Cardiovascular Outcomes With Tirzepatide Versus GLP‐1 Receptor Agonists in Overweight or Obesity: A Systematic Review and Meta‐Analysis

ABSTRACT

Aims

To compare cardiovascular outcomes associated with tirzepatide versus glucagon‐like peptide‐1 receptor agonists (GLP‐1 RAs) in adults with overweight or obesity.

Materials and Methods

PubMed/MEDLINE, Embase and the Cochrane Central Register of Controlled Trials were searched through 30 March 2026. The study was registered in PROSPERO (CRD420261354941). Randomized and observational studies comparing tirzepatide with GLP‐1 receptor agonists and reporting cardiovascular outcomes were included. Data were extracted on study characteristics and outcomes using adjusted hazard ratios when available. Risk of bias was assessed using RoB 2 and ROBINS‐I.

Results

Eight studies (one randomized and seven observational) including 323 439 patients were analysed. Tirzepatide was not associated with a statistically significant reduction in major adverse cardiovascular events (MACE) compared with GLP‐1 receptor agonists (HR 0.85, 95% CI 0.70–1.04; I ² = 90.4%). Substantial heterogeneity was observed. Effect estimates were directionally consistent across secondary outcomes, including all‐cause mortality (HR 0.90, 95% CI 0.79–1.02) and cardiovascular mortality (HR 0.88, 95% CI 0.76–1.00). In exploratory subgroup analyses, patients with type 2 diabetes showed lower estimated risks of all‐cause mortality (HR 0.85, 95% CI 0.76–0.94) and heart failure (HR 0.75, 95% CI 0.58–0.97).

Conclusions

Tirzepatide was not associated with a statistically significant reduction in MACEs. Although effect estimates were directionally consistent across analyses, substantial heterogeneity and the predominance of observational studies limit causal interpretation. Further randomized trials are needed to clarify the comparative cardiovascular effects of tirzepatide.

1. Introduction

Obesity has become a major global health burden and is a key driver of cardiovascular disease, contributing substantially to morbidity and mortality worldwide [1, 2]. The association between excess adiposity and cardiovascular risk is largely mediated by metabolic dysfunction, including insulin resistance, systemic inflammation, hypertension and dyslipidemia [3, 4].

In this context, incretin‐based therapies—particularly glucagon‐like peptide‐1 receptor agonists (GLP‐1RAs)—have emerged as a major therapeutic advance, demonstrating significant effects on weight reduction and cardiometabolic risk [5, 6, 7]. Large‐scale cardiovascular outcomes trials have shown that GLP‐1RAs, such as semaglutide, dulaglutide and liraglutide, reduce the risk of major adverse cardiovascular events (MACE), including cardiovascular death, myocardial infarction and stroke [8, 9, 10, 11]. More recently, these benefits have been extended to individuals without diabetes [12].

Tirzepatide, a dual glucose‐dependent insulinotropic polypeptide (GIP) and GLP‐1 receptor agonist, has demonstrated greater efficacy in weight loss and metabolic improvement compared with selective GLP‐1RAs, and emerging but inconclusive evidence suggests potential cardiovascular benefit [13, 14, 15]. However, whether these enhanced metabolic effects translate into incremental cardiovascular benefit remains uncertain.

Recent network meta‐analyses in patients with type 2 diabetes have suggested that tirzepatide may provide cardiovascular benefits comparable to or numerically greater than those observed with established GLP‐1 receptor agonists [16, 17]. However, these analyses relied predominantly on indirect comparisons, were largely driven by a single cardiovascular outcomes trial of tirzepatide and were restricted to diabetic populations at high cardiovascular risk. As a result, the comparative cardiovascular effects of tirzepatide versus GLP‐1 receptor agonists across broader populations with overweight or obesity remain uncertain. In parallel, direct comparative studies evaluating tirzepatide versus GLP‐1 receptor agonists have yielded heterogeneous and sometimes conflicting results, with some suggesting greater benefit with GLP‐1RAs, others showing similar effects and some favouring tirzepatide [18, 19, 20, 21]. These discrepancies, together with variability in study design, populations and outcome definitions, hinder clear clinical interpretation and underscore the need for a systematic synthesis of the evidence.

