The cardiovascular safety of incretin-based therapies: a review of the evidence

A literature search of established sources, including congress abstracts, was performed to identify recent publications dealing with the topic for review. Relevant articles were selected for review and discussion in light of the author’s knowledge of the field and clinical judgement. The European label for lixisenatide [25] was consulted for additional information. The official clinical trials registry at http://​www.​clinicaltrials.​gov was also searched to identify long-term CV outcomes clinical trials with incretin-based therapies in individuals with T2D.

Following nutrient intake, two major incretin hormones, GLP-1 and glucose-dependent insulinotropic polypeptide (GIP), are secreted from the gastrointestinal tract into the circulation [26]. It is now recognised that GLP-1 and GIP are responsible for 50–70% of total insulin secretion after oral glucose intake in healthy individuals [26], augmenting insulin release in a glucose-dependent manner, and leading to reductions in average blood glucose levels. GLP-1 exerts an additional glucose-lowering effect through actions at pancreatic alpha cells which result in an inhibition of glucagon secretion [26]. There is evidence to suggest that the incretin system is functionally impaired in people with T2D, although this is not universally observed [27‐31]. This impairment may contribute to the characteristic postprandial hyperglycaemia observed in affected individuals [27]. Native GLP-1 has been shown to reduce hyperglycaemia in T2D; however, as GLP-1 is rapidly (within 2–3 minutes) degraded by the enzyme dipeptidyl peptidase-4 (DPP-4), continuous GLP-1 infusion would be required to maintain optimal glycaemic control [32]. Therefore, incretin-based therapies have been developed either to raise endogenous levels of incretin hormones (DPP-4 agonists) or to mimic GLP-1 effects (GLP-1 receptor agonists) over more extended timeframes.

Pharmacological inhibition of the DPP-4 enzyme with DPP-4 inhibitors increases plasma incretin hormone levels. Several different oral DPP-4 inhibitors are clinically available, including sitagliptin, saxagliptin, linagliptin, alogliptin, vildagliptin (in the EU and Australia), teneligliptin and analagliptin (both available only in Japan) and gemigliptin (available only in Korea). Subcutaneous injection of GLP-1 receptor agonists mimics the effects of native GLP-1 through stimulation of the G_s-protein coupled GLP-1 receptor. Two GLP-1 receptor agonists are currently widely available: liraglutide (a GLP-1 analogue with 97% homology to native human GLP-1) and exenatide (53% homology to native human GLP-1). Exenatide was introduced to the market in 2007, having been developed from the peptide exendin-4 in the venom of the Gila monster, and is administered twice daily (BID). A major advantage of liraglutide when introduced in 2009 was its once daily administration. An extended release once-weekly formulation of exenatide (also known as exenatide long-acting release [LAR]) is also available in both the European Union [33] and in the U.S. [34]. Lixisenatide, a novel GLP-1 receptor agonist, was recently launched in several regions, including the European Union [25], and a number of GLP-1 receptor agonists are currently in clinical development, including albiglutide.

The effects of native GLP-1 on contractility, heart rate and/or blood pressure have been investigated [35‐37], but are not well-established. Differing effects of native GLP-1 on these parameters have been observed, presenting a complex picture. These varying effects may arise from differences in species, GLP-1 concentrations and experimental methods used.

Native GLP-1 exerts its effects via the GLP-1 receptor; but may also have GLP-1 receptor-independent effects [38]. The GLP-1 receptor is ubiquitously expressed, located not only in pancreatic cells but also in the lungs, kidneys, intestines, and peripheral and central nervous systems reviewed in [26, 37]. In addition, GLP-1 receptors are expressed in the CV system. In mice, GLP-1 receptors are distributed in myocardial tissue (left and right ventricles, septum and, to a lesser extent, atria), endocardium, microvascular endothelium, and coronary smooth muscle cells [38]. GLP-1 receptor expression in the human heart, including the coronary artery endothelial cells, has also been demonstrated [39, 40].

