Cardiovascular outcomes with glucagon-like peptide-1 receptor agonists in patients with type 2 diabetes mellitus: A systematic review and meta-analysis

Introduction

A total of 30 million adults in the United States were estimated to have diabetes mellitus (DM) in 2015 with cardiovascular disease reported as the major cause of mortality among them.¹³ Glucagon-like peptide-1 receptor agonists (GLP1RAs) are relatively newer anti-hyperglycemic agents, which have been shown to have benefits of reduction in glycated hemoglobin level, a modest decrease in body weight, blood pressure, serum lipid levels, low risk of hypoglycemia with potential cardiovascular benefits.^4,–6 The results from randomized control trials (RCTs) evaluating cardiovascular outcomes of GLP1RAs have been varied. Some studies showed a statistically significant reduction in the primary composite cardiovascular outcomes^7,,–10 while others did not.^11,–13 Prior meta-analyses evaluating the effect of GLP1RAs on cardiovascular outcomes have shown a statistically significant reduction in all-cause mortality and cardiovascular mortality, while most of these studies showed no significant reduction in non-fatal myocardial infarction (MI) or the rate of hospitalization for heart failure.^14,,–17 The last reported meta-analysis showed a significant difference in mortality, major adverse cardiac events (MACE) but did not perform a per patient-year analysis to account for the variation in study duration between RCTs, which ranged from 17 to 64 months.¹⁸ It is also unclear if the cardiovascular benefits of GLP1RAs extend to DM patients with chronic kidney disease (CKD). Therefore, we performed a comprehensive meta-analysis to evaluate the effects of GLP1RAs on cardiovascular outcomes in patients with type 2 DM and also sought to evaluate the benefit of GLP1RAs in patients with CKD and type 2 DM.

Methods

The meta-analysis was performed in accordance with the preferred reporting items for systematic reviews and meta-analysis (PRISMA) guidelines.¹⁹ We searched MEDLINE, Cochrane, ClinicalTrials.gov databases systematically from inception to 15 July 2019, for relevant published articles that assessed the cardiovascular outcomes of GLP1RAs compared with placebo in patients with type 2 DM. The search involved the following keywords: ‘Glucagon-like peptide-1 receptor agonist’ OR ‘GLP1-RA’ OR ‘Dulaglutide’ OR ‘Exenatide’ OR ‘Liraglutide’ OR ‘Lixisenatide’ OR ‘Semaglutide’ OR ‘albiglutide’ AND ‘mortality’ or ‘myocardial infarction’ or ‘stroke’ or ‘cardiovascular outcomes’ AND ‘diabetes’ or ‘diabetes mellitus’. We also reviewed the reference lists in the retrieved articles, relevant reviews and prior meta-analyses for additional published data.

Inclusion criteria consisted of RCTs that met the following criteria: (a) studies that included patients with type 2 DM; (b) studies that compared GLP1RAs with placebo; (c) studies that reported outcomes including but not limited to mortality, primary composite endpoint, non-fatal MI and non-fatal stroke. Articles published in languages other than English, abstracts, case reports, clinical guidelines, and consensus documents were excluded. Studies not performed on humans were excluded. The authors (VPP and PP) screened articles for inclusion independently, retrieved potentially relevant articles and determined their eligibility. Relevant full-text articles were finally included in the analysis.

Using a standardized protocol, all relevant information from each article was extracted by two authors (VPP and PP). We recorded the following study characteristics: name of the study, name of the first author, year of publication, study design, intervention drug, number of subjects in each group, age, gender distribution, baseline characteristics, baseline comorbidities, duration of follow-up and outcomes including, but not limited to, all-cause mortality, cardiovascular mortality, primary composite endpoint, non-fatal MI and non-fatal stroke. The quality of studies used in the analysis was assessed using the Cochrane Collaboration risk of bias tool.²⁰ Two authors (VPP and VR) independently assessed study quality. Any disagreements were resolved by consensus among the authors.

All-cause mortality and cardiovascular mortality were included in the analysis as reported by the studies. The primary composite endpoint was defined to include cardiovascular mortality, non-fatal MI and non-fatal stroke as reported by the studies. Non-fatal MI and non-fatal stroke were included as defined and reported by the individual studies. If the non-fatal MI and stroke numbers were not specified separately, then the number of MI and stroke as reported was included in the analysis. CKD included two subgroups of patients with estimated glomerular filtration rate (eGFR) less than 60 ml/min/1.73 m² and eGFR less than 30 ml/min/1.73 m².

