#heart-failure #HFpEF #SGLT2-inhibitors #MRA #GLP-1-receptor-agonists #obesity-HFpEF #heart-failure-management #guidelines

HFpEF Management

Definition

HFpEF management is organised by an absence: no pharmacotherapy has ever reduced mortality in this phenotype. Every positive trial improved heart-failure hospitalisation, symptoms or both, and the mortality columns stayed null. That single fact explains the shape of the pathway — two foundational drugs rather than four, no device therapy with randomised mortality evidence, and comorbidity control carrying as much of the burden as the prescription.

What changed in 2026 is that there are now two foundational drugs rather than one. An SGLT2 inhibitor and an MRA are both Class I, Level A — the SGLT2i on EMPEROR-Preserved and DELIVER, the MRA on a pooled TOPCAT plus FINEARTS-HF meta-analysis rather than on any positive spironolactone trial. (sources/HF-ESC-2026, rating: very high)

What the phenotype is — definition, diagnostic scores, the ASE algorithm, latent HFpEF — is on entities/HFpEF. The shared syndrome is on entities/Heart-Failure, and the HFrEF pathway, which shares the first two drugs and nothing after them, is on concepts/HFrEF-Management.

Key Concepts

The Outcome Trials

Trial	Agent	n	LVEF entry	Primary endpoint	Result	CV death	All-cause death
PARAGON-HF	sacubitril-valsartan vs valsartan	4,796	≥45%	Total HF hospitalisations + CV death	RR 0.87 (0.75–1.01); P=0.06	HR 0.95 (NS)	—
TOPCAT	spironolactone	3,445	≥45%	CV death, aborted arrest, HF hospitalisation	HR 0.89 (0.77–1.04); P=0.14	HR 0.90 (NS)	HR 0.91 (NS)
EMPEROR-Preserved	empagliflozin	5,988	>40%	CV death or first HF hospitalisation	HR 0.79 (0.69–0.90); NNT 31	HR 0.91 (NS)	HR 1.00
DELIVER	dapagliflozin	6,263	>40%	Worsening HF or CV death	HR 0.82 (0.73–0.92)	HR 0.88 (NS)	—
FINEARTS-HF	finerenone	6,001	≥40%	Total worsening HF events + CV death	Rate ratio 0.84 (0.74–0.95); P=0.007	HR 0.93 (NS)	HR 0.93 (NS)
STEP-HFpEF	semaglutide	529	≥45%, BMI ≥30, no diabetes	KCCQ-CSS and weight	+7.8 pts; −10.7 pp	—	—
STEP-HFpEF DM	semaglutide	616	≥45%, BMI ≥30, T2DM	KCCQ-CSS and weight	+7.3 pts; −6.4 pp	—	—
SUMMIT	tirzepatide	731	≥50%, BMI ≥30	CV death or worsening HF; KCCQ-CSS	HR 0.62 (0.41–0.95); +6.9 pts	HR 1.58 (NS)	HR 1.25 (NS)

(Table: sources/arni-paragon-hf-nejm-2019, sources/spironolactone-hfpef-topcat-nejm-2014, sources/empagliflozin-hfpef-nejm-2021, sources/dapagliflozin-deliver-nejm-2022, sources/finerenone-hfpef-fineartshf-nejm-2024, sources/semaglutide-stephfpef-nejm-2023, sources/semaglutide-stephfpef-dm-nejm-2024, sources/tirzepatide-hfpef-summit-nejm-2025, rating: very high)

Two patterns run through the table: no CV death endpoint has been met by any agent, and every positive trial reduced HF events rather than death — which is why guidelines say "HF hospitalisation or CV death" while the mortality columns stay null.

Guideline Positions

ESC 2026: SGLT2i and MRA both Class I/A; ACE-I/ARB/ARNI IIb/C only, useful mainly when there is a coexisting indication such as hypertension, diabetes or CKD. The rest of the Class I list is short — dynamic diuretic dosing for congestion, and treatment of aetiology and comorbidity. (sources/HF-ESC-2026, rating: very high)
AHA 2022: BP control COR 1; AF management COR 2a; ARB and ARNi COR 2b at the lower LVEF boundary; MRA COR 2b. Nitrates and PDE5 inhibitors are COR 3: No Benefit — NEAT-HFpEF and RELAX both negative. (sources/HF-AHA-2022)
Pooled SGLT2i: HR 0.80 for CV death or first HF hospitalisation, HR 0.74 for HF hospitalisation, HR 0.88 for CV death alone (P=0.052). (sources/HF-update-ESC-2023)

The two guidelines differ by two recommendation steps on MRA, and that gap is not a timing artefact — the argument is set out under Contradictions.

