Comparative Cardiovascular Outcomes of Tirzepatide and Glucagon‐Like Peptide‐1 Receptor Agonists in Patients With Type 2 Diabetes and Atherosclerotic Cardiovascular Disease

Abstract

Background

To evaluate the association between tirzepatide use and 1‐year risk of major adverse cardiovascular events in patients with type 2 diabetes and atherosclerotic cardiovascular disease, compared with GLP‐1 (glucagon‐like peptide‐1) receptor agonists.

Methods

We conducted a retrospective, propensity score‐matched cohort study using the TriNetX network. A total of 16 402 patients with type 2 diabetes and atherosclerotic cardiovascular disease initiating tirzepatide or GLP‐1 receptor agonists between January 1, 2022, and March 31, 2025 were matched 1:1. The primary outcome was 1‐year major adverse cardiovascular events; secondary outcomes included all‐cause mortality, acute myocardial infarction, major adverse limb events, and tissue plasminogen activator use.

Results

Tirzepatide was associated with a lower risk of major adverse cardiovascular events (hazard ratio [HR], 0.75 [95% CI, 0.63–0.91]) and reduced risks of all‐cause mortality (HR, 0.69 [95% CI, 0.53–0.90]), major adverse limb events (HR, 0.59 [95% CI, 0.39–0.88]), and acute myocardial infarction (HR, 0.70 [95% CI, 0.53–0.93]), compared with GLP‐1 receptor agonists. Results were robust across subgroups and sensitivity analyses.

Conclusions

Among patients with type 2 diabetes and atherosclerotic cardiovascular disease, tirzepatide was associated with a significantly reduced 1‐year risk of major adverse cardiovascular events and other cardiovascular outcomes compared with GLP‐1 receptor agonists. These findings support the cardiometabolic potential of tirzepatide, warranting further prospective validation.

Nonstandard Abbreviations and Acronyms

GIP

glucose‐dependent insulinotropic polypeptide

GLP‐1 RA

glucagon‐like peptide‐1 receptor agonist

HFE

heart failure exacerbation

MACE

major adverse cardiovascular events

MALE

major adverse limb events

T2D

type 2 diabetes

Clinical Perspective.

What Is New?

• Tirzepatide was associated with a lower 1‐year risk of major adverse cardiovascular events compared with GLP‐1 (glucagon‐like peptide‐1) receptor agonists in patients with type 2 diabetes and atherosclerotic cardiovascular disease.

• The observed associations were consistent across secondary outcomes and multiple sensitivity analyses, supporting robustness of the findings.

What Are the Clinical Implications?

• These results are hypothesis generating and may inform future randomized trials evaluating dual incretin‐based therapies for cardiovascular risk reduction.

According to the International Diabetes Federation, an estimated 537 million individuals were living with type 2 diabetes (T2D) in 2021, exerting substantial economic pressure on health care systems worldwide. ¹ , ² T2D is associated with a 2‐ to 6‐fold increased risk of cardiovascular mortality compared with individuals without diabetes ³ and is recognized as a key contributor to the development of atherosclerotic cardiovascular disease (ASCVD), encompasses coronary artery disease, cerebrovascular disease, and peripheral artery disease. ⁴ Recent clinical trials have demonstrated that GLP‐1 (glucagon‐like peptide‐1) receptor agonists (RAs) improve cardiovascular outcomes in patients with both ASCVD and T2D, ⁵ , ⁶ with consistent benefits observed across various subgroups. ⁷ The benefits of GLP‐1 RAs extend beyond glycemic control. ⁸ Their effects on weight reduction and inflammation modulation may also contribute to cardiovascular risk reduction in patients with T2D and ASCVD. ⁸

Tirzepatide, a dual agonist of both GIP (glucose‐dependent insulinotropic polypeptide) and GLP‐1 receptors, has demonstrated superior metabolic efficacy, including greater weight reduction and anti‐inflammatory effects, compared with GLP‐1 RAs alone. ⁹ , ¹⁰ One study demonstrated that, in patients with T2D, treatment with tirzepatide was associated with a reduced risk of all‐cause mortality, major adverse cardiovascular events (MACE), acute kidney injury, and other adverse renal outcomes compared with GLP‐1 RAs. ¹¹ Another study revealed that tirzepatide significantly reduced the risk of major adverse limb events (MALE) in patients with peripheral artery disease and T2D. ¹² Nevertheless, earlier studies have predominantly examined patients with T2D and have overlooked the broader spectrum of ASCVD phenotypes. In addition, the ongoing SURPASS‐CVOT (Study of Tirzepatide Compared With Dulaglutide on Major Cardiovascular Events in Participants With Type 2 Diabetes) trial, which directly compares tirzepatide with dulaglutide in patients with T2D at risk of cardiovascular disease, is expected to provide pivotal randomized evidence on MACE incidence. ¹³ Although results are anticipated soon, real‐world data remain essential to complement trial findings by reflecting broader patient populations and clinical practice.

To bridge this knowledge gap, we performed a real‐world analysis using the TriNetX database to investigate the association between tirzepatide use and 1‐year MACE in patients with T2D and ASCVD, covering the full range of ASCVD phenotypes.

METHODS

Data Availability

The aggregated data sets generated and analyzed during this study were obtained from the TriNetX platform. Due to TriNetX's data‐sharing policies, we do not have access to individual‐level data; only deidentified, aggregated data were available to us for analysis. Access to the TriNetX data is restricted as it contains protected health information. Researchers wishing to access these data may apply through the TriNetX platform. Approval for access typically requires demonstration of appropriate credentials, a clear research purpose, and compliance with applicable privacy regulations. The application and approval process may require several weeks to complete, subject to the applicant's qualifications and the specifics of the research proposal. For more information on data access, please visit the TriNetX website: https://trinetx.com, or contact TriNetX directly via their support email at support@trinetx.com.

