Effects of GLP-1 receptor agonists and SGLT-2 inhibitors on cardiac structure and function: a narrative review of clinical evidence

Beyond its metabolic effects, GLP-1—an endogenous hormone secreted by intestinal endocrine cells—shows interesting direct effects on the myocardium. In human non-failing hearts, GLP-1R is expressed in cardiomyocytes of both atria and, to a negligible extent, ventricles [7]. Acute GLP-1 infusion in humans produces a 40% increase in myocardial microvascular blood flow, both in lean and obese subjects [8]. In a pilot study in 12 patients without T2D and with HFrEF, a 5-week infusion of GLP-1 significantly improved LVEF (from 21 ± 3 to 27 ± 3%, p < 0.01), peak oxygen uptake (VO_2peak from 10.8 ± 0.9 to 13.9 ± 0.6 mL/min/kg, p < 0.001), and 6-min walk distance (from 232 ± 15 to 286 ± 12 m, p < 0.001) when compared to standard therapy [9]. This was not confirmed by later placebo-controlled trials in subjects without diabetes with stable HFrEF (FE 30 ± 2%): no amelioration of LV indices was seen after 48-h infusion of GLP-1, while hypoglycemia, tachycardia, and an increase in diastolic blood pressure occurred raising some concerns [10]. A recent meta-analysis of four studies with short term GLP-1 infusion in subjects with HFrEF showed a modest effect on LVEF (+ 4.4%, 95% CI [1.36–7.44]), with no significant change in brain natriuretic peptide (NT-proBNP) levels [11].

GLP-1Ra mimic the “incretin” effect on weight loss and blood pressure (the latter mostly due to afterload reduction and natriuresis [12]); the infusion of exenatide for 6 h in 20 subjects with T2D and decompensated HFrEF induced a small decrease in pulmonary capillary wedge pressure (14.8 to 12.6 mmHg) and an increase in cardiac output from 1.8 to 2.1 L/min, largely due to an increase in heart rate, without affecting NT-proBNP levels [13]. The activation of GLP-1R can provide protection against ischemia–reperfusion injury, probably due to its ability to stimulate myocardial glucose uptake during postischemic contractile dysfunction, although this metabolic shift was never demonstrated in humans [14]. Acute GLP-1 infusion in subjects with acute myocardial infarction both with preserved [14] and reduced LVEF < 40% [15] was associated with an improved LVEF recovery irrespective of diabetic status. In subjects with myocardial infarction and ST segment elevation (STEMI), exenatide administration 15 min before revascularization caused a modest, non-significant reduction in infarct size (dependent on tissue viability and significant [− 30%] in those who received an earlier treatment [16]), with no consequences on left ventricle (LV) function nor clinical events at 30 days [16]. Later, the treatment with either exenatide or liraglutide administered at the time of primary angioplasty in acute myocardial infarction [17, 18] has been consistently proven effective in reducing infarct size by approximately 25% with no effect on LV function, while pre-treatment with i.v. GLP-1 was associated with an improved LV dysfunction during elective angioplasty for single-vessel disease [14]. Interestingly, the protective effect on infarct size appears to be time-dependent, so that delayed intervention is associated with no protection [19]. Negative results were also reported by a large, double-blind, placebo-controlled trials on both patients with preserved LVEF receiving exenatide at the time of percutaneous coronary intervention during acute myocardial infarction [20].

Negative results were seen in studies involving exenatide infusion at the time of coronary artery bypass in patients with and without diabetes undergoing elective procedure for coronary artery disease with either reduced [21] or normal LVEF [22]. Interestingly, however, the experimental group experienced lesser arrhythmic events and better hemodynamic stability in the perioperative period. Whether the anti ischemia–reperfusion injury action of GLP-1Ra is also observed in subjects on chronic treatment remains an open question, although a sub analysis of the LEADER study does not support this hypothesis. In the cohort of subjects undergoing myocardial infarction, no difference was observed in the subtype distribution (ST vs. non-ST, symptomatic vs. silent), in death rate or in-hospital troponin levels [23].