SGLT2 Inhibitors

Class I/A in ESC 2026, across the whole LVEF ≥50% range. EMPEROR-Preserved gave HR 0.79 (NNT 31) and DELIVER HR 0.82 for worsening HF or CV death; neither met its CV-death endpoint individually. (sources/HF-ESC-2026, rating: very high; sources/empagliflozin-hfpef-nejm-2021, sources/dapagliflozin-deliver-nejm-2022, rating: very high)
The benefit is hospitalisation, not survival. The pooled CV-death estimate is HR 0.88 (P=0.052), and neither trial was powered for mortality given a CV death rate about half that of HFrEF (3.8 vs 7.9 per 100 patient-years). (sources/HF-update-ESC-2023)
The two trials disagree at the top of the EF range — DELIVER found no attenuation at LVEF ≥60% whereas EMPEROR-Preserved suggested possible attenuation at ≥65%, which determines whether benefit can be assumed in the oldest, most comorbid patients. See Contradictions.

MRA — TOPCAT and FINEARTS-HF

Guideline status changed decisively in 2026. An MRA — steroidal or non-steroidal — is now Class I, Level A foundational therapy for symptomatic HFpEF, on a pooled meta-analysis of TOPCAT and FINEARTS-HF. AHA 2022 still assigns COR 2b. (sources/HF-ESC-2026, rating: very high)

TOPCAT's headline result hides a stratum interaction (P=0.01): hospitalisation-stratum HR 1.01 versus BNP-stratum HR 0.65 (P=0.003). Hyperkalaemia doubled (18.7% vs 9.1%) and about a third discontinued. (sources/spironolactone-hfpef-topcat-nejm-2014, rating: high)
FINEARTS-HF is the only positive MRA trial in this population, and its benefit was consistent in patients already on an SGLT2i — suggesting complementary mechanism rather than overlap. Hyperkalaemia 3.0% vs 1.4%; kidney composite numerically worse. See entities/Finerenone. (sources/finerenone-hfpef-fineartshf-nejm-2024, rating: very high)

Diuretics and Congestion

Dynamic, individualised loop-diuretic dosing according to volume status is Class I, Level A — to relieve symptoms, improve exercise capacity and reduce HF hospitalisation. The Task Force graded it A despite the absence of a dedicated trial because effectiveness is considered beyond doubt. Use the lowest dose that maintains euvolaemia; the dose required drifts over time. (sources/HF-ESC-2026, rating: very high)
Diuretics have no demonstrated disease-modifying effect. Their effect on morbidity and mortality has never been tested in an adequately powered trial, and every HFpEF disease-modifier trial ran on a high background of loop diuretic use. Chronic thiazide use in stable patients is discouraged because it causes electrolyte disturbance that goes undetected outside hospital; short-term addition to a loop diuretic is Class IIa for resistant oedema. (sources/HF-ESC-2026)
Recurrent congestion despite an escalating loop dose: concepts/Diuretic-Resistance. The decompensation pathway is on concepts/Decompensated-HF-Management.

Obesity-HFpEF — GLP-1 and Dual GIP/GLP-1

The KCCQ gain is the largest of any HFpEF therapy, by an order of magnitude. Semaglutide's +7.8 points is 4–15× the 0.5–2.3 points seen with SGLT2i, ARNi or spironolactone — and STEP-HFpEF almost halved serious adverse events (13.3% vs 26.7%), driven by fourfold fewer cardiac events. (sources/semaglutide-stephfpef-nejm-2023, rating: very high)
The effect is not purely weight loss. In STEP-HFpEF DM the weight reduction was ~40% smaller than in STEP-HFpEF but the symptom benefit was similar, and NT-proBNP fell despite weight loss — implicating decongestion and inflammation alongside adiposity. (sources/semaglutide-stephfpef-dm-nejm-2024, rating: very high)
Tirzepatide is the first agent with a hard-outcome benefit in obesity-HFpEF (CV death or worsening HF HR 0.62), and it did not require elevated natriuretic peptides for entry — an explicit acknowledgement that adiposity suppresses them. See entities/Tirzepatide. (sources/tirzepatide-hfpef-summit-nejm-2025, rating: very high)
ESC 2026 now recommends them: semaglutide or tirzepatide IIa/B1 for symptomatic HF with LVEF ≥45% and BMI ≥30 kg/m², regardless of diabetes status; bariatric surgery IIb/C at BMI ≥35 when structured lifestyle change plus weight-reducing medication has failed. AHA 2022 has no equivalent. See concepts/NuSH-Therapies and concepts/GLP-1-Receptor-Agonists. (sources/HF-ESC-2026)