Data Source

This retrospective cohort investigation used the TriNetX platform, a global federated health research network providing access to electronic health records from ∼150 million patients across 135 health care organizations. These electronic records encompass diagnoses, procedures, medication administration, laboratory findings, and genomic information. The Chi Mei Hospital institutional review board approved this study (approval number: 11402‐E02) and waived informed consent requirements as the analysis employed only de‐identified, aggregated statistical data. We conducted this research in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology guidelines. We obtained access to the Global Collaborative Network database in April 2025. Detailed information on medication dose, titration schedules, or cumulative exposure was not consistently available within the TriNetX platform; therefore, treatment exposure was defined based on initiation of the index therapy rather than dose intensity.

Study Design

This study included adult patients (≥18 years) diagnosed with T2D and ASCVD who were prescribed either tirzepatide or GLP‐1RAs for the first time between January 1, 2022, and March 31, 2025. We selected this time frame to correspond with the US Food and Drug Administration's approval of tirzepatide in 2022. T2D was identified using the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD‐10‐CM) code E11. ASCVD was defined as having diagnoses of ischemic heart diseases (ICD‐10‐CM codes I20–I25), cerebral infarction (I63), or peripheral vascular diseases (I73).

Participants were divided into 2 groups: the tirzepatide group, comprising individuals who began tirzepatide treatment after their T2D and ASCVD diagnoses, and the GLP‐1RA group, comprising individuals who began GLP‐1RA treatment after their T2D and ASCVD diagnoses. We designated the index date for each participant as the date of their first prescription for either tirzepatide or GLP‐1RAs. To ensure an incident user design, we excluded patients with any record of tirzepatide or GLP‐1RAs usage before their index date, as well as those receiving medications from both categories concurrently. To ensure an incident new‐user design, we excluded patients with any record of tirzepatide or GLP‐1 RA use at any time before their index date. Thus, all patients were required to be naïve to both drug classes before the initiation of the index therapy, preventing inclusion of individuals who may have previously received or survived prior treatment with the comparator drug. Furthermore, we excluded patients who experienced any prespecified outcomes before the index date (MACE components, including acute myocardial infarction [AMI], MALE, or death) to prevent misclassification of incident outcomes. The detailed algorithms used for defining patient characteristics, clinical diagnoses, procedures, medications, and laboratory values are provided in Table S1.

Covariates and Propensity Score Matching

Following the establishment of study cohorts, index dates, outcomes, and potential confounding factors, we developed a covariate matrix using individual patient data from the 12‐month period before the index date. We calculated propensity scores through logistic regression to determine each participant's likelihood of assignment to the second cohort based on their baseline characteristics. Employing a 1:1 matching strategy, we used a greedy nearest‐neighbor algorithm with a caliper width of 0.1 pooled SDs to pair participants from the smaller cohort with those having comparable propensity scores in the larger cohort. We considered covariate balance to be adequate when the standardized mean difference was <0.1. ¹⁴ The propensity score model included demographic characteristics (age, sex, race), comorbidities (hypertension, obesity, chronic kidney disease, cerebrovascular disease, ischemic heart disease, heart failure, atrial fibrillation, chronic liver disease, chronic lung disease, human immunodeficiency virus, neoplasms, and diabetes‐related complications), medication use (antihypertensive agents, lipid‐lowering drugs, antiplatelets/anticoagulants, and antidiabetic agents including insulin, metformin, sodium–glucose cotransporter 2 inhibitors, dipeptidyl peptidase‐4 inhibitors, sulfonylureas, and thiazolidinediones), and laboratory measurements (hemoglobin A1c [HbA1c], estimated glomerular filtration rate, low‐density lipoprotein cholesterol). Covariates were selected based on their known or plausible relationships with cardiovascular outcomes and treatment assignment. ¹² , ¹⁵ , ¹⁶ Comprehensive definitions and operational codes for all covariates are detailed in Table S2.

Outcomes and Follow‐Up

Our primary outcome was MACE. Secondary outcomes included MALE, all‐cause mortality, and AMI. MACE was defined as a composite of tissue plasminogen activator (tPA) use, AMI, or all‐cause mortality, and MALE was identified as amputation of lower limb at any level. ¹² , ¹⁵ , ¹⁷ We used tPA administration as a proxy for acute ischemic stroke, as reliance on ICD‐10 codes alone may introduce misclassification due to coding variability and residual diagnostic codes. ¹⁸ TPA use therefore offers a more specific indicator of clinically significant ischemic stroke requiring reperfusion therapy.

We included all‐cause mortality rather than cardiovascular‐specific mortality because the TriNetX database does not provide adjudicated cause of death, consistent with prior real‐world studies using administrative data. ¹⁵ This pragmatic definition was adopted to comprehensively capture fatal events while minimizing potential misclassification bias. The primary analysis followed an intention‐to‐treat–like framework, in which participants were analyzed according to the initial treatment assignment at the index date regardless of subsequent treatment changes during follow‐up. Additionally, we specified negative outcome controls including skin cancer, sensorineural hearing loss, lumbar radiculopathy, traumatic brain injury, and hernia, selected because they have no known association with incretin‐based therapies, occur with sufficient incidence in real‐world data to provide adequate power, and are unrelated to T2D or ASCVD. These outcomes were used to assess potential residual confounding and coding bias. ¹¹ , ¹⁹ We also evaluated safety outcomes that are clinically relevant for incretin‐based therapies. These included diarrhea, hypoglycemia, pancreatitis, and major adverse kidney events. Major adverse kidney events were defined as a composite of end‐stage kidney disease, urgent dialysis initiation, or dialysis dependence. Detailed definitions and coding criteria for all outcomes are provided in Table S3.