Data from animal models and pilot clinical studies have indicated that native GLP-1 may have cardioprotective effects in the setting of ischaemia, or following ischaemic injury [37, 38, 41, 42]. For example, in a study of 21 individuals with acute MI and severe systolic dysfunction after successful primary angioplasty, GLP-1 infusion (in addition to standard care) significantly improved left ventricular function compared with control individuals receiving standard care (including aspirin, clopidogrel, heparin, glycoprotein IIb/IIIa inhibitors, beta-blockers, angiotensin converting enzyme (ACE) inhibitors, and/or statins) [41]. Moreover, studies using the technique of brachial artery flow-mediated vasodilation (measured by ultrasonography) during GLP-1 infusion in people with T2D and stable coronary artery disease suggest that GLP-1 may improve endothelial function in some individuals with T2D [43]. The effects of GLP-1 on endothelial function have been reviewed by Sjöholm [44].

Interestingly, in preclinical studies, cardioprotective effects were apparent in GLP-1 receptor knockout mice, supporting the hypothesis that certain cardioprotective actions of GLP-1 are mediated through GLP-1 receptor-independent pathways [38].

People with T2D who are overweight or obese, are hypertensive or have dyslipidaemia are at increased risk of adverse CV events. Incretin-based therapies have been shown to have an impact on these CV risk factors; however, differences in these effects between DPP-4 inhibitors and GLP-1 receptor agonists have been noted. These differences may arise from differences in mechanism of action, or levels of GLP-1 receptor activation produced by these individual drug classes [12].

Obesity increases the risk of CVD mortality in T2D [45]. As more than half of individuals with T2D are obese [46], weight control is an important aspect of CV risk management in T2D. Encouragingly, lifestyle and certain pharmacological interventions that promote weight loss in people with T2D can improve CV risk profiles [47, 48].

DPP-4 inhibitors do not have major effects on weight [12]. Sitagliptin has been associated with changes in body weight of between 0.0 and −1.5 kg [49‐52]. Clinical studies examining saxagliptin have shown either dose-independent, numerical decreases (−0.11 to −1.8 kg) or increases (+0.5 to +0.8 kg) in body weight [53‐56]. Meanwhile, vildagliptin treatment has been associated with minimal weight gain of 0.2–1.3 kg [57].

In randomized controlled trials, use of the GLP-1 receptor agonists exenatide, liraglutide, or lixisenatide has been associated with body weight reductions. Exenatide BID treatment has been associated with average weight losses of up to 2.8 kg [58‐60], and exenatide LAR with mean weight losses of up to 3.7 kg [61‐66]. Similarly, in the GetGoal phase 3 development programme, lixisenatide was associated with mean body weight reductions of up to 3.0 kg [67‐75], although not all observed weight changes were statistically different from comparators (placebo, exenatide) [67, 69, 70, 75]. In the Liraglutide Effect and Action in Diabetes (LEAD) phase 3 clinical trials, consistent, significant improvements in body weight were observed with liraglutide use versus comparator arms, with average weight losses of up to 3.2 kg [76‐81].

Head-to-head trials comparing exenatide LAR [62, 64] or liraglutide [82‐84] with sitagliptin have confirmed significantly greater changes in weight with the GLP-1 receptor agonists.

Hypertension is a risk factor for adverse CV events, raising the risk of coronary heart disease by approximately 2-fold and the risk of stroke by 2- to 4-fold [85, 86].

Relatively few studies have examined the effects of DPP-4 inhibitors on blood pressure in people with T2D. Limited evidence suggests that sitagliptin may reduce systolic blood pressure (SBP) [83, 87, 88], although this effect has not been observed across all studies [89]. Pooled data from six phase 3 clinical trials indicated similar small decreases in SBP and DBP with linagliptin 5 mg once daily and placebo [90].

The clinically available GLP-1 receptor agonists reduce SBP [12, 25]. In phase 3, placebo-controlled studies of lixisenatide, reductions of up to 2.1 mmHg were observed [25]. In the case of exenatide BID, reductions in SBP of 3.5 mmHg have been observed, with the greatest improvements seen in those who lost most weight (−8.1 mmHg in 25% of subjects) [91]. Liraglutide treatment was associated with reductions in SBP of up to 6.7 mmHg across the LEAD phase 3 development programme [76‐82]. Interestingly, the reductions in SBP observed with liraglutide were observed prior to major weight loss, suggesting that SBP reduction does not occur solely via weight reduction [92]. As would be predicted, meta-analysis of the liraglutide phase 3 trials showed that individuals with the highest baseline SBP (>140 mmHg) exhibited the greatest reductions in SBP (−11.4 mmHg) [93], and that these reductions in SBP were independent of antihypertensive treatment [94].

