SURPASS-CVOT Explained: What the Big Tirzepatide Heart Trial Actually Showed

SURPASS-CVOT is the largest and longest tirzepatide trial that has reported: 13,299 participants, 640 sites, 30 countries, a median of four years of follow-up. Lilly announced topline results on 31 July 2025, presented them at the European Association for the Study of Diabetes meeting in Vienna on 18 September 2025, and published them in the New England Journal of Medicine that December [1][2][3].

It is also the most misreported tirzepatide result, because the headline number people repeat — a 16% lower death rate — is not the trial’s primary finding and comes with a real statistical caveat.

This article explains what was tested, what was found, and what it does and does not mean. It is not medical advice.

What question was SURPASS-CVOT built to answer?

Regulators require large cardiovascular outcome trials for diabetes drugs. The question is usually: does this drug increase heart risk, and ideally, does it reduce it?

SURPASS-CVOT enrolled adults aged 40 or older who had type 2 diabetes and established atherosclerotic cardiovascular disease — meaning they had already had a heart attack, stroke, or documented artery disease. They needed an HbA1c between 7.0% and 10.5% and a BMI of at least 25 [3].

The population was sicker and older than a typical obesity trial: average age 64.1 years, 29.0% women, average BMI 32.6, average HbA1c 8.4%, and an average of 14.7 years living with diabetes.

Participants were randomized 1:1 to either tirzepatide at their maximum tolerated dose (5 mg, 10 mg or 15 mg) or dulaglutide 1.5 mg, sold as Trulicity. Both were weekly injections, and the trial was double-blind with sham dose escalation so nobody could tell which drug they were on from the titration schedule [2].

The primary endpoint was MACE-3: the time to a first occurrence of cardiovascular death, heart attack or stroke.

Why compare it with Trulicity instead of a placebo?

This is the single most important design decision in the trial, and it shapes how every result should be read.

Dulaglutide had already been shown in the REWIND trial to reduce cardiovascular events compared with placebo. Once a treatment with proven benefit exists for a population at high risk, running a placebo arm in that population raises ethical problems — you would be knowingly leaving some participants without a treatment known to help.

So Lilly used what is called an active comparator. The trade-off is precision of claim. A placebo trial can say “this drug reduces heart attacks by X% compared with nothing.” An active-comparator trial can only say “this drug is at least as good as, or better than, that other drug.”

Stephen Nicholls of Monash University, who presented the results, put it this way at EASD: the trial “sought to use an active comparator and not a placebo, and by doing that we have clearly established the cardioprotective nature of tirzepatide” [2]. Independent commentators were more cautious, pointing out that dulaglutide’s own benefit in REWIND was modest and that the “preservation of effect” framework has not previously been used to grant a cardiovascular indication [2].

What was the primary result?

A primary endpoint event occurred in 801 of 6,586 participants (12.2%) on tirzepatide and 862 of 6,579 (13.1%) on dulaglutide.

• Hazard ratio 0.92, 95.3% confidence interval 0.83 to 1.01
• Non-inferiority: met, p=0.003
• Superiority: not met, p=0.09 (reported as p=0.086 in the EASD presentation)

Two technical notes. The confidence interval is 95.3%, not 95% — a consequence of how the trial spent its statistical error budget across interim analyses. And non-inferiority was defined in advance as the upper limit of that interval falling below 1.05. It came in at 1.01, so it passed comfortably. Superiority would have required the upper limit to drop below 1.00, and it missed by 0.01 [1][3].

What did the secondary endpoints show?

This is where the widely quoted numbers come from. All of them were prespecified, and all sit below the primary endpoint in the statistical hierarchy [2][4]:

Endpoint	Tirzepatide	Dulaglutide	Hazard ratio (95% CI)
All-cause death	8.6%	10.2%	0.84 (0.75 to 0.94)
CV death, MI, stroke or revascularization	16.5%	18.5%	0.88 (0.81 to 0.96)
Cardiovascular death	5.6%	6.2%	0.89 (0.77 to 1.02)
Myocardial infarction	4.7%	5.4%	0.86 (0.74 to 1.00)
Stroke	3.5%	3.8%	0.91 (0.76 to 1.09)
CV death or heart failure hospitalization/urgent visit	7.8%	8.5%	0.91 (0.81 to 1.03)

The all-cause mortality result is the one that traveled. Lilly’s own release led with it: a 16% lower rate of death from any cause [1].