Therefore, we conducted a systematic review and meta‐analysis to compare cardiovascular outcomes associated with tirzepatide versus GLP‐1 receptor agonists in adults with overweight or obesity, with or without diabetes.

2. Methods

This systematic review and meta‐analysis were conducted and reported in accordance with the Cochrane Collaboration Handbook for Systematic Reviews of Interventions and the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses statement guidelines (Methods S1 and S2) [22, 23]. The study protocol was prospectively registered in the International Prospective Register of Systematic Reviews (PROSPERO; CRD420261354941).

2.1. Data Source and Strategy

A comprehensive literature search was performed in PubMed/MEDLINE, Embase and the Cochrane Central Register of Controlled Trials (CENTRAL) from database inception through March 30, 2026. The search strategy combined controlled vocabulary terms (MeSH and Emtree) and free‐text keywords related to tirzepatide, GLP‐1 RAs, obesity and overweight. The complete search strategies for each database are provided in the Methods S3.

All records were imported into Zotero (Corporation for Digital Scholarship) for deduplication and subsequently uploaded to Rayyan (Qatar Computing Research Institute) for screening. Two reviewers (J.P.M.R.J.S. and D.L.S.) independently screened titles and abstracts, followed by full‐text assessment of potentially eligible studies. Disagreements were resolved by consensus or adjudication by a third reviewer.

2.2. Eligibility Criteria

We included randomized controlled trials (RCTs) and observational studies with an active comparator design, including studies using propensity score matching, inverse probability of treatment weighting or target trial emulation methods that compared tirzepatide with any approved GLP‐1 RA.

Eligible studies enrolled adults (≥ 18 years) with overweight or obesity (body mass index [BMI] ≥ 27 kg/m²). Studies including populations with lower BMI thresholds (e.g., ≥ 25 kg/m²) were also eligible provided that the study population was predominantly obese, defined as a mean BMI ≥ 30 kg/m². Studies were required to report at least one prespecified cardiovascular endpoint and to have a minimum follow‐up duration of 3 months.

We excluded studies with placebo or no active comparator, those reporting only surrogate outcomes (e.g., weight loss, glycated haemoglobin or lipid levels) and non‐original studies (including reviews, editorials, case reports and case series). In cases of overlapping populations, the most comprehensive or recent dataset was included.

2.3. Endpoints

The primary endpoint was MACE. Definitions of MACE across included studies are summarized in Table S1. Secondary endpoints included all‐cause mortality, cardiovascular mortality, myocardial infarction, acute coronary syndrome, stroke and hospitalization for heart failure or incident heart failure.

2.4. Subgroups and Sensitivity Analysis

Prespecified subgroup analyses were conducted for the primary endpoint (MACE) according to age (≥ 65 vs. < 65 years), sex, BMI (≥ 30 vs. < 30 kg/m²) and diabetes status. Additional exploratory subgroup analyses were conducted for prior heart failure, obstructive sleep apnoea and according to baseline atherosclerotic cardiovascular disease (ASCVD) status. For secondary outcomes, subgroup analyses according to diabetes status were also performed.

Sensitivity analyses were conducted to evaluate the robustness of the findings across all outcomes. These included leave‐one‐out analyses, type of GLP‐1 receptor agonist comparator, exclusion of industry‐funded studies and analyses stratified by study design. For the primary endpoint, additional analyses were performed by restricting to studies using three‐component definitions of MACE (death, myocardial infarction and stroke) and further to those defining MACE as a composite of all‐cause mortality, myocardial infarction and stroke. An additional exploratory sensitivity analysis was conducted according to new‐user design status to evaluate whether inclusion of studies enrolling prevalent users influenced the primary outcome estimate.

2.5. Data Extraction

Data extraction was performed independently by two pairs of reviewers using a standardized, pilot‐tested data collection form. Extracted variables included study characteristics, baseline population data and outcomes of interest. Discrepancies were resolved through discussion and, when necessary, adjudication by a third reviewer. Consistency checks were performed prior to statistical analysis.