A synthesis of the available studies is provided in Table 1. In both a single-arm [24] and a small placebo-controlled RCT [25], chronic treatment with liraglutide in T2D subjects without established cardiovascular disease (CVD) improved diastolic function by reducing filling pressures by 20% (reduced E-wave and E/e′ ratio). This effect is possibly related to the reduction in BMI. On the contrary, no significant effect on diastolic function was observed in two studies comparing liraglutide vs. metformin in newly diagnosed T2D [26] and vs. glimepiride in T2D patients with subclinical HF [27]. Similarly, no difference was seen in two RCTs with GLP-1Ra on HFrEF patients with or without T2D [28, 29]. Liraglutide was essentially neutral on 2D indices of LV systolic function and cardiac output (with a − 3% decline in EF counterbalanced by a slight increase in HR) both in T2D without established CVD [25] and in HFrEF with or without T2D [28, 29]. No significant amelioration of either functional capacity (6-min walk time), LVEF, or natriuretic peptides was evident with other GLP-1Ra on T2D subjects without HF [25] and with HF [30], wherein some concerns about cardiovascular safety (more arrhythmias and ischemic events, possibly secondary to the higher heart rate) were raised in HFrEF [28, 29]. However, in newly diagnosed T2D subjects with subclinical systolic dysfunction as expressed by reduced LV global longitudinal strain (GLS)—an early, less load-dependent systolic index—liraglutide induced an 8% improvement in GLS, which was not entirely justified by the decrease in weight and HbA_1c [26].

In conclusion, whether the activation of GLP-1R signaling is associated with a selective amelioration of LV structure and/or function in subjects with or without ischemic heart disease or overt HF remains unclear. Although GLP-1 increases heart rate and thus cardiac output, evidence for independent effects of GLP-1 on ventricular function is inconclusive [31].

Although not powered to specifically detect differences in single CV events, all the major CV outcomes trials with GLP-1Ra reported data on HHF. Nevertheless, despite being of different duration and different inclusion criteria, the results altogether provide insight on this outcome. In the LEADER trial [32], T2D subjects with CVD or at high risk for CVD randomized to liraglutide vs. placebo experienced a 13% reduction in the risk of HHF, which was not statistically significant (12 vs. 14 per 1000 person-years; RR: 0.87 [0.73–1.05]). A subsequent analysis showed no difference in HF-related endpoints (hospitalization and CV death) and no safety concerns in patients with HF at baseline [33]. Interestingly, there was no difference in those with or without HF at baseline also in terms of kidney protection (RR 0.77 vs. 0.78) as evaluated by the composite endpoint of new onset microalbuminuria, doubling creatinine and eGFR, renal replacement therapy, and death of renal disease. The SUSTAIN 6 trial [34] recruited a similar population, and despite the positive effects of semaglutide on weight loss (3–4 kg), glycate hemoglobin (HbA_1c 0.7–1.0%), renal outcomes (RR 0.64) and MACE (RR 0.74), the rate of HHF was 17.6 vs. 16.1 per 1000 person-years in the semaglutide and placebo arm, respectively (RR 1.11 [0.76–1.61]). No substantial differences were observed related to sex, age, or between patients in primary or secondary CV prevention [35].

The population recruited in the ELIXA study [36] was at higher risk of HF, having all participants experienced a recent coronary event and 20% of them a diagnosis of heart failure at entry. Indeed, the incidence of HHF was 19.9 per 1000 person-years, but the treatment with lixisenatide for 24 months did not affect the risk for HHF (RR 0.96 [0.75–1.23]). Similarly, in the HARMONY [37], PIONEER 6 [33], and EXSCEL trials [38], involving T2D subjects at intermediate risk of HHF (10–12 events per 1,000 person-years in the placebo arms), neither albiglutide (RR for CV deaths or HHF 0.85 [0.70–1.04]) nor oral semaglutide (RR 0.86 [0.48–1.55]) nor 1-weekly exenatide (RR 0.94 [0.78–1.13]) significantly reduced the risk of HHF. The REWIND study [39] recruited T2D subjects at a lower risk of HF, being the rate of hospitalization plus urgent visit for HF being 8.9 per 1000 person-years in the placebo arm. Also in this population, the treatment with dulaglutide was not associated with a significant HHF risk reduction (RR 0.93 [0.77–1.12]). The very recent AMPLITUDE-O trial [40] makes an exception since the treatment with efpeglenatide was associated with a 39% reduction in HHF (RR 0.61 [0.38–0.98]). Of note, in terms of MACE and renal outcomes, as well as in terms of HbA1c, blood pressure, heart rate and body weight, the results of AMPLITUDE-O were similar to SUSTAIN 6, LEADER and REWIND trials. Although the population of the study AMPLITUDE-O was relatively more enriched with patients with CVD (90%) and with kidney disease (31%), as shown in Fig. 1, the absolute risk of HHF in the study was similar to other studies and no convincing trend of HHF risk reduction emerges in relation to the absolute risk of the event of the different trials. Why efpeglenatide differs with respect to other GLP-1Ra in terms of HHF prevention remains to be explained.

