SUMMIT is the trial in this class that was actually built to test whether people with heart failure and a preserved ejection fraction end up in hospital less often. That distinguishes it from the semaglutide trials in the STEP-HFpEF article, whose two primary endpoints were a questionnaire and a scale and which counted heart-failure events as exploratory outcomes. Here the event composite was a primary endpoint, it was adjudicated, and it moved. What the two heart-failure syndromes are and why this class touches one of them is set out in the heart failure article; this page is SUMMIT’s own machinery, and the places where it does not resemble its neighbor at all.
Two primary endpoints, and only one of them is a questionnaire
731 patients were randomly assigned 1:1, double-blind, to tirzepatide titrated up to 15 mg subcutaneously once weekly or to placebo, on top of background therapy, for at least 52 weeks. Entry required heart failure in New York Heart Association class II to IV, an ejection fraction of at least 50%, and a body-mass index of at least 30. The two preregistered primary endpoints were a time-to-first-event composite of adjudicated cardiovascular death or a worsening heart-failure event, and the change from baseline to week 52 in the Kansas City Cardiomyopathy Questionnaire clinical summary score.[1][2]
Median follow-up was 104 weeks (first quartile 66, third quartile 126), twice the 52 weeks its neighbor ran.[3] That is the other structural difference: an event endpoint needs time to accumulate events, and the trial was designed to give it some.
Who could get in
The eligibility funnel was narrow and physiological. Beyond the ejection fraction and body-mass index floors, entrants needed an NT-proBNP above 200 pg/mL, or above 600 pg/mL with atrial fibrillation, plus either left atrial enlargement or elevated left ventricular filling pressure; an estimated glomerular filtration rate below 70 mL/min/1.73 m² or a heart-failure decompensation within the previous twelve months; a six-minute walk distance between 100 and 425 meters; and a questionnaire score of 80 or below. Cardiac amyloidosis, severe valvular disease, a glycated hemoglobin of 9.5% or higher, prior pancreatitis and dialysis were all exclusions.[2]
The cohort that produced was sick. Mean age was 65.2 years, 393 of 731 (53.8%) were women, mean body-mass index was 38.2, the mean questionnaire score was 53.5 and the mean six-minute walk distance 302.8 meters — and 388 (53%) had already had a worsening heart-failure event in the preceding year.[3] Unlike its neighbor, diabetes was not excluded; it was a stratification variable for randomization.[4]
The composite, and the component that ran backwards
Cardiovascular death or a worsening heart-failure event occurred in 36 of 364 patients (9.9%) on tirzepatide and 56 of 367 (15.3%) on placebo — hazard ratio 0.62 (95% CI, 0.41 to 0.95; P = 0.026). Worsening heart-failure events alone occurred in 29 (8.0%) against 52 (14.2%), a hazard ratio of 0.54 (95% CI, 0.34 to 0.85).[1]
Now the part that rarely travels with the headline. Adjudicated cardiovascular death occurred in 8 patients on tirzepatide and 5 on placebo — hazard ratio 1.58 (95% CI, 0.52 to 4.83).[1] Death from any cause was 19 against 15, a hazard ratio of 1.245 (95% CI, 0.633 to 2.452), and the analysis counting total heart-failure events and all-cause deaths returned 61 against 82, a rate ratio of 0.72 (95% CI, 0.46 to 1.14).[2]
Those intervals are wide enough to contain almost anything, which is the honest reading: thirteen cardiovascular deaths across 731 people settles nothing in either direction. But a composite that reaches significance while its mortality component points the other way is a specific shape, and the whole benefit here is carried by hospitalizations and urgent visits avoided. That is a real thing to want. It is not the same claim as living longer, and the difference is worth holding onto.
Everything measured alongside the score
The questionnaire score rose 19.5 points on tirzepatide against 12.7 on placebo, a between-group difference of 6.9 (95% CI, 3.3 to 10.6; P < 0.001). An expanded analysis reports the rest: six-minute walk distance improved 18.3 meters (95% CI, 9.9 to 26.7), the EQ-5D-5L health state index 0.06 (95% CI, 0.03 to 0.09), a Patient Global Impression of Severity shift with a proportional odds ratio of 1.99 (95% CI, 1.44 to 2.76) and a New York Heart Association class shift of 2.26 (95% CI, 1.54 to 3.31), with fewer heart-failure medications required (P = 0.015). The hierarchical composite returned a win ratio of 1.63 (95% CI, 1.17 to 2.28).[3] Weight fell 13.85% against 2.24%, and high-sensitivity C-reactive protein 38.8% against 5.9%.[2]
The stratum that lost less weight and did just as well
Diabetes was a randomization stratum, so the comparison inside it is a randomized one. Patients with type 2 diabetes lost 10.4% of body weight (95% CI, 8.7 to 12.2) against 12.9% (95% CI, 11.2 to 14.6) in those without (P for interaction = 0.04) — and got the same event reduction: hazard ratio 0.64 (95% CI, 0.35 to 1.15) with diabetes and 0.61 (95% CI, 0.33 to 1.10) without (P for interaction = 0.95). Questionnaire score, walk distance, quality of life and functional class did not differ between the strata either, and nor did the reduction in paracardiac fat or left ventricular mass.[4]
Two and a half percentage points of weight bought nothing measurable on the endpoint that counts hospitalizations. The trial’s own authors put it as the possibility that heart-failure benefit in this class may not be faithfully estimated by the size of the weight change.[4] A related dissociation appears in the semaglutide trials, but it appears there as a contrast between two trials on a symptom score. Here it sits inside one trial, across a stratum the randomization protects, and lands on a clinical event.
