A claims-based cohort study of 52,971 adults with type 2 diabetes and established atherosclerotic cardiovascular disease found that initiating tirzepatide was associated with a 32% lower hazard of major adverse cardiovascular events than initiating sitagliptin, driven mainly by reductions in…
Adding tirzepatide to standard diabetes care was associated with a 32% lower hazard of major adverse cardiovascular events MACE than adding sitagliptin in adults with type 2 diabetes and established atherosclerotic cardiovascular disease, according to a population-based cohort study published in The BMJ in 2026. The report, "Tirzepatide and the risk of atherosclerotic cardiovascular events," appears in volume 394 as article e100011, with the DOI 10.1136/bmj-2026-100011. N. Krger is the first author; S. Schneeweiss and S.V. Wang are co-authors.
The study drew on two US administrative claims databases covering May 2022 to May 2025 and was built as an emulation of SURPASS-CVOT , the cardiovascular outcomes trial that showed tirzepatide was non-inferior to dulaglutide for MACE. Head-to-head trials answer whether one drug beats another; they do not quantify how much benefit a drug adds on top of contemporary standard care. By comparing tirzepatide with sitagliptin, an oral dipeptidyl peptidase-4 inhibitor with neutral cardiovascular effects that serves as a validated placebo proxy, the analysis aimed to estimate exactly that incremental benefit. Earlier trial analyses had linked tirzepatide to reduced heart failure hospitalization and all-cause mortality; this study extends the question to routine clinical care, and it is the claims-based estimate that supplies the incremental number.
The result in brief: a one-year weighted MACE risk of 2.9% with tirzepatide versus 4.4% with sitagliptin, an absolute risk difference of 1.4 percentage points, and a hazard ratio of 0.68. The association was driven mainly by lower hazards of myocardial infarction HR 0.67 and all-cause mortality HR 0.55 . Stroke moved little, with a hazard ratio of 0.91.
The study cohort included 52,971 individuals: 35,353 patients who initiated tirzepatide and 17,618 who initiated sitagliptin. After propensity score overlap weighting , the effective sample size was 7,442 patients per group. The weighted population was 50.8% women, with an average age of approximately 70 years. Background therapy was contemporary and balanced: 85% of patients in each group had documented statin use, 18% used insulin, and 21% used an SGLT2 inhibitor.
The MACE difference was not spread evenly across components. Myocardial infarction and all-cause mortality carried the signal; stroke did not. A combined outcome of myocardial infarction or stroke excluding death also favored tirzepatide, and an expanded composite adding unstable angina and coronary revascularization modestly favored tirzepatide as well.
Two further findings broaden the clinical picture. The tirzepatide group had a markedly lower risk of infections requiring hospitalization and lower infection-related mortality, and infection-related deaths contributed to the all-cause mortality difference. Gastrointestinal adverse events did not differ between groups. Negative control outcomes , abdominal hernia and lumbar radiculopathy, showed no association, a check that reduces concern that the results reflect unmeasured differences in general health rather than the treatment itself. Results were broadly consistent across sensitivity analyses and patient subgroups.
The gastrointestinal null result deserves scrutiny. Tirzepatide's tolerability profile, dominated by nausea, vomiting, and diarrhea during dose escalation, would lead most prescribers to expect more gastrointestinal harm than sitagliptin produces. Claims data, however, capture only events that generate a diagnosis code; routine symptoms managed at home or handled by dose adjustment are largely invisible. The null result therefore does not establish equal tolerability, only the absence of a coded difference in this population.
Eligibility mirrored SURPASS-CVOT: patients aged 40 years or older with type 2 diabetes, atherosclerotic cardiovascular disease prior ischemic stroke, myocardial infarction, or peripheral, coronary, or carotid artery disease , and a body mass index of at least 25 kg/m2. Exclusion criteria covered pregnancy, active liver disease, advanced heart failure, end-stage kidney disease, conditions predisposing to incretin-related adverse effects, and recent use of a GLP-1 receptor agonist. Follow-up ran up to one year per patient, ending earlier at an outcome, health-plan disenrollment, or treatment discontinuation or switch; the median on-treatment follow-up was less than six months.
Sitagliptin was chosen as the comparator for a specific reason. In dedicated cardiovascular outcome trials, DPP-4 inhibitors have repeatedly shown a neutral effect on MACE, neither increasing nor decreasing risk relative to placebo. That neutrality is what makes sitagliptin a validated placebo proxy: the difference between the two groups can be read as the incremental effect of tirzepatide on top of standard care, rather than as harm from the comparator. Sitagliptin is also the clinically relevant alternative in this setting, the incretin-based drug a prescriber might choose when concerned about cost, injection burden, or the gastrointestinal effects of an injectable agonist.
The choice of all-cause mortality as the third component of the MACE composite is a deliberate design decision. Cardiovascular mortality is the conventional component of MACE, but death certificates in claims data are unreliable, and cardiovascular death can be misclassified as non-cardiovascular and vice versa. All-cause mortality avoids that misclassification entirely, at the cost of including deaths that no vascular drug could plausibly prevent. The trade is acceptable in a hypothesis-generating claims study, but it means the headline 0.68 hazard ratio is not purely a vascular effect.
