IJCA-24963; No of Pages 8 International Journal of Cardiology xxx (2017) xxx–xxx

Contents lists available at ScienceDirect

International Journal of Cardiology

journal homepage: www.elsevier.com/locate/ijcard

Effects of -like -1 receptor on mortality and cardiovascular events: A comprehensive meta-analysis of randomized controlled trials

Matteo Monami a,StefaniaZannonia, Laura Pala a, Antonio Silverii a, Francesco Andreozzi b, Giorgio Sesti b, Edoardo Mannucci a,⁎ a Diabetology, Azienda Ospedaliero-Universitaria Careggi and University of Florence, Florence, Italy b Department of Medical and Surgical Sciences, University Magna Graecia of Catanzaro, Catanzaro. Italy article info abstract

Article history: Introduction: The publication of the results of LEADER and SUSTAIN-6 trials suggested a possible beneficial effect Received 7 January 2017 of the class of GLP-1 receptor agonists on cardiovascular morbidity and mortality. The aim of the present meta- Received in revised form 4 March 2017 analysis is to collect and synthetize all available evidence on the effect of GLP-1 receptor agonists on cardiovas- Accepted 29 March 2017 cular events and mortality. Available online xxxx Methods: A Medline search for GLP-1 receptor agonists (, , lixisenatide, , , or ) was performed, collecting all randomized clinical trials with a duration Keywords: N11 weeks, enrolling patients with , and comparing a GLP-1 receptor with placebo or Metanalysis Cardiovascular events any other non-GLP-1 receptor agonist drug. The principal outcome of this analysis was the effect of GLP-1 recep- Mortality tor agonists on all-cause and cardiovascular mortality, overall (fatal plus nonfatal) myocardial infarction, stroke, Glucagon-like peptide-1 receptor agonists and heart failure. Results: Out of 113 trials fulfilling inclusion criteria (mean duration 41.7 ± 38.2 weeks), 32, 25, 48, 43 and 32 re- ported at least one event for all-cause and cardiovascular mortality, overall (fatal plus nonfatal) myocardial in- farction, stroke, and heart failure, respectively. In GLP-1 receptor agonist-treated patients, all-cause mortality, cardiovascular mortality, and myocardial infarction were significantly lower than in comparators (MH-OR [95% CI] 0.88 [0.79–0.97], p = 0.015, 0.84 [0.74–0.96], p = 0.009, and 0.90 [0.80–1.00], p = 0.050, respectively), whereas no beneficial effect was observed for stroke and heart failure (MH-OR [95% CI] 0.90 [0.81–1.00]. p = 0.059. 0.89 [0.76–1.04]. p = 0.15. and 0.92 [0.81–1.06]. p = 0.25. respectively). Conclusions: Overall, the agents of this class appear to reduce all-cause mortality, cardiovascular mortality, and the incidence of myocardial infarction at mid-term follow up. © 2017 Elsevier B.V. All rights reserved.

1. Introduction major cardiovascular events was significantly reduced in the active treatment group. In addition, the trial with liraglutide showed a signifi- Three cardiovascular outcome studies with different Glucagon-Like cant reduction in all-cause and cardiovascular mortality, whereas that Peptide-1 (GLP-1) receptor agonists were recently published [1–3].All with semaglutide reported a significant reduction in the incidence of the three trials reached their principal endpoint, i.e. the demonstration stroke [1–3]. of non-inferiority versus placebo with respect to major cardiovascular These results raised important questions about the possibility of a events, thus confirming the safety of the experimental drugs. However, class effect of GLP-1 receptor agonists on cardiovascular risk. In fact, in one [1] of the trials (with lixisenatide), no difference across treatment the populations enrolled in the three trials were notably different, groups was observed for the principal endpoint or any pre-defined sec- with trials with liraglutide and semaglutide including a majority of sub- ondary endpoint, whereas in the other two studies the incidence of jects with established (non-recent) cardiovascular disease, and the study with lixisenatide enrolling patients with a recent coronary event [1–3]. On the other hand, the three molecules differ for kinetic and chemical structure: lixisenatide is a short-acting analogue of exenatide, ⁎ Corresponding author at: Diabetology, Azienda Ospedaliero-Universitaria Careggi, Via with a low homology to human GLP-1 [4], whereas both liraglutide and delle Oblate 4, 50141 Florence, Italy. E-mail addresses: matteo.monami@unifi.it (M. Monami), edoardo.mannucci@unifi.it semaglutide are long-acting GLP-1 receptor agonists, with an amino (E. Mannucci). acid sequence almost identical to that of human GLP-1 [4].

http://dx.doi.org/10.1016/j.ijcard.2017.03.163 0167-5273/© 2017 Elsevier B.V. All rights reserved.

Please cite this article as: M. Monami, et al., Effects of glucagon-like peptide-1 receptor agonists on mortality and cardiovascular events: A comprehensive meta-analysis of rand..., Int J Cardiol (2017), http://dx.doi.org/10.1016/j.ijcard.2017.03.163 2 M. Monami et al. / International Journal of Cardiology xxx (2017) xxx–xxx

The aim of the present meta-analysis is to collect and synthetize all 3. Results available evidence on the effect of GLP-1 receptor agonists on cardiovas- cular events and mortality, including data from randomized trials with Out of 1.147 and 532 items identified through MEDLINE/Embase, non-cardiovascular endpoints, in order to improve our insight on the www.clinicaltrials.gov and FDA/EMA websites, respectively, 113 trials cardiovascular effects of these molecules. were selected, as summarized in Fig. 1 of Supplementary materials. The quality of trials (all with intention-to-treat analysis) was generally fi 2. Materials and methods good (Table 1 Supplementary materials). The trials ful lling the inclu- sion criteria enrolled 33,167 and 26,683 patients in GLP-1 receptor ago- This analysis is part of a larger systematic review, the protocol of which nist and comparator arms, respectively, with a mean duration of (CRD42015020245) was published on the University of York (Centre for Reviews and treatment of 41.7 ± 38.2 weeks. The main characteristics of the selected Dissemination) website [5]. trials, and the outcomes of interest in each study, are reported in Table 1and Table 1 of Supplementary materials. The search of www. 2.1. Data sources and searches clinicaltrials.gov website allowed the identification of 26 unpublished