2.6. Quality Assessment

Risk of bias in RCTs was assessed using the Cochrane Risk of Bias 2 (RoB 2) tool, whereas non‐randomized studies were evaluated using the Risk of Bias in Non‐Randomized Studies of Interventions (ROBINS‐I) tool [24, 25]. Assessments were conducted independently by two reviewers (J.P.M.R.J.S. and O.I.d.M.B.), with disagreements resolved by consensus or adjudication by a third reviewer.

With fewer than 10 studies available for most outcomes, formal testing for small‐study effects (e.g., Egger's regression) was not performed, in accordance with recommendations from the Cochrane Handbook [26, 27]. Funnel plots were generated for visual inspection (Figure S7A–G).

2.7. Statistical Analysis

Effect estimates were reported as hazard ratios (HRs) with corresponding 95% confidence intervals (CIs). When available, adjusted effect estimates were preferentially extracted.

Pooled analyses were conducted using a random‐effects model with the restricted maximum likelihood (REML) estimator and the generic inverse variance method. Heterogeneity was assessed using Cochran's Q test and the I ² statistic. A two‐sided p value < 0.05 was considered statistically significant.

To explore potential sources of between‐study heterogeneity, random‐effects meta‐regression analyses were performed using prespecified study‐level covariates, including mean BMI, proportion of patients with diabetes, mean age and follow‐up duration. Additional exploratory covariates included prevalence of hypertension and background use of sodium–glucose cotransporter‐2 inhibitors, insulin and metformin. Models were estimated using the DerSimonian–Laird method with Knapp–Hartung adjustment. Given the limited number of studies, only univariable models were fitted, and analyses were conducted only when at least four studies were available for a given covariate. Results were interpreted cautiously.

All analyses were performed using R software version 4.3.3 (R Foundation for Statistical Computing, Vienna, Austria) with the ‘meta’ [28], ‘metafor’ [29] and ‘dmetar’ [30] packages.

3. Results

3.1. Study Selection and Baseline Characteristics

The initial search yielded 1881 records. After removing duplicates and ineligible studies, 17 studies underwent full‐text assessment for eligibility and 8 studies were included, comprising 1 RCT [19] and 7 observational cohort studies [18, 20, 21, 31, 32, 33, 34], totaling 323 439 patients (Figure S1). Table S1 summarizes the key characteristics of the included studies and the baseline characteristics of the included participants are provided in Table S2. In the pooled population, the mean age was 56.7 years, 57.5% were women and the mean BMI was 38.1 kg/m². The prevalence of type 2 diabetes and hypertension was 71.3% and 73.0%, respectively. Overall, these characteristics indicate a population with a high cardiometabolic risk profile. Seven of the eight included studies employed a new‐user design, whereas one study included prevalent users. Follow‐up duration ranged from 6 to 48 months across studies.

3.2. Pooled Analyses

In the pooled analysis for MACE (Figure 1), given the substantial between‐study heterogeneity (I ² = 90.4%), pooled estimates are presented primarily to quantify consistency of direction rather than to provide a precise magnitude of effect; caution is warranted in interpretation. Tirzepatide and GLP‐1 receptor agonists showed similar risks for the primary endpoint (HR 0.85, 95% CI 0.70–1.04; p = 0.12). A subgroup analysis according to study design showed similar effect estimates in observational studies (HR 0.84, 95% CI 0.66–1.07) and the RCT (HR 0.92, 95% CI 0.83–1.01), with no evidence of subgroup differences (p for interaction = 0.51).

FIGURE 1.

Pooled analysis of major adverse cardiovascular events (MACE). Forest plot comparing tirzepatide versus GLP‐1 receptor agonists for major adverse cardiovascular events (MACE). Effect estimates are presented as hazard ratios (HRs) with 95% confidence intervals (CIs) using random‐effects models. Studies are stratified according to study design (observational studies and randomized controlled trials). The vertical line indicates no effect (HR = 1). Heterogeneity (I ²) and the test for subgroup differences are reported.

