SGLT2 Inhibition and cardiovascular events: why did EMPA-REG Outcomes surprise and what were the likely mechanisms?

The four trials were not designed as glucose-lowering trials per se since background glucose-lowering medications (other than incretin therapy) were generally allowed to be changed freely in both arms, in line with usual care. DPP-4 inhibitors are glucose-lowering agents that neither increase weight nor cause hypoglycaemia and that have negligible effects on lipids or blood pressure (Table 1). That noted, meta-analyses of shorter-term DPP-4 inhibitor studies seemed to suggest that cardiovascular risk might be significantly lowered in line with as yet unknown ‘pleiotropic’ effects.

The results of the first two DPP-4 inhibitor trials (Saxagliptin Assessment of Vascular Outcomes Recorded in Patients with Diabetes Mellitus [SAVOR]–Thrombolysis in Myocardial Infarction [TIMI] 53 [6] and Examination of Cardiovascular Outcomes with Alogliptin versus Standard of Care [EXAMINE] [7], testing saxagliptin and alogliptin, respectively, as add-on therapies) were presented at the European Society of Cardiology annual meeting in September 2014. These two trials were conducted largely in high-risk patients (secondary prevention or acute coronary syndrome patients) to allow rapid accumulation of cardiovascular events and, consequently, were short in duration (Table 2).

The trials reported modest HbA_1c differences between treatment arms and demonstrated non-inferiority for cardiovascular events (SAVOR–TIMI 53, HR 1.00 [95% CI 0.89, 1.12]; EXAMINE, HR 0.96 [upper boundary of one-sided repeated CI 1.16]). One concern was a slight but significant increase in risk of heart failure hospitalisation (HFH) seen with saxagliptin in the SAVOR−TIMI study. EXAMINE also showed a higher risk of HFH [7] and, although not significant, the pooled results of SAVOR−TIMI and EXAMINE did raise a suspicion that this class of drugs may be associated with an increase in the risk of heart failure. This point has since been heavily debated and further studies or trials with these agents are being planned specifically to examine effects on cardiac structure and function.

The third DPP-4 inhibitor trial was the Trial Evaluating Cardiovascular Outcomes with Sitagliptin (TECOS), which tested sitagliptin in patients with established cardiovascular disease [8]. The duration of this trial was somewhat longer and the difference in HbA_1c between treatment arms was again small (Table 2). The results showed neutrality with respect to cardiovascular outcomes (HR 0.98 [95% CI 0.88, 1.09]) and there was no evidence for an increase in HFH. The fourth trial published in this series of new diabetes agents tested lixisenatide, a short acting GLP-1 receptor agonist, in patients with recent acute coronary syndrome [9]. Given that GLP-1 receptor agonists lower not only glucose but also weight and blood pressure [10], it was hoped that the results would demonstrate cardiovascular benefit. Again, the trial demonstrated non-inferiority for cardiovascular events (HR 1.02 [95% CI 0.89, 1.17]) with no suggestion of benefit. Similarly, HFH and mortality were unchanged.

While the availability of new glucose-lowering agents that have been shown to be safe in robust trials has substantially expanded the treatment options available, it appears that demonstrating not only non-inferiority but actual clinical benefit, in the context of falling cardiovascular event rates, would require even larger studies with longer follow-up than have yet been considered. Although assessing safety is critically important, trials of agents with modest effects on HbA_1c which are conducted in specific high-risk populations (e.g. those with recent acute coronary syndrome, arguably limiting the ability to impact upon major cardiovascular events within the first year) over relatively short durations seem unlikely to have the capacity to show benefit, even if the agent in question is superior to standard practice. It is important to stress that the results of Liraglutide Effect and Action in Diabetes: Evaluation of Cardiovascular Outcome Results (LEADER; testing liraglutide; ClinicalTrial.gov registration no. NCT01179048 [11]) and Exenatide Study of Cardiovascular Event Lowering Trial (EXSCEL; testing once-weekly exenatide; ClinicalTrial.gov registration no. NCT01144338) are eagerly awaited since they involve GLP-1 receptor agonists that have a longer duration of action than lixisenatide and have stronger glucose- and weight-lowering effects [12]. Hence, the effect of GLP-1 receptor agonists on cardiovascular outcomes should remain an open question for now.