Subgroup and Sensitivity Analysis

We conducted prespecified subgroup analyses stratified by sex (male versus female), body mass index (BMI; <30 kg/m² versus ≥30 kg/m²), ASCVD subtype (ischemic heart disease, cerebral infarction, or peripheral vascular disease), baseline estimated glomerular filtration rate (<45 versus ≥45 mL/min per 1.73m²), HbA1c (<9% versus ≥9%), type of GLP‐1 RA, and the presence of heart failure, obesity, and obstructive sleep apnea. In addition, we examined CI overlap across subgroups to evaluate whether apparent differences in hazard ratios (HRs) were statistically meaningful. ²⁰

To manage time‐dependent confounding factors, we conducted landmark analyses beginning 3 months after the index date and continuing through 1 year of follow‐up. Also, we performed sensitivity analyses extending the follow‐up windows beyond the primary analysis period, including assessments from day 1 to 2 years and day 1 to 3 years after the index date. In addition, we evaluated multiple cohorts defined by varying initiation windows, which referred to different allowable intervals between the dual diagnosis of T2D and ASCVD and the initiation of either tirzepatide or GLP‐1RAs therapy. Specifically, we restricted treatment initiation to occur within 1 month, 3 months, or 6 months after the dual diagnosis, with the index date set at the time of the first prescription. This approach did not alter the exposure definition itself but rather limited the eligible population to those initiating therapy within narrower time frames, allowing us to examine whether temporal proximity between diagnosis and treatment initiation influenced the observed outcomes. ²¹ , ²²

Furthermore, we implemented a treatment‐persistence proxy (per protocol‐like) analysis. Because the TriNetX platform does not provide prescription stop dates or support censoring at treatment switching/discontinuation, we approximated persistence by restricting to participants who received at least one additional prescription of the index drug between 6 and 12 months after the index date and who had no recorded prescription of the alternative drug during follow‐up. This approach reduces contamination and approximates an on‐treatment effect. ²³ This analysis was conducted to evaluate the potential impact of treatment switching or crossover on the primary findings. We restricted the cohort to individuals with documented baseline use of SGLT2 (sodium‐glucose cotransporter 2) inhibitors (SGLT2is) during the 12 months before the index date. This analysis was performed to evaluate whether differential background SGLT2i exposure between treatment groups influenced the observed association with cardiovascular outcomes, given the established cardioprotective and renoprotective effects of SGLT2is. In an additional sensitivity analysis, ischemic stroke was alternatively defined using ICD‐10‐CM code I63 to evaluate whether outcome definitions based on thrombolytic therapy influenced the observed associations.

Additional Analysis

We examined heart failure exacerbation (HFE) as a supplementary end point, defined as a composite of pulmonary edema or escalation of decongestive therapy (injectable loop or thiazide diuretics, or metolazone). Follow‐up and censoring mirrored the primary analysis (1 year from the index date). HRs were estimated with Cox models. We stratified results by the presence of HF status in the 12 months before index. Because encounter‐level information in TriNetX does not reliably distinguish inpatient from outpatient care, HFE events identified through diagnostic codes or escalation of decongestive therapy reflect a composite of decompensations across all clinical settings. In an exploratory extension, we additionally evaluated MACE among individuals without diabetes to assess whether tirzepatide confers cardioprotective effects independent of glycemic status.

Statistical Analysis

We presented continuous variables as means±SDs and categorical variables as frequencies and percentages. To determine HRs and 95% CIs for all outcomes, we employed Cox proportional hazards regression models. We assessed the proportional hazards assumption through examination of Schoenfeld residuals. ²⁴ Between‐group survival comparisons were conducted using Kaplan–Meier curves and log‐rank tests. Additionally, we calculated the E‐value, which estimates the minimum strength of association that an unmeasured confounder would need to have with both the treatment and the outcome to fully explain the observed association, conditional on the measured covariates. ²⁵ We excluded participants with missing data or those who were lost to follow‐up to preserve analytical integrity. All statistical tests were conducted with 2‐sided significance, with P<0.05 considered statistically significant. As dose‐level information was unavailable, dose–response analyses could not be performed. We performed all statistical analyses using the TriNetX platform.

Consent to Participate

This study was conducted using deidentified data from the TriNetX research network. The requirement for informed consent was waived, as the study involved only anonymized patient records and no direct patient intervention. The study was conducted in accordance with the Declaration of Helsinki.

RESULTS

Patient Selection

This study used data from the TriNetX research network, which included 135 health care organizations and a total of 150 057 084 individuals as of April 6, 2025. We identified 64 669 548 patients who visited health care organizations between January 1, 2022, and March 31, 2025. After applying exclusion criteria—including age under 18 years, presence of any prespecified outcomes before the index date, prior use of study drugs, lack of concurrent diagnoses of ASCVD and T2D, receipt of the comparator medication, or loss to follow‐up—a total of 98 643 eligible patients who received either tirzepatide or GLP‐1RAs were included. Among these, 8204 were new users of tirzepatide and 90 439 were new users of GLP‐1RA. After 1:1 propensity score matching, 2 matched cohorts of 8201 patients each were established (Figure 1). The median (Q1–Q3) follow‐up duration was 236 (103–365) days in the tirzepatide group and 365 (347–365) days in the GLP‐1RA group.

Figure 1. Study cohort process.

ASCVD indicates atherosclerotic cardiovascular disease; GLP‐1 RA, glucagon‐like peptide‐1 receptor agonist; HCO, health care organizations; and T2D, type 2 diabetes.

Study Population Characteristics

Before matching, the tirzepatide group (n=8204) was younger (mean age 63.6 versus 64.5 years), had a higher BMI (36.3 versus 34.2 kg/m²), and were more frequently White (79.0% versus 65.2%) compared with the GLP‐1RA group (n=90 439). They also had lower rates of chronic conditions such as cerebrovascular disease, atrial fibrillation, and chronic lower respiratory disease and were less likely to be receiving insulin or antihypertensive therapies. Differences in HbA1c levels (mean 7.7% versus 8.3%) and other metabolic profiles such as low‐density lipoprotein cholesterol were also observed, with a standardized mean difference >0.2 in several variables (Table 1). After matching, 8201 patients in each group with balanced distributions across most baseline characteristics. Standardized mean differences for all covariates were <0.1. The mean age of both groups was ~63.8 years, and the proportion of women was 46.0% in the GLP‐1RA group and 45.8% in the tirzepatide group. BMI remained higher in the tirzepatide group (36.3 versus 34.9 kg/m²), with a modest standardized difference (0.179). The prevalence of comorbidities, medication use, HbA1c, estimated glomerular filtration rate, and lipid profiles were comparable between the matched groups (Table 1).

Table 1.