A Medline/Embase search for GLP-1 receptor agonists (exenatide, liraglutide, and undisclosed, although completed, trials (Table 2 Supplementary lixisenatide, albiglutide, dulaglutide, or semaglutide) was performed, collecting all ran- materials). domized clinical trials on humans published in English up to September 15th. 2016. The identification of relevant abstracts, the selection of studies based on the criteria described 3.1. All-cause mortality below, and the subsequent data extraction were performed independently by two of the fl authors (S.Z., M.M.), and con icts resolved by the third investigator (E.M.). Completed fi but still unpublished trials were identified through a search of www.clinicaltrials.gov Of the 113 trials ful lling the inclusion criteria, 14 did not report website, using the same keywords. In addition, for approved drugs, Medical Reviews information on all-cause mortality, whereas 67reported zero events in were retrieved from the Food and Drug Administration (FDA) website [6], and the all treatment groups. The principal analysis was therefore performed Summary of Product Characteristics from the European Medicines Agency (EMA) website on 32 trials, enrolling 20,280 and 16,939 patients in GLP-1 receptor ag- [7], for the identification of further unpublished and otherwise undisclosed trials. onist and comparator arms, respectively. The number of reported deaths was 720 (3.6%) for GLP-1 receptor agonists and 785 (4.6%) for 2.2. Study selection comparators. I2 was b0.001, suggesting no relevant heterogeneity. Funnel plot A meta-analysis was performed including all randomized clinical trials with a dura- tion of treatment of at least 12 weeks, enrolling patients with type 2 diabetes, comparing analysis (Fig. 2 Supplementary materials) and Kendall's tau a GLP-1 receptor agonist with placebo or any other non-GLP-1 receptor agonist drug, pro- (−0.08; p = 0.51) did not suggest any relevant publication bias. vided that concurrent treatment was the same for all treatment arms, and that the doses of In GLP-1 receptor agonist-treated patients, all-cause mortality GLP-1 receptor agonist were among those approved by FDA and/or EMA. was significantly lower than in comparators (MH-OR [95% CI] 0.88 [0.79–0.97]. p = 0.015; Fig. 3 Supplementary materials). This result 2.3. Data extraction and quality assessment was confirmed in the sensitivity analysis with continuity correction (MH-OR 0.88 [0.79–0.96]. p = 0.012). When trials with different Results of trials were retrieved from the primary publication and, if needed, from other publications referring to the same trial. When information were unavailable on pub- molecules were analysed separately, the difference in mortality versus lished papers, data were retrieved (in this hierarchical order) from FDA Medical Reviews comparators was significant only with dulaglutide and liraglutide; EMA Summaries of Product Characteristics, www.clinicaltrials.gov study results, and however, when the trials with a cardiovascular endpoint were exclud- trial results on manufacturers' company websites. Data retrieval was performed by two ed, the result for liraglutide was no longer statistically significant fl of the investigators (A.S. and L.P.), and con icts resolved by a third investigator (E.M.). Re- (Fig. 1). trieved data included all outcomes reported below, plus the main features of each trial (concurrent therapy, principal endpoint, baseline characteristics of enrolled patients A subgroup analysis was performed for trials with different compar- (age, BMI, duration of diabetes, HbA1c), and effects of treatment on HbA1c and BMI. The ators (Fig. 2), showing a significant reduction of all-cause mortality only quality of trials was assessed using the Cochrane Collaboration's Tool for Assessing Risk in placebo-controlled trials, driven by the cardiovascular outcome stud- of Bias in randomized controlled trials; quality was not used as a criterion for the selection ies. When trials with a duration of treatment ≥52 weeks (n =20)were of trials, but only for descriptive purposes. analysed separately, MH-OR was 0.88[0.79–0.98], p =0.023. Meta-regression analyses did not detect a significant effect on all- 2.4. Data synthesis and analysis cause mortality of any of the putative moderators explored (Table 3 Supplementary materials). The principal outcome of this analysis was the effect of GLP-1 receptor agonists, com- pared with placebo or other active drugs, on all-cause and cardiovascular mortality, overall (fatal plus nonfatal) myocardial infarction, stroke, and heart failure (HF). For the latter out- 3.2. Cardiovascular mortality come, hospitalization for HF was considered whenever available; when that information was not reported, heart failure reported as serious treatment-emergent adverse Information on cardiovascular mortality was available for 83 trials, event was considered. The composite endpoint of major cardiovascular events (MACE) 2 b was not considered, because it is not usually reported as such in trials with non- 25 of which with at least one event. I was 0.001, suggesting no cardiovascular principal endpoints, which were the majority of available studies. Hetero- relevant heterogeneity. Funnel plot analysis (Fig. 4 Supplementary geneity (on all-cause mortality) was assessed by using I2 statistics. In order to estimate materials) and Kendall's tau (−0.05; p = 0.73) did not suggest any possible publication/disclosure bias we used funnel plots and the Begg adjusted rank cor- relevant publication bias. relation test [8,9], including published and unpublished, but disclosed, trials. Considering The analysis was therefore performed on 25 trials, enrolling 16,656 the differences across trials in molecules, treatment schedules, inclusion criteria, and length of follow-up, a random-effects model was applied, calculating Mantel-Haenszel and 15,175 patients in GLP-1 receptor agonist and comparator arms, re- odds ratio with 95% Confidence Interval (MH-OR) for all the events defined above, on an spectively. The number of reported cardiovascular deaths was 438 intention-to-treat basis, excluding trials with zero events. For all the principal endpoints, (2.6%) for GLP-1 receptor agonists and 514 (3.4%) for comparators. Car- a sensitivity analysis was performed with continuity correction, in order to avoid distor- diovascular mortality was significantly reduced by GLP-1 receptor ago- tions due to the exclusion of trials with zero events. Subgroup analyses were performed – for all endpoints for different drugs of the class, different classes of comparators, and trials nists (MH-OR [95% CI] 0.84 [0.74 0.96]. p = 0.009; Fig. 5 Supplementary with cardiovascular and non-cardiovascular endpoints. A post-hoc analysis was per- materials). This result was confirmed by the sensitivity analysis with formed on all principal endpoints selecting only trials with a duration of at least continuity correction (MH-OR [95% CI] 0.84 [0.74–0.95]. p = 0.007). 52 weeks. In addition, a post-hoc meta-regression analysis was performed, exploring the In separate analyses for different GLP-1 receptor agonists, none of moderating effect of mean age, duration of diabetes, HbA1c, fasting glucose, and BMI at the molecules of the class reached a statistically significant effect on car- study entry, as well as reduction of HbA1c and BMI versus comparators. All analyses were performed using Comprehensive Meta-analysis Version 2.Biostat. (Englewood. NJ. diovascular mortality (Fig. 1). Differences from any class of active com- USA).The meta-analysis was reported following the PRISMA checklist [10]. parators did not reach statistical significance (Fig. 2). On the other hand,

Please cite this article as: M. Monami, et al., Effects of glucagon-like peptide-1 receptor agonists on mortality and cardiovascular events: A comprehensive meta-analysis of rand..., Int J Cardiol (2017), http://dx.doi.org/10.1016/j.ijcard.2017.03.163 M. Monami et al. / International Journal of Cardiology xxx (2017) xxx–xxx 3

All-cause mortality Cardiovascular mortality

0.01 0.10 1.0 10.0 N trials MH-OR, 95% CI p 0.01 0.10 1.0 10.0 100.0 N trials MH-OR, 95% CI p

Overall 32 0.88 [0.79; 0.98] 0.017 24 0.84 [0.74; 0.96] 0.012

Including CVOTs Albiglutide 5 0.95 [0.50; 1.79] 0.87 6 0.74 [032; 1.71] 0.48

Dulaglutide 4 0.31 [0.10; 0.97] 0.044 4 0.29 [0.07; 1.25] 0.096

Exenatide b.i.d. 5 1.01 [0.39; 2.62] 0.98 5 0.83 [0.22; 3.17] 0.79

Exenatide OW 1 0.30 [0.01; 7.30] 0.45 1 2.97 [0.12; 73.4] 0.57

Liraglutide 8 0.84 [0.73; 0.96] 0.014 3 0.48 [0.15; 1.60] 0.23 Semaglutide 1 1.03 [0.72; 1.49] 0.85 1 0.96 [0.63; 1.45] 0.83