Results were similar across secondary outcomes (Figure 2). Estimates for all‐cause mortality (HR 0.90, 95% CI 0.79–1.02; I ² = 22.8%), cardiovascular mortality (HR 0.88, 95% CI 0.76–1.00; I ² = 0.0%) and heart failure (HR 0.86, 95% CI 0.72–1.03; I ² = 55.0%) did not reach statistical significance. No significant associations were observed for myocardial infarction, acute coronary syndrome or stroke. Detailed forest plots for each outcome are provided in Figure S2A–F.

FIGURE 2.

Cardiovascular outcomes. Forest plot comparing tirzepatide versus GLP‐1 receptor agonists for cardiovascular outcomes, including major adverse cardiovascular events (MACE), all‐cause mortality, cardiovascular mortality, myocardial infarction, acute coronary syndrome, stroke and heart failure. Effect estimates are presented as hazard ratios (HRs) with 95% confidence intervals (CIs) using random‐effects models. The vertical line indicates no effect (HR = 1). Heterogeneity (I ²) and p values are reported for each outcome.

3.3. Subgroup Analyses

In subgroup analyses for MACE (Figure 3), no statistically significant interactions were observed across predefined categories. Estimates were similar across age groups (< 65 years: HR 0.97, 95% CI 0.81–1.16; ≥ 65 years: HR 0.97, 95% CI 0.82–1.14; p for interaction = 0.98; k = 3) and between sexes (males: HR 0.88, 95% CI 0.72–1.07; females: HR 0.82, 95% CI 0.61–1.09; p = 0.68; k = 4).

FIGURE 3.

Subgroup analyses for MACE. Forest plot of subgroup analyses for major adverse cardiovascular events (MACE) comparing tirzepatide versus GLP‐1 receptor agonists according to predefined study‐level and patient‐level characteristics. Effect estimates are presented as hazard ratios (HRs) with 95% confidence intervals (CIs) using random‐effects models. The vertical dashed line indicates no effect (HR = 1). p values for interaction are reported to assess differences between subgroups.

No statistically significant interaction was observed by BMI (p = 0.84; k = 3). Estimates were HR 0.69 (95% CI 0.53–0.89) for BMI ≥ 30 kg/m² and HR 0.72 (95% CI 0.50–1.05) for BMI < 30 kg/m². Similarly, no statistically significant interaction was observed according to diabetes status (p = 0.28). Estimates were HR 0.80 (95% CI 0.63–1.00; k = 5) in patients with type 2 diabetes and HR 1.04 (95% CI 0.68–1.60; k = 2) in those without diabetes.

In exploratory analyses of additional comorbidities, estimates remained close to neutrality among patients with prior heart failure (HR 0.94, 95% CI 0.83–1.08; k = 2). In analyses restricted to populations with obstructive sleep apnoea, the pooled estimate numerically favoured tirzepatide (HR 0.74, 95% CI 0.55–0.99; I ² = 88.1%; k = 3). However, this finding should be interpreted cautiously given the limited number of studies, substantial heterogeneity and the exploratory nature of these analyses.

An additional exploratory analysis according to baseline ASCVD status was performed. The pooled estimate was HR 0.91 (95% CI 0.59–1.38; k = 3) among studies restricted to patients with established ASCVD and HR 0.75 (95% CI 0.58–0.98; k = 3) among studies without established ASCVD. The test for subgroup differences was not statistically significant (p = 0.0515). Although the interaction test approached nominal statistical significance, the limited number of studies and the exploratory nature of this analysis preclude definitive conclusions regarding differential treatment effects according to ASCVD status.

For secondary outcomes, analyses restricted to patients with type 2 diabetes demonstrated generally directionally favourable estimates for tirzepatide across most endpoints (Figure 4). Lower estimated risks were observed for all‐cause mortality (HR 0.85, 95% CI 0.76–0.94; I ² = 33.3%; k = 4), heart failure outcomes (HR 0.75, 95% CI 0.58–0.97; I ² = 11.4%; k = 2) and acute coronary syndrome (HR 0.60, 95% CI 0.39–0.93; I ² = 58.4%; k = 2). No statistically significant differences were observed for cardiovascular mortality, myocardial infarction or stroke.

FIGURE 4.

