Background

Heart failure (HF) has significant morbidity and is often a result of impaired left ventricular myocardial function [1]. HF with preserved ejection fraction (HFpEF) involves impaired myocardial function with normal left ventricle size and ejection fraction; in contrast, HF with reduced ejection fraction (HFrEF) involves an enlarged left ventricle size and reduced ejection fraction. Evidence-based HF treatment reduces morbidity and mortality in HFrEF [2]. HFpEF prevalence is rising due to an ageing population, however, there are no treatments which reduce morbidity and mortality [3]. Diagnosing HFpEF is often confounded by the occurrence of similar symptoms in patients with multiple medical comorbidities [3]. The most prevalent risk factor for HFpEF is hypertension [3]. Several RCTs have explored the benefits of β-blockers [4], ARBs [5], ACEi [6], and mineralocorticoid receptor antagonists (MRAs) [7] in HFpEF and identified trends towards reduced cardiovascular morbidity and mortality [8]. The lack of strong evidence in HFpEF treatment has led to considerable treatment variation [9].

MRAs can be selective (e.g., eplerenone) or nonselective (e.g., spironolactone). Eplerenone was synthesized through chemical modification of spironolactone in order to enhance binding of mineralocorticoid receptors while reducing off-target binding to progesterone or androgen receptors [10]. Eplerenone is associated with lower rates of impotence, gynecomastia or breast pain in comparison to spironolactone [11, 12].

MRAs found initial use in HF exacerbations as diuretics in patients’ refractory to combined ACEi and loop diuretic therapy [13]. However, spironolactone at doses with no significant diuretic effect was found to reduce cardiovascular mortality [14]. This effect was presumably due to a reduction in myocardial and vascular fibrosis [14]. This effect may arise from spironolactone blocking aldosterone’s ability to stimulate collagen synthesis at the myocardial level [15]. Spironolactone and eplerenone have demonstrated significant mortality benefit in HFrEF [11, 12]. In contrast, MRAs in HFpEF do not reduce all-cause mortality, however, they do reduce hospitalizations, improve quality of life, and improve echocardiographic measurements of diastolic function [16].

Chronically elevated aldosterone levels contribute towards structural changes in the heart which promote water retention, myocardial fibrosis, and increased arrhythmogenicity [17]. MRAs in HFpEF improved echocardiographic and biochemical measures of diastolic function [16, 18]. However, a large prospective RCT in HFpEF patients treated with spironolactone did not demonstrate a significant benefit in terms of cardiovascular outcomes [7].

Objectives

Our objectives were to evaluate the risks and benefits of MRA usage in adults with HF. We were particularly interested in differences between selective and nonselective MRAs in HFpEF and HFrEF in terms of cardiovascular outcomes and adverse effects.

Methods

Our systematic review and meta-analysis complies with the PRISMA statement [19].

Eligibility criteria

We included randomized controlled trials (RCTs) of MRAs vs. placebo or standard therapy in adults (≥18 years old) with HFpEF or HFrEF. Included trials evaluated nonselective MRAs (e.g., canrenone, spironolactone), and selective MRAs (e.g., eplerenone, finerenone). Included trials contained at least one outcome of interest: mortality (all-cause or cardiovascular), cardiovascular hospitalizations, hyperkalemia, or gynecomastia.

Literature search

We searched the Cochrane Central Register of Controlled Trials (The Cochrane Library Issue 1, 2016), MEDLINE (January 1995 to January 29, 2016), and EMBASE (January 1995 to January 29, 2016) for articles meeting our inclusion criteria. Our search strategy for Ovid MEDLINE and EMBASE is in Appendix 1 and our search strategy for the Cochrane Register of Controlled Trials is in Appendix 2. Our search did not have any language restrictions. We excluded reviews, editorials, and conferences but not unpublished studies or abstracts.

Study selection

We entered the retrieved citations into Reference Manager (v12.0.3), and duplicate records were removed. One investigator (NB) screened citations for relevance based on their title and abstract. Both investigators reviewed the full text articles of relevant articles for study inclusion. Cohen’s kappa statistic was used to quantify chance-corrected agreement between the investigators. Disagreements on study inclusion were resolved through a consensus process of having a discussion between the two investigators.

Data collection and analysis

Both investigators extracted data independently from included articles. We resolved disagreements during data extraction by consensus. If data were incomplete or unclear we attempted to contact trial authors. We extracted the following items from each study: population (type of heart failure, study size), intervention (MRA type), control (placebo, none, other), and outcomes (all-cause mortality, cardiovascular mortality, hospitalizations, hyperkalemia, and gynecomastia/breast pain). We used each study’s definition of these outcomes.