The analysis used propensity score overlap weighting, weighted Kaplan-Meier survival methods, and Cox proportional hazards models. The endpoints were prespecified and are worth enumerating:
Overlap weighting explains the arithmetic of the cohort. The 52,971 patients were reduced to an effective sample of 7,442 per group because the method downweights patients whose propensity scores place them far from equipoise: the tirzepatide patients who would never plausibly receive sitagliptin, and the sitagliptin patients who would never plausibly receive tirzepatide. The 14,884 effective patients are the ones in whom the comparison is most credible. The price of that comparability is precision. The effective sample is roughly a quarter of the raw cohort, and the short follow-up narrows the number of events available for the component analyses.
Trial emulation has advantages that ordinary observational analyses lack. Explicit eligibility criteria, a defined follow-up window, and a precommitted endpoint list reduce researcher degrees of freedom, and overlap weighting focuses the comparison where confounding is smallest. What the design cannot do is eliminate confounding entirely. Claims data do not capture body mass index trajectories, frailty, or physician judgment, and treatment was not randomly assigned. The study shows that patients who began tirzepatide in routine US care had a lower hazard of MACE than comparable patients who began sitagliptin. That is an association, though a carefully constructed one.
Tirzepatide is a single peptide that activates two receptors: glucagon-like peptide-1 GLP-1 and glucose-dependent insulinotropic polypeptide GIP . GLP-1 receptor activation is the established mechanism behind the cardiovascular benefit seen across this drug class. It lowers glucose in a glucose-dependent manner, reduces body weight, lowers blood pressure, and improves lipid profiles. It also acts on the vessel wall directly: GLP-1 receptors are expressed on endothelial cells, vascular smooth muscle cells, and cardiomyocytes, and activation reduces vascular inflammation, improves endothelial function, and may stabilize atherosclerotic plaque.
The biology of the GIP half of the molecule has been the subject of a reversal in thinking. Early obesity research treated GIP primarily as a fat-storage hormone: GIP receptors are abundant in adipose tissue, GIP signaling promotes postprandial triglyceride uptake, and GIP receptor antagonism was once tested as a weight-loss strategy. Tirzepatide, a dual agonist that nonetheless produces greater weight loss than pure GLP-1 receptor agonists, overturned that expectation in human trials, and the current interpretation is that the two incretin signals act synergistically rather than additively. GIP agonism is now thought to improve adipose tissue insulin sensitivity, modulate lipid partitioning, and contribute to glucose-dependent insulin secretion. Combined GIP and GLP-1 activation produces greater weight loss and glycemic improvement than GLP-1 receptor agonism alone, and a strong myocardial infarction signal is consistent with a drug that reduces atherothrombotic risk through multiple metabolic and vascular pathways.
The stroke result is the component that does not fit neatly. Myocardial infarction and ischemic stroke share the same fundamental pathology, atherosclerosis and thrombosis, and a drug that reduced one would be expected to reduce the other. The 0.91 hazard ratio for stroke may reflect a true differential effect; stroke in this older population is frequently cardioembolic, arising from atrial fibrillation rather than unstable plaque, and incretin-based therapies have no established effect on cardiac embolism. Alternatively, the study may simply have recorded too few stroke events in a one-year window to detect a modest effect. The study cannot distinguish those explanations; the confidence interval around the stroke estimate would be the first piece of evidence needed to do so.
The mortality component requires particular care. The composite included all-cause death, which captures non-cardiovascular deaths as well. The exploratory finding that infection-related deaths contributed to the mortality difference is biologically coherent: GLP-1 receptors are expressed on immune cells, and incretin signaling modulates inflammatory responses. But the authors flagged infection-related mortality as a post hoc analysis, and claims data cannot reliably separate true infection prevention from the characteristics of patients prescribed each drug. The mechanism is plausible; the evidence is not yet confirmatory.
The Peptide Atlas registry for tirzepatide lists 251 registered clinical trials. The phase breakdown on file covers 5 Phase 2 trials, 2 Phase 4 trials, and 1 Phase 3 trial, with 10 trials currently recruiting. The distribution, with Phase 4 studies nearly as numerous as Phase 2 studies, captures a molecule already past its pivotal development program whose sponsors and academic collaborators are testing new populations and new combinations. The named trials show how far the molecule has spread beyond its original indication. NCT06180616 is a Phase 2 trial of tirzepatide for type 1 diabetes with overweight or obesity. NCT07468552 is a Phase 2 trial for cannabis use disorder. NCT07027969 is a Phase 4 trial linking metabolic surgery to atrial fibrillation research, and NCT07630454 is a Phase 4 trial testing…
Peptides referenced: Tirzepatide, Dulaglutide, Glucagon, GLP-1.
Related reading: Gedeon Richter and Adalvo Expand Partnership to Tirzepatide Therapies, Meta-Analyses Question Long-Term Value of GLP-1 Weight-Loss Drugs, Lilly Lifts 2026 Guidance as Tirzepatide Sales Surge, Peptide Rivals Split, MHRA Yellow Card data link GLP-1 drugs to 153 fatality reports.