Cardiovascular outcomes in patients with diabetes. Forest plot comparing tirzepatide versus GLP‐1 receptor agonists for cardiovascular outcomes, including major adverse cardiovascular events (MACE), all‐cause mortality, cardiovascular mortality, myocardial infarction, acute coronary syndrome, stroke and heart failure among patients with diabetes. Hazard ratios (HRs) with 95% confidence intervals (CIs) are shown using random‐effects models. The vertical dashed line represents no effect (HR = 1). Heterogeneity (I ²) and p values are provided for each outcome.

Overall, although several subgroup analyses were prespecified, these analyses should be interpreted as exploratory and hypothesis‐generating, particularly given the limited number of studies available for several comparisons, substantial residual heterogeneity and the absence of adjustment for multiple testing.

3.4. Sensitivity Analyses

Multiple sensitivity analyses were performed and yielded findings consistent with the primary results. Restriction to studies using three‐component MACE definitions produced similar effect estimates, regardless of whether death was defined as all‐cause or cardiovascular (HR 0.95, 95% CI 0.72–1.26; I ² = 79%) or restricted to all‐cause mortality, myocardial infarction and stroke (HR 0.96, 95% CI 0.61–1.49; I ² = 83%) (Figure S3A,B).

Analyses stratified by study characteristics were also consistent with the main findings. Restriction to semaglutide as the comparator (HR 0.97, 95% CI 0.83–1.14; I ² = 75.8%) and to studies employing a new‐user design (HR 0.89, 95% CI 0.73–1.09) yielded effect estimates similar in direction and magnitude to the primary analysis. In contrast, exclusion of industry‐funded studies was associated with a statistically significant reduction in MACE (HR 0.78, 95% CI 0.63–0.98; I ² = 92.5%) (Figure S3C–E).

For secondary outcomes, sensitivity analyses restricted to semaglutide comparators, non–industry‐funded studies and observational designs showed consistent results without material changes in statistical significance. A more pronounced reduction in heart failure outcomes was observed in analyses excluding industry‐funded studies (HR 0.80, 95% CI 0.72–0.90), whereas other outcomes remained largely unchanged (Figure S4A–C).

Leave‐one‐out analyses further supported the robustness of the findings (Figure S5A–G). No single study materially influenced the overall estimate for MACE, although exclusion of Wilson et al. resulted in a statistically significant reduction (HR 0.81, 95% CI 0.67–0.98), indicating a modest influence of this study. Similar patterns were observed for secondary outcomes.

Overall, heterogeneity patterns were preserved across analyses, suggesting that between‐study variability was not driven by any single study. Collectively, these analyses reinforce the robustness and consistency of the observed treatment effects.

3.5. Quality Assessment

Risk of bias assessments are presented in Figure S6A,B. The randomized trial was judged to be at low risk of bias, whereas most observational studies were rated as having serious risk of bias, primarily driven by residual confounding and selective reporting. Risk of bias was generally lower for outcome measurement and participant selection, consistent with the expected limitations of non‐randomized designs.

Visual inspection of funnel plots did not suggest substantial asymmetry across outcomes (Figure S7A–G). For MACE, although some dispersion was observed, the overall distribution did not indicate a consistent pattern of publication bias. Across other outcomes, variability in study distribution was likely driven by the limited number of studies and between‐study heterogeneity. Formal statistical testing for publication bias was not performed due to the small number of studies per analysis. Overall, these findings do not suggest a major influence of publication bias on the results.

3.6. Meta‐Regression

Random‐effects meta‐regression analyses were conducted as exploratory analyses to investigate potential sources of between‐study heterogeneity. No consistent associations were observed between treatment effects and prespecified study‐level covariates, including mean BMI, diabetes prevalence, mean age, follow‐up duration, hypertension prevalence and metformin use. An apparent association between mean age and cardiovascular mortality was observed, although this finding should be interpreted cautiously given the limited number of available studies and the low statistical power of meta‐regression analyses in this setting. Borderline associations were noted for mean BMI and diabetes prevalence in analyses of myocardial infarction and stroke, but these did not reach statistical significance. Detailed results are provided in the Supporting Information (Tables S3–S8 and Figures S8–S13).

