Literature search and study characteristics

The flowchart detailing the systematic literature search and selection process is presented in Fig. 1. The search identified 3,697 relevant articles across PubMed (n = 338), Embase (n = 2,160), Web of Science (n = 369), and The Cochrane Library (n = 830). After removing duplicates, 2,696 articles remained for title and abstract screening. The full texts of 19 articles were reviewed, resulting in the exclusion of 5 conference abstracts, 1 post-hoc analysis of an RCT, 3 articles with duplicate data, and 1 article that did not fit the population criteria. In total, 9 studies were included in the meta-analysis [16,17,18,19,20,21,22,23,24].

Fig. 1

Flowchart of the systematic search and selection process

Tables 1 and 2 present the characteristics of the included studies and patient populations. All 9 studies were multicenter, randomized, double-blind, controlled trials, encompassing 63,613 patients in total (GLP-1RAs group: 32,461; control group: 31,152). Sample sizes ranged from 3,183 to 14,752 patients, with the average age of participants between 60.1 and 66.6 years. The proportion of male participants ranged from 38 to 70%, and the median follow-up duration ranged from 1.3 to 5.4 years. The GLP-1RAs evaluated included albiglutide, dulaglutide, lixisenatide, liraglutide, semaglutide, exenatide, and efpeglenatide. Among the 9 studies, 1 study incorporated standard treatment for T2DM in both the treatment and control groups [20], and another study included SGLT2 inhibitors in part of the treatment and control groups [22]. The remaining 7 studies were placebo-controlled trials of GLP-1RAs monotherapy.

Table 1 Baseline characteristics of the included studiesTable 2 Patients’ demographic and clinical characteristicsQuality Assessment results

The Cochrane Collaboration’s Risk of Bias tool was used to evaluate the bias risk in all included studies. The assessment indicated that all studies had a low risk of bias in key areas, including random sequence generation, allocation concealment, and blinding. Overall, the methodological quality was rated as high (Supplementary Figure S1).

Primary composite outcome

Eight studies, encompassing 60,080 patients in total, were analyzed to evaluate the impact of GLP-1 RAs on the primary composite outcome. During the follow-up period, 3,136 of 30,694 patients in the GLP-1 RAs group (10.2%) experienced a primary composite outcome, compared to 3,443 of 29,386 patients in the control group (11.7%). The forest plot (Fig. 2A) displays the HRs and their 95% CIs for each study. The pooled analysis revealed that GLP-1 RAs notably lowered the risk of the primary composite outcome (HR: 0.86; 95% CI: 0.80–0.92, P < 0.0001), with no significant heterogeneity was detected (I² = 46%, P = 0.08). The funnel plot (Supplementary Figure S2A) and Egger’s test indicated no significant publication bias (P = 0.154).

Fig. 2

Forest plot of pooled HRs for cardiovascular outcomes: (A) Primary composite outcome, (B) Cardiovascular death, (C) All-cause death, and (D) Myocardial infarction

Cardiovascular death

Nine studies, encompassing 63,613 patients in total, were analyzed to evaluate the impact of GLP-1 RAs on cardiovascular death. During the follow-up period, 1,413 of 32,461 patients in the GLP-1 RAs group (4.4%) experienced cardiovascular death, compared to 1,588 of 31,152 patients in the control group (5.1%). The forest plot (Fig. 2B) displays the HRs and their 95% CIs for each study. The pooled analysis demonstrated that GLP-1 RAs significantly reduced cardiovascular death risk (HR: 0.85; 95% CI: 0.78–0.93, P = 0.0002), with no significant heterogeneity detected (I² = 26%, P = 0.22). Funnel plot analysis (Supplementary Figure S2B) and Egger’s test indicated no significant publication bias (P = 0.293).

All-cause death

Nine studies, encompassing 63,613 patients in total, were analyzed to assess the effect of GLP-1 RAs on all-cause death. During the follow-up period, 2,266 of 32,461 patients in the GLP-1 RAs group (7.0%) experienced all-cause death, compared to 2,456 of 31,152 patients in the control group (7.9%). The forest plot (Fig. 2C) displays the HRs and their 95% CIs for each study. The pooled analysis demonstrated that GLP-1 RAs significantly reduced the risk of all-cause death (HR: 0.87; 95% CI: 0.82–0.93, P < 0.00001), with no significant heterogeneity detected (I² = 10%, P = 0.35). Analysis using the funnel plot (Supplementary Figure S2C) and Egger’s test showed no significant publication bias (P = 0.410).

Myocardial infarction

Nine studies, encompassing 63,613 patients in total, were analyzed to assess the effect of GLP-1 RAs on myocardial infarction. During the follow-up period, 1,667 of 32,461 patients in the GLP-1 RAs group (5.1%) experienced myocardial infarction, compared to 1,790 of 31,152 patients in the control group (5.8%). The forest plot (Fig. 2D) displays the HRs and their 95% CIs for each study. The pooled analysis demonstrated that GLP-1 RAs significantly reduced myocardial infarction risk (HR: 0.90; 95% CI: 0.82–0.98, P = 0.02), with no notable heterogeneity observed (I² = 28%, P = 0.19). Funnel plot analysis (Supplementary Figure S2D) and Egger’s test showed no significant publication bias (P = 0.350).