The picture is not uniform, though, and the same trial supplies the counterweight. Across increasing tertiles of baseline body-mass index the walk-distance advantage grew from 9.9 to 26.3 to 37.5 meters (P = 0.025) and the weight advantage from 10.7 to 11.8 to 14.4 percentage points (P = 0.006), while the event endpoint showed no heterogeneity by body-mass index at all. Among those on tirzepatide, greater weight loss was associated with larger gains in walk distance and questionnaire score — an association within the treated arm, which randomization does not protect.[7]
What the heart and the kidneys looked like
A cardiac magnetic resonance substudy imaged 175 participants at baseline and analyzed 106 with adequate paired images at 52 weeks. Its prespecified primary endpoint, left ventricular mass, fell 11 g (95% CI, −19 to −4; P = 0.004) against placebo, and paracardiac adipose tissue fell 45 mL (95% CI, −69 to −22; P < 0.001). The change in mass tracked the change in weight (P < 0.02).[5] Fifty patients on drug against 56 on placebo is a small denominator, and the same caution applies that applies to every substudy in the SURMOUNT-1 article.
The renal analysis is mostly a warning about instruments. Estimated glomerular filtration rate by cystatin C sat roughly 9 mL/min/1.73 m² below the creatinine-based estimate at baseline, and tirzepatide produced an apparent decline at 12 weeks by creatinine that did not appear by cystatin C. Chronic kidney disease did not blunt the treatment effect, and the absolute risk reduction was numerically larger in those who had it.[6] Why body composition distorts both markers is covered in the kidney article.
The safety column that does not match its neighbor
In STEP-HFpEF the placebo arm carried roughly twice the serious adverse events, because in this syndrome a serious event is usually the disease arriving at a hospital. That inversion does not reproduce in SUMMIT. Serious adverse events occurred in 96 of 364 on tirzepatide and 94 of 367 on placebo — even — with 19 deaths against 15.[2] Cardiac failure as a reported event did run in the drug’s favor, 11 against 20, and hypotension against it, 22 against 10.[2]
Gastrointestinal events behaved as expected: nausea 62 against 23, diarrhea 65 against 23, vomiting 38 against 7, constipation 54 against 22. Adverse events led to discontinuation of the trial drug in 23 patients (6.3%) against 5 (1.4%), while overall completion was high at 332 of 364 and 331 of 367.[1][2] Anyone carrying a tolerability expectation across from the other trial should carry the numbers, not the pattern: a longer trial in a different population on a different molecule produced a different safety column.
What it was not built to show, and who it describes
It was not powered to test mortality, and its mortality figures say so. It says nothing about ejection fractions between 40% and 50%, which its neighbor partly covers and this trial excluded. It enrolled nobody with a glycated hemoglobin of 9.5% or higher, nobody with cardiac amyloidosis, and nobody whose questionnaire score was above 80 — so the mildly symptomatic are outside it by design. And it tested one titration schedule to a single ceiling dose, not a dose-response.
Its population is also not the one most readers picture. Only 151 of 731 participants enrolled in the United States; Argentina contributed 204, Brazil 105 and Mexico 81, with 131 participants recorded as Asian, 47 as American Indian or Alaska Native and 36 as Black or African American.[2] That is a materially different base from the predominantly White North American and European cohorts of the semaglutide trials.
Every participant arrived with an echocardiogram, a natriuretic peptide level, a walk test and a cardiologist’s diagnosis, and received branded, FDA-approved tirzepatide under supervision for two years. None of those exists on a telehealth intake form. Sellers on the tirzepatide board dispense compounded preparations, which are not FDA-approved and are not reviewed by the FDA for safety, efficacy or quality before they are dispensed; the distance between the two is set out in the compounded-versus-brand article.