Metformin and heart failure: never say never again

Abstract

Metformin represents the cornerstone of treatment for type 2 diabetes mellitus. Traditionally, heart failure (HF) was considered a contraindication to its use. However, more recent evidence has shown that this should no longer be the case. Indeed, studies have demonstrated that metformin may even reduce the risk of incident HF and mortality in diabetic patients, while improving up to 2-year survival rates in those with HF. In addition, it appears to exert cardioprotective actions. Although longer follow-up data and more explicit information about the situation in patients with very advanced HF are needed, the cardiac safety of metformin has profound clinical implications and may be anticipated to further encourage its widespread use.

Keywords:

• contraindications
• diabetes mellitus
• heart failure
• metformin
• oral hypoglycaemic agents

1. Introduction

Metformin has long been established as the mainstay of treatment for type 2 diabetes mellitus (T2DM) [1,2]. Not only does it reduce hyperglycaemia but it is not associated with unwanted hypoglycaemia (unless used along with insulin or excessive exercise), and it has favourable actions on body weight and serum lipids [3]. More importantly, based on an observational study of participants in a large randomised controlled trial (RCT), it confers long-term protection from all T2DM-related endpoints, myocardial infarction and death from any cause [4]. Traditionally, the use of metformin has been subject to contraindications [5-7]. These have mainly included renal failure and conditions predisposing to hypoxia and lactate accumulation, notably liver disease, heart failure (HF) and severe pulmonary disease [5-7]. However, the validity of such contraindications has been questioned [8-10]. Indeed, there is now evidence that metformin: i) reduces the incidence of new HF and cardiovascular mortality; and ii) improves survival in patients with HF.

2. Reduced incidence of new heart failure and cardiovascular mortality

In a Canadian health database, metformin use was associated with reduced all-cause mortality [odds ratio (OR) 0.60, 95% confidence interval (CI) 0.49 – 0.74] and deaths from cardiovascular disease (OR 0.64, 95% CI 0.49 – 0.84) compared with sulfonylureas (Table 1) [11]. The same beneficial effect was seen with metformin + sulfonylurea combination therapy compared with sulfonylurea alone (OR 0.66, 95% CI 0.58 – 0.75 for all-cause mortality; OR 0.64, 95% CI 0.54 – 0.77 for cardiovascular deaths) [11].

Table 1. Main studies involving metformin and heart failure (HF).

Study [Ref.]	Design	Follow-up	Main findings
Reduced incidence of HF
McAlister et al. [12]	Retrospective observational cohort study	4.65 years (mean)	Significantly (p < 0.001) reduced incidence of HF with metformin
Tzoulaki et al. [14]	Retrospective cohort study based on a general practice database	7.1 years (mean)	Compared with metformin, first- and second-generation sulfonylureas increased congestive HF (adjusted HR 1.46 and 1.30, respectively)
Pantalone et al. [15]	Cohort study based on electronic health records	72 months	Metformin reduced HF (HR 0.76) and mortality (HR 0.54)
Reduced mortality in patients with HF
Eurich et al. [16]	Retrospective cohort study of health service data	2.5 years (mean)	Metformin reduced all-cause mortality (adjusted HR 0.70)
Masoudi et al. [17]	Retrospective cohort study of health service data	1 year	Metformin reduced crude mortality (OR 0.86)
Inzucchi et al. [18]	Retrospective cohort study of health service data	1 year	Metformin reduced mortality (HR 0.92)
Karter et al. [19]	Retrospective cohort study of health service data	10.2 months (mean)	Metformin reduced hospitalisation for HF (HR 0.70)
Eurich et al. [20]	Pooled data analysis	1 year	Metformin reduced all-cause hospital admissions (OR 0.85)
MacDonald et al. [21]	Nested case-control study	2.8 years (mean)	Metformin reduced mortality (adjusted OR 0.65)
Shah et al. [22]	Cohort study of diabetic patients with advanced systolic HF	2 years	Metformin increased survival rates (p = 0.007)
Andersson et al. [23]	Nationwide retrospective cohort study	844 days (median)	Metformin reduced all-cause mortality (adjusted HR 0.85)
Roussel et al. [24]	Cohort study	2 years	Metformin reduced mortality rates (adjusted HR 0.69)
Evans et al. [25]	Cohort study of health service data	1 year	Metformin reduced mortality (OR 0.59)
Aguilar et al. [26]	Nationwide cohort study	2 years	Metformin reduced mortality (HR 0.76) and hospitalisation due to HF (HR 0.93)

HF: Heart failure; HR: Hazard ratio; OR: Odds ratio.