The principal summary effects measures were the odds ratio (OR) and corresponding 95% confidence intervals (CIs) estimated by using Mantel–Haenszel random effects model.²¹ A two-sided P value less than 0.05 was considered statistically significant. We conservatively used a priori, the Mantel–Haenszel random effects model assuming substantial variability in the treatment effect size across studies.²¹ The presence of statistical heterogeneity was evaluated by Cochrane’s Q test I² statistic: when I² values were greater than 50%, we planned to explore individual study characteristics.²² Publication bias was assessed using the funnel plot.²³ We also planned a priori subgroup analysis of type 2 DM with eGFR less than 60 ml/min/1.73 m² and eGFR less than 30 ml/min/1.73 m². We also performed a per patient-year analysis with the use of the generic inverse variance method to account for variation in the duration of follow-up between the studies.²⁰ Statistical analysis was performed using Review Manager (RevMan), version 5.3, the Nordic Cochrane Centre, the Cochrane Collaboration, Copenhagen, 2014.

Results

We retrieved a total of 456 articles from the title and abstract analysis. Following further review, a total of 180 articles were excluded due to studies published in languages other than English and studies not performed in humans. The remaining 276 articles were carefully screened, and seven articles were deemed relevant for the purpose of our meta-analysis as shown in Figure 1.

Figure 1.

PRISMA flow chart for the selection of studies included in the meta-analysis.

Open in new tabDownload slide

Seven RCTs^7,,,,,–13 including a total of 56,004 adult patients with type 2 DM were included in this meta-analysis. Among these patients, 27,977 received GLP1RAs and 28,027 received placebo. Baseline characteristics were variable among the included studies as shown in Table 1. The ELIXA trial¹¹ included patients with type 2 DM and recent acute coronary syndrome. Most of the studies, except for ELIXA,¹¹ defined the primary composite endpoint as an outcome of cardiovascular mortality, non-fatal MI, and non-fatal stroke. The HARMONY trial⁸ reported all MI and all strokes, including both fatal and non-fatal together, but did not provide an individual breakdown. HARMONY,⁸ ELIXA,¹¹ PIONEER-6¹² and EXSCEL¹³ excluded patients with eGFR less than 30 ml/min/1.73 m² while the REWIND trial¹⁰ excluded patients with eGFR less than 15 ml/min/1.73 m². We noted another study, the FREEDOM-CVO (ITCA) trial,²⁴ which compared ITCA 650 (exenatide) with placebo. The preliminary data of FREEDOM-CVO (ITCA)²⁴ suggested the non-inferiority of the drug but the trial could not be included in our meta-analysis as the study was not published.

In the primary pooled analysis, GLP1RAs demonstrated a statistically significant reduction in the outcomes of all-cause mortality (OR 0.88, 95% CI 0.82–0.95; P < 0.001), cardiovascular mortality (OR 0.88, 95% CI 0.81–0.96; P = 0.004), the primary composite endpoint (OR 0.86, 95% CI 0.80–0.91; P < 0.001) and non-fatal stroke (OR 0.86, 95% 0.77–0.95; P = 0.004). There was no statistical difference in the outcome of non-fatal MI (OR 0.92, 95% CI 0.83–1.01; P = 0.09). Sensitivity analysis demonstrated the robustness of the results. The significant difference in the outcomes of all-cause mortality, cardiovascular mortality, primary composite endpoint, and non-fatal stroke did not dissipate with the sequential exclusion of individual trials. The results were similar to primary pooled analysis during sequential exclusion of the ELIXA trial¹¹ which used short-acting GLP1RAs and PIONEER 6¹² which included the oral formation of GLP1RAs. The forest plot of the effect on the outcomes is shown in Figures 2 and 3. There was no significant heterogeneity in the outcomes of non-fatal stroke, all-cause mortality, and cardiovascular mortality. There was moderate heterogeneity with an I² estimate of 30–50% for outcomes of the primary composite endpoint and non-fatal MI. For the outcome of non-fatal MI, heterogeneity resolved after removal of data from the HARMONY trial⁸ (OR 0.98, 95% CI 0.88–1.03; P = 0.27).

Figure 2.

Forest plot of the effect of glucagon-like peptide-1 receptor agonists (GLP1RAs) versus placebo in patients with type 2 diabetes mellitus. (a) All cause mortality, (b) Caridovascular mortality and (c) Primary composite end point*-Major adverse cardiac events.

Data are events in each group and weighted risk ratios. The horizontal line is the 95% confidence interval (CI). The diamond shape is the pooled mean difference of all studies.

Open in new tabDownload slide

Figure 3.

Forest plot of the effect of glucagon-like peptide-1 receptor agonists (GLP1RAs) versus placebo in patients with type 2 diabetes mellitus. (a) Non fatal stroke (b) Non fatal myocardial infarction, (c) CKD patients with GFR <60 – Primary composite endpoint* and (d) CKD patients with GFR <30 –30 Primary composite endpoint*.