The aforementioned neutral results of trials might explain why, when EMPA-REG Outcomes data were reported at the EASD annual meeting in Stockholm in 2015, the audience were highly surprised with the reductions in cardiovascular mortality, all-cause mortality and HFH for empagliflozin vs placebo. Could these results have been predicted based on what we knew about the mode of action? At the level of commonly considered risk factor changes, the answer is no—see Table 1. While SGLT2 inhibitors lower not only glucose but also weight and blood pressure, their effects on lipids are mixed with parallel rises in LDL-cholesterol and HDL-cholesterol [13]. Such risk factor patterns therefore made it impossible to predict the net effect of SGLT2 inhibitors on cardiovascular outcomes.

EMPA-REG Outcomes tested empagliflozin at two doses (10 and 25 mg daily) vs placebo in patients with existing cardiovascular disease over about 3 years. In keeping with what is known about the drug’s efficacy, active treatment compared with placebo led to reductions in HbA_1c (by ~0.3–0.5%), weight (by ~2 kg) and systolic blood pressure (by ~3 mmHg) without any compensatory increase in heart rate [5]. There were also slight rises in LDL-cholesterol and HDL-cholesterol levels as expected. The primary outcome of this trial was the ‘MACE’ (major adverse cardiac events) composite of death from cardiovascular causes, non-fatal myocardial infarction or non-fatal stroke, recommended in the US Food and Drug Administration’s guidance on evaluating new glucose-lowering agents and used in most of the other trials discussed earlier. While the primary outcome was reduced by empagliflozin (HR 0.86 [95 CI 0.74, 0.99]), myocardial infarction was not significantly reduced (HR 0.87 [95% CI 0.70, 1.09]), although it was directionally concordant, and stroke was non-significantly increased (HR 1.18 [95% CI 0.89, 1.56]) despite the fall in blood pressure [5]. The pattern of results from other pre-specified outcomes revealed something rather different. Empagliflozin significantly lowered death from cardiovascular causes (by 38%), HFH (by 35%) and death from any cause (by 32%) [5]. Although one must be cautious in interpreting these results on the basis that they were secondary outcomes and while further trial evidence is crucial, such a substantial reduction in all-cause death in particular gives confidence in their validity. It was the reduction in cardiovascular death that primarily drove the reduction in the primary endpoint. These benefits on mortality and HFH emerged very rapidly, almost immediately. Reassuringly, other than the expected increase in genital infections with empagliflozin (6.4% vs 1.8%), there was no increase in rates of diabetic ketoacidosis, fractures or hypoglycaemia [5]. There is now considerable interest in determining the potential underlying mechanisms for the EMPA-REG Outcomes findings.

A non-atherothrombotic explanation for the benefits is suggested by two important aspects from the EMPA-REG Outcomes trial: first, by substantial reductions in HFH and cardiovascular mortality rather than any clear effect on myocardial infarction or stroke risk; second, by the very rapid emergence of benefit as shown by the separation of Cumulative Incidence curves for cardiovascular mortality and HFH following randomisation [5]. To better understand what may have happened, we need to revisit the pathophysiology of heart failure.

Heart failure is defined by the inability of the heart to deliver sufficient oxygen to peripheral organs and clinically as a syndrome defined by the presence of symptoms such as ankle swelling, dyspnoea and fatigue and signs such as elevated venous jugular pressure and pulmonary crackles. It can arise as a result of almost any abnormality of the structure, mechanical function or electrical activity of the heart and is the common, end-stage, manifestation of many cardiovascular diseases including myocardial infarction. Once heart failure ensues, reduced systolic (contraction) and/or diastolic (relaxation) function may lead to sodium and fluid retention, due to renal haemodynamic changes and as a result of activation of the renin—angiotensin—aldosterone system, sympathetic nervous system (SNS) and other neurohumoral systems. SNS activation occurs in an attempt to maintain cardiac output, through increased contractility (stroke volume) and heart rate, and redistribute blood flow (but at the expense of enhanced systemic vasoconstriction). Heart failure is characterised by a trajectory of deteriorating cardiac output and declining renal function leading to fluid retention, peripheral oedema and pulmonary congestion, which if severe may lead to hospitalisation and treatment with intravenous diuretic. Of course there is a strong link between heart failure and sudden death: dilated and hypertrophied cardiac chambers fail and fibrillate, so patients may die from either pump failure or sudden arrhythmia.

At baseline, 10% of EMPA-REG Outcomes participants were clinically recorded as having heart failure, and it is likely that a further proportion of this high cardiovascular risk population had subclinical cardiac dysfunction. The results of the trial suggest that empagliflozin somehow interrupts one or more of the mechanisms involved in the progression of cardiac dysfunction mentioned above. We postulate that the relevant effect or effects of empagliflozin may involve the mechanisms included in the text box below.