More recently, the data on the favourable effect of metformin have been strengthened. After a mean follow-up of 4.65 years, a retrospective observational Canadian cohort study of all adults without HF treated with oral hypoglycaemic agents [12] found that the incidence of HF was higher among patients using sulfonylurea monotherapy (4.4 cases per 100 treatment years) than those taking metformin monotherapy (3.3 cases per 100 years, p < 0.001). Sulfonylurea treatment had an unadjusted hazard ratio (HR) of 1.32 (95% CI 1.13 – 1.54, p < 0.001) in comparison to metformin. However, after adjustment for baseline differences between the groups, the risk of HF associated with sulfonylurea was no longer significant (adjusted HR 1.16, 95% CI 0.96 – 1.41) [12]. Still, high-dose sulfonylurea conferred a risk of incident HF compared with high-dose metformin (HR 1.38, 95% CI 1.20 – 1.60) [12].

In a US cross-sectional analysis of military health service data [13], comparison of cardiovascular outcomes by antidiabetic prescription drug classes revealed that the lowest average annual incidence of acute myocardial infarction was seen with metformin (33.09/10,000) and the highest with insulin (51.67/10,000) and sulfonylureas (48.58/10,000); the p value was not provided. Similar results were observed for the annual incidence of HF (metformin 67.82/10,000, insulin 114.10/10,000, sulfonylureas 104.59/10000; p value not provided) [13].

In 2009, two large, retrospective cohorts provided similar results [14,15]. In a retrospective cohort study based on a general practice database in the UK [14], after adjustment for potential confounders, first-generation sulfonylureas increased the risk of myocardial infarction (HR 1.37, 95% CI 1.15 – 1.62, p = 0.0003 and HR 1.27, 95% CI 1.07 – 1.50, p = 0.007, depending on the adjustment model), compared with metformin. Second-generation sulfonylureas also increased the risk of myocardial infarction (HR 1.31, 95% CI 1.21 – 1.43, p < 0.001 and HR 1.25, 95% CI 1.15 – 1.36, p < 0.001, depending on the adjustment model) [14]. Compared with metformin, first-generation sulfonylureas increased the risk of a first episode of congestive HF (HR 1.46, 95% CI 1.32 – 1.63, p < 0.001 and HR 1.29, 95% CI 1.17 – 1.44, p < 0.001, depending on the adjustment model) [14]. Likewise, second-generation sulfonylureas increased the risk of a first episode of congestive HF (HR 1.30, 95% CI 1.22 – 1.38, p < 0.001 and HR 1.19, 95% CI 1.12 – 1.27, p < 0.001, depending on the adjustment model) [14]. In a US cohort [15], multivariate analysis showed that metformin administration was associated with a reduced risk of HF (HR 0.76, 95% CI 0.64 – 0.91) and mortality (HR 0.54, 95% CI 0.46 – 0.64) compared with sulfonylurea therapy.

3. Improved survival in patients with heart failure

In a Canadian database including 12,272 subjects newly embarking on oral hypoglycaemic agents between the years 1991 and 1996 [16], 1833 subjects with incident HF were identified. Metformin monotherapy and metformin + sulfonylurea combination therapy were shown in multivariate regression analysis to reduce all-cause mortality both at 1 year (adjusted OR 0.66, 95% CI 0.44 – 0.97 and adjusted OR 0.54, 95% CI 0.42 – 0.70, respectively) and after a mean follow-up of 2.5 years (adjusted OR 0.70, 95% CI 0.54 – 0.91 and adjusted OR 0.61, 95% CI 0.52 – 0.72, respectively) [16]. Although this work showed the beneficial effect of metformin in HF, its limitations include the retrospective observational design and the absence of data on severity of HF and on the presence or otherwise of renal failure [16].

In the US, Masoudi et al. [17] carried out a retrospective cohort study of 16,417 diabetic Medicare beneficiaries discharged after hospitalisation for HF from April 1998 to March 1999 or from July 2000 to June 2001. In this cohort, crude 1-year mortality was significantly (p = 0.0001) lower in metformin-treated patients (24.7%) than in those not receiving insulin-sensitising agents [17]. Again, metformin administration was linked with reduced mortality (OR 0.86, 95% CI 0.78 – 0.97) [17]. The study's limitations are a potential selection bias (aged > 65 years), short follow-up and uncertain drug exposure, given that the analysis is based on hypoglycaemic agent prescription at discharge but no verification of its continued prescription [17].