4. Discussion

In this meta‐analysis, tirzepatide was not associated with a statistically significant reduction in MACE compared with GLP‐1 receptor agonists (HR 0.85, 95% CI 0.70–1.04). Although effect estimates were directionally consistent across outcomes and analyses, the absence of statistical significance and the substantial between‐study heterogeneity limit the strength of inference. Nonetheless, the concordance of findings across multiple endpoints and analytical approaches reduces the likelihood that these results are entirely spurious, with imprecision likely driven by between‐study variability.

Recent network meta‐analyses by Shokravi et al. [16] and Pham et al. [17] suggested that tirzepatide provides cardiovascular benefit at least comparable to GLP‐1 receptor agonists in patients with type 2 diabetes, largely based on indirect comparisons from randomized cardiovascular outcomes trials. In contrast, our study evaluated direct active‐comparator evidence and incorporated real‐world observational data in broader overweight and obesity populations, including individuals without diabetes. Although the overall direction of effect was generally consistent with prior literature, our pooled analysis showed greater heterogeneity and did not demonstrate a statistically significant reduction in MACE. These findings suggest that any incremental cardiovascular benefit of tirzepatide over established GLP‐1 receptor agonists may be smaller or more context‐dependent in real‐world settings.

One important factor that may explain the absence of statistical significance is the use of GLP‐1 receptor agonists as active comparators. These agents are independently associated with cardiovascular benefit, which may attenuate observable differences and reduce the likelihood of detecting incremental effects [8, 9, 10, 11, 12]. As such, this comparison represents a high therapeutic benchmark, in which demonstrating incremental benefit is inherently more challenging.

A key consideration is the availability of the SURPASS‐CVOT trial [19], a large, double‐blind randomized study comparing tirzepatide with dulaglutide in patients with type 2 diabetes and established cardiovascular disease. In that trial, tirzepatide was noninferior but not superior for MACEs (HR 0.92, 95% CI 0.83–1.01). In this context, the present meta‐analysis does not aim to refine the precision of this randomized estimate but rather to contextualize it across broader and more heterogeneous populations, including individuals without diabetes, real‐world settings and additional cardiovascular outcomes such as heart failure. While the randomized trial provides the most robust estimate of effect in patients with type 2 diabetes, the present analysis extends its interpretation by incorporating complementary evidence on effectiveness and potential variability across populations and study designs.

Beyond these considerations, sensitivity analyses provided additional insight into potential sources of variability. Exclusion of industry‐funded studies was associated with a statistically significant reduction in MACE risk (HR 0.78, 95% CI 0.63–0.98). However, substantial heterogeneity persisted after exclusion of industry‐funded studies (I ² = 92.5%), suggesting that sponsorship alone is unlikely to explain the observed between‐study variability. Notably, this shift appeared to be driven predominantly by exclusion of the Wilson et al. study, which was both industry‐funded and among the largest real‐world cohorts included in the analysis. Consistently, leave‐one‐out analyses demonstrated that the exclusion of Wilson et al. alone resulted in a statistically significant reduction in MACE risk (HR 0.81, 95% CI 0.67–0.98), highlighting the substantial influence of this study on the pooled estimate.

Several characteristics may explain the influence of Wilson et al., including its large sample size, observational real‐world design, inclusion of patients with established ASCVD but without diabetes and relatively short follow‐up duration. Importantly, these findings should not be interpreted as evidence of systematic sponsorship bias, but rather as illustrating the influence that individual large observational studies may exert on pooled estimates in clinically heterogeneous evidence bases with limited randomized data.

From a clinical perspective, subgroup analyses suggested differences in effect estimates across populations with higher metabolic burden, including those with higher BMI, type 2 diabetes and obstructive sleep apnoea (Figures 3 and 4). Given the well‐established relationship between adiposity, insulin resistance and cardiovascular risk, the greater weight reduction associated with tirzepatide provides a plausible biological mechanism underlying these findings [3, 4, 13]. In addition, exploratory analyses according to baseline ASCVD status showed numerically more favourable estimates in studies enrolling patients without established ASCVD than in those restricted to patients with pre‐existing ASCVD, although the test for subgroup differences was not statistically significant. These observations raise the possibility that treatment effects may vary according to baseline cardiometabolic and cardiovascular risk profiles. However, given the exploratory nature of subgroup analyses and the limited number of contributing studies, this interpretation should be considered hypothesis‐generating.