Risk of bias

Our risk of bias assessment was completed using the Cochrane Risk of Bias Tool. It evaluates individual studies for several biases: selection, performance, detection, attrition and reporting. We evaluated the quality of evidence for each outcome using GRADE criteria [20], which evaluates an outcome across studies based on risk of bias, inconsistency, indirectness, imprecision and publication bias.

Statistical analysis

We obtained the relative risk for each outcome from the original study and used RevMan (version 5.3.5) and R [21, 22] to analyze data and generate figures. We used the Mantel-Haenszel method with a 95% confidence interval, and a random effects model to pool results. We quantified statistical heterogeneity using the I2 statistic. We interpreted an I2 value of 0–25% as low heterogeneity, 25–50% as moderate heterogeneity, and >50% as high heterogeneity. A priori we established two hypotheses to explain potential heterogeneity: HF type (HFpEF and HFrEF), and MRA type (selective, or nonselective). We assessed for publication bias using funnel plots for each outcome.

Results

Trial selection

We screened 2566 citations, and selected 36 for full text review. Of these, 15 articles [7, 11, 12, 18, 23–33] met our inclusion criteria and were included in our systematic review (see Fig. 1). Overall, there was excellent agreement on trial eligibility (Cohen’s kappa 94%). We excluded articles from the systematic review because of treatment in a non-HF setting (N = 4), lack of relevant outcomes (N = 13), study duplication (N = 3), and not an RCT design (N = 1).

Fig. 1
figure 1

Study selection flow diagram. Overview of process used to identify studies for inclusion in the systematic review. Three databases (MEDLINE, EMBASE, Cochrane) were searched for relevant articles. After identification, studies were screened against our inclusion criteria. Included studies were used in our meta-analysis

Full size image

Trial characteristics

Table 1 reports the trial characteristics of the 15 RCTs containing 16321 patients. The patients had either HFpEF (N = 4027) or HFrEF (N = 12294) and the MRA treatment group was either nonselective, e.g., canrenone, spironolactone, N = 11 RCTs, 6678 patients; or selective, e.g., eplerenone, N = 4 RCTs, 9643 patients. Studies had an average length of follow-up of 15 months.

Table 1 Overview of trials meeting systematic review inclusion criteria
Full size table

Risk of bias within included trials

Table 2 reports the quality of included studies. Five trials had unclear or absent allocation concealment [23, 25, 26, 28, 30]. Two studies had inadequate blinding and were of single-blind design [23, 32]. Two large studies were terminated early due to meeting pre-defined benefit criteria [11, 33]. Another two studies did not use intention-to-treat analysis. Overall, loss-to-follow-up was low with a range of 0 to 6.6%.

Table 2 Risk of bias summary for each study included in the meta-analysis
Full size table

Results of meta-analysis

Table 3 reports a summary of findings. We included outcomes for cardiovascular death (7 RCTs), all-cause mortality (12 RCTs), cardiac hospitalization (10 RCTs), hyperkalemia (15 RCTs), and gynecomastia (N = 11 RCTs). Quality of evidence for cardiovascular death, all-cause mortality, and cardiac hospitalization were rated moderate; hyperkalemia and gynecomastia were rated high using GRADE guidelines [20]. For each outcome, HFrEF evidence was of high quality, but the quality of evidence for HFpEF was of moderate quality for all-cause mortality, cardiovascular death, and cardiac hospitalizations.

Table 3 Summary of findings for the effect of mineralocorticoid receptor antagonists in treating Heart Failure
Full size table

Meta-analysis of cardiovascular death (see Fig. 2) revealed a significant risk reduction, RR 0.81 [0.75–0.87], I2 0% (low heterogeneity). Our analysis of cardiovascular death by HF type indicated only a single trial of HFpEF (TOPCAT) which had no significant reduction in cardiovascular death [7]. Using either selective or nonselective MRA had a similar reduction in cardiovascular death (Additional file 1: Figure S1).

Fig. 2
figure 2

Forest plot of cardiovascular death with MRA use in HF. Seven trials reported cardiovascular death rates when using MRAs in HF compared to control. Our Forest plot has been subdivided according to HF type

Full size image

Meta-analysis of all-cause mortality (see Fig. 3) revealed a significant risk reduction, RR 0.83 [0.77–0.88], I2 0% (low heterogeneity). HF type subgroups indicated the benefit was limited to HFrEF. Use of either a selective or nonselective MRA had a similar reduction in all-cause mortality (Additional file 2: Figure S2).