In conclusion, in individuals with T2D at high-very high CV risk, GLP-1Ra are extremely safe with respect to HF development or decompensation, showing either a neutral or a small beneficial effect. While two metanalyses reported a neutral effect on HHF [41, 42], two other independent meta-analyses [43, 44] estimated a 9% risk reduction of HHF of marginal statistical significance (95% CI 1–17%), which is close to the mean RR of 12% calculated including also the more recent trials (Fig. 1). The data with efpeglenatide are very intriguing and await confirmation, possibly in populations at very high risk of HHF. In cohorts with T2D and HFrEF data is scarce and not encouraging, a statistically non-significant (p = 0.09) 30% increase in HHF was reported in one study [28] and a statistically significant excess of serious adverse CV events (10 vs. 3%) was shown in another report [29], rising possible safety concerns in these more fragile patients.

SGLT-2i have proved effective in reducing congestion without worsening renal function in acute decompensated HFrEF since the first day of administration [45]. A reduction of cardiac preload can be seen secondary to SGLT-2i induced natriuretic effect, which is in part osmotic (proportional to plasma glucose levels and with a phasic, meal-related pattern) and in part direct from the inhibition of proximal tubular reabsorption of sodium, wherein glucose and sodium are co-transported in a molar ratio of 1:1. Therefore, when SGLT-2 is inhibited, a higher sodium load reaches the distal tubule, translating into a mild (10–20%) increase in sodium excretion which, however, is short lasting (1–3 days) possibly because of some degree of activation of the renin–angiotensin–aldosterone system (RAAS) [46]. The reduction in plasma volume (-9% vs. placebo) appears to be transient, too, as it is clearly detectable at 1 week but not after 12 weeks of treatment [47]. It therefore appears unlikely that the effect of SGLT-2i on heart function is mainly explained by their natriuretic properties alone. Possibly, this effect only emerges in interaction with specific conditions, especially a sequential nephron blockade and/or a blockade of the RAAS. In fact, in patients with T2D the natriuretic effect of SGLT-2i was still evident after 5 days only when administered on top of hydrochlorothiazide therapy (25 mg), which by itself had no effect on sodium excretion [48]. This synergistic effect has been also described with loop diuretics: in healthy subjects, when dapagliflozin was given for 7 days after 7 days of bumetanide (1 mg), its first-dose natriuretic effect was twofold greater than when given alone. Moreover, when bumetanide was given after dapagliflozin, its natriuretic effect was 40% greater [49]. This natriuretic effect is coupled with lesser RAAS activation and a more favorable volume redistribution with respect to other diuretics, as has been elegantly demonstrated by the use of bioimpedance spectroscopy [50] as well as modeling data of free water vs. electrolyte-free water clearances [51]. The result is a gentler reduction of plasma volume and a more effective mobilization of interstitial fluids than seen with usual diuretics. Whether in subjects with RAAS blockade or in edematous conditions this effect would be larger and persistent in unknown—but appears plausible. Indeed, a small recent pilot study (EMPA-RESPONSE-AHF trial) [52] has shown that in subjects with or without diabetes presenting with acute heart failure, viz a similar cumulative furosemide dose (320 vs. 300 mg), the initiation of empagliflozin 10 mg was associated with a larger urine output over 4 days (+ 3449, [578–6321] mL) and a more favorable net fluid balance (− 2701, [+ 586 to − 8988] mL) with respect to placebo. Of note, the treatment was continued for 30 days and was associated with a major improvement in 60-days prognosis, being 10 vs. 33% the incidence of the composite outcome (i.e., in-hospital worsening of HF, readmission for HF, and death).