Baseline Characteristics of Patients Before and After Matching

Variables	Before matching			After matching
	Tirzepatide group (n=8204)	GLP‐1 RA group (n=90 439)	Standardized difference	Tirzepatide group (n=8201)	GLP‐1 RA group (n=8201)	Standardized difference
Age at index, y
Mean±SD	63.6±10.3	64.5±10.7	0.079	63.6±10.3	63.9±10.7	0.021
Sex, n (%)
Female	3762 (45.9)	37 646 (41.6)	0.085	3759 (45.8)	3771 (46.0)	0.003
Male	4435 (54.1)	52 737 (58.3)	0.086	4435 (54.1)	4420 (53.9)	0.004
Body mass index, kg/m2
Mean±SD	36.3±7.6	34.2±7.5	0.279	36.3±7.6	34.9±7.6	0.179
≥30, n (%)	4591 (56.0)	49 081 (54.3)	0.034	4589 (56)	4687 (57.2)	0.024
Race, n (%)
White	6479 (79)	58 972 (65.2)	0.311	6476 (79.0)	6478 (79.0)	0.001
Black	768 (9.4)	13 795 (15.3)	0.180	768 (9.4)	765 (9.3)	0.001
Asian	329 (4.0)	4695 (5.2)	0.056	329 (4.0)	309 (3.8)	0.013
Comorbidities, n (%)
Nicotine dependence	1063 (13)	11 395 (12.6)	0.011	1062 (13.0)	1018 (12.4)	0.016
Hypertension	6687 (81.5)	69 550 (76.9)	0.114	6684 (81.5)	6728 (82.0)	0.014
Obesity	2260 (27.5)	21 778 (24.1)	0.079	2258 (27.5)	2282 (27.8)	0.007
T2D related kidney complications	1969 (24.0)	21 878 (24.2)	0.004	1968 (24)	1962 (23.9)	0.002
T2D related ophthalmic complications	815 (9.9)	8935 (9.9)	0.002	814 (9.9)	796 (9.7)	0.007
T2D related neurological complications	2022 (24.6)	20 891 (23.1)	0.036	2019 (24.6)	2015 (24.6)	0.001
T2D related circulatory complications	1184 (14.4)	12 493 (13.8)	0.018	1183 (14.4)	1141 (13.9)	0.015
Cerebrovascular diseases	1416 (17.3)	17 762 (19.6)	0.061	1416 (17.3)	1424 (17.4)	0.003
Heart failure	2049 (25.0)	22 084 (24.4)	0.013	2047 (25.0)	2051 (25.0)	0.001
Atrial fibrillation and flutter	1453 (17.7)	14 371 (15.9)	0.049	1452 (17.7)	1495 (18.2)	0.014
Chronic kidney disease	2087 (25.4)	23 279 (25.7)	0.007	2086 (25.4)	2113 (25.8)	0.008
Chronic lower respiratory diseases	1042 (12.7)	10 736 (11.9)	0.025	1040 (12.7)	1049 (12.8)	0.003
Fibrosis and cirrhosis of liver	235 (2.9)	2239 (2.5)	0.024	235 (2.9)	223 (2.7)	0.009
Human immunodeficiency virus disease	28 (0.3)	507 (0.6)	0.033	28 (0.3)	32 (0.4)	0.008
Neoplasms	28 (0.3)	507 (0.6)	0.033	28 (0.3)	32 (0.4)	0.008
Lipid‐lowering medications, n (%)
3‐hydroxy‐3‐methylglutaryl‐coenzyme A reductase inhibitors	4821 (58.8)	56 282 (62.2)	0.071	4821 (58.8)	4796 (58.5)	0.006
Ezetimibe	735 (9.0)	6117 (6.8)	0.082	733 (8.9)	681 (8.3)	0.023
Fibrates	337 (4.1)	3972 (4.4)	0.014	337 (4.1)	335 (4.1)	0.001
Antihypertensives, n (%)
Angiotensin‐converting enzyme inhibitors	1763 (21.5)	24 650 (27.3)	0.135	1763 (21.5)	1808 (22)	0.013
Angiotensin receptor blockers	2327 (28.4)	24 748 (27.4)	0.022	2325 (28.4)	2322 (28.3)	0.001
Alpha blockers	848 (10.3)	10 637 (11.8)	0.045	848 (10.3)	848 (10.3)	0.000
Beta blockers	4127 (50.3)	47 982 (53.1)	0.055	4124 (50.3)	4075 (49.7)	0.012
Calcium channel blockers	2414 (29.4)	29 148 (32.2)	0.061	2412 (29.4)	2366 (28.9)	0.012
Diuretics	3217 (39.2)	37 332 (41.3)	0.042	3214 (39.2)	3237 (39.5)	0.006
Anticoagulants, n (%)
Aspirin	2541 (31.0)	36 030 (39.8)	0.186	2541 (31)	2480 (30.2)	0.016
Clopidogrel	1263 (15.4)	17 252 (19.1)	0.098	1263 (15.4)	1246 (15.2)	0.006
Ticagrelor	239 (2.9)	3131 (3.5)	0.031	239 (2.9)	261 (3.2)	0.016
Prasugrel	86 (1.0)	1188 (1.3)	0.025	86 (1.0)	78 (1.0)	0.010
Antidiabetic agents, n (%)
Biguanides	2707 (33.0)	36 090 (39.9)	0.144	2707 (33.0)	2721 (33.2)	0.004
Sulfonylureas	924 (11.3)	16 476 (18.2)	0.197	924 (11.3)	936 (11.4)	0.005
Thiazolidinediones	248 (3.0)	3247 (3.6)	0.032	248 (3.0)	251 (3.1)	0.002
Alpha glucosidase inhibitors	15 (0.2)	469 (0.5)	0.057	15 (0.2)	15 (0.2)	0.000
Dipeptidyl peptidase 4 inhibitor	569 (6.9)	11 523 (12.7)	0.196	569 (6.9)	522 (6.4)	0.023
Sodium‐glucose cotransporter‐2 inhibitor	2120 (25.8)	17 874 (19.8)	0.145	2117 (25.8)	2113 (25.8)	0.001
Insulin	3330 (40.6)	43 071 (47.6)	0.142	3329 (40.6)	3329 (40.6)	0.000
Hemoglobin A1c, %
Mean±SD	7.7±1.8	8.3±2.0	0.292	7.7±1.8	8±1.8	0.140
≥9, n (%)	1649 (20.1)	23 090 (25.5)	0.130	1648 (20.1)	1645 (20.1)	0.001
Estimated glomerular filtration rate, mL/min per 1.73 m2
Mean±SD	69.2±25.9	67.9±27.8	0.049	69.2±25.9	68.6±26.1	0.024
<15, n (%)	244 (3.0)	5179 (5.7)	0.135	244 (3.0)	208 (2.5)	0.027
Cholesterol in low‐density lipoprotein, mg/dL
Mean±SD	77.3±37.2	79.2±37.9	0.050	77.3±37.2	79±37.5	0.044
≥130, n (%)	665 (8.1)	7253 (8.0)	0.003	664 (8.1)	670 (8.2)	0.003

GLP‐1 RA indicates glucagon‐like peptide‐1 receptor agonist; and T2D, type 2 diabetes.