Whether the blood pressure-lowering effects of empagliflozin alone can explain the observed benefit is worthy of consideration since reducing blood pressure (even by small amounts) can influence heart failure risks, although not necessarily as rapidly as seen in EMPA-REG Outcomes. Empagliflozin also induces a considerable diuresis, with early loss of urinary glucose in particular and subsequently of sodium as reflected by a sizeable increase in the haematocrit (~4%) compared with placebo, and certainly adequate dosage of diuretics can rapidly reduce heart failure risks, or risks for HFH [14]. Notably, in the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT), in which just over one-third of patients had type 2 diabetes, use of the thiazide diuretic chlorthalidone was associated with a 28% (95% CI 20, 34) lower risk of heart failure compared with amlodipine despite only a 0.8 mmHg difference in achieved systolic blood pressure [15]. That noted, there were no differences in other vascular endpoints.

We consider that the cardiovascular benefit of empagliflozin is related to the manner in which it induces diuresis (both glucose and sodium losses), notably with a reduction in the progression to renal failure and with a slowing in the deterioration of renal function. Type 2 diabetes is typified by upregulated SGLT2 tubular transporters and increased tubular glucose reabsorption, along with sodium reabsorption in the proximal renal tubules, leading to decreased sodium delivery to the macula densa, vasodilatation of the afferent arteriole with vasoconstriction of the efferent arteriole and resultant intraglomerular hypertension. The combination of the diuretic effect and increased sodium delivery to the macula densa [16], thereby helping to address the maladaptive arteriolar responses, may explain the finding of both cardiac and renal benefit of SGLT2 inhibition. In essence, improving renal sodium and glucose handling, with subsequent reductions in fluid burden especially in individuals with or susceptible to cardiac dysfunction, may have been the key driver underlying the benefits seen in EMPA-REG Outcomes (Fig. 1). There is a need for detailed mechanistic studies incorporating state of the art imagining techniques to better determine the full cardiovascular actions of empagliflozin and other SGLT2 inhibitors in patients with diabetes. Such studies should involve a range of diabetes patients with and without cardiovascular disease and with and without existing heart failure. In addition, major trials will be required to establish whether SGLT2 inhibitor therapy, with its proposed novel diuretic-like action, might provide clinically meaningful benefit in lower-risk populations.

There have been secular decreases in myocardial infarction and stroke rates in patients with diabetes, reflecting considerable improvements in lipid and blood pressure management over the last two decades. Indeed, in EMPA-REG Outcomes, more than 75% of patients were on statins and around 95% on antihypertensive agents at baseline, and baseline LDL-cholesterol and blood pressure results would be considered excellent in most clinical settings. However, heart failure remains an under-recognised complication of diabetes and one which carries a poor prognosis [17]. In a recent observational analysis of patients with diabetes, heart failure was found to be the second-most common type of first cardiovascular disease presentation (after peripheral vascular disease) [18]. In diabetes, the pathogenesis of heart failure is likely to be multi-factorial given the presence of risk factors including dysglycaemia, coronary heart disease, hypertension, obesity, renal dysfunction and others. It is therefore increasingly apparent that therapeutic approaches that also reduce the development and progression of heart failure (i.e. beyond atherothrombotic targeting) are crucial in patients with diabetes.

Why stroke risk in EMPA-REG Outcomes did not decline on empagliflozin, when the risk of stroke is normally very sensitive to reductions in blood pressure, is not clear. Meta-analysis of trials suggests that stroke risks can be significantly lowered even when baseline systolic blood pressure is <140 mmHg [19] and so further investigations are clearly required. That the haemoconcentration effect of SGLT2 inhibition may have balanced out the blood pressure-lowering effect is plausible, and results from other major ongoing trials will yield valuable data.

The effects of empagliflozin on important secondary endpoints in EMPA-REG Outcomes, namely HFH, cardiovascular and all-cause death, were unexpected. The rapid emergence of these benefits points strongly towards non-atherothrombotic mechanisms, perhaps principally haemodynamic effects. Further mechanistic studies will be needed to identify these mechanisms and further clinical trials of SGLT2 inhibitors in patients with and without heart failure, with and without cardiovascular disease and, potentially, with and without diabetes, will likely ensue. Finally, we will also shortly know whether these findings are common to other drugs in the class. It appears that a new era in diabetes–cardiovascular research has emerged.