Furthermore, SGLT-2i have been shown to improve endothelial function and reduce arterial stiffness. However, these effects appear modest in size and transient as they are clearly detectable after 2 days of treatment with dapagliflozin [53] but not after 4 weeks [54]. Also, SGLT-2i are known to raise free fatty acids (FFA) and ketone bodies, substrates avidly and efficiently used by the myocardium for energy production [55]. Exogenous β-hydroxybutyrate (β-HB) infusion causes a dramatic 75% increase in myocardial blood flow in normal subjects [56], while in patients with HF it increases cardiac output by 40%, due to an improvement in both stroke volume and heart rate [57]. However, these remarkable effects were observed at plasma β-HB concentrations of 1.6 and 3.3 mM, which are much higher than those achieved in patients on chronic SGLT-2i treatment (0.5–0.8 mM). In addition, these values are reached only in the fasting state, markedly declining in the postabsorptive condition [58]. In the study by Nielsen et al., a more physiologic infusion was tested, yielding a negligible (7%), though statistically significant, increase in cardiac output at a plasma concentration of 0.7 mM [57]. Furthermore, the effect on myocardial blood flow was observed during the simultaneous infusion of insulin, a non-physiological condition.

The available data (Table 2) suggest a significant beneficial effect of SGLT-2i on several variables closely related to cardiovascular outcomes, such as LV systolic and diastolic function as well as LV remodelling. Functional and structural cardiac improvements can be found in T2D patients both with and without overt HF, and across HF subgroups. Yet, a gradient of efficacy and relevance might be detected, with the HFrEF population experiencing greater benefit than the other HF subgroups (HFmrEF and HFpEF), and still lesser changes observed in the non-HF population. In this setting, the effect on LV systolic function expressed as 2D LV EF ranges from 2 to 10% in T2D without HF to 20–30% in T2D with HF either with reduced EF, mildly reduced EF [59] and preserved EF [59–61] show an intermediate 5–10% increase in LV EF. Intriguingly, the improvements in GLS appear to be present across all HF phenotypes with a comparable gradient [59, 62], and to be recognizable despite no increase in LVEF [59]. This pattern is also consistent with the reduction in natriuretic peptides observed with SGLT-2i, more evident in patients with HFrEF [62] and in those with higher baseline peptide values [61, 63]. With regard to diastolic function, there appears to be a more homogeneous 10–15% improvement throughout the different classes of LV EF, more pronounced in subjects with worse baseline values [60].

Chronic (6 months) treatment with SGLT-2i is associated with a consistent 5–10% reduction in LV mass index (LVMi). However, it is worth noting that different studies reported discrepant effects on LV linear dimensions and volumes (Table 1), and two trials with the same drug, dose, and duration of treatment recorded opposite results in LVMi and LV end-diastolic volume (LVEDV) as assessed by cardiac magnetic resonance imaging [64, 65]. Hitherto, it is uncertain whether definite LV remodelling occurs with SGLT-2i, at least within a 6-month treatment period.