Primary Outcome

During the follow‐up period, patients in the tirzepatide group exhibited a lower MACE incidence rate compared with those in the GLP‐1RA group (3.3 versus 3.7 events per 100 person‐years). The tirzepatide group experienced 173 MACE events, whereas the GLP‐1RA group registered 301 such events. Treatment with tirzepatide was associated with a significantly reduced MACE risk, yielding a HR of 0.75 (95% CI, 0.63–0.91; P=0.003). The calculated E‐value was 2.0, with a lower confidence limit of 1.4 (Table 2). Assessment of the proportionality assumption revealed no violations (Schoenfeld test P=0.515). The Kaplan–Meier analysis similarly demonstrated a significantly lower cumulative MACE incidence in the tirzepatide group compared with the control group (log‐rank P=0.003, Figure 2).

Table 2.

Hazard Ratio of Outcomes Between Tirzepatide and GLP‐1 RA Groups

Outcome	Tirzepatide group (n=8201)		GLP‐1 RA group (n=8201)		HR (95% CI)	P value	E‐value (95% LCL)
	Events (n)	Incidence rate per 100 person‐y	Events (n)	Incidence rate per 100 person‐y
Primary outcome
Major adverse cardiovascular event	173	3.3	301	3.7	0.75 (0.63–0.91)	0.003	2.0 (1.4)
Secondary outcomes
Major adverse limb event	33	0.6	76	0.9	0.59 (0.39–0.88)	0.010	2.8 (1.5)
All‐cause mortality	81	1.5	158	1.9	0.69 (0.53–0.90)	0.007	2.3 (1.5)
Tissue plasminogen activator use	65	1.2	94	1.1	0.88 (0.64–1.20)	0.410	1.5 (1.0)
Acute myocardial infarction	72	1.4	141	1.7	0.70 (0.53–0.93)	0.014	2.2 (1.4)

GLP‐1 RA indicates glucagon‐like peptide‐1 receptor agonist; and HR, hazard ratio.

Figure 2. Kaplan–Meier time‐to‐event free curves of the primary outcome.

GLP‐1 RA indicates glucagon‐like peptide‐1 receptor agonist.

Secondary Outcomes

Tirzepatide use was associated with a significantly lower risk of several secondary outcomes compared with GLP‐1RAs. The incidence rate of MALE was 0.6 per 100 person‐years in the tirzepatide group and 0.9 in the GLP‐1RA group, corresponding to an HR of 0.59 (95% CI, 0.39–0.88; P=0.010), with an E‐value of 2.8 (95% lower confidence limit, 1.5). Similarly, all‐cause mortality was lower among tirzepatide users (1.5 versus 1.9 per 100 person‐years), with an HR of 0.69 (95% CI, 0.53–0.90; P=0.007; E‐value, 2.3 [lower confidence limit, 1.5]). The risk of AMI was also reduced in the tirzepatide group (1.4 versus 1.7 per 100 person‐years), with an HR of 0.70 (95% CI, 0.53–0.93; P=0.014), and an E‐value of 2.2 (lower confidence limit, 1.4). In contrast, there was no significant difference between the groups in the incidence of tPA use (1.2 versus 1.1 per 100 person‐years; HR, 0.88 [95% CI, 0.64–1.20]; P=0.410, Table 2).

Subgroup Analysis

The beneficial association between tirzepatide use and reduced MACE risk compared with GLP‐1RAs remained consistent across most examined subgroups (Figure 3). Tirzepatide was associated with a reduced risk of MACE in men (HR, 0.80 [95% CI, 0.70–0.92]; P=0.002), whereas a similar but nonsignificant trend was observed in women (HR, 0.86 [95% CI, 0.74–1.09]; P=0.063). No significant treatment–sex interaction was detected (P‐interaction=0.569), suggesting that the observed difference. When stratified by baseline BMI, tirzepatide was associated with a neutral effect on MACE among patients with BMI <30 kg/m² (HR, 0.98 [95% CI, 0.80–1.22]; P=0.876), whereas a numerically larger risk reduction was observed in those with BMI ≥30 kg/m² (HR, 0.88 [95% CI, 0.78–1.01]; P=0.059). The treatment–BMI interaction was not statistically significant (P‐interaction=0.413), indicating no strong evidence of effect modification by BMI despite a trend toward greater benefit in patients with obesity. When stratified by ASCVD subtype, tirzepatide was associated with a significantly lower MACE risk in patients with ischemic heart diseases (HR, 0.79 [95% CI, 0.64–0.99]; P=0.038), whereas the reductions in patients with cerebral infarction (HR, 0.66 [95% CI, 0.37–1.19]; P=0.165) and peripheral vascular disease (HR, 0.85 [95% CI, 0.58–1.26]; P=0.429) did not reach statistical significance. In the subgroup with HF, tirzepatide administration was associated with a significantly lower MACE risk compared with GLP‐1RA therapy (HR, 0.66 [95% CI, 0.47–0.93]; P=0.016), whereas this difference did not reach statistical significance in patients without heart failure. Similarly, patients with HbA1c levels <9% showed a significant reduction in MACE risk with tirzepatide treatment (HR, 0.72 [95% CI, 0.54–0.96]; P=0.025), whereas this effect was not statistically significant in those with HbA1c levels ≥9%. When analyzing by specific GLP‐1RA comparators, tirzepatide demonstrated significantly lower MACE hazards compared with both liraglutide (HR, 0.77 [95% CI, 0.61–0.97]; P=0.027) and dulaglutide (HR, 0.72 [95% CI, 0.60–0.88]; P<0.001) but not when compared with semaglutide. However, no statistically significant differences were observed across subgroups defined by obesity status, presence of obstructive sleep apnea, or estimated glomerular filtration rate (<45 or ≥45 mL/min per 1.73m²). Interaction testing did not reveal any statistically significant interactions between treatment group and baseline variables (all P for interaction >0.05), indicating that the association between tirzepatide and reduced MACE risk was consistent across subgroups.