A further retrospective report included 24,953 diabetic Medicare beneficiaries discharged after hospitalisation for acute myocardial infarction between April 1998 and March 1999 or between July 2000 and June 2001 [18]. Again, it was shown that 1-year mortality was insignificantly lower in patients receiving metformin (HR 0.92, 95% CI 0.81 – 1.06) but significantly reduced in those treated with metformin + thiazolidinedione (HR 0.52, 95% CI 0.34 – 0.82). Mortality was not increased in patients admitted with HF or pulmonary oedema (HR 0.96, 95% CI 0.78 – 1.19) nor in those re-admitted for HF (HR 0.87, 95% CI 0.73 – 1.04) [18]. The limitations of this work include a potential selection bias (subjects > 65 years following acute myocardial infarction), short follow-up and uncertain drug exposure (the analysis is based on hypoglycaemic agent prescribed at discharge but its continued prescription during the 1-year period could not be verified) [18]. Interestingly, an enquiry into the cardiac risk associated with pioglitazone initiation [19], reported a significantly lower incidence (HR 0.70, 95% CI 0.49 – 0.99) of hospital admissions for HF among metformin users, although it might be criticised for its relatively short follow-up (mean 10.2 months).

In 2007, a systematic review including pooled data analysis [20], found that metformin reduced all-cause hospital admissions at 1 year compared with other treatments (OR 0.85, 95% CI 0.76 – 0.95, p = 0.004). The authors concluded that metformin was the only hypoglycaemic agent not associated with harm in patients with HF and, indeed, led to reduced all-cause mortality in two individual studies [20].

A nested, case-control study of subjects ≥ 35 years of age and newly diagnosed with diabetes and HF (as well as matched controls) [21] showed that metformin monotherapy/combination therapy reduced mortality in the UK. After a mean duration of diabetes and HF of 2.8 years, adjusted OR for mortality with metformin monotherapy was 0.65 (95% CI 0.48 – 0.87) and with metformin in general (i.e., monotherapy or combination therapy) adjusted OR for mortality was 0.72 (95% CI 0.59 – 0.90) [21]. A further cohort study included diabetic patients with advanced systolic HF [mean ejection fraction 24%, 42% of patients classed as New York Heart Association (NYHA) III and 45% as NYHA IV] [22]. Metformin-treated patients exhibited significantly (p = 0.007) improved 1-year survival rates (91%) vs. those not receiving metformin (76%) with (HR 0.37, 95% CI 0.18 – 0.76) [22]. At 2 years, survival rates with and without metformin were 78% and 63%, respectively (p = 0.007). After multivariate adjustment for potential confounders, metformin was associated with an insignificant trend for improved survival: adjusted HRs for mortality at 1 and 2 years were 0.63 (95% CI 0.21 – 1.89, p = 0.40) and 0.79 (95% CI 0.36 – 1.71, p = 0.54), respectively [22].

A nationwide Danish retrospective cohort study of diabetic patients with HF and median observational time 844 days [23] demonstrated reduced all-cause mortality with metformin monotherapy (adjusted HR 0.85, 95% CI 0.75 – 0.98, p = 0.02) and metformin + sulfonylurea combination therapy (adjusted HR 0.89, 95% CI 0.82 – 0.96, p = 0.003), using sulfonylurea monotherapy as reference.

In a further investigation of diabetic patients with established atherothrombosis [24], 2-year mortality rates were 6.3% with metformin and 9.8% without metformin (adjusted HR 0.76, 95% CI 0.65 – 0.89, p < 0.001). Of particular note, the association with lower mortality was also seen among patients with known congestive HF (adjusted HR 0.69, 95% CI 0.54 – 0.90, p = 0.006) [24].

Furthermore, a Scottish study including diabetic patients with newly diagnosed HF receiving oral hypoglycaemic agents [25] confirmed reduced mortality in those receiving metformin, with or without sulfonylureas, both at 12 months (OR 0.59, 95% CI 0.36 – 0.96) and long term (OR 0.67, 95% CI 0.51 – 0.88).