Weight loss (discussed above), natriuresis and vasodilation [95] are likely to contribute to the SBP-lowering effects of GLP-1 receptor agonists. A recent mouse model study has linked GLP-1 receptor stimulation by liraglutide to cAMP/Epac2-dependent atrial natriuretic peptide release, which would be predicted to increase natriuresis and vasodilation, thereby lowering blood pressure [95]. Consistent with these observations, a GLP-1 receptor-dependent, nitric oxide- and endothelium-independent vasodilatory effect of GLP-1 has been identified in rat femoral artery preparations [96].

Elevated fasting triglycerides (TG) and low density lipoprotein cholesterol (LDL-C), and reduced high density lipoprotein cholesterol (HDL-C), are associated with increased risk of CVD [97].

DPP-4 inhibitors have been shown to have either no effects or minor effects on fasting lipid levels in people with T2D [90, 98, 99]; however, they may have an effect on postprandial lipids. Indeed, significant reductions in postprandial TG, apolipoprotein B-48 (ApoB-48, a LDL), very low density lipoprotein cholesterol (VLDL-C) and free fatty acids (FFAs) have been reported following sitagliptin treatment [postprandial area under the curve (AUC) reduced by 9.4, 7.8, 9.3 and 7.6%, respectively, p < 0.05 for all measurements vs. placebo] [100]. Postprandial TG and ApoB-48 reductions and have also been reported following vildagliptin treatment [incremental AUC −2.0 ± 0.8 (p = 0.011) and −0.6 ± 0.3 (p = 0.008), respectively] [101] and alogliptin treatment [incremental AUC −3.4 ± 0.7 (p < 0.001) and −0.9 ± 0.2 (p = 0.028), respectively] [102]. The effects of incretin-based therapies on postprandial hyperlipidaemia have been reviewed elsewhere by Ansar and colleagues [103].

While no lipid profile data have yet been published for lixisenatide, meta-analyses have shown overall improvements in fasting lipids with exenatide (BID and LAR) and liraglutide therapy. Liraglutide reduces LDL, FFA and TG levels (by 0.2, 0.09 and 0.2 mmol/L, respectively, p < 0.01) and exenatide BID decreases LDL levels (by 0.15 mmol/L, p < 0.05 [104]). Trends towards reductions in FFA and TG were also evident with exenatide, but these did not reach statistical significance [104]. Small decreases in HDL reported for both liraglutide and exenatide (0.04 and 0.05 mmol/L, respectively, p < 0.01) may not be clinically meaningful. No reductions in VLDL-C were observed for either GLP-1 receptor agonist [104].

Improvements in surrogate CV risk biomarkers, such as high sensitivity C-reactive protein (hsCRP), B-type natriuretic peptide (BNP), and plasminogen activator inhibitor-1 (PAI-1) have also been observed following treatment with liraglutide or exenatide BID [105‐108].

Substantial resting heart rate elevation is associated with increased CV mortality, and drugs that prolong cardiac repolarisation carry a risk of provoking adverse CV events [109, 110]. What are the effects of incretin-based therapies on these parameters?

Consistent with the findings outlined above regarding a cardioprotective effect of native GLP-1 under ischaemic conditions, data from rodent and porcine models indicate that GLP-1 receptor agonists may limit infarct size following ischaemia-reperfusion injury [119‐122]. A possible cardioprotective effect of exenatide following myocardial infarction has also been identified in a clinical setting. In people with ST-segment elevation MI undergoing primary percutaneous coronary intervention, intravenous exenatide administration at the time of reperfusion led to a reduction in the area of myocardium at risk, as assessed by cardiac magnetic resonance 3 months after intervention [123]. A post hoc analysis tested the hypothesis that this effect would be more pronounced in those with short duration of ischaemia. Data from 148 participants were stratified according to median time from ambulance call or first contact with healthcare system to first balloon inflation (132 min). In subjects with short duration ischaemia (≤132 min), exenatide reduced final infarct size by 30% and increased myocardial salvage index by 14%; in contrast, no such effects were observed in those longer duration ischaemia (>132 min) [124]. Taken together, these findings highlight a potential cardioprotective effect of GLP-1 receptor agonists in the setting of myocardial ischaemia-reperfusion injury, provided they are administered early.