Figure 3. Subgroup analysis for the risk of the primary outcome.

ASCVD indicates atherosclerotic cardiovascular disease; BMI, body mass index; eGFR, estimated glomerular filtration rate; and GLP‐1 RA, glucagon‐like peptide‐1 receptor agonist.

Negative Outcome Controls

No significant differences were observed between the tirzepatide and GLP‐1RA groups in the risk of skin cancer, sensorineural hearing loss, lumbar radiculopathy, traumatic brain injury, or hernia (Table S4).

Sensitivity Analysis

Although BMI remained slightly imbalanced after matching (standardized mean difference=0.179), the absolute difference in mean BMI values was small and the prevalence of obesity (BMI >30 kg/m²) did not differ significantly between groups. In subgroup analyses stratified by obesity status, the direction of association consistently favored tirzepatide, although the differences were not statistically significant.

In a landmark analysis restricted to events occurring between 3 months and 1 year after treatment initiation, tirzepatide remained associated with a significantly lower risk of MACE compared with GLP‐1RA (Table S5). Further analyses varying the time window between diagnosis of T2D and ASCVD and the initiation of tirzepatide consistently demonstrated significant cardiovascular benefit. Specifically, when tirzepatide was initiated within 1 month, 3 months, or 6 months after dual diagnosis, consistent results were observed (Table S6). An as‐treatment analysis comparing patients who continued receiving tirzepatide versus GLP‐1RAs throughout the follow‐up also yielded consistent findings, with a significantly lower risk of MACE in the tirzepatide group (Table S7). These consistent findings in the treatment‐persistence proxy analysis suggest that potential crossover between tirzepatide and GLP‐1 RAs is unlikely to have materially influenced the primary results.

Safety Outcomes

No significant differences were observed between tirzepatide and GLP‐1RAs in the risk of diarrhea (HR, 0.99 95% CI, 0.92–1.07; P=0.827), hypoglycemia (HR, 1.05 95% CI, 0.89–1.23; P=0.569), or pancreatitis (HR, 1.03 95% CI, 0.81–1.32; P=0.814). In contrast, tirzepatide use was associated with a significantly lower risk of major adverse kidney events (HR, 0.90 95% CI, 0.82–0.99; P=0.041, Table S8).

Additional Analysis

Using HFE as the end point, tirzepatide was associated with a lower 1‐year risk versus GLP‐1RAs (HR, 0.83 [95% CI, 0.74–0.94]; P=0.002). Stratified by baseline HF status, the estimates were HR 0.89 (95% CI, 0.72–1.11; P=0.309) among patients with baseline HF and HR 0.84 (95% CI, 0.73–0.96; P=0.010) among those without baseline HF. By HF phenotype, results were HR 1.19 (95% CI, 0.85–1.67; P=0.313) for HF with reduced ejection fraction and HR 0.92 (95% CI, 0.68–1.24; P=0.566) for HF with preserved ejection fraction. (Table S9). In an exploratory analysis restricted to individuals without diabetes, tirzepatide was associated with a significantly lower risk of MACE compared with GLP‐1RA therapy (HR, 0.76 [95% CI, 0.64–0.91]; P=0.003) (Table S10). In a sensitivity analysis restricted to patients with baseline SGLT2i use, the association between tirzepatide and lower MACE risk remained directionally consistent with the primary analysis but did not reach statistical significance (HR, 0.87 [95% CI, 0.70–1.09]; P=0.221) (Table S11). In sensitivity analyses using extended follow‐up windows, tirzepatide remained associated with a lower risk of MALE compared with GLP‐1 receptor agonists. During follow‐up from day 1 to 2 years, the HR for MALE was 0.65 (95% CI, 0.54–0.78; P<0.001). Similar findings were observed with follow‐up extended to 3 years (HR, 0.73 [95% CI, 0.61–0.87]; P<0.001; Table S12). In an additional sensitivity analysis using ICD‐10‐CM code I63 to define ischemic stroke, results remained consistent with the primary analysis. Tirzepatide was associated with a lower risk of MACE (HR, 0.86 [95% CI, 0.79–0.94]; P<0.001) and ischemic stroke (HR, 0.86 [95% CI, 0.76–0.98]; P=0.028) compared with GLP‐1 RAs (Table S13).

DISCUSSION

This cohort study, which included 16 402 patients, demonstrated that in individuals with T2D and ASCVD, treatment with tirzepatide was significantly associated with a lower 1‐year incidence of MACE compared with GLP‐1 RAs. For secondary outcomes, tirzepatide was associated with lower cumulative incidences of all‐cause mortality, MALE, and AMI. Subgroup analyses showed consistent results across most strata, further supporting the potential benefit of tirzepatide in individuals with both T2D and ASCVD. In sex‐stratified analyses, the association between tirzepatide and lower MACE risk was directionally similar in both men and women. Although statistical significance was reached in men but not in women, the CIs overlapped substantially, and the interaction test was not significant (P for interaction=0.569), indicating no evidence of sex‐specific modification of the treatment association. These findings suggest that the observed cardiovascular association is consistent across sexes, although future randomized trials will be needed to determine whether biological sex influences treatment response. Notably, before matching, patients who initiated tirzepatide had lower baseline HbA1c levels compared with those who initiated GLP‐1RAs, despite tirzepatide being more potent for glycemic lowering. This pattern likely reflects real‐world prescribing practices, where clinicians may preferentially select tirzepatide for patients with obesity and acceptable glycemic control rather than indicating prior GLP‐1 RA exposure. Importantly, our strict new‐user design excluded patients with any history of tirzepatide or GLP‐1 RA use before the index date, thereby minimizing potential survivor bias. After matching, HbA1c levels were balanced across groups, further reducing the likelihood that these baseline differences biased the results.