Lixisenatide 9 0.93 [0.77; 1.13] 0.47 5 0.97 [0.78; 1.22] 0.82

Excluding CVOTs

Liraglutide 7 0.75 [0.26; 2.17] 0.59 2 0.13 [0.02; 1.05] 0.056

Lixisenatide 8 0.68 [0.20; 2.23] 0.52 4 0.44 [0.07; 2.79] 0.38

Favours Favours Favours Favours GLP-1RA comparators GLP-1RA comparators

Fig. 1. MH-OR for all-cause and cardiovascular mortality with 95% CI with different molecules, considering all randomized clinical trials with and without cardiovascular endpoints. cardiovascular mortality was significantly reduced by GLP-1 receptor 3.4. Stroke agonists in placebo-controlled trials; the difference was significant in trials with non-cardiovascular endpoints, but not in those with cardio- Information on stroke was reported by 92 studies, 43 of which with vascular endpoints (Fig. 2). at least one event. I2 was b0.001, suggesting no relevant heterogeneity. When trials with a duration of treatment ≥52 weeks (n =14)were Funnel plot analysis (Fig. 10 Supplementary materials) and Kendall's analysed separately, MH-OR was 0.85[0.74–0.97], p = 0.017. Meta- tau (−0.09; p = 0.40) did not suggest any relevant publication bias. regression analyses did not reveal any relevant effect of any of the The principal analysis was therefore performed on 43 trials, enrolling moderators listed above (Table 3 Supplementary materials), with the 22,114 and 19,436 patients in GLP-1 receptor agonist and comparator exception of a non-significant trend for BMI at enrolment (Fig. 6 Supple- arms, respectively. The number of reported stroke was 313 (1.4%) for mentary materials). GLP-1 receptor agonists and 340 (1.7%) for comparators. The incidence of stroke with GLP-1 receptor agonists was not signif- 3.3. Myocardial infarction icantly different from that of comparator groups, neither in the main analysis (MH-OR [95% CI] 0.89 [0.76–1.04], p = 0.16; Fig. 11 Supple- Information on myocardial infarction was reported for 91 trials, 48 of mentary materials) nor in the sensitivity analysis with continuity cor- which with at least one event.I2 was b0.001, suggesting no relevant het- rection (MH-OR [95% CI] 0.88 [0.76–1.02], p = 0.088). When trials erogeneity. Funnel plot analysis (Fig. 7 Supplementary materials) and with a duration of treatment ≥52 weeks (n = 20) were analysed sepa- Kendall's tau (0.13; p = 0.20) did not suggest any relevant publication rately, MH-OR was 0.89[0.75–1.04], p =0.14. bias. Trials reporting at least one event enrolled 22,990 and 19,412 pa- No significant differences were observed in subgroup analyses with tients in GLP-1 receptor agonist and comparator arms, respectively. different molecules (with the only exception of semaglutide) or with The number of reported myocardial infarction was 674 (2.9%) for GLP- different comparators (Figs. 3 and 4). No significant effect of moderators 1 receptor agonists and 725 (3.7%) for comparators. was detected in meta-regression analyses (Table 3 Supplementary In the principal analysis, the effect of GLP-1 receptor agonists on the materials). incidence of myocardial infarction reached a marginal statistical signif- icance (MH-OR [95% CI] 0.90 [0.80–1.00], p = 0.050; Figure 8 Supple- mentary materials); a significant difference was observed in the 3.5. Heart failure sensitivity analysis with continuity correction (MH-OR [95% CI] 0.89 [0.80–0.99]. p = 0.039). None of the molecules of the class, when Information on heart failure was reported by 80 studies, 32 of which analysed separately, was associated with a significant reduction in the with at least one event. I2 was b0.001, suggesting no relevant heteroge- incidence of myocardial infarction (Fig. 3). A non significant trend neity. Funnel plot analysis (Fig. 12 Supplementary materials) and toward lower incidence of myocardial infarction with GLP-1 receptor Kendall's tau (−0.04; p = 0.77) did not suggest any relevant publica- agonists was observed in placebo-controlled studies; however, tion bias. Trials reporting at least one event enrolled 18,861 and between-group difference did not reach statistical significance neither 16,082 patients in GLP-1 receptor agonist and comparator arms, respec- in studies with non-cardiovascular endpoints, nor in those with cardio- tively. The number of reported heart failure was 427 (2.3%) for GLP-1 re- vascular endpoint. In addition, no significant difference was detected in ceptor agonists and 452 (2.8%) for comparators. comparison with any class of active comparators (Fig. 4). In a separate GLP-1 receptor agonists did not appear to produce any effect on analysis, including only trials with a duration of treatment ≥52 weeks heart failure (MH-OR [95% CI] 0.92 [0.81–1.06]. p = 0.25; Fig. 13 Supple- (n = 21) were analysed separately, MH-OR was 0.90 [0.81–1.01], p = mentary materials); this result was confirmed by the analysis with con- 0.083. tinuity correction (MH-OR [95% CI] 0.92 [0.81–1.06], p = 0.24). In a In meta-regression analyses, a higher mean BMI at enrolment was separate analysis, including only trials with a duration of treatment associated with a greater beneficial effect of GLP-1 receptor agonists ≥52 weeks (n = 19) were analysed separately, MH-OR was 0.92 on myocardial infarction (Fig. 9 Supplementary materials); non- [0.80–1.06], p =0.25. significant trends were observed also for higher HbA1c at baseline and No significant differences were observed in subgroup analyses with greater HbA1c reduction (Table 3 Supplementary materials). different molecules or with different comparators (Figs. 3 and 4). No

Please cite this article as: M. Monami, et al., Effects of glucagon-like peptide-1 receptor agonists on mortality and cardiovascular events: A comprehensive meta-analysis of rand..., Int J Cardiol (2017), http://dx.doi.org/10.1016/j.ijcard.2017.03.163 4 M. Monami et al. / International Journal of Cardiology xxx (2017) xxx–xxx

Table 1 Description of the outcomes considered in the present meta-analysis.

First author (year)a Comparator Dose GLP-1RA Trial duration Number of All cause Cardiovascular Myocardial Stroke Heart patients death death infarction failure