In parallel, the translation of metabolic improvements into reductions in hard cardiovascular outcomes is not immediate. Cardiovascular risk reduction is a cumulative process, and the relatively short follow‐up durations of several included studies may have limited the ability to detect differences in clinical events (Table S1). This temporal disconnect between improvements in surrogate markers and hard outcomes is an important consideration when interpreting these findings.

Substantial heterogeneity was observed across analyses, particularly for MACE (I ² = 90.4%), which substantially limits the interpretability of pooled estimates. This variability likely reflects the combined influence of multiple clinical and methodological differences across studies rather than any single identifiable source. Notably, substantial heterogeneity persisted across multiple subgroup and sensitivity analyses, including restrictions by study design, comparator type, MACE definition, diabetes status and baseline ASCVD status (Figure 3 and Figure S3A–E). Although sensitivity analyses restricted to more homogeneous MACE definitions yielded similar effect estimates, heterogeneity remained considerable, suggesting that between‐study variability is multifactorial. Nonetheless, the overall direction of effect remained generally consistent across analyses, supporting consistency in treatment effect direction despite uncertainty regarding its magnitude.

To further explore this variability, meta‐regression analyses were conducted but did not identify consistent study‐level modifiers of treatment effect. This finding should be interpreted in the context of the limited number of studies, restricted statistical power and the use of aggregated data, which may have reduced the ability to detect true study‐level modifiers of treatment effect. Accordingly, the absence of identified modifiers does not exclude the presence of clinically relevant effect modification.

4.1. Strengths and Limitations

This study has several strengths, including the comprehensive evaluation of multiple clinically relevant outcomes, the use of extensive sensitivity analyses and the integration of randomized and real‐world evidence. Together, these elements provide a robust and balanced assessment of treatment effects across diverse clinical contexts.

Despite these strengths, several limitations should be acknowledged. The number of included studies was limited, particularly for some outcomes and substantial heterogeneity was observed. The predominance of observational studies introduces the potential for residual confounding. Although most observational studies employed a new‐user design, residual treatment‐selection and indication bias cannot be excluded. In addition, multiple subgroup and secondary analyses were performed without formal adjustment for multiple testing, increasing the possibility of chance findings. Furthermore, meta‐regression analyses were constrained by the small number of studies and the use of aggregated data, limiting the ability to fully explain observed variability. Accordingly, these findings should be interpreted in light of the underlying study designs.

From a clinical perspective, these findings do not demonstrate a statistically significant difference in cardiovascular outcomes between tirzepatide and GLP‐1 receptor agonists. However, the directionally consistent estimates observed across analyses, in the context of an active comparator with established efficacy, support the need for further investigation. Future large‐scale randomized trials with longer follow‐up will be essential to confirm these findings and to better identify patient populations most likely to benefit.

Author Contributions

J.P.M.R.J.S. contributed to study conception and design, statistical analysis, data interpretation and drafting of the manuscript. D.L.S. contributed to study design, data screening, data collection, analysis and critical revision of the manuscript. B.V.N., J.R.C.d.S., L.B.A.S. and O.I.d.M.B. contributed to data collection, analysis and critical revision of the manuscript. W.F.G. contributed to study conception, supervision, data interpretation and critical revision of the manuscript. All authors approved the final version of the manuscript. J.P.M.R.J.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Funding

The authors have nothing to report.

Ethics Statement

The authors have nothing to report.

Conflicts of Interest

W.F.G. reports proctorship honoraria from Abbott (Navitor transcatheter aortic valve) and Meril Life Sciences (MyVal transcatheter aortic valve) and speaker honoraria from Libbs Farmacêutica (ticagrelor). He has received travel and meeting support from Teleflex Incorporated for attendance at the Triple‐i CVI Summit 2026. He reports no financial or non‐financial relationships with Eli Lilly, Novo Nordisk or any other manufacturer of tirzepatide or GLP‐1 receptor agonists and no relationships relevant to the subject matter of this manuscript. The other authors declare no conflicts of interest.

Supporting information

Method S1 PRISMA Main Checklist.