Although some subgroups did not reach statistical significance, a favorable trend was observed. Sensitivity analyses further reinforced the robustness of these findings. These results underscore the potential of tirzepatide as an effective therapeutic option for improving clinical outcomes in patients with T2D across a broad spectrum of ASCVD phenotypes. Because the primary analysis followed an intention‐to‐treat–like framework based on initial treatment assignment, the observed associations should be interpreted as reflecting the comparative effectiveness of treatment initiation strategies rather than strict on‐treatment pharmacologic effects. The attenuated association observed among patients with baseline SGLT2is use may reflect a ceiling effect from concurrent cardioprotective therapy. Because SGLT2is themselves substantially reduce cardiovascular risk, the incremental benefit attributable to tirzepatide compared with GLP‐1 RAs may be partially diluted in this subgroup, leading to smaller relative differences despite a consistent direction of effect.

Our findings build upon and extend the existing literature regarding cardiovascular outcomes with incretin‐based therapies. Several large‐scale randomized trials have firmly established the cardiovascular benefits of GLP‐1 RAs in patients with T2D. The LEADER (Liraglutide Effect and Action in Diabetes: Evaluation of Cardiovascular Outcome Results) trial demonstrated that liraglutide reduced cardiovascular mortality, ²⁶ SUSTAIN‐6 (Trial to Evaluate Cardiovascular and Other Long‐Term Outcomes With Semaglutide in Subjects With Type 2 Diabetes) showed semaglutide significantly lowered the risk of stroke, ⁵ and REWIND (Researching Cardiovascular Events With a Weekly Incretin in Diabetes) indicated that dulaglutide conferred MACE reduction even in a population with relatively lower baseline cardiovascular risk. ²⁷ Similarly, the Harmony Outcomes trial confirmed MACE reduction with albiglutide. ²⁸ Collectively, these studies underscore the class effect of GLP‐1 RAs in reducing MACE rate. A meta‐analysis of 8 randomized trials involving 60 080 patients demonstrated that GLP‐1 RAs significantly reduced the risk of MACE in patients with T2D (HR, 0.86 [95% CI, 0.80–0.93]; P < 0.0001). ²⁹ Another meta‐analysis of 9 studies involving 63 709 participants with ASCVD, most of whom had T2D, demonstrated that GLP‐1 RAs significantly reduced the risk of MACE (risk ratio, 0.85 [95% CI, 0.80–0.91]; P < 0.001). ³⁰ Although these above trials have established the cardiovascular benefits of GLP‐1 RAs in patients with both ASCVD and T2D, evidence for tirzepatide remains comparatively limited but emerging. In the SURPASS‐4 (Study of Tirzepatide Once a Week Versus Insulin Glargine Once a Day in Participants With Type 2 Diabetes and Increased Cardiovascular Risk) trial, tirzepatide demonstrated a numerical reduction in MACE compared with insulin glargine, although cardiovascular outcomes were exploratory end points. ³¹ Beyond glycemic control, the SURMOUNT‐1 (Efficacy and Safety of Tirzepatide Once Weekly Versus Placebo in Participants Who Are Either Obese or Overweight With Weight‐Related Comorbidities) trial in individuals with obesity showed marked weight reduction and improvements in cardiometabolic risk factors, highlighting potential mechanisms for cardiovascular protection. ³² , ³³ Furthermore, a retrospective cohort study demonstrated that tirzepatide significantly reduced the risk of MALE in patients with peripheral artery disease and T2D (HR, 0.44 [95% CI, 0.33–0.59], P<0.001). ¹² Recent trials have also suggested that tirzepatide may be superior to GLP‐1 RAs in reducing glycated hemoglobin levels and achieving greater weight loss. ³⁴ , ³⁵ Additionally, an observational study involving 14 834 patients with T2D showed that tirzepatide was associated with a significantly lower risk of MACE compared with GLP‐1 RAs (HR, 0.80 [95% CI, 0.71–0.91]). ¹¹ However, previous studies have primarily focused on the general T2D population, and evidence remains limited regarding whether tirzepatide offers additional prognostic benefits over GLP‐1 RAs specifically in patients with both T2D and ASCVD.

The ongoing SURPASS‐CVOT trial, directly comparing tirzepatide with dulaglutide on MACE incidence, is expected to provide pivotal randomized evidence. ¹³ Until those results become available, our real‐world analysis provides timely complementary data, suggesting that tirzepatide may offer additional cardiovascular benefit over GLP‐1 RAs, particularly in patients with established ASCVD and T2D.

Both GLP‐1 and GIP are part of the incretin system, which regulates hormone secretion in response to carbohydrate intake. ³⁶ GIP acts within the central nervous system to mitigate the nausea and vomiting commonly associated with GLP‐1 receptor agonism, without compromising satiety or reducing energy intake. ³⁷ This action enhances the overall tolerability of treatment. Clinical trials have shown that tirzepatide, a dual GIP and GLP‐1 RA, is more effective than selective GLP‐1 RAs in stimulating insulin secretion, improving insulin sensitivity, enhancing meal tolerance, and suppressing glucagon secretion, ³⁸ while achieving similar reductions in appetite and caloric intake. ³⁹ Beyond its metabolic effects, GIP may exert cardiovascular benefits through mechanisms independent of nutrition. GIP pathway activation in response to inflammatory stimuli has been shown to stabilize atherosclerotic plaques by inhibiting monocyte and macrophage activation ⁴⁰ and may also protect cardiac tissue by preventing cardiomyocyte hypertrophy, apoptosis, and interstitial fibrosis. ⁴¹ The potential superiority of tirzepatide over GLP‐RAs observed in our findings for MACE reduction suggests that combined incretin receptor activation may provide additional cardiometabolic benefits.