Albiglutide Drug Mg weeks ID Comp. ID Comp. ID Comp. ID Comp. ID Comp. ID Comp

Ahrén 2014 (1) 50 104 302 307 4 6 1 1 5 2 2 0 1 0 Placebo 302 101 4 1 1 1 5 2 2 1 1 0 302 302 4 2 1 2 5 2 2 0 1 0 Rosenstock 2014 (2) Lispro 30 26 285 281 2 1 1 0 3 1 2 0 0 0 Seino 2014 (3) Placebo 30 16 54 52 0 0 0 0 0 0 0 1 0 0 Lepore 2016 (4) Placebo 15–30 12 40 30 0 0 0 0 0 0 0 0 0 1 NCT01733758 (5) Placebo 30–50 26 310 77 3 0 0 0 0 0 0 0 0 0 Leiter 2014 (6) Sitagliptin 30 52 249 246 4 4 2 2 1 1 2 4 0 1 Reusch 2014 (7) Placebo 30 156 150 151 3 3 1 2 2 1 0 0 0 1 Nauck 2016 (8) Placebo 30–50 52 200 101 4 0 0 0 0 2 1 1 2 2 Weissman 2014 (9) Glargine 30 156 504 241 12 5 3 3 6 2 2 0 2 2 Home 2015 (10) 30–50 156 271 277 3 6 2 2 1 2 0 4 0 4 Placebo 30–50 156 271 115 3 3 2 1 1 1 0 2 0 1 Dulaglutide Blonde 2015 (11) Glargine 0.75–1.5 52 588 296 2 3 0 3 2 6 5 2 1 1 NCT01648582 (12) Glargine 0.75–1.5 52 516 254 1 0 NR NR 0 0 2 1 1 0 Miyagawa 2015 (13) Placebo 0.75 52 280 70 0 0 0 0 0 0 0 0 0 0 Araki 2015 (14) Glargine 0.75 26 181 180 0 0 0 0 1 0 2 0 0 0 NCT01064687 (15) Placebo 0.75–1.5 26 560 141 2 0 1 0 1 0 1 0 1 0 Giorgino 2015 (16) Glargine 0.75–1.5 78 545 262 1 2 1 1 2 1 2 2 3 1 Umpierrez 2014 (17) 0.75–1.5 52 539 268 0 0 0 0 1 0 1 0 1 2 Weinstock 2015 (18) Sitagliptin 0.75 1.5 104 606 177 1 2 1 1 3 2 1 0 0 0 NCT01769378 (19) Placebo 1.5 24 240 60 nr NR NR NR 0 0 1 0 0 0 NCT01644500 (20) Glimepiride 0.75–1.5 26 536 271 1 0 NR NR 0 0 0 0 0 0 NCT01001104 (21) Placebo 0.75 12 35 37 0 0 0 0 0 0 0 0 0 0 Grunberger 2012 (22) Placebo 1.5 12 29 32 0 0 0 0 0 0 0 0 0 0 NCT01149421 (23) Placebo 0.75–1.5 26 505 250 NR NR NR NR 0 0 1 1 0 0 Exenatide Chaudhuri 2012 (24) Placebo 20 12 12 12 0 0 0 0 NR NR NR NR NR NR Gudipaty 2014 (25) Glimepiride 20 26 17 17 0 0 0 0 0 0 0 0 0 0 Sitagliptin 20 26 17 13 0 0 0 0 0 0 0 0 0 0 Diamant 2014 (26) Lispro 20 30 247 263 1 0 1 0 0 2 2 1 1 1 Bergenstal 2009 (27) BiAsp 20 24 124 248 0 1 0 1 NR NR NR NR NR NR Xu 2015 (28) 10 or 20 48 142 138 0 0 0 0 1 1 0 0 0 0 Pioglitazone 142 136 0 0 0 0 1 1 0 0 0 0 Derosa 2010 (29) 20 52 63 65 0 0 0 0 0 0 0 0 0 0 Derosa 2011 (30) Glimepiride 20 52 52 49 0 0 0 0 0 0 0 0 0 0 Gallwitz 2012 (31) Glimepiride 20 208 490 487 5 5 NR NR 1 4 3 0 1 0 DeFronzo 2005 (32) Placebo 10 or 20 30 223 113 0 0 0 0 NR NR NR NR NR NR Buse 2004 (33) Placebo 10 or 20 30 248 123 0 0 0 0 1 2 0 0 NR NR Bunck 2009 (34) Glargine 20 64 36 33 0 0 0 0 0 0 0 0 0 0 Kendall 2005 (35) Placebo 10 or 20 30 486 247 0 1 0 1 NR NR NR NR NR NR Nauck 2007 (36) BiAsp 20 52 253 248 2 1 2 1 2 0 0 1 0 0 Moretto 2008 (37) Placebo 10 or 20 24 155 77 0 0 0 0 0 0 0 0 NR NR Heine 2005 (38) Glargine 20 26 282 267 0 0 0 0 1 0 0 0 0 0 Sathyanarayana 2011 (39) None 20 52 12 12 0 0 0 0 0 0 0 0 0 0 Jaiswal 2015 (40) Glargine 20 78 22 24 0 0 0 0 0 1 0 0 1 1 NCT00701935 (41) Placebo 20 26 43 37 0 0 0 0 0 1 0 0 0 1 Kadowaki 2009 (42) Placebo 20 12 37 40 0 0 0 0 0 0 0 0 0 0 Davis 2007 (43) Insulina 10 or 20 16 33 16 NR NR NR NR 0 0 0 0 0 0 Barnett 2007 (44) Glargine 20 16 138 138 0 0 0 0 0 0 0 0 0 0 Gao 2009 (45) Placebo 20 16 234 232 0 0 0 0 0 0 0 1 0 0 Gill 2010 (46) Placebo 20 12 28 26 0 0 0 0 0 0 0 0 0 0 Liutkus 2010 (47) Placebo 20 26 111 54 0 0 0 0 0 0 0 0 0 0 Gallwitz 2011 (48) Aspart 30/70 20 26 181 173 NR NR NR NR 1 0 0 0 0 0 NCT00375492 (49) Placebo 20 24 98 98 0 0 0 0 0 0 0 0 0 0 Davies 2009 (50) Glargine 20 26 118 116 NR NR NR NR 1 0 0 0 0 1 Kadowaki 2011(51) Placebo 10 or 20 24 145 36 0 0 0 0 0 0 0 0 0 0 NCT00667732 (52) Placebo 20 24 17 17 0 0 0 0 0 0 0 0 0 0 Apovian 2010 (53) Placebo 20 24 96 98 0 0 0 0 0 0 0 0 0 0 Gurkan 2014 (54) Glargine 20 26 17 17 NR NR NR NR NR NR NR NR NR NR Wu 2011 (55) Placebo 20 16 12 11 NR NR NR NR NR NR NR NR NR NR Yuan 2012 (56) Metformin 20 26 33 26 0 0 0 0 0 0 0 0 0 0 Liang 2013 (57) Placebo 20 52 34 36 0 0 0 0 0 0 0 0 0 0 Buse 2011 (58) Placebo 20 30 137 122 0 1 0 1 1 1 0 0 0 0 Zinman 2007 (59) Placebo 10 or 20 16 121 112 0 0 0 0 0 0 0 0 0 0 DeFronzo 2010 (60) None 20 20 47 45 0 0 0 0 0 0 0 0 0 0 45 45 0 0 0 0 0 0 0 0 0 0 Derosa 2013 (61) Placebo 20 52 86 85 0 0 0 0 NR NR NR NR NR NR Exenatide once-weekly Kim 2007 (62) Placebo 2 15 15 14 0 0 0 0 0 0 0 0 0 0 Bergenstal 2010 (63) Pioglitazone 2 26 160 165 0 0 0 0 0 2 0 1 0 0 Sitagliptin 160 166 0 0 0 0 0 0 0 1 0 0

Please cite this article as: M. Monami, et al., Effects of glucagon-like peptide-1 receptor agonists on mortality and cardiovascular events: A comprehensive meta-analysis of rand..., Int J Cardiol (2017), http://dx.doi.org/10.1016/j.ijcard.2017.03.163 M. Monami et al. / International Journal of Cardiology xxx (2017) xxx–xxx 5

Table 1 (continued)

First author (year)a Comparator Dose GLP-1RA Trial duration Number of All cause Cardiovascular Myocardial Stroke Heart patients death death infarction failure