Although the CV and kidney outcomes trials with SGLT-2i were not all designed to detect difference in HHF, having also different duration and recruiting subjects with different characteristics, they provide overall a large amount of data and offer interesting insights. An extended analysis of the EMPA-REG OUTCOME trial [66] has provided the following important information regarding the effect of empagliflozin on outcomes: (a) it consistently reduced incident HHF by 30–40%, together with death from HF, investigator-reported HF or edema (as adverse events) and the introduction of loop diuretics, (b) the effect was not influenced by the presence of HF at baseline, sex, race, metabolic control, kidney function, background metabolic and cardiovascular therapy, (c) it was independent of the dose, and (d) it was not associated with more serious adverse events. A similar analysis performed on the CANVAS data [67] extends to canagliflozin most of the above mentioned information related to empagliflozin, apart from the following aspects: the effect on CV death + HHF was more evident in subjects with HF at baseline (RR: 0.61 [0.46–0.80] vs. 0.87 [0.62–1.06], p = 0.02 for interaction), in those with body mass index ≥ 30 kg/m², with HbA_1c ≥ 8.0%, or on any diuretic. These findings, however, were not confirmed by a preliminary analysis of the CREDENCE trial data [68], which, however, confirmed the efficacy and safety of canagliflozin also in patients with reduced eGFR (44–30 mL/min/1.73 m²). Although unable to fully exploit the effect of the drug on glucose control, patients with eGFR 30–60 mL/min/1.73 m² are likely to benefit more from SGLT-2i treatment both in relative (RR: 0.62 [0.51–0.75] vs. 0.71 [0.61–0.82]) and absolute terms (HHF incidence is twofold higher) than those with preserved kidney function. The only study that accurately stratified participants in HF subgroups based on LVEF is the DECLARE-TIMI58 trial [69] showing that the effect of dapagliflozin on reducing the risk of HHF was similar in patients without and with HF, and irrespective of their LVEF value. In contrast, the effect on CV death prevention was particularly evident (almost entirely concentrated) in the subgroup with HFrEF. The DAPA-HF trial in HFrEF demonstrated that dapagliflozin is effective on HHF and cardiovascular death prevention also in subjects without T2D and regardless of either the etiology (ischemic vs. non-ischemic) or the severity of HF (in terms of EF, NT-proBNP levels, and background therapy). In synthesis, as confirmed by a very recent meta-analysis [70], there is very little heterogeneity in terms of HHF prevention among the different subgroups of patients with T2D undergoing treatment with SGLT-2i. Possibly, SGLT-2i could reveal as the only effective treatment in patients with HFpEF [71], wherein several promising drugs proved negative on major endpoints. In fact, the very recent publication of the EMPEROR-Preserved trial has revealed that empagliflozin administration can significantly reduce the number of HHF (HR 0.73; 95% CI 0.61–0.88) in subjects with HFpEF, regardless of the presence or absence of diabetes [72].

In an attempt to compare the results of major clinical trials with SGLT-2i according to the preexisting heart function of the recruited subjects, we exploited the data of 4 trials for which Kaplan Meier curves of HHF were available and performed the PISA analysis, which provides homogeneous, time-dependent, clinically and economically meaningful estimates of treatment effects in positive clinical trials [73]; a brief description of the method is provided in Additional file 1: Appendix S1. As shown in Fig. 2, the kinetics of the gain in months free of HHF is very different in the cohorts with no or low prevalence of heart failure (EMPA-REG, CANVAS, and DECLARE-noHF: 10, 15 and 0%, respectively) with respect to those with HF (DECLARE, DAPA-HF). In the 3 low-risk cohorts, the difference is largely attributable to the prevalence of cardiovascular disease (EMPA-REG 100%, CANVAS 70% and DECLARE-noHF 36%). In pharmacoeconomic terms, the number needed to treat (NNT) to achieve 1 year free of HHF displays very different kinetics, such that after 3 years of treatment it is 36 NNT in EMPA-REG, 51 in CANVAS, and 159 in DECLARE-noHF; however, the difference attenuates with time. The gain in terms of time free of HHF is dramatically greater in the cohorts with HF (Fig. 2, right panels) showing a gradient according to the baseline impairment in LVEF (DECLARE-HFrEF > DECLARE-HF norEF), but no gradient whether T2D is present in 40% or 100% of the population (DAPA-HF vs. DECLARE rEF cohort). The NNT to achieve 1 year free of HHF after 3 years of treatment is 8.4, 8.5 and 22 in these three cohorts. As expected, from a pharmaco economic perspective, the impact of SGLT-2i on HHF prevention will be in proportion of the individual risk of the event, with a substantial difference also between subjects with preserved or reduced EF. The PISA analysis also indicates that in the long term the treatment is likely to be cost effective in most of the different cohorts.

Not applicable.

All the authors gave their consent to publication.

Natali A. has received a grant for an investigator initiated clinical study from Boheringer-Ingelheim and onoraria as speakers from Eli-Lilly, Servier, Astra Zeneca. Ferranini E. has served on the Advisory Board of Boehringer Ingelheim/Lilly&Co., Servier, Lexicon, and Oramed, has received research grants from Boehringer Ingelheim/Lilly&Co and Janssen, and speaker fees from Boehringer Ingelheim/Lilly&Co. and MSD. LN, DT, and AM have no conflict of interest to disclose.

Not applicable.