Data are events in each group and weighted risk ratios. The horizontal line is the 95% confidence interval (CI). The diamond shape is the pooled mean difference of all studies.

Open in new tabDownload slide

Per patient-year analysis performed using the generic inverse variance method yielded results similar to the primary pooled analysis. In comparison to placebo, GLP1RAs continued to demonstrate a statistically significant reduction for the outcomes of all-cause mortality (OR 0.88, 95% CI 0.82–0.95; P = 0.001), cardiovascular mortality (OR 0.89, 95% CI 0.82–0.97; P = 0.005), primary composite endpoint (OR 0.87, 95% CI 0.82–0.93; P < 0.001) and non-fatal stroke (OR 0.86, 95% CI 0.78–0.95; P = 0.004). There was no difference for the outcome of non-fatal MI (OR 0.92, 95% CI 0.84–1.02; P = 0.11). The results of the per patient-year analysis are shown in Figures 4 and 5.

Figure 4.

Forest plot of per patient-year analysis comparing the effect of glucagon-like peptide-1 receptor agonists (GLP1RAs) versus placebo in the patients with type 2 diabetes mellitus. (a) All cause mortality and (b) Cardiovascular mortality.

Data are the natural logarithm of risk rate ratio and standard error. The horizontal line is the 95% confidence interval (CI). The diamond shape is the pooled mean difference of all studies.

Open in new tabDownload slide

Figure 5.

Forest plot of per patient-year analysis comparing the effect of glucagon-like peptide-1 receptor agonists (GLP1RAs) versus placebo in the patients with type 2 diabetes mellitus. (a) Primary composite end point* -Major adverse cardiac events, (b) Non fatal myocardial infarction and (c) Non fatal stroke.

Data are the natural logarithm of risk rate ratio and standard error. The horizontal line is the 95% confidence interval (CI). The diamond shape is the pooled mean difference of all studies.

Open in new tabDownload slide

The subgroup analysis of patients with eGFR less than 60 ml/min/1.73 m² included 4682 patients on GLP1RAs and 4684 patients on placebo. There was no statistically significant difference in the outcome of the primary composite endpoint between GLP1RAs and placebo (OR 0.85, 95% CI 0.70–1.04; P = 0.12). The subgroup analysis of patients with eGFR less than 30 ml/min/1.73 m² included a total of 171 patients on GLP1RAs and 174 patients on placebo. There was no statistical difference in the outcome of the primary composite endpoint between GLP1RAs and placebo (OR 0.82, 95% CI 0.48–1.39; P = 0.46). There was significant heterogeneity with I² estimate greater than 50% for the outcome of CKD with eGFR less than 60 ml/min/1.73 m². Exploratory analysis revealed that the heterogeneity resolved when results of the LEADER trial⁷ were removed from the analysis (OR 0.97, 95% CI 0.84–1.11; P = 0.65). The LEADER trial⁷ included a higher number of patients with eGFR less than 30 ml/min/1.73 m² compared to other studies. There was no significant heterogeneity for the primary composite endpoint in CKD with eGFR less than 30 ml/min/1.73 m². The forest plot of the effect on the outcomes in CKD is shown in Figure 3.

The Cochrane risk of bias assessment of RCTs included in this meta-analysis is shown in Supplementary Figure 1. The funnel plot demonstrated evidence of publication bias, probably due to the inclusion of pertinent RCTs only to improve the quality of the meta-analysis, as shown in Supplementary Figure 2.

Discussion

Our meta-analysis demonstrated a statistically significant reduction in the outcomes of all-cause mortality, cardiovascular mortality, primary composite endpoint, and non-fatal stroke with GLP1RAs compared to placebo. There was no statistical difference in the outcome of non-fatal MI. In the subgroup analysis of patients with CKD, there was no statistical difference in the primary composite endpoint outcomes. Our study is the most comprehensive meta-analysis to date on GLP1RAs incorporating per patient-year analysis of seven RCTs. Moreover, no meta-analysis in the current literature evaluated the cardiovascular outcomes of GLP1RAs in type 2 DM patients with eGFR less than 60 ml/min/1.73 m² and eGFR less than 30 ml/min/1.73 m².

There was a 12% relative reduction in all-cause mortality and cardiovascular mortality with GLP1RAs compared with placebo, while the absolute risk reduction was 0.84% and 0.55%, respectively, which is consistent with prior meta-analyses.^6,14,15 In a study by Monami et al.,¹⁶ which included data from 113 trials including RCTs with non-cardiovascular endpoints, GLP1RAs showed a reduction in all-cause mortality, cardiovascular mortality, and incidence of MI. The recent meta-analysis by Giugliano et al.¹⁸ which included the seven RCTs as in our study showed similar results in the primary pooled analysis.