Here is the caveat that belongs with it every time. That endpoint was not controlled for multiplicity-adjusted type 1 error. Lilly’s release says so explicitly. When a trial tests many endpoints, some will look significant by chance unless the statistics are adjusted for that. The primary endpoint carried the protection; the secondary ones did not.

An independent cardiologist commenting on the EASD presentation noted the additional oddity: a clear mortality signal without a matching reduction in cardiovascular death or in MACE itself is hard to explain mechanically, and could reflect non-cardiovascular causes, or chance [2].

In absolute terms, the difference was 1.6 percentage points over four years. That is not nothing — but it is a smaller-sounding number than “16% lower.”

What is the imputed-placebo analysis, and should you trust it?

Because the trial had no placebo arm, researchers ran a prespecified indirect analysis to estimate what tirzepatide would have looked like against nothing.

The method: take the hazard ratio for tirzepatide versus dulaglutide from SURPASS-CVOT, multiply it by the hazard ratio for dulaglutide versus placebo from REWIND, and adjust for differences between the two trial populations using propensity scoring. They included all 13,165 SURPASS-CVOT participants and 2,055 of the 9,901 REWIND participants who would have qualified for SURPASS-CVOT [5].

The estimates versus a “putative placebo”:

• MACE-3: hazard ratio 0.72 (95% CI 0.55 to 0.94)
• Cardiovascular death or heart failure events: hazard ratio 0.70 (95% CI 0.51 to 0.96)
• All-cause death: hazard ratio 0.61 (95% CI 0.45 to 0.82)

Sensitivity analyses using unadjusted data, the full REWIND cohort, or a broader GLP-1 meta-analysis pointed the same way.

How much weight to put on this depends on how comfortable you are with chained cross-trial comparisons. REWIND ran years earlier, in a different population, against a background of different standard care. The analysis is clever and it was prespecified, but it is a statistical construction, not a randomized result. The confidence intervals are wide for a reason.

What about the kidneys?

A prespecified exploratory analysis published in The Lancet Diabetes & Endocrinology in 2026 looked at major kidney events across the whole SURPASS-CVOT population, which included 2,948 people with high-risk chronic kidney disease at baseline [6].

The composite kidney outcome — persistent macroalbuminuria, a persistent 50% or greater drop in kidney filtration, end-stage kidney disease, or kidney death — occurred in:

• 396 of 6,586 (6.0%) on tirzepatide
• 498 of 6,579 (7.6%) on dulaglutide
• Hazard ratio 0.77 (95% CI 0.68 to 0.88), p=0.0002

The benefit held in both risk strata: hazard ratio 0.70 (95% CI 0.58 to 0.84) in low-to-moderate-risk chronic kidney disease and 0.79 (95% CI 0.64 to 0.96) in high-risk. It was driven mainly by fewer new cases of persistent macroalbuminuria — protein leaking into the urine — rather than by fewer people progressing to dialysis.

Lilly’s topline release also reported that in participants at high or very high kidney risk, tirzepatide slowed the decline in estimated filtration rate by 3.5 mL/min/1.73 m2 at three years versus dulaglutide (95.0% CI 2.57 to 4.50) [1].

Worth noting: the authors describe this as exploratory. And a separate analysis from the SUMMIT heart failure trial cautions that measuring kidney function during rapid weight loss is genuinely difficult, because the two standard blood-based formulas — one using creatinine, one using cystatin C — are both distorted by changes in muscle and fat, and often disagree for the same patient [7].

Does this apply to Zepbound users without diabetes?

No, and this is the most common misreading.

Every participant in SURPASS-CVOT had type 2 diabetes and established atherosclerotic cardiovascular disease. The result cannot be transferred to a person taking Zepbound for weight management who does not have diabetes.

The trial designed for that population is SURMOUNT-MMO. It has 15,374 participants with obesity and no diabetes, with or at risk of cardiovascular disease. Its primary endpoint is a five-component composite: non-fatal heart attack, non-fatal stroke, coronary revascularization, heart failure events, or death from any cause. As of 14 September 2026 it is active and not recruiting, and its registry record lists an estimated primary completion date of October 2027 [8].

Secondary coverage claiming a 2026 readout is wrong.

Until SURMOUNT-MMO reports, semaglutide is the only drug in this class with randomized evidence of reduced heart events in people who have obesity but not diabetes, from the SELECT trial.

What about heart failure?

Tirzepatide does have a heart failure result, from a different trial.