ARNi and the Renin-Axis Question

The primary composite of total HF hospitalisations plus CV death did not reach significance, nor did either component. Exploratory signals after the primary failed — NYHA improvement (OR 1.45), halved worsening renal function (HR 0.50), KCCQ +1.0 — are hypothesis-generating only, as is the suggested benefit at LVEF 45–57% and in women. ESC 2026 rates ARNi IIb/C alongside ACE-I and ARB; AHA 2022 gives COR 2b. See entities/Sacubitril-Valsartan. (sources/arni-paragon-hf-nejm-2019, rating: very high; sources/HF-ESC-2026)

Blood Pressure

Hypertension is the leading modifiable cause of HFpEF. Prefer RAAS inhibition (ARNi > ARB > ACEi), adding an MRA for both blood pressure and HFpEF benefit; SGLT2i lower pressure modestly; diuretics are needed for volume control before other agents work. Goal SBP <130 mmHg. See entities/Hypertension. (sources/HT-AHA-2025, rating: very high)

Iron Deficiency and Anaemia

Iron deficiency is slightly more prevalent in HFpEF than in HFrEF, and both iron deficiency and anaemia independently predict reduced functional capacity, worse quality of life and worse prognosis. (sources/HF-ESC-2026, rating: very high)
The treatment recommendation stops at HFrEF. IV ferric carboxymaltose is Class I/B1 for symptoms and quality of life and IIa/B1 for HF hospitalisation — but only in symptomatic HFrEF. Every supporting trial enrolled LVEF ≤45% and the guideline states that data in HFpEF are scarce, so there is nothing to extrapolate from. (sources/HF-ESC-2026; sources/fe-hf-heartfid-nejm-2023, rating: high)
Erythropoiesis-stimulating agents should be avoided for anaemia in HF — RED-HF showed no reduction in death or hospitalisation and an excess of thromboembolic events. (sources/HF-ESC-2026)

Beta-Blockers — the Inverse of HFrEF

Not foundational therapy in HFpEF, and not recommended for blood-pressure control. Negative chronotropy impairs diastolic filling and removes heart-rate reserve that these patients depend on for cardiac output during exertion. ESC 2026 places them among agents that "may be used particularly for comorbidities," with uncertainty about any effect on morbidity and mortality — not among drugs to withdraw. (sources/HF-ESC-2026, rating: very high; sources/HT-AHA-2025)
Coexisting COPD is the setting where avoidance is argued. Chronotropic impairment matters more when bronchial obstruction already limits exercise, and this is the narrow basis for not starting a beta-blocker for HFpEF alone. Where there is another indication — angina, AF, post-MI — they are continued. This is the exact inverse of concepts/HFrEF-Management, where bisoprolol is the preferred agent. See concepts/HF-COPD-Comorbidity. (sources/hf-copd-hfreview-2025, rating: high)

Aetiology-Specific Therapy

Four causes inside the HFpEF population have their own treatment, and none of it is HFpEF-specific:

ATTR cardiac amyloidosis — 5–13% of HFpEF. Work-up on entities/Heart-Failure; treatment and prognosis on entities/ATTR-Amyloidosis and imaging on concepts/Cardiac-Amyloidosis-Imaging. (sources/echo-hfpef-ase-2025, rating: very high)
Anderson–Fabry disease — SGLT2 inhibitors are recommended, and cardiac myosin inhibitors must not be used, since mavacamten is approved only for sarcomeric obstructive HCM. See concepts/Fabry-Cardiomyopathy. (sources/fabry-ehj-2024, rating: high)
Post-capillary pulmonary hypertension — affects at least half of symptomatic HFpEF, and no approved therapy targets its pulmonary vascular component: PDE5 inhibitors are ineffective and riociguat failed. The one positive signal is procedural — pulmonary-artery denervation — but it was open-label without sham control. See concepts/Pulmonary-Hypertension and concepts/Pulmonary-Artery-Denervation. (sources/HF-AHA-2022, rating: very high; sources/padn-hf-ph-nejm-2026, rating: high)
Atrial fibrillation — a different NT-proBNP rule-out threshold applies in AF, and AF drives decompensation. See entities/Atrial-Fibrillation and concepts/Atrial-Cardiomyopathy.