Methods S2. PRISMA Abstract Checklist.

Methods S3. Details of the Search Strategy According to the Database.

Table S1: Characteristics of included studies.

Table S2: Baseline Characteristics of Included Participants.

Table S3: Meta‐regression Analysis for MACE.

Table S4: Meta‐regression Analysis for All‐cause Mortality.

Table S5: Meta‐regression Analysis for Cardiovascular Mortality.

Table S6: Meta‐regression Analysis for Myocardial Infarction.

Table S7: Meta‐regression Analysis for Stroke.

Table S8: Meta‐regression Analysis for Heart Failure outcomes.

Figure S1: PRISMA flow diagram.

Figure S2: Forest plots for secondary outcomes.

Figure S2: (A) All‐cause mortality. (B) Cardiovascular mortality. (C) Myocardial infarction. (D) Acute coronary syndrome. (E) Stroke. (F) Heart failure outcomes.

Figure S3: Sensitivity Analysis for MACE. (A) MACE restricted to studies with three‐component definitions (all‐cause or cardiovascular mortality, myocardial infarction and stroke). (B) MACE restricted to studies defining the outcome as all‐cause mortality, myocardial infarction and stroke. (C) MACE restricted to semaglutide comparator. (D) MACE excluding industry‐funded studies. (E) MACE restricted to studies employing a new‐user design.

Figure S4: Sensitivity analyses for secondary outcomes. (A) Secondary outcomes restricted to semaglutide comparator. (B) Secondary outcomes excluding industry‐funded studies. (C) Secondary endpoints restricted to observational studies.

Figure S5: Leave‐one‐out analyses. (A) MACE. (B) All‐cause mortality. (C) Cardiovascular mortality. (D) Myocardial infarction. (E) Acute coronary syndrome. (F) Stroke. (G). Heart failure outcomes.

Figure S6: Risk of bias assessment of included studies. (A) Risk of bias in randomized controlled trials (RoB 2). (B) Risk of bias in observational studies (ROBINS‐I).

Figure S7: Publication bias assessment. (A) MACE. (B) All‐cause mortality. (C) Cardiovascular mortality. (D) Myocardial infarction.

(E) Acute coronary syndrome. (F) Stroke. (G) Heart failure outcomes.

Figure S8: Meta‐regression plots for MACE. (A) Mean BMI. (B) Diabetes prevalence. (C) Mean age. (D) Follow‐up duration. (E) Hypertension prevalence. (F) Prevalence of metformin use.

Figure S9: Meta‐regression plots for all‐cause mortality. (A) Mean BMI. (B) Diabetes prevalence. (C) Mean age. (D) Follow‐up duration. (E) Hypertension prevalence. (F) Prevalence of metformin use.

Figure S10: Meta‐regression plots for cardiovascular mortality. (A) Mean BMI. (B) Diabetes prevalence. (C) Mean age. (D) Follow‐up duration. (E) Hypertension prevalence.

Figure S11: Meta‐regression plots for myocardial infarction. (A) Mean BMI. (B) Diabetes prevalence. (C) Mean age. (D) Follow‐up duration. (E) Hypertension prevalence.

Figure S12: Meta‐regression plots for stroke. (A) Mean BMI. (B) Diabetes prevalence. (C) Mean age. (D) Follow‐up duration. (E) Hypertension prevalence. (F) Prevalence of metformin use.

Figure S13: Meta‐regression plots for heart failure outcomes. (A) Mean BMI. (B) Diabetes prevalence. (C) Mean age. (D) Follow‐up duration. (E) Hypertension prevalence.

Silva J. P. M. R. J., Nogueira B. V., Sartori D. L., et al., “Cardiovascular Outcomes With Tirzepatide Versus GLP‐1 Receptor Agonists in Overweight or Obesity: A Systematic Review and Meta‐Analysis,” Diabetes, Obesity and Metabolism 28, no. 9 (2026): 8627–8636, 10.1111/dom.71051.

Data Availability Statement

All data analysed in this study were extracted from publications cited in the manuscript and Supporting Information. The study protocol was prospectively registered in PROSPERO (CRD420261354941) and is available at: https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD420261354941.