This study leveraged a large, multicenter real‐world database and employed propensity score matching to enhance statistical power and reduce the impact of measured confounders. Our results demonstrate that, in patients with T2D and ASCVD, the use of tirzepatide is significantly associated with a lower risk of 1‐year MACE compared with GLP‐1 RAs. Although the observed HRs (~0.70–0.80) may appear modest in isolation, it is important to recognize that the comparator group consisted of GLP‐1 RAs, which themselves are established cardioprotective therapies with MACE reductions of ~12% to 20% versus placebo in prior randomized controlled trials. Demonstrating an additional benefit over an active agent with proven cardiovascular efficacy therefore represents a clinically relevant incremental effect. Unlike previous tirzepatide versus GLP‐1 RA trials, whose end points were primarily glycemic and weight outcomes, our study provides early evidence of comparative cardiovascular effectiveness on hard outcomes. Given the high baseline cardiovascular risk in T2D with ASCVD, even modest relative reductions translate into meaningful absolute population‐level benefit, supporting the potential added value of tirzepatide beyond established GLP‐1 RA therapy. These findings were further supported by multiple sensitivity analyses, reinforcing the robustness of the observed associations. Collectively, our results provide promising real‐world evidence supporting the clinical potential of tirzepatide in the management of patients with T2D and ASCVD. Nevertheless, further prospective studies are warranted to validate these findings and assess long‐term cardiovascular outcomes.

Limitations

This study has several limitations. First, as the TriNetX database is registry based, there may be issues of misidentification or underrepresentation, particularly for milder cases or individuals less engaged with the health care system, which could affect the generalizability of our findings. Additionally, the reliance on diagnostic codes for identifying clinical variables and outcomes may result in misclassification, potentially introducing bias. To mitigate this, we performed a negative control analysis using unrelated outcomes, which showed no significant differences associated with tirzepatide use, suggesting minimal registration bias. Second, individual‐level information on treatment adherence, prescription duration, dose titration, and over‐the‐counter medication use was not available, and exposure was defined based on prescription records rather than confirmed consumption. Detailed dose intensity and titration schedules were not available within TriNetX; therefore, potential dose–response relationships between tirzepatide or GLP‐1 RAs and cardiovascular outcomes could not be evaluated. In addition, treatment switching or crossover during follow‐up could not be fully captured because detailed prescription stop dates are not available within the TriNetX platform. Although this may introduce exposure misclassification, a treatment‐persistence proxy (as‐treated–like) sensitivity analysis yielded consistent results, suggesting that crossover is unlikely to have materially influenced the primary findings. Third, although propensity score matching balanced a wide array of clinical characteristics—including smoking status, which was available and incorporated into matching, some relevant confounders were not measurable within TriNetX. These include socioeconomic status, alcohol intake, diabetes duration, medication adherence, and HbA1c variability. Such factors may influence real‐world cardiovascular risk and could not be directly controlled. To estimate the potential magnitude of residual confounding, we calculated E‐values (2.0 for MACE), indicating that an unmeasured confounder would require ≥2‐fold associations with both treatment assignment and cardiovascular outcomes, beyond all adjusted covariates, to fully account for our findings. Although this supports the robustness of the observed association, residual confounding cannot be fully excluded. Fourth, it should be noted that our definition of MACE included all‐cause mortality, as cause‐specific death was not available within the TriNetX platform. Although this differs from conventional trial‐based definitions, it reflects the real‐world nature of administrative databases and has been similarly applied in previous observational studies. Future prospective trials with adjudicated cardiovascular deaths are warranted to confirm these findings. Fifth, BMI remained imbalanced after matching (standardized mean difference, 0.179), likely reflecting real‐world prescribing patterns favoring tirzepatide in individuals with higher adiposity. Subgroup analyses did not show significant interaction by obesity status, but residual BMI‐related confounding remains possible. Sixth, the definition of HFE may be nonspecific and does not differentiate inpatient from outpatient events; further, HF phenotype could not be incorporated into the primary propensity score matching model due to the lack of echocardiographic left ventricular ejection fraction data and because the study was designed to evaluate ASCVD‐related MACE rather than HF‐specific prognostic differences. Accordingly, the HFE analysis should be interpreted as exploratory. Seventh, median follow‐up was shorter in the tirzepatide cohort due to recent clinical adoption. Censoring rules were applied identically in both groups, and consistent findings across landmark and treatment‐persistence sensitivity analyses reduce concern for surveillance bias, though modest influence is still possible.

Finally, competing risks regression could not be performed because the TriNetX platform does not provide cause‐specific mortality data or patient‐level time‐to‐event information required for Fine–Gray or other competing risk models. As a result, death from noncardiovascular causes could not be analyzed separately, and our MACE definition relied on all‐cause mortality to avoid informative censoring. Although multiple sensitivity analyses yielded consistent results, residual competing risk bias cannot be fully excluded. In addition, our use of tPA administration as a proxy for ischemic stroke was chosen to minimize misclassification risk inherent in ICD‐10 coding; although this approach increases specificity, it may underestimate the total number of stroke events, as patients not treated with tPA or managed with mechanical thrombectomy were not captured. In addition, because we excluded patients with prespecified outcomes (MACE, AMI, MALE, or death) before the index date to ensure an incident outcome design, our study population primarily represents individuals with ASCVD diagnoses but without very recent acute events. This may limit the generalizability of our findings to patients with more advanced or unstable disease who have experienced prior events. Additionally, future real‐world studies directly comparing tirzepatide with individual GLP‐1RAs such as liraglutide and dulaglutide may further clarify potential heterogeneity within the class.

CONCLUSIONS

Tirzepatide was significantly associated with a lower 1‐year incidence of MACE in patients with T2D and ASCVD compared with GLP‐1 RAs. These findings offer important insights into the management of this high‐risk population, particularly given the rising prevalence of T2D and its associated cardiovascular burden. However, the observational design of this study highlights the need for further prospective trials to validate these results and to investigate long‐term cardiovascular outcomes.

Sources of Funding

This research was conducted without external funding or financial support, ensuring full independence in study design, data collection, analysis, interpretation, and articlepreparation.

Disclosures

The authors declare that there is no conflict of interest regarding the publication of this article.

Supporting information

Tables S1–S7

STROBE Statement