Fig. 3
figure 3

Forest plot of all-cause mortality with MRA use in HF. Twelve trials reported all-cause mortality rates with MRA use in HF compared to control. Our Forest plot has been subdivided according to HF type

Full size image

Meta-analysis of cardiac hospitalizations (see Fig. 4) revealed a significant risk reduction, RR 0.80 [0.70–0.92], I2 58.4% (high heterogeneity). Our a priori subgroup analysis partially explained the heterogeneity within this outcome, as a significant reduction in cardiac hospitalizations was found in the HFrEF and nonselective MRA subgroups (Additional file 3: Figure S3).

Fig. 4
figure 4

Forest plot of cardiovascular hospitalizations with MRA use in HF. Ten trials reported cardiovascular hospitalization rates with MRA use in HF compared to control. Our Forest plot has been subdivided according to HF type

Full size image

Hyperkalemia was significantly more common with MRA use, RR 2.03 [1.78–2.31], I2 0% (low heterogeneity), see Fig. 5. Subgroup analysis by MRA or HF type did not significantly influence the rate of hyperkalemia (Additional file 4: Figure S4).

Fig. 5
figure 5

Forest plot of hyperkalemia with MRA use in HF. Fifteen trials reported hyperkalemia rates with MRA use in HF compared to control. Our Forest plot has been subdivided according to HF type

Full size image

Gynecomastia was significantly more common with MRA use, RR 3.28 [1.18–9.10], I2 81.7% (high heterogeneity), see Fig. 6. MRA type explained this heterogeneity as selective MRAs did not produce significant amounts of gynecomastia (RR 0.74 [0.43–1.27], I2 0%) while nonselective MRAs did (RR 7.37 [4.42–12.30], I2 0%).

Fig. 6
figure 6

Forest plot of gynecomastia with MRA use in HF. Eleven trials reported gynecomastia rates with MRA use in HF compared to control. Our Forest plot has been subdivided according to MRA type

Full size image

Our analysis of funnel plots for each outcome except gynecomastia revealed no significant asymmetry (Additional file 5: Figure S5, Additional file 6: Figure S6, Additional file 7: Figure S7, Additional file 8: Figure S8 and Additional file 9: Figure S9) and suggested the absence of publication bias. Two MRA subgroups within the funnel plot for gynecomastia explained the asymmetry of the plot (Additional file 9: Figure S9).

Discussion

Summary of evidence

15 trials evaluated the use of MRAs compared to placebo or no treatment for HF. MRA use in patients with heart failure was associated with a significant reduction in adverse cardiovascular outcomes: cardiovascular death (RR 0.81 [0.75–0.87], I2 0%), all-cause mortality (RR 0.83 [0.77–0.88], I2 0%), and cardiac hospitalizations (RR 0.80 [0.70–0.92], I2 58.4%). Our a priori specified subgroup analysis demonstrated that the benefits of MRAs are limited to HFrEF. Both selective and nonselective MRAs increase the risk of hyperkalemia (RR 2.03 [1.78–2.31], I2 0%), but gynecomastia is limited to nonselective MRAs (nonselective MRAs RR 7.37 [4.42–12.30], I2 0% vs. selective MRAs RR 0.74 [0.43–1.27], I2 0%RR 7.37 [4.42–12.30).

Strengths and limitations

Our systematic review has strengths including adherence to PRISMA reporting guidelines. In addition, our conclusions are based on evidence of moderate and high quality (GRADE). HFpEF evidence was of moderate quality, and HFrEF evidence was of high quality for cardiovascular death and all-cause mortality. The quality of evidence for cardiovascular death and all-cause mortality was reduced due the evidence for MRA use in HFpEF being limited to a single trial with large effect size [7], and several smaller trials with confidence intervals crossing unity [18, 27, 28]. For cardiovascular hospitalizations, the quality of evidence was reduced by confidence intervals in HFpEF and HFrEF studies crossing unity [7, 33]. Evidence for hyperkalemia and gynecomastia with MRA usage was of high quality. Overall, the evidence supporting MRA use in HFrEF is based on a larger number of trials with significant effect sizes for reducing adverse cardiac events. In contrast, the evidence for MRA use in HFpEF is based on a smaller number of trials, only one of which had a significant