Heart failure is one phrase covering two diseases, and this drug class treats them in opposite directions. In the version where the heart squeezes normally but cannot fill — preserved ejection fraction — a GLP-1 produced the largest symptom improvements ever recorded there. In the version where the muscle is weakened, two randomized trials found nothing and a safety signal. The atherosclerotic outcome trials that produced the familiar hazard ratios are a separate body of work, set out in the cardiovascular article, and they do not settle either half of this one.
Why preserved ejection fraction is the obesity phenotype
Heart failure with preserved ejection fraction, HFpEF, is a stiff, congested heart with a normal pumping fraction. A large share of it is driven by excess weight directly: fat around the heart restricts filling, plasma volume rises, and low-grade inflammation runs alongside. That is why a drug that removes 10% of body weight is not treating a symptom of the disease here, it is acting on part of its cause. Heart failure with reduced ejection fraction, HFrEF, is damaged contractile muscle, and weight is not what broke it.
The symptom score, and what a point of it means
These trials measure symptoms with the Kansas City Cardiomyopathy Questionnaire clinical summary score, which runs from 0 to 100 with higher scores meaning fewer symptoms and fewer physical limitations.
STEP-HFpEF randomized 529 patients with HFpEF and a body-mass index of 30 or higher to once-weekly semaglutide 2.4 mg or placebo for 52 weeks. The score rose 16.6 points on semaglutide and 8.7 on placebo, an estimated difference of 7.8 points (95% CI, 4.8 to 10.9; P < 0.001). Weight fell 13.3% against 2.6%, a difference of 10.7 percentage points (95% CI, −11.9 to −9.4). The hierarchical composite of death, heart failure events and changes in score and walk distance gave a win ratio of 1.72 (95% CI, 1.37 to 2.15).[1]
Two things are worth reading in that pair of numbers. The placebo group improved 8.7 points on its own, which is what a trial of a symptom score in a congested patient tends to do. And the treatment difference still cleared it by 7.8 points, with a lower confidence bound of 4.8 — a separation few therapies in this syndrome have produced at all.
Adding diabetes damped the effect
STEP-HFpEF DM ran the same design in 616 participants who had HFpEF, a body-mass index of 30 or more, and type 2 diabetes. The score rose 13.7 points against 6.4 on placebo (P < 0.001), and weight fell 9.8% against 3.4%, a difference of 6.4 percentage points (95% CI, −7.6 to −5.2).[2]
Both numbers are smaller than in the non-diabetes trial, on both the symptom score and the weight. The walk distance moved less too: an estimated between-group difference of 14.3 m (95% CI, 3.7 to 24.9; P = 0.008), against 21.5 m versus 1.2 m in the first trial.[1][2] Inflammation still fell, with a treatment ratio for C-reactive protein of 0.67 (95% CI, 0.55 to 0.80), and the win ratio was 1.58 (95% CI, 1.29 to 1.94).[2] Which indication a person holds changes the size of the answer, a split covered for weight generally in the two-approvals article.
Exercise capacity tracked weight, not time
A prespecified secondary analysis pooled both trials, 1,145 randomized patients, and followed the six-minute walk. Semaglutide increased walk distance by 14.6 m against placebo at 20 weeks (95% CI, 8.6 to 20.7; P < 0.0001), and 17.1 m at 52 weeks (95% CI, 9.2 to 25.0; P < 0.0001), with no significant interaction across any relevant subgroup.[3]
The dose-response is the interesting part. Each one-unit fall in body-mass index on treatment was associated with a 4.1 m increase in walk distance (95% CI, 2.4 to 5.7; P < 0.0001).[3] Patients who walked least at baseline were older, more inflamed and more congested, and they gained as much as everyone else. That is a mechanical result rather than a metabolic one: less body to carry, more distance covered. Whether that arithmetic holds in people whose exertion is not limited by weight is a separate question, and the pooled body-composition evidence behind it sits in the muscle article.