More recently, Aguilar et al. [26] have used propensity score-matched samples to explore the relationship between metformin use and the risk of death or of hospitalisation in a cohort of diabetic patients with HF treated in ambulatory clinics at Veteran Affairs medical centres. At 2 years, significantly (p < 0.001) fewer deaths occurred in metformin-treated patients (15.8%) than in those not receiving metformin (25.5%) [26]. Metformin was linked with significantly reduced risk of death (HR 0.76, 95% CI 0.63 – 0.92; p < 0.01) and slightly reduced risk of hospitalisation due to HF (HR 0.93, 95% CI 0.74 – 1.18) [26].

3.1 Proposed mechanisms

The potential mechanisms underlying the beneficial effect of metformin are now receiving attention [27-30]. Three mechanisms have hitherto been proposed. Based on experimental evidence [30-35], the main mechanism appears to be the activation of adenosine monophosphate-activated protein kinase (AMPK) [36]. In experimental models of diabetes and HF, metformin activates AMPK, leading to its increased phosphorylation, which, in turn, results in increased phosphorylation of endothelial nitric oxide synthase (eNOS) [30-33,35]. Thus, plasma and myocardial levels of nitric oxide increase [30,31], ultimately resulting in improved endothelial function, preservation of left ventricular ejection fraction [30,31,33,35], reduced left ventricular dilatation [30,33,35], improved left ventricular end-diastolic pressure [31,35] and also reduced myocardial autophagy [34] and apoptosis [31]. Importantly, the favourable actions of metformin are abolished in mice lacking functional AMPK and/or eNOS [30].

The second mechanism appears to be reduced cardiac fibrosis [37]. In a mouse HF model, metformin reduced pressure overload-induced cardiac fibrosis by inhibiting the action of transforming growth factor (TGF)-β1 in cardiac fibroblasts [37]. This inhibition was mediated by suppressing the phosphorylation of Smad3 [37], the downstream effector of TGF-β1, which is known to play a pivotal role in cardiac remodelling [38]. This second mechanism needs further elucidation and confirmation, given that other experimental work showed that metformin reduced triacylglycerol content in the myocardium but had no effect on fibrosis [39]. Interestingly, there may be a relationship between these two mechanisms, as increased AMPK and eNOS phosphorylation have been shown to reduce TGF-β1 and basic fibroblast growth factor (bFGF) in both the circulation and the myocardium [35].

The third mechanism is increased glucose utilisation via its oxidation in myocardial cells [40,41]. Indeed, metformin has been reported to increase myocardial GLUT1 and GLUT4 cell surface transporters, facilitating intracellular glucose uptake and utilisation [41]. By restoring glucose oxidation, metformin improves myocardial energy supply and inhibits free fatty acid (FFA) peroxidation [40,41]. Circulating FFAs may be increased especially in hyperadrenergic states, notably HF, and exert cardiotoxic actions [40].

3.2 Critical interpretation and implications for clinical practice

The limitations of current knowledge necessitate caution in the interpretation of promising findings from the above-mentioned studies. Indeed, the bulk of evidence is based on retrospective observational cohort studies [11-14,17,20-24]. Clearly, an RCT of metformin vs. placebo would provide more robust data but, as has already been reported, such a trial is virtually impossible [42]. The reason why such a trial is not feasible lies in the current absence of any clinical uncertainty about the efficacy and safety of metformin in HF, which was ultimately shown to undermine the utility of the trial and precluded randomisation. The authors of the report suggested that a comprehensive Phase IV prospective evaluation of metformin in HF might provide the most conclusive answer [42]. Second, selection bias in the subjects included cannot always be excluded. A third important limitation is the relatively short follow-up, usually 1 year [15,17,19,20] or 2 years [22-24,26], although mean follow-up was 4.65 years in the McAlister et al. study [12]. Fourth, sustained exposure to metformin was not absolutely certain in some studies [18-20]. Moreover, there is almost no information on severity of HF (e.g., according to NYHA) and whether this influences outcomes. Furthermore, although some workers have performed multivariate analyses to exclude potential confounding factors [20-26], the latter have not always been entirely convincingly accounted for. Fifth, and final, most of the information on the mechanisms of action for metformin are currently derived from experimental research [30-35,37,39] and need further clinical study.