Albiglutide Drug Mg weeks ID Comp. ID Comp. ID Comp. ID Comp. ID Comp. ID Comp

Diamant 2012 (64) Glargine 2 84 233 223 0 0 0 0 1 0 1 0 0 0 NCT01652729 (65) Placebo 2 28 181 61 NR NR NR NR 0 1 0 0 0 0 Sitagliptin 181 122 NR NR NR NR 0 0 0 0 0 0 Davies 2013 (66) Detemir 2 26 111 105 0 0 0 0 0 0 0 0 0 0 Inagaki 2012 (67) Glargine 2 52 215 212 1 0 1 0 0 0 1 0 1 0 Russell-Jones 2012 (68) Metformin 2 26 248 409 0 0 0 0 0 0 0 1 0 0 Liraglutide Pratley 2011 (69) Sitagliptin 1.2–1.8 52 446 219 1 2 0 2 2 1 0 0 2 0 Mathieu 2014 (70) Aspart 0.6–1.8 28 88 89 0 0 0 0 0 0 0 1 0 0 D'Alessio 2015 (71) Glargine 0.6–1.8 24 489 489 NR NR NR NR 0 0 2 2 0 0 De Wit 2014 (72) None 0.6–1.8 52 26 24 0 0 0 0 0 0 0 0 0 0 Lind 2015 (73) Placebo 0.6–1.8 24 64 60 0 0 0 0 0 0 0 0 0 1 Marre 2009 (74) Placebo 0.6–1.8 26 695 114 0 0 0 0 3 2 0 0 0 0 Rosiglitazone 695 232 0 0 0 0 3 0 0 0 0 0 Nauck 2009 (75) Glimepiride 0.6–1.8 104 724 242 1 0 0 0 5 3 2 0 2 0 Placebo 724 121 1 0 0 0 5 0 2 0 2 0 Garber 2011 (76) Glimepiride 1.2–1.8 104 498 248 0 1 0 0 2 2 0 0 1 0 Zinman 2009 (77) Placebo 1.2–1.8 26 355 175 0 0 0 0 1 0 0 0 1 0 Russell-Jones 2009 (78) Glargine 1.8 26 230 232 1 2 0 2 1 2 0 0 NR NR Placebo 230 114 1 2 0 2 1 2 0 1 NR NR Marso 2016 (79) Placebo 1.8 198 4668 4672 381 447 219 278 292 339 173 199 218 248 Davies 2016 (80) Placebo 1.8 26 140 139 4 1 NR NR 2 1 1 0 2 0 Forst 2012 (81) None 1.8 12 21 19 NR NR NR NR NR NR NR NR NR NR Seino 2010 (82) Glibenclamide 0.9 52 268 132 1 0 0 0 3 1 0 2 0 0 Kaku 2010 (83) Placebo 0.6 52 88 88 0 0 0 0 0 0 0 0 0 0 NCT00614120 (84) Glimepiride 0.6–1.8 16 698 231 0 0 0 0 1 0 1 0 1 0 NCT01617434 (85) Placebo up to 1.8 26 225 225 0 0 0 0 1 0 0 0 0 0 NCT01907854 (86) Sitagliptin 0.6–1.8 26 202 204 0 0 0 0 0 0 0 1 1 0 Zang 2016 (87) Sitagliptin 1.8 26 183 184 0 0 0 0 0 0 0 1 0 0 Yang 2011 (88) Glimepiride 0.6–1.8 16 698 231 NR NR NR NR NR NR NR NR NR NR Li 2012 (89) Insulin 0.6 12 42 42 0 0 0 0 NR NR NR NR NR NR Davies 2015 (90) Placebo 1.8 56 211 212 0 0 0 0 NR NR 2 1 0 0 Nandy 2014 (91) Placebo 1.8 12 16 16 0 0 0 0 0 0 0 0 0 0 Glimepiride 16 17 0 0 0 0 0 0 0 0 0 0 Idorn 2016 (92) Placebo 1.8 12 25 22 0 0 0 0 0 0 0 0 0 0 vonScholten 2016 (93) Placebo 1.8 12 32 32 0 0 0 0 1 0 0 0 0 0 Retnakaran 2014 (94) Placebo 1.8 48 26 25 0 0 0 0 0 0 0 0 0 0 Vanderheiden 2016 (95) Placebo 1.8 26 35 36 0 0 0 0 NR NR NR NR NR NR Kumarathurai 2016 (96) Placebo 1.8 12 39 39 0 0 0 0 2 0 0 0 0 0 Charbonnel 2013 (97) Sitagliptin 1.2 26 327 326 0 1 0 0 0 1 0 0 0 0 Smits 2016 (98) Placebo 1.8 12 19 17 0 0 0 0 0 0 0 0 0 0 Sitagliptin 19 19 0 0 0 0 0 0 0 0 0 0 Lixisenatide Ratner 2010 (99) Placebo 10 or 20 13 107 109 0 0 0 0 NR NR NR NR NR NR NCT00976937 (100) Sitagliptin 20 24 158 161 NR NR NR NR 0 0 0 0 0 0 Pfeffer 2015 (101) Placebo 20 108 3034 3034 211 223 156 158 270 261 67 60 122 127 Bolli 2014 (102) Placebo 20 24 322 160 1 0 NR NR NR NR NR NR NR NR Seino 2012 (103) Placebo 20 24 154 157 0 1 0 0 NR NR 2 0 NR NR Riddle 2013 (104) Placebo 20 24 328 167 1 0 1 0 2 3 0 0 1 0 Yu Pan 2014 (105) Placebo 20 24 196 194 0 0 0 0 1 0 1 1 0 0 Ahrén 2013 (106) Placebo 20 24 510 170 1 0 NR NR NR NR NR NR NR NR Fonseca 2012 (107) Placebo 20 12 239 122 NR NR NR NR NR NR NR NR NR NR NCT01798706 (108) Placebo 20 24 176 174 NR NR NR NR 0 1 0 1 0 0 Pinget 2013 (109) Placebo 20 24 323 161 0 1 0 1 0 1 1 0 0 0 Riddle 2013 (110) Placebo 20 24 223 223 0 2 0 1 0 2 1 0 1 0 Rosenstock 2014 (111) Placebo 20 24 574 285 2 0 NR NR NR NR NR NR NR NR Rosenstock 2016 (135) Glargine 20 30 234 467 1 3 1 2 0 0 1 1 0 2 Semaglutide Marso 2016 [136] Placebo 0.5–1.0 109 1648 1649 62 60 44 46 47 64 27 44 59 54

a See references in Supplementary materials; GLP-1RA: Glucagon-like Peptide-1 Receptor Agonists; ID: Investigational Drug (GLP-1RA); Comp.: Comparator; NR: Not reported.

significant effect of moderators was detected in meta-regression suggested a possible beneficial effect of the class of GLP-1 receptor ago- analyses (Table 3 Supplementary materials). nists on cardiovascular morbidity and mortality (Section 2.3). In the present meta-analysis, the incidence of major cardiovascular events 4. Discussion (MACE), which was the principal endpoint of both those trials, was not considered, because time to first event cannot be reliably assessed The publication of the results of the Liraglutide Effect and Action in when including studies with non-cardiovascular endpoints. However, Diabetes: Evaluation of cardiovascular outcome Results (LEADER) and a significant effect was detected for two of the components of MACE, the Trial to Evaluate Cardiovascular and Other Long-term Outcomes i.e., cardiovascular mortality and myocardial infarction. This result is with Semaglutide in Subjects with Type 2 Diabetes (SUSTAIN-6) trials not surprising, considering that a relevant proportion (approximately

Please cite this article as: M. Monami, et al., Effects of glucagon-like peptide-1 receptor agonists on mortality and cardiovascular events: A comprehensive meta-analysis of rand..., Int J Cardiol (2017), http://dx.doi.org/10.1016/j.ijcard.2017.03.163 6 M. Monami et al. / International Journal of Cardiology xxx (2017) xxx–xxx

All-cause mortality Cardiovascular mortality 0.01 0.10 1.0 10.0 0.01 0.10 1.0 10.0 100.0 N trials MH-OR,95% CI p N trials MH-OR,95% CI p

Placebo 21 0.88 [0.79; 0.9ì9] 0.027 13 0.85 [0.74; 0.97] 0.016

CVOTs 3 0.89 [0.80; 0.99] 0.035 3 0.87 [0.74; 1.04] 0.12

Non CVOTs 18 0.80 [0.44; 1.48] 0.48 10 0.39 [0.15; 0.98] 0.046

Insulin 11 0.92 [0.49; 1.73] 0.80 10 0.73 [0.32; 1.67] 0.46

DPP4-Inhibitors 5 0.69 [0.27; 1.73] 0.42 4 0.45 [0.13; 1.52] 0.20

Sulfonylureas 6 0.81 [0.37; 1.77] 0.60 1 1.02 [0.06; 16.3] 0.99

Glitazones 1 0.51 [0.12; 2.04] 0.34 1 1.02 [0.14; 7.31] 0.98

Favours Favours Favours Favours GLP-1RA comparators GLP-1RA comparators

Fig. 2. MH-OR with 95% CI for all-cause and cardiovascular mortality for GLP-1 receptor agonists versus placebo or other active comparators.