SUMMIT enrolled 731 adults with heart failure with preserved ejection fraction and a BMI of 30 or higher, treated for a median of 104 weeks. Cardiovascular death or worsening heart failure occurred in 9.9% on tirzepatide versus 15.3% on placebo, a hazard ratio of 0.62 (95% CI 0.41 to 0.95, p=0.026). Symptom scores on the Kansas City Cardiomyopathy Questionnaire improved by 19.5 points versus 12.7 [9].

One honest wrinkle: cardiovascular deaths in SUMMIT were numerically higher in the tirzepatide arm (2.2% versus 1.4%), with wide confidence intervals in a trial not powered to detect a difference in that outcome.

SUMMIT was a placebo-controlled trial, so it does support a direct claim. But it was in a narrow population — obesity-related HFpEF — and there is no FDA heart failure indication for tirzepatide.

What do real-world studies say about tirzepatide versus semaglutide for heart events?

They disagree, which is itself the finding.

One analysis of Komodo US claims data, called STEER, matched 10,625 patients in each arm — adults aged 45 and older with overweight or obesity and established cardiovascular disease but no diabetes. It found semaglutide associated with a 29% lower risk of a 3-point MACE composite (hazard ratio 0.71, p=0.046) and 22% lower for a 5-point composite (hazard ratio 0.78, p=0.040) [10].

A different analysis using TriNetX data, in 47,804 propensity-matched pairs of older adults with type 2 diabetes and obesity, found tirzepatide associated with lower MACE (3.7% versus 4.1%) and lower all-cause mortality (0.2% versus 0.4%) [11].

Different populations, different databases, opposite directions. Neither is randomized. No randomized head-to-head cardiovascular trial of tirzepatide and semaglutide exists, and none is registered.

The bottom line

SURPASS-CVOT did what it set out to do: it showed tirzepatide is not worse than a drug with proven cardiovascular benefit, in people with type 2 diabetes and established heart disease. It also showed favorable movement on all-cause death, kidney function, HbA1c and weight — findings that are real but statistically unprotected and, in the kidney case, labeled exploratory.

What it did not do is show superiority over dulaglutide, or say anything about people with obesity and no diabetes. That question is SURMOUNT-MMO’s, and it is not expected to answer before late 2027.

Decisions about heart risk and which medication fits belong with a healthcare provider who knows your history.

Sources

1. Lilly topline release on SURPASS-CVOT, 2025-07-31 — https://investor.lilly.com/news-releases/news-release-details/lillys-mounjaro-tirzepatide-gipglp-1-dual-agonist-demonstrated
2. EASD 2025 coverage of SURPASS-CVOT, Healio, 2025-09-19 — https://www.healio.com/news/endocrinology/20250919/tirzepatide-similar-to-dulaglutide-for-cv-outcomes-but-tied-to-lower-rates-of-death
3. SURPASS-CVOT primary publication, New England Journal of Medicine 2025 — https://pubmed.ncbi.nlm.nih.gov/41406444/
4. EASD 2025 CVOT Symposium presentation deck, Eli Lilly — https://assets.ctfassets.net/cuupgdmwy210/6zk2qBHBfZ9nxOtxk0vM73/3074992f852cd866d3b55943fd340e24/EASD25_CVOT_Symposium_Lilly_Deck_for_MDD_WITH_NOTES_FINAL_18SEP2025.pdf
5. Imputed-placebo analysis of SURPASS-CVOT using REWIND, Diabetes Care 2026 — https://pubmed.ncbi.nlm.nih.gov/41940793/
6. SURPASS-CVOT kidney analysis, The Lancet Diabetes & Endocrinology 2026 — https://pubmed.ncbi.nlm.nih.gov/42114520/
7. SUMMIT chronic kidney disease analysis, Journal of the American College of Cardiology 2025 — https://pubmed.ncbi.nlm.nih.gov/40162940/
8. SURMOUNT-MMO registry record, ClinicalTrials.gov — https://clinicaltrials.gov/study/NCT05556512
9. SUMMIT, New England Journal of Medicine 2025 — https://pubmed.ncbi.nlm.nih.gov/39555826/
10. STEER real-world cardiovascular comparison, Diabetes Obesity and Metabolism 2026 — https://pubmed.ncbi.nlm.nih.gov/41491349/
11. TriNetX tirzepatide versus semaglutide cohort, Diabetes & Vascular Disease Research 2026 — https://pubmed.ncbi.nlm.nih.gov/42376874/