A further important point to consider is renal failure, which is frequently encountered in HF. Renal failure may increase the concentration of metformin and the risk of untoward effects, including lactic acidosis [5,6,43]. Unfortunately, studies in HF have so far provided little information on the presence of renal insufficiency. However, clinicians need to assess renal failure before prescribing metformin in HF and exclude severe renal failure.

The practical implications may be outlined as follows. The presence of HF no longer represents a contraindication to metformin use [17,27,29]. Indeed, the US Food and Drug Administration (FDA) has removed HF as an absolute contraindication to metformin use [44,45]. Furthermore, the Canadian Diabetes Association 2008 clinical practice guidelines recommend metformin as first-line therapy in HF [46]. The 2011 American Diabetes Association (ADA) standards of medical care suggest that metformin may be used in stable HF, provided that renal function is normal [47]. In this context, it is relevant that diabetes severity in the individual patient confers a higher risk of HF rather than any particular hypoglycaemic agent [48,49]. Thus, particular caution is needed in unstable patients, in whom an acute decompensation is very likely, in which case metformin may become temporarily less safe to use [5,6]. Of greater importance, the safety of metformin in patients with very advanced HF and poor condition must be demonstrated more explicitly.

In conclusion, HF may no longer be considered an absolute contraindication for metformin [17,27,29]; conversely, there is ample evidence to suggest that this agent might reduce the risk of incident HF and mortality in T2DM [11-16,18,19] and also improve 1-year survival rates in diabetic patients with HF [14,21-26]. In addition, metformin may exert cardioprotective actions, possibly by AMPK activation and reduced cardiac fibrosis [30-35,37]. However, we need longer follow-up data and more information as to the role of the severity of HF. New knowledge on the cardiac safety of metformin will further enhance its importance as the cornerstone of treatment for T2DM.

4. Expert opinion

Metformin is, and should remain, the first-line therapy for T2DM [1,2]. It reduces hyperglycaemia, exerts beneficial actions on body weight and serum lipids and reduces long-term T2DM-related morbidity and mortality [3,4]. It was previously believed that HF represented a contraindication to metformin use [5,7]. This, however, is no longer the case. Conversely, studies have shown that metformin reduces the risk of developing HF in DM [12-16]. More impressively, in case of HF it may prolong survival for up to 2 years [17]. This is important, given that 10 – 25% of T2DM patients receiving metformin may suffer from HF [50-52].

The clinical implications are that metformin should no longer be withheld from patients with HF. In such patients, metformin treatment may indeed be anticipated to contribute to some reduction in the dosage of sulfonylureas and/or insulin and to some improvement in body weight [1,2]. These widely known beneficial effects will be added to the above-mentioned favourable effect on mortality [1,2,17]. Thus, the argument for metformin prescription in HF is quite strong. However, renal failure is not infrequent in HF, increasing the frequency of adverse effects, most notoriously acidosis [5,6,43]. Hence, while prescribing metformin in HF, clinicians need to be diligent about controlling renal function and ruling out renal failure.

There is evidence that the cardioprotective actions of metformin are largely due to AMPK activation [36]; reduced cardiac fibrosis [37] and increased myocardial glucose utilisation [41] also appear to play an important role. Researchers are further exploring these mechanisms and we expect to improve our knowledge in the near future. Although the precise underlying mechanisms may be of little relevance to the average clinician, they enhance the argument for metformin administration in HF, providing a pathophysiological basis for this argument.

In conclusion, T2DM patients with HF stand to benefit from metformin, and treatment with this agent should – generally – not be interrupted because of HF [44-47]. What remains to be further clarified is whether the benefit in mortality conferred by metformin in HF is sustainable in the longer term and which patient category will benefit most.

Declaration of interest

This editorial was written independently. No company or institution supported it financially. N Papanas has been an advisory board member of TrigoCare International, has participated in sponsored studies by Novo Nordisk and Novartis, has received honoraria as a speaker for Novo Nordisk and Pfizer and has attended conferences sponsored by TrigoCare International, Novo Nordisk, sanofi-aventis and Pfizer. E Maltezos has participated in sponsored studies by Novo Nordisk and Novartis, and has attended conferences sponsored by Wyeth, Pfizer and Bayer. DP Mikhailidis has been an advisory board member and speaker for Merck Sharp & Dohme (MSD) and has attended conferences sponsored by AstraZeneca, sanofi-aventis and MSD.