34 and 53% of cardiovascular deaths and incident myocardial infarc- It is conceivable that multiple mechanisms are involved in the re- tions, respectively) of reported events derived from the LEADER and duction of cardiovascular risk during treatment with GLP-1 receptor ag- SUSTAIN-6 trials. onists. The improvement of glucose control [14] and of other risk The relative risk reduction for myocardial infarction and cardiovas- factors, such as blood pressure, body weight and lipid profile [15,16], cular mortality (10 and 14%, respectively) may seem relatively small. could play a major role. However, direct effects of GLP-1 agonists on However, it should be considered that the majority of events was ob- the cardiovascular system are also plausible [17]. served in cardiovascular outcome trials, in which a large proportion of One of the main issues in the interpretation of cardiovascular out- the patients enrolled was already treated with other drugs for cardio- come trials with GLP-1 receptor agonists is the identifications of reasons vascular protection, such as statins, antiplatelet agents, beta blockers, underlying differences in results between the Evaluation of Lixisenatide and ACE inhibitors or angiotensin receptor blockers [1–3]. Therefore, in Acute Coronary Syndrome (ELIXA) study [1] and the two other trials the protective effect of GLP-1 receptor agonists appears to be additive cited above (Section 2.3). In the cardiovascular trial with lixisenatide, to that of current cardiovascular treatments. In addition, the protocol patients were enrolled within 180 days from an acute coronary syn- of cardiovascular outcome studies is designed in such a way as to min- drome; a majority of those patients had undergone a coronary revascu- imize between-group differences in glycemic control; in fact, investiga- larization shortly before the initiation of treatment [1]. Those patients tors were invited to adjust concurrent hypoglycemic treatments with represent a population at very high cardiovascular risk, which could the aim of reaching and maintaining good glycemic control in all pa- be less sensitive to some therapeutic interventions for longer-term pre- tients [1–3]. This particular design may lead to an underestimation of vention of cardiovascular disease. Therefore, the lack of beneficial ac- the actual beneficial effects of experimental drugs. tions on MACE in the ELIXA trial could be the effect of the selection of Overall, the results of the present meta-analysis are consistent with population, rather than of the lack of efficacy of the drug. In this respect, those of pooled analyses of patient-level data on cardiovascular events data from other trials with lixisenatide, in different patient populations, observed in phase 3 trials. Such analysis showed either no effect [11] could be very interesting. Unfortunately, the number of events observed or a trend toward a reduction of risk [12,13],whichdidnotreachstatis- in the trials with lixisenatide different from ELIXA is too small to draw tical significance due to the limited size of the samples studied. In fact, any conclusion on this point. the number of events observed in non-cardiovascular, phase 3 trials is On the other hand, differences between cardiovascular trials with too small to allow any definitive conclusion. liraglutide and semaglutide, on one side, and lixisenatide on the other,

Myocardial infarction Stroke Heart failure

0.10 1.0 10.0 N trials MH-OR, 95% CI p 0.10 1.0 10.0 N trials MH-OR, 95% CI p 0.10 1.0 10.0 100.0 N trials MH-OR, 95% CI p

Overall 45 0.90 [0.81; 1.00] 0.059 42 0.89 [0.76; 1.04] 0.15 45 0.90 [0.81; 1.00] 0.059

Including CVOTs Albiglutide 4 1.34 [0.65; 2.78] 0.42 7 0.86 [0.31; 2.40] 0.78 4 1.34 [0.65; 2.78] 0.42

Dulaglutide 6 0.48 [0.19; 1.22] 0.12 9 0.96 [0.42; 2.23] 0.93 6 0.48 [0.19; 1.22] 0.12

Exenatide b.i.d. 10 0.77 [0.32; 1.86] 0.56 4 1.32 [0.31; 5.60] 0.70 10 0.77 [0.32; 1.86] 0.56

Exenatide OW 3 0.72 [0.12; 4.43] 0.72 4 1.14 [0.24; 5.57] 0.87 3 0.72 [0.12; 4.43] 0.72

Liraglutide 14 0.86 [0.73; 1.01] 0.060 11 0.86 [0.70; 1.05] 0.15 14 0.86 [0.73; 1.01] 0.060

Semaglutide 1 0.73 [0.50; 1.07] 0.10 1 0.61 [0.37; 0.99] 0.044 1 0.73 [0.50; 1.07] 0.10

Lixisenatide 6 1.01 [0.85; 1.21] 0.87 7 1.15 [0.82; 1.61] 0.42 6 1.01 [0.85; 1.21] 0.87

Excluding CVOTs

Liraglutide 10 0.90 [0.45; 1.79] 0.76 10 0.78 [0.31; 1.91] 0.58 10 0.90 [0.45; 1.79] 0.76

Lixisenatide 5 0.38 [0.11; 1.24] 0.11 6 1.50 [0.38; 5.93] 0.56 5 0.38 [0.11; 1.24] 0.11

Favours Favours Favours Favours Favours Favours GLP-1RA comparators GLP-1RA comparators GLP-1RA comparators

Fig. 3. MH-OR with 95% CI for acute myocardial infarction, stroke, and heart failure with different molecules, considering all randomized clinical trials with and without cardiovascular endpoints.

Please cite this article as: M. Monami, et al., Effects of glucagon-like peptide-1 receptor agonists on mortality and cardiovascular events: A comprehensive meta-analysis of rand..., Int J Cardiol (2017), http://dx.doi.org/10.1016/j.ijcard.2017.03.163 M. Monami et al. / International Journal of Cardiology xxx (2017) xxx–xxx 7

Myocardial infarction Stroke Heart failure

0.101.0 10.0 N trials MH-OR,95%CI p 0.010.10 1.0 10.0 N trials MH-OR,95%CI p 0.01 0.1 1.0 10.0 N trials MH-OR,95%CI p

Placebo 25 0.90 [0.80;1.00] 0.055 20 0.87 [0.74;1.03] 0.10 16 0.92 [0.80;1.06] 0.24

Insulin 14 0.98 [0.50;1.91] 0.95 13 1.30 [0.65;2.61] 0.45 9 0.81 [0.33;2.00] 0.65

DPP4-Inhibitors 5 0.97 [0.37;2.55] 0.96 6 0.62 [0.21;1.85] 0.39 4 1.67 [0.34;8.11] 0.53

Sulfonylureas 6 0.81 [0.37;1.79] 0.61 5 1.45 [0.32;6.52] 0.63 5 1.87 [0.45;7.70] 0.39

Glitazones 4 0.69 [0.17;2.73] 0.59 2 0.19 [0.02;1.61] 0.13 1 0.11 [0.01;2.09] 0.14

Favours Favours Favours Favours Favours Favours GLP-1RA comparators GLP-1RA comparators GLP-1RA comparators