Stroke

Nine studies, encompassing 63,613 patients in total, were analyzed to evaluate the impact of GLP-1 RAs on stroke. During the follow-up period, 821 of 32,461 patients in the GLP-1 RAs group (2.5%) experienced a stroke, compared to 939 of 31,152 patients in the control group (3.0%). The forest plot (Fig. 3A) displays the HRs and their 95% CIs for each study. Pooled analysis revealed that GLP-1 RAs significantly lowered stroke risk (HR: 0.85; 95% CI: 0.77–0.95, P = 0.003), with no notable heterogeneity observed (I² = 16%, P = 0.30). Funnel plot analysis (Supplementary Figure S3A) and Egger’s test showed no evidence of significant publication bias (P = 0.931).

Fig. 3

Forest plot of pooled hrs for cardiovascular outcomes: (A) stroke, (B) HF hospitalization, and (C) UA hospitalization

HF hospitalization

Nine studies, encompassing 63,613 patients in total, were analyzed to evaluate the impact of GLP-1 RAs on HF hospitalization. During the follow-up period, 1,218 of 32,461 patients in the GLP-1 RAs group (3.8%) experienced HF hospitalization, compared to 1,329 of 31,152 patients in the control group (4.3%). The forest plot (Fig. 3B) displays the HRs and their 95% CIs for each study. Pooled analysis revealed that GLP-1 RAs significantly lowered HF hospitalization risk (HR: 0.90; 95% CI: 0.83–0.97, P = 0.005), with no notable heterogeneity observed (I² = 0%, P = 0.54). Funnel plot analysis (Supplementary Figure S3B) and Egger’s test showed no significant publication bias (P = 0.587).

UA hospitalization

Seven studies, encompassing 50,617 patients in total, were analyzed to evaluate the impact of GLP-1 RAs on UA hospitalization. During the follow-up period, 861 of 25,963 patients in the GLP-1 RAs group (3.3%) experienced UA hospitalization, compared to 820 of 24,654 patients in the control group (3.3%). The forest plot (Fig. 3C) displays the HRs and their 95% CIs for each study. The pooled analysis indicated that UA hospitalization risk was not significantly different between the GLP-1 RAs and control groups (HR: 1.04; 95% CI: 0.95–1.15, P = 0.38), with no notable heterogeneity observed (I² = 0%, P = 0.84). Funnel plot analysis (Supplementary Figure S3C) and Egger’s test showed no significant publication bias (P = 0.698).

Subgroup analysis

We performed a subgroup analysis on cardiovascular outcomes, categorizing by intervention type, patient characteristics, control group type, follow-up duration, and sample size. The results indicated that monotherapy led to a significant decrease in the risk of all-cause death and stroke, but its effects on cardiovascular death and myocardial infarction were limited. In contrast, combination therapy demonstrated more pronounced risk reductions in cardiovascular death, all-cause death, and HF hospitalization. Among patients with chronic kidney disease, GLP-1 RAs notably decreased the risk of cardiovascular death, all-cause death, and HF hospitalization. In studies with long-term follow-up (≥ 2 years), GLP-1 RAs showed more significant effects in reducing the risks of all-cause death, HF hospitalization, and stroke. Additionally, studies with larger sample sizes demonstrated that GLP-1 RAs provided significant protective effects across multiple cardiovascular outcomes (Table 3).

Table 3A Subgroup analysis of cardiovascular outcomes: primary composite outcome, cardiovascular death, all-cause death, myocardial infarctionTable 3B Subgroup analysis of cardiovascular outcomes: HF Hospitalization, Stroke, UA HospitalizationSensitivity analysis

Sensitivity analyses were conducted for each cardiovascular outcome to assess the influence of individual studies on the pooled HRs. This was achieved by excluding one study at a time and observing the changes in the pooled HRs. The results demonstrated that the exclusion of any single study did not significantly alter the pooled HRs for the composite outcome, cardiovascular death, all-cause death, stroke, HF hospitalization, or UA hospitalization, indicating the robustness of the findings. For myocardial infarction, sensitivity analysis showed that excluding the studies by Hernandez 2018 and Marso 2016a had some impact on the overall effect and heterogeneity, but the overall stability of the results remained consistent. This indicates that GLP-1 RAs generally have consistent effects on myocardial infarction across studies, despite some variability in specific trials (Supplementary Figures S4-S5).

GRADE assessment

In the GRADE assessment, the evidence quality for UA hospitalization was downgraded to “Moderate” due to imprecision, as the HR effect was not statistically significant. For the other cardiovascular outcomes, including the primary composite outcome, all-cause death, cardiovascular death, myocardial infarction, HF hospitalization, and stroke, no serious risk of bias, inconsistency, indirectness, or imprecision was detected, and no publication bias was identified. Reasonable confounding factors did not significantly affect the validity of the results, and the evidence quality for these outcomes was rated as “High” (Table 4).

Table 4 GRADE evidence profile