The present study adhered to the standards of the preferred reporting items for systematic review and meta-analysis (PRISMA) for a meta-analysis and systematic review of randomized clinical trials  (Supplementary Appendix). This meta-analysis and systematic review was prospectively registered in PROSPERO (CRD42019132137).
Compliance with Ethics Guidelines
This article is based on previously conducted studies and does not contain any studies with human participants or animals performed by any of the authors.
Inclusion criteria: (1) randomized controlled clinical trials (RCTs) performed in adults (age ≥ 18 years) with T2DM, of any diabetes duration, geographical distribution, countriy, race, follow-up, sample, and any background treatment administrated to participants; (2) including at least one intervention group and one more control; (3) treatment with long-acting GLP-1RAs as the intervention groups; and (4) designed to evaluate the cardiovascular outcome of GLP-1RAs, such as three-point major adverse cardiovascular events (MACE; cardiovascular death, non-fatal myocardial infarction, or non-fatal stroke) and four-point MACE (including cardiovascular death, myocardial infarction, stroke, or hospitalization for unstable angina death) for the primary outcome.
Exclusion criteria: (1) non-human studies, such as animal studies; (2) studies with no results of cardiovascular outcomes; (3) studies with insufficient data (without data of subpopulation analysis) or with incomplete trials; and (4) reviews, editorials, commentaries, opinion articles, or conference abstracts without original data.
Study Identification and Selection
PubMed, Cochrane Library, Embase, Clinical Trials, Science Direct, and all databases in Web of Science were scrupulously searched up to March 31, 2020 without limits of language. Full search strategies are listed in the Supplementary Appendix. Two reviewers (LH and HZ) independently conducted systematic literature retrievals for RCTs, and disagreements were settled through discussion with a third reviewer. The major search terms and were as follows: “diabetes mellitus, diabetes,” “GLP-1 receptor agonists, exenatide, liraglutide, semaglutide, albiglutide, dulaglutide, taspoglutide,” and “cardiovascular outcome, cardiovascular events, cardiovascular diseases, major adverse cardiovascular events, MACE.”
On the basis of inclusion and exclusion criteria, two reviewers (LH and NY) screened the titles and abstracts of the retrieved literatures independently. The selection process was repeated twice by each reviewer. The references of the identified literatures were manually searched and screened as an important supplement. Discrepancies were resolved through discussions with other team members until a consensus was reached.
Data Extraction and Quality Assessment
Data were extracted independently by two reviewers (LH and HZ), using standardized predefined data extraction forms. The primary efficacy outcome for this analysis was three-point MACE (cardiovascular mortality, non-fatal myocardial infarction, non-fatal stroke), and cardiovascular mortality, non-fatal myocardial infarction, non-fatal stroke, all-cause mortality, and hospitalization for heart failure (HF) were also analyzed individually. Importantly, the characteristics of interest for the stratifications of subpopulations were as follows: age, sex, region, race, diabetes mellitus (DM) duration, history of HF, history of established CVD, body mass index (BMI), glycated hemoglobin levels, and estimated glomerular filtration rate (GFR). The safety endpoints comprised severe hypoglycemia, pancreatitis, pancreatic cancer, serious gastrointestinal events, retinopathy, adverse events (AEs) leading to drug discontinuation, and serious adverse events (SAE). The definition of primary safety endpoints is outlined in the Supplementary Appendix.
The quality of the enrolled RCTs was assessed with the Cochrane risk of bias assessment tool . All trials met the criteria for being performed well, and had a low risk of bias according to the Cochrane tool for assessing risk of bias in randomized clinical trials (Supplementary Appendix).
Hazard ratios (HRs) with 95% confidence intervals (CIs) were synthesized for cardiovascular efficacy outcomes with random-effects models. Meta-regression analyses were performed using the residual maximum likelihood method and Hartung–Knapp adjustment for tests of interactions to assess the differences in cardiovascular efficacy across the subpopulations. Overall odds ratios (ORs) with 95% CIs were applied for safety endpoints, which were obtained through inverse-variance weights for the weighted mean logOR estimate, and then exponentiated. Heterogeneity was quantified with Cochrane Q statistic and I2 statistic, which was judged as low (I2 ≤ 25%), moderate (25% < I2 ≤ 50%), or high (I2 ≥ 75%). Statistical analyses were conducted with Stata statistical software version 15.0 (Stata Corp, College Station, TX, USA), and P values less than 0.05 were considered statistically significant.