Fig. 4. MH-OR with 95% CI for acute myocardial infarction, stroke, and heart failure for GLP-1 receptor agonists versus placebo or other active comparators. could be determined by diversities in the characteristics of the drugs. heart failure reported as serious adverse events; those include all epi- Lixisenatide, despite its use as a once-daily drug, is a short-acting sodes requiring hospitalization, plus those cases in which the survival agent, with a half-life of 2–4h[18], whereas the other two agents war- of the patient was at risk, regardless of hospital admission. Although rant steadier circulating drug concentration across the 24 h [19]. If GLP- the event of a life-threatening condition treated at home is unlikely, it 1 receptor agonists have protective effects on the cardiovascular system, is possible that the number of events in non-cardiovascular trials was as suggested by many experimental studies [20], those actions could be overestimated. greater with longer-acting agents. Unfortunately, cardiovascular events The mean duration of the studies included in the meta-analysis is in trials with exenatide b.i.d., the other available short-acting agent, relatively short; therefore, the cardiovascular effects of GLP-1 receptor which had already been previously summarized in a pooled-analysis agonists in the longer term deserve further investigation. Notably, dif- of patient-level data [13], are too few to allow any speculation on differ- ferences in cardiovascular events between the active drug and placebo ences between short- and long-acting agents. in both LEADER and SUSTAIN-6 trials [2,3] were evident only after sev- Lixisenatide differs from both liraglutide and semaglutide also for its eral months of treatment. Therefore, it is possible that the present meta- chemical structure. In fact, lixisenatide is homologous to exenatide, analysis, which includes a large number of shorter-term trials, underes- whereas the other two molecules are very similar to human GLP-1. It timates the actual beneficial effects of GLP-1 receptor agonists. On the has been suggested that some of the actions of human GLP-1, including other hand, when trials with a duration of one year or more were those on the cardiovascular system [21,22] could be mediated by a re- analysed separately, the results did not differ from those of the main ceptor different from the “classical” GLP-1 receptor; the affinity of indi- analysis; this suggests that the distortion introduced by the inclusion vidual GLP-1 receptor agonists for other putative GLP-1 binding sites is of shorter-term trials is not very relevant. unknown. Among long-acting GLP-1 receptor agonists, four (liraglutide, A further limitation of this meta-analysis is the lack of adjudication semaglutide, dulaglutide, and albiglutide) have a high homology to of cardiovascular events and cardiovascular mortality in many of the human GLP-1, whereas exenatide long-acting release (LAR) is a differ- smaller studies, with the possibility of misclassifications. In addition, de- ent formulation of exenatide. In this meta-analysis, a trend toward a spite the efforts at collecting all available evidence, including that de- reduction of events was observed with dulaglutide; this result is consis- rived from the so-called “grey literature”, data from some completed tent with that of a recent pooled analysis of patient-level data [12].On trials could not be retrieved. The possibility of distortions determined the other hand, a pooled analysis of trials with albiglutide did not by selective disclosure cannot be entirely ruled out. suggest any reduction of risk [11] Results on cardiovascular morbidity Meta-regression analyses were performed post-hoc, as a tool for de- and mortality of dulaglutide and albiglutide do not seem to be very tecting possible moderators of the cardiovascular effects of GLP-1 recep- different from those of liraglutide and semaglutide, after the exclusion tor agonists. Interestingly, this class of drugs appeared to produce a of cardiovascular outcome trials; on the other hand, the number of greater reduction of cardiovascular mortality and of the incidence of events reported in trials with exenatide LAR is too small to draw any myocardial infarction in patients with a higher BMI. This suggests that conclusion. subjects with a higher degree of obesity could benefitmorefromGLP- In the SUSTAIN-6 trial, semaglutide was associated with a significant 1 receptor agonist treatment, although this difference was not detected reduction in the incidence of stroke [3]. This result should be interpreted in the LEADER and SUSTAIN-6 trials (Section 2.3). The interpretation of with caution, because of the small number of observed events. Notably, this result should be very cautious because of the risk of ecological falla- when GLP-1 receptor agonists are considered as a class, such a protec- cy, i.e. the possibility of confounding factors, unaccounted for in this tion from stroke is not detectable. Although a specificactionof analysis, associated with a higher BMI: for example, we did not consider semaglutide cannot be completely ruled out, a casual result could be ethnicity, which has a relevant impact on body mass index, and which more plausible. could moderate the biological effects of GLP-1 receptor agonists. Some molecules of the class of Dipeptidyl Peptidase 4 (DPP4) inhib- itors have been associated with an increased risk of hospitalization for 4.1. Conclusions heart failure [23–25], although the issue is controversial [25–27].Inall the three cardiovascular trials with GLP-1 receptor agonists, the hospi- Overall, the agents of this class appear to reduce all-cause and car- talization rate for heart failure was not different with the experimental diovascular mortality, and the incidence of myocardial infarction at drug than with placebo [1–3]. In the present analysis, with the addition mid-term follow up. Available data suggest differences across molecules of studies with non-cardiovascular endpoints, no signal of risk was de- of the class, but they are insufficient to verify whether those differences tected. This result should be interpreted with caution, because the num- are due to kinetics or other features of individual drugs. The possibility ber of patients undergoing hospitalization for heart failure is not usually that patients with a higher degree of obesity have greater cardiovascular available in non-cardiovascular trials. As a proxy, we collected cases of benefits with GLP-1 receptor agonists deserves further investigation.

Please cite this article as: M. Monami, et al., Effects of glucagon-like peptide-1 receptor agonists on mortality and cardiovascular events: A comprehensive meta-analysis of rand..., Int J Cardiol (2017), http://dx.doi.org/10.1016/j.ijcard.2017.03.163 8 M. Monami et al. / International Journal of Cardiology xxx (2017) xxx–xxx