Contradictions / Open Questions

A June 2026 institutional handbook still lists MRA in HFpEF as optional. The Hospital Authority handbook treats HFmrEF and HFpEF together under "diuretics for volume overload, SGLT2i," with ACE-I or ARB, ARNI and MRA as "other medications to be considered" — the AHA 2022 position, against this page's ESC 2026 Class I/A. It also keeps a single NT-proBNP <125 pg/mL rule-out rather than the age-adjusted thresholds, and its HFpEF definition carries no "never previously <50%" clause. (sources/Handbook-HA-2024, rating: high)
MRA in HFpEF: ESC 2026 Class I/A versus AHA 2022 COR 2b. Not a timing artefact like the SGLT2i gap. The ESC grade rests on a meta-analysis pooling TOPCAT — neutral overall, with a data-integrity problem in Russia/Georgia — and FINEARTS-HF, which used finerenone at LVEF >40%. No trial of spironolactone in HFpEF has ever been positive, so the strongest recommendation now sits on the weakest trial base of any foundational therapy. (sources/HF-ESC-2026, rating: very high; sources/spironolactone-hfpef-topcat-nejm-2014, rating: high)
The two SGLT2i trials disagree at the top of the EF range. DELIVER found no attenuation at LVEF ≥60% whereas EMPEROR-Preserved suggested possible attenuation at ≥65% — a difference that determines whether benefit can be assumed in the oldest, most comorbid patients. (sources/dapagliflozin-deliver-nejm-2022, sources/empagliflozin-hfpef-nejm-2021)
EMPEROR-Preserved's 23% discontinuation rate may bias hard outcomes toward the null, higher than DAPA-HF (10.5%) and DELIVER. (sources/empagliflozin-hfpef-nejm-2021)
No HFpEF pharmacotherapy has reduced mortality — SGLT2i, finerenone, ARNi, spironolactone, semaglutide and tirzepatide all failed on death. For SGLT2i the pooled CV-death estimate is HR 0.88 (P=0.052), and neither trial was powered for mortality given a CV death rate about half that of HFrEF (3.8 vs 7.9 per 100 patient-years). Either mortality is driven by mechanisms these drugs do not touch (cancer, non-cardiovascular death, frailty), or the disease is too heterogeneous for one agent. (sources/finerenone-hfpef-fineartshf-nejm-2024, sources/HF-update-ESC-2023)
TOPCAT is unreliable in two independent ways, and the ESC Class I rests on it anyway. A post-hoc metabolite analysis found canrenone in only ~38% of Russian "active treatment" participants versus >90% in the Americas, indicating placebo dispensing in both arms; excluding those sites, the composite shows benefit (Americas HR ~0.82; P=0.026). Separately, the BNP stratum benefited (HR 0.65) while the hospitalisation stratum did not (HR 1.01), suggesting entry by biochemical evidence selected a more volume-overloaded, MRA-responsive population. (sources/spironolactone-hfpef-topcat-nejm-2014, rating: high)
PARAGON-HF: near-miss or true negative? P=0.06 against alpha 0.048 raises whether a placebo comparator would have reached significance, since valsartan's own ARB activity may have attenuated the ARNI effect. (sources/arni-paragon-hf-nejm-2019, rating: very high)
ARNI at the lower EF boundary versus SGLT2i across the whole range. PARAGON-HF suggests ARNI benefit at LVEF 45–57%; SGLT2i benefits across >40%. Whether ARNI is complementary in lower-EF HFpEF or practically irrelevant now that SGLT2i is Class I cannot be resolved without a combination trial. (sources/arni-paragon-hf-nejm-2019, sources/HF-update-ESC-2023)
Finerenone versus spironolactone has never been compared, and its kidney signal reverses in HFpEF. The apparent divergence (rate ratio 0.84 significant versus HR 0.89 not) reflects total-versus-time-to-first endpoints, different background therapy (13.6% SGLT2i versus none) and TOPCAT's integrity problem — so finerenone cannot be called superior. Separately, the kidney composite was numerically worse (HR 1.33) in a low-albuminuria population, cautioning against assuming the nephroprotection seen in diabetic CKD. (sources/finerenone-hfpef-fineartshf-nejm-2024, sources/spironolactone-hfpef-topcat-nejm-2014)
Semaglutide's benefit shrinks in patients already on SGLT2i. In STEP-HFpEF DM the KCCQ-CSS difference was 5.3 points (−0.2 to 10.7) in SGLT2i users versus 8.3 points (4.5–12.1) in non-users. The interaction was prespecified but not powered, so whether it adds meaningfully on top of SGLT2 inhibition is unresolved. (sources/semaglutide-stephfpef-dm-nejm-2024)
Semaglutide's KCCQ gain is 4–15× larger than any prior HFpEF drug and the mechanism is unexplained. Weight loss alone cannot account for it — the NT-proBNP fall and 43.5% CRP reduction implicate direct haemodynamic and anti-inflammatory effects — and separating the two would need a matched weight-loss control arm that no trial has. (sources/semaglutide-stephfpef-nejm-2023)
Tirzepatide reduces HF events while deaths trend higher in the same trial. Composite HR 0.62, yet all-cause deaths 19 vs 15 (HR 1.25) and CV deaths 10 vs 5 (HR 1.58), both non-significant; 11 of 15 CV deaths were not preceded by worsening HF. Whether tirzepatide defers HF events without altering the mortality trajectory is unresolved. (sources/tirzepatide-hfpef-summit-nejm-2025)
SUMMIT did not require elevated natriuretic peptides, revising its endpoints about a year before completion after FDA discussion. That challenges the premise that raised peptides are needed to start GLP-1-based therapy, but also means it enrolled patients who may not meet guideline HFpEF definitions. (sources/tirzepatide-hfpef-summit-nejm-2025)
Intentional weight loss refutes the obesity paradox in HFpEF. Observational data implied weight loss worsened HF prognosis and discouraged it; STEP-HFpEF shows intentional pharmacologic weight loss producing large symptomatic and event benefits. The reconciliation is that observational studies captured unintentional cardiac cachexia — a different phenomenon. (sources/semaglutide-stephfpef-nejm-2023)
Beta-blocker advice is stated more strongly in the literature than in the guideline. Reviews argue for avoidance in HFpEF generally, conflating three different claims: absence of benefit, demonstrated harm in COPD, and harm from chronotropic impairment. ESC 2026 permits beta-blockers for comorbidity indications and expresses only uncertainty. No trial has tested withdrawal, so neither position is evidenced. (sources/HF-ESC-2026, rating: very high; sources/hf-copd-hfreview-2025, rating: high)
PADN in HFpEF-associated PH: a large unblinded effect, and the reason to wait. It halved clinical worsening where every drug has failed, with the larger effect in the LVEF >40% and IpcPH subgroups — where a pulmonary-vascular mechanism is least plausible, inviting unblinded endpoint assessment as the alternative explanation. The sham-controlled trial decides it. (sources/padn-hf-ph-nejm-2026, rating: high)

Connections

Part of entities/HFpEF — the phenotype: definition, diagnostic scores and provocation
Part of entities/Heart-Failure — the parent overview
The other phenotype's pathway: concepts/HFrEF-Management — two shared pillars, then divergence; beta-blocker advice inverted
Decompensation and pulmonary oedema: concepts/Decompensated-HF-Management
Drug pages: entities/Sacubitril-Valsartan · entities/Finerenone · entities/Semaglutide · entities/Tirzepatide
Drug classes: concepts/GLP-1-Receptor-Agonists · concepts/NuSH-Therapies · concepts/SGLT2-Inhibitors-in-CKD
Aetiology-specific: entities/ATTR-Amyloidosis · concepts/Fabry-Cardiomyopathy · concepts/Pulmonary-Hypertension · concepts/Pulmonary-Artery-Denervation · entities/Atrial-Fibrillation
Comorbidity: concepts/HF-COPD-Comorbidity · entities/COPD · entities/Obesity · concepts/Visceral-Adiposity · concepts/Obesity-Paradox
Congestion and kidney: concepts/Diuretic-Resistance · concepts/Cardiorenal-Syndrome · concepts/Iron-Deficiency-in-HF
Non-pharmacological: concepts/Cardiac-Rehabilitation-HF · concepts/Sleep-Disordered-Breathing · concepts/Cardiopulmonary-Exercise-Testing