We also demonstrated a 14% relative reduction and 1.43% absolute risk reduction in the primary composite endpoint. There was a statistically significant reduction in non-fatal stroke; however, the difference in non-fatal MI did not reach statistical significance. The results of prior meta-analyses have been variable in this respect. In the study by Jia et al.¹⁵ there was no significant difference in MACE, non-fatal MI, non-fatal stroke, hospitalization for heart failure, and coronary revascularization. However, studies by Bethel et al.⁶ and Zhang et al.¹⁷ showed a reduction in the composite adverse cardiovascular outcome. We propose that the significant difference detected by our analysis is likely to be due to the increased power from the inclusion of additional newer RCTs. We did not include the results of the ELIXA trial¹¹ in the analysis of primary composite endpoint due to heterogeneity in the definition of the outcome. In spite of this exclusion, our results continue to be consistent with the meta-analysis by Giugliano et al.,¹⁸ which evaluated all seven trials. The results of our per patient-year analysis, which was congruous with the primary pooled analysis, also add to the robustness of our study, making it unique from the previously performed meta-analyses.

Our study demonstrated the lack of a significant difference in the primary composite endpoint in patients with CKD. Although Zhang et al.¹⁷ showed that GLP1RAs numerically reduced the rates for nephropathy, studies evaluating cardiovascular outcomes with GLP1RA use in patients with type 2 DM and CKD are limited. A Cochrane review by Lo et al.²⁵ included only two trials with a total of 283 patients in comparison to the 9366 patients included in our analysis. The study by Lo et al.²⁵ did not show a statistical difference in the outcomes of all-cause mortality and MI, but a meaningful conclusion could not be drawn due to the small sample size. However, our results suggest the use of caution while prescribing GLP1RAs for the purpose of cardiovascular outcome benefit in patients with CKD. It is unclear if the pharmacological characteristics of GLP1RAs make them unfavorable in patients with CKD. It is also possible that the lack of benefit seen in CKD patients may be due to inadequate sample size in the included studies. There was also significant heterogeneity which could have affected the results. These conclusions are hypothesis generating, advocating further trials to evaluate the role of GLP1RAs in patients with CKD.

Although our results demonstrated a cardiovascular outcome benefit of GLP1RAs in patients with type 2 DM, it is unclear if the benefit is a class effect or specific to individual agents. Among the studies reviewed in this meta-analysis, short-acting GLP1RA lixisenatide showed no significant difference in MACE in the ELIXA trial.¹¹ Long-acting GLP1RAs were used in the remaining six RCTs. The analysis of the six RCTs after removal of ELIXA¹¹ yielded similar results in all outcomes compared to the primary pooled analysis. The evaluation of the above studies suggests that the cardiovascular benefit of GLP1RAs may be a class effect rather than being an effect of individual drug profile.

We acknowledge several limitations. The baseline characteristics of the patients and follow-up duration in the included studies were heterogeneous. Results were analyzed on trial-level data, but not on individual patient-level data. We were not able to include the FREEDOM-CVO (ITCA) trial²⁴ due to the unavailability of the complete study results which could have biased our results. PIONEER-6¹² evaluated an oral formulation of GLP1RAs while the remaining studies used subcutaneous formulations. There was variation in follow-up duration between the trials; however, per patient-year analysis yielded similar results to the primary pooled analysis. There was evidence of publication bias, which was expected in a study including only RCTs. There was significant heterogeneity in the outcomes of the primary composite endpoint in CKD patients with eGFR less than 60 ml/min/1.73 m² which might have biased the results. Differences in the trials including the design, analysis, outcome definitions, exclusion of CKD patients in some trials, the pharmacokinetics of various GLP1RAs, and to what extent these affect the results is unclear.

Conclusion

Our meta-analysis of seven major RCTs comparing GLP1RAs with placebo in adult type 2 DM patients showed a significant reduction in the outcomes including all-cause mortality, cardiovascular mortality, primary composite endpoint, and non-fatal stroke. There was no statistical difference observed in the outcomes of non-fatal MI in all patients and primary composite endpoint outcomes in patients with CKD including eGFR less than 60 ml/min/1.73 m² and eGFR less than 30 ml/min/1.73 m². GLP1RAs have an overall favorable cardiovascular outcome along with glucose-lowering properties in adult patients with type 2 DM. Further studies are needed to evaluate the cardiovascular benefits in patients with type 2 DM and CKD.

References