Declaration of interests investigators. Liraglutide and cardiovascular outcomes in type 2 diabetes, N. Engl. J. Med. 375 (2016) 311–322. [3] S.P. Marso, S.C. Bain, A. Consoli, F.G. Eliaschewitz, E. Jódar, L.A. Leiter, I. Lingvay, J. Matteo Monami has received speaking fees from Bristol Myers Rosenstock, J. Seufert, M.L. Warren, V. Woo, O. Hansen, A.G. Holst, J. Pettersson, T. Squibb, Eli-Lilly, Merck, Novonordisk, Merck, and Takeda, and research Vilsbøll, SUSTAIN-6 investigators. Semaglutide and cardiovascular outcomes in pa- tients with type 2 diabetes, N. Engl. J. Med. (2016) (Epub ahead of print). grants from Bristol Myers Squibb. [4] A. Lund, F.K. Knop, T. Vilsbøll, Glucagon-like peptide-1 receptor agonists for the Stefania Zannoni, Laura Pala, Antonio Silverii, and Francesco treatment of type 2 diabetes: differences and similarities, Eur. J. Intern. Med. 25 Andreozzi have no conflicts of interest. (2014) 407–414. Giorgio Sesti has received consultancy fees from Servier, Intarcia, [5] http://www.crd.york.ac.uk/PROSPERO/display_record.asp?ID=CRD42015020245 (last accessed on 3rd November 2016). , Janssen, BoehringerIngelheim, Eli Lilly, Astra Zeneca, [6] http://www.ema.europa.eu/ema/ (last accessed on 3rd November 2016). MSD Italy, Sanofi,Pfizer, and Abbott, and speaking fees from Novo [7] http://www.fda.gov (last accessed on 3rd November 2016). Nordisk, MSD Italy, Boehringer Ingelheim, Eli Lilly, Janssen, Astra [8] C.B. Begg, M. Mazumdar, Operating characteristics of a rank correlation test for publication bias, Biometrics 50 (1994) 1088–1101. Zeneca, TherasLifetech and Takeda. [9] M. Egger, S.G. Davey, M. Schneider, C. Minder, Bias in meta-analysis detected by a Edoardo Mannucci has received consultancy fees from Merck and simple graphical test, BMJ 315 (1997) 629–634. Novartis, speaking fees from Astra Zeneca, Bristol Myers Squibb, [10] D. Moher, A. Liberati, J. Tetzlaff, D.G. Altman, Preferred reporting items for systemat- ic reviews and meta-analyses: the PRISMA statement, Ann. Intern. Med. 151 (2009) Merck, and Novartis, and research grants from Merck, Novartis, and 264–269. Takeda. [11] M. Fisher, M.C. Petrie, P.D. Ambery, J. Donaldson, J. Ye, J.J. McMurray, Cardiovascular safety of albiglutide in the harmony programme: a meta-analysis, Lancet Diabetes Endocrinol. 3 (2015) 697–703. Contributor statements [12] K.C. Ferdinand, F.T. Botros, C.M. Atisso, P.T. Sager, Cardiovascular safety for once- weekly dulaglutide in type 2 diabetes: a pre-specified meta-analysis of prospective- Matteo Monami was involved in each of the following points: ly adjudicated cardiovascular events, Cardiovasc. Diabetol. 15 (2016) 38. [13] R. Ratner, J. Han, D. Nicewarner, I. Yushmanova, B.J. Hoogwerf, L. Shen, Cardiovascu- 1. Design lar safety of exenatide BID: an integrated analysis from controlled clinical trials in 2. Data Collection participants with type 2 diabetes, Cardiovasc. Diabetol. 10 (2011) 22. [14] E. Mannucci, M. Monami, A. Ceriello, C.M. Rotella, A back to glycemic control: an al- 3. Analysis ternative look at the results of cardiovascular outcome trials in type 2 diabetes, Nutr. 4. Writing manuscript Metab. Cardiovasc. Dis. (2017) (Epub ahead of print). Stefania Zannoni, Laura Pala, Antonio Silverii, and Francesco [15] J.E. Potts, L.J. Gray, E.M. Brady, K. Khunti, M.J. Davies, D.H. Bodicoat, The effect of glucagon-like peptide 1 receptor agonists on weight loss in type 2 diabetes: a sys- Andreozziwere involved in each of the following points: tematic review and mixed treatment comparison meta-analysis, PLoS One 10 (6) (2015 Jun 29), e0126769. 1. Data Collection [16] M. Monami, I. Dicembrini, N. Marchionni, C.M. Rotella, E. Mannucci, Effects of 2. Manuscript revision glucagon-like peptide-1 receptor agonists on body weight: a meta-analysis, Exp. Giorgio Sesti was involved in each of the following points: Diabetes Res. 2012 (2012) 672658. [17] M. Luconi, G. Cantini, A. Ceriello, E. Mannucci, Perspectives on cardiovascular effects 1. Design of incretin-based drugs: from bedside to bench, return trip, Int. J. Cardiol. (2017) (Epub-ahead of print). 2. Reviewing manuscript [18] U. Werner, H. Haschke, A.W. Herling, W. Kramer, Pharmacological profile of Edoardo Mannucci was involved in each of the following points: lixisenatide: a new GLP-1 receptor agonist for the treatment of type 2 diabetes, Regul. Pept. 164 (2010) 58–64. 1. Design [19] A. Uccellatore, S. Genovese, I. Dicembrini, E. Mannucci, A. Ceriello, Comparison 2. Data Collection review of short-acting and long-acting glucagon-like peptide-1 receptor agonists, Diabetes Ther. 6 (2015) 239–256. 3. Analysis [20] E. Mannucci, I. Dicembrini, Incretin-based therapies and cardiovascular risk, Curr. 4. Writing manuscript Med. Res. Opin. 28 (2012) 715–721. All the authors approved the final version of this manuscript. [21] G. Cantini, E. Mannucci, M. Luconi, Perspectives in GLP-1 research: new targets. New receptors, Trends Endocrinol. Metab. 27 (2016) 427–438. [22] K. Ban, M.H. Noyan-Ashraf, J. Hoefer, S.S. Bolz, D.J. Drucker, M. Husain, Acknowledgements and statements Cardioprotective and vasodilatory actions of glucagon-like peptide 1 receptor are mediated through both glucagon-like peptide 1 receptor-dependent and – This research was performed independently of any funding, as part -independent pathways, Circulation 117 (2008) 2340 2350. [23] B.M. Scirica, D.L. Bhatt, E. Braunwald, P.G. Steg, J. Davidson, B. Hirshberg, P. Ohman, of the institutional activity of the investigators. R. Frederich, S.D. Wiviott, E.B. Hoffman, M.A. Cavender, J.A. Udell, N.R. Desai, O. Human and animal rights. Mosenzon, D.K. McGuire, K.K. Ray, L.A. Leiter, I. Raz, SAVOR-TIMI 53 steering com- Informed consent was not necessary because of no experimentation mittee and investigators. and cardiovascular outcomes in patients with type 2 diabetes mellitus, N. Engl. J. Med. 369 (2013) 1317–1326. with human subjects was performed. [24] W.B. White, C.P. Cannon, S.R. Heller, S.E. Nissen, R.M. Bergenstal, G.L. Bakris, A.T. Perez, P.R. Fleck, C.R. Mehta, S. Kupfer, C. Wilson, W.C. Cushman, F. Zannad, E.X.A.M.I.N.E. Investigators, after acute coronary syndrome in patients with type 2 diabetes, N. Engl. J. Med. 369 (2013) 1327–1335. Appendix A. Supplementary data [25] M. Monami, I. Dicembrini, E. Mannucci, Dipeptidyl peptidase-4 inhibitors and heart failure: a meta-analysis of randomized clinical trials, Nutr. Metab. Cardiovasc. Dis. 24 Supplementary data to this article can be found online at http://dx. (2014) 689–697. [26] J.B. Green, M.A. Bethel, P.W. Armstrong, J.B. Buse, S.S. Engel, J. Garg, R. Josse, K.D. doi.org/10.1016/j.ijcard.2017.03.163. Kaufman, J. Koglin, S. Korn, J.M. Lachin, D.K. McGuire, M.J. Pencina, E. Standl, P.P. Stein, S. Suryawanshi, F. Van de Werf, E.D. Peterson, R.R. Holman, TECOS Study References Group, Effect of Sitagliptin on cardiovascular outcomes in type 2 diabetes, N. Engl. J. Med. 373 (2015) 232–242. [1] M.A. Pfeffer, B. Claggett, R. Diaz, K. Dickstein, H.C. Gerstein, L.V. Køber, F.C. Lawson, L. [27] G.P. Fadini, A. Avogaro, L. Degli Esposti, P. Russo, S. Saragoni, S. Buda, G. Rosano, S. Ping, X. Wei, E.F. Lewis, A.P. Maggioni, J.J. McMurray, J.L. Probstfield, M.C. Riddle, S.D. Pecorelli, L. Pani, OsMed Health-DB Network. Risk of hospitalization for heart failure Solomon, J.C. Tardif, E.L.I.X.A. Investigators, Lixisenatide in patients with type 2 dia- in patients with type 2 diabetes newly treated with DPP-4 inhibitors or other oral betes and acute coronary syndrome, N. Engl. J. Med. 373 (2015) 2247–2257. glucose-lowering medications: a retrospective registry study on 127.555 patients [2] S.P. Marso, G.H. Daniels, K. Brown-Frandsen, P. Kristensen, J.F. Mann, M.A. Nauck, from the Nationwide OsMed Health-DB Database, Eur. Heart J. 36 (2015) S.E. Nissen, S. Pocock, N.R. Poulter, L.S. Ravn, W.M. Steinberg, M. Stockner, B. 2454–2462. Zinman, R.M. Bergenstal, J.B. Buse, LEADER steering committee; LEADER trial

Please cite this article as: M. Monami, et al., Effects of glucagon-like peptide-1 receptor agonists on mortality and cardiovascular events: A comprehensive meta-analysis of rand..., Int J Cardiol (2017), http://dx.doi.org/10.1016/j.ijcard.2017.03.163