was non-significant.
Did BMI affect how well dapagliflozin worked in DAPA-HF?
Patients with BMI <30 and ≥30 kg/m² had nearly identical hazard ratios (0.73 and 0.76), with a non-significant interaction P-value of 0.78. Weight-related caloric loss from glycosuria does not explain the cardiovascular benefit.
Were Black patients adequately represented in DAPA-HF subgroup analyses?
Black patients comprised only 4.7% of the trial (n = 225). The point estimate favored dapagliflozin (HR 0.73), but the wide confidence interval (0.38 to 1.38) limits reliable conclusions for this group.
Does dapagliflozin add benefit on top of sacubitril-valsartan (ARNI)?
In the ARNI subgroup (~500 patients), the HR was 0.75, consistent with the overall trial effect. The confidence interval was wide due to small sample size, but no evidence of attenuation was observed.
How do DAPA-HF subgroup findings compare with EMPEROR-Reduced?
EMPEROR-Reduced (empagliflozin) showed a nearly identical subgroup pattern. A pooled meta-analysis confirmed class-level consistency across diabetes status, age, sex, and renal function for SGLT2 inhibitors in HFrEF.
What do the 2022 AHA/ACC/HFSA guidelines say about SGLT2 inhibitors for HFrEF?
SGLT2 inhibitors (dapagliflozin and empagliflozin) carry a Class I (strongest) recommendation for HFrEF, regardless of diabetes status. This recommendation was directly informed by DAPA-HF and EMPEROR-Reduced subgroup data.
Should clinicians select heart failure patients for dapagliflozin based on subgroup characteristics?
No. The consistent benefit across all 14 pre-specified subgroups, with no significant interaction, means dapagliflozin should be offered to all eligible HFrEF patients. Restricting use to specific subgroups would deny benefit without evidence-based justification.
References
- McMurray JJV, Solomon SD, Inzucchi SE, et al. Dapagliflozin in patients with heart failure and reduced ejection fraction. N Engl J Med. 2019;381(21):1995-2008. PubMed
- Petrie MC, Verma S, Docherty KF, et al. Effect of dapagliflozin on worsening heart failure and cardiovascular death in patients with heart failure with and without diabetes. Lancet Diabetes Endocrinol. 2020;8(10):P800-P810. PubMed
- Packer M, Anker SD, Butler J, et al. Cardiovascular and renal outcomes with empagliflozin in heart failure (EMPEROR-Reduced). N Engl J Med. 2020;383(15):1413-1424. PubMed
- Heidenreich PA, Bozkurt B, Aguilar D, et al. 2022 AHA/ACC/HFSA guideline for the management of heart failure. J Am Coll Cardiol. 2022;79(17):e263-e421. PubMed
- Zannad F, Ferreira JP, Pocock SJ, et al. SGLT2 inhibitors in patients with heart failure with reduced ejection fraction: a meta-analysis. Lancet. 2020;396(10254):819-829. PubMed
- FDA. Farxiga (dapagliflozin) prescribing information. Label