Incretin-based therapies were amongst the first T2D treatments for which detailed evaluation of long-term CV safety was encouraged under the 2008 FDA guidance [11]. This guidance recommends that, with regard to the risk ratio for major adverse cardiovascular events (MACE), pre-approval clinical trials should demonstrate that the upper boundary of the two-sided 95% CI is less than 1.8 versus the control group, with subsequent outcomes trials indicating a two-sided 95% CI for risk ratio of less than 1.3 [11].

MACE analyses carried out for sitagliptin [125], saxagliptin [111] and alogliptin [126] have not indicated increased CV risk. Consistent with these findings, a recent meta-analysis identified no increased risk of CV events with a pooled group of DPP-4 inhibitors added to metformin versus metformin monotherapy [relative risk (95% CI) 0.54 (0.25-1.19), p = 0.13] [127]. Indeed, DPP-4 monotherapy was associated with a significantly lower risk of CV events than metformin monotherapy [relative risk (95% CI) 0.36 (0.15–0.85), p = 0.0.2], indicating a possible CV benefit of DPP-4 inhibitors [127]. Additionally, linagliptin has been associated with significantly fewer CV events than glimepiride over a 2-year period [both in combination with metformin ± one additional oral antidiabetic drug; relative risk (95% CI): 0.46 (0.23–0.91), p = 0.0213], despite similar reductions in mean HbA_1c[128].

MACE analyses conducted for GLP-1 receptor agonists have similarly reported values within the FDA predefined safety limits for diabetes therapies. One meta-analysis of results from randomised trials of GLP-1 receptor agonists (exenatide BID, exenatide LAR, liraglutide, albiglutide) for the period up to November 2010 indicated the following cardiovascular safety margins [Mantel-Haenszel odds ratio for MACE (95% CI)]: all GLP-1 receptor agonists 0.74 (0.50–1.08), p = 0.12; exenatide 0.85 (0.50-1.45), p = 0.55; liraglutide 0.69 (0.40–1.22), p = 0.20 [129]. An as-yet-unpublished meta-analysis of eight, phase 3 studies of lixisenatide indicated a hazard ratio (95% CI) for adjudicated MACE of 1.03 (0.64, 1.66) versus placebo [25]. For liraglutide, retrospective MACE analysis of pooled data from all completed phase 2 and phase 3 randomised trials to date, plus all open-label trial extensions, indicated that the incidence ratio for broad/serious adjudicated MACE versus comparator drugs (metformin, glimepiride, rosiglitazone, insulin glargine) and placebo was 0.73 (95% CI 0.38, 1.41) [130].

These pooled data, mostly from phase 2 and 3 trials, are complemented by reassuring observational data. For example, a retrospective analysis of the U.S. LifeLink™ database of medical and pharmaceutical insurance claims (for the period June 2005 to March 2009) indicated that exenatide BID was not associated with an excess CV risk compared to other glucose-lowering therapies [131]. Consistent with this, another retrospective analysis found no elevated risk of MACE with exenatide BID compared to insulin for ≤1 year [132].

The long-term CV safety and efficacy of many incretin-based therapies are being prospectively evaluated in randomised control trials (Table 1). With the exception of the EXSCEL study, all include people with pre-existing CV disease and/or people with high risk for CV disease [16‐24]. As of September 2013, a body of evidence is beginning to emerge on the long-term CV safety profile of incretin-based therapies in individuals at high risk of CV events: the results of the first two trials (SAVOR-TIMI 53 [saxagliptin] [22] and EXAMINE (alogliptin [24]), published as this article went to press, meet the FDA criteria for non-inferiority of these agents over placebo but provide no positive evidence of CV risk reduction. More evidence is eagerly awaited to inform clinical decision-making when seeking to minimise the risk of CVD in people with T2D.