SUMMIT reduced events and did not reduce deaths
SUMMIT randomized 731 patients with heart failure, an ejection fraction of at least 50% and a body-mass index of at least 30 to tirzepatide at up to 15 mg weekly or placebo, with a median follow-up of 104 weeks. Adjudicated cardiovascular death or a worsening heart-failure event occurred in 36 patients (9.9%) on tirzepatide against 56 (15.3%), a hazard ratio of 0.62 (95% CI, 0.41 to 0.95; P = 0.026). The symptom score rose 19.5 points against 12.7, a difference of 6.9 points (95% CI, 3.3 to 10.6).[4]
The composite is not a mortality result, and the trial says so in its own numbers. Death from cardiovascular causes occurred in 8 patients (2.2%) on tirzepatide and 5 (1.4%) on placebo, a hazard ratio of 1.58 (95% CI, 0.52 to 4.83).[4] Thirteen deaths cannot establish anything in either direction, which is exactly the point: the benefit lived in worsening-heart-failure events, and the death column pointed the other way with an interval far too wide to read. Adverse events led to discontinuation in 6.3% on tirzepatide against 1.4% on placebo,[4] the same tolerability pattern described in the side-effect article. Nothing here justifies quoting a tirzepatide plan on the tirzepatide boards as a heart drug.
In reduced ejection fraction the trials failed
FIGHT randomized 300 patients recently hospitalized with heart failure and reduced ejection fraction — median ejection fraction 25% — to daily liraglutide advanced to 1.8 mg or placebo for 180 days. The primary endpoint was a global rank across death, heart-failure rehospitalization and change in NT-proBNP.[5]
Liraglutide had no significant effect: a mean rank of 146 against 156 for placebo (P = .31). Deaths numbered 19 (12%) against 16 (11%), a hazard ratio of 1.10 (95% CI, 0.57 to 2.14). Rehospitalizations for heart failure numbered 63 (41%) against 50 (34%), a hazard ratio of 1.30 (95% CI, 0.89 to 1.88).[5] Neither reaches significance. Both point the wrong way.
LIVE randomized 241 stable patients with an ejection fraction of 45% or less, on optimal heart failure treatment, to liraglutide 1.8 mg daily or placebo for 24 weeks. Ejection fraction did not change: a mean difference of −0.8% (95% CI, −2.1 to 0.5; P = 0.24). Heart rate rose by 7 beats per minute (95% CI, 5 to 9; P < 0.0001). Serious cardiac events occurred in 12 patients (10%) on liraglutide against 3 (3%) on placebo, P = 0.04.[6] Its authors wrote that this raises concern about using liraglutide in chronic heart failure with reduced left ventricular function.
Why the two syndromes behave differently
The heart-rate finding is the plausible hinge. These drugs raise resting heart rate by several beats per minute in nearly every population studied. In a stiff, over-filled heart that tolerates it, and the decongestion that comes with losing a tenth of body weight dominates. In a dilated, weakened heart running on a fixed stroke volume, a persistently faster rate is a cost rather than a side note.
Trial size is the honest caveat on the negative side. FIGHT and LIVE together randomized 541 patients over six months and 24 weeks respectively, which is small enough that a modest benefit could have been missed. What they do establish is that no positive result exists in this population, and that the one safety difference that reached significance went against the drug.[5][6]
What none of this licenses
Every participant above had a diagnosis made by a cardiologist, with an ejection fraction measured on an echocardiogram and a symptom score administered in a clinic. An intake form cannot produce any of those, and someone who has been told their heart is weak is in the group where the randomized evidence is negative. That distinction is not something to resolve from a website, particularly at the ages where heart failure is common, as the article on older adults sets out.
The trials also used branded product at labeled doses under supervision. Most sellers in this market dispense compounded semaglutide or tirzepatide, and compounded drugs are not FDA-approved and are not reviewed by the FDA for safety, efficacy or quality before they are dispensed — the distinction drawn throughout the compounding article. A win ratio earned in a 529-patient cardiology trial does not travel to a vial by sharing a molecule name, and the way a figure here is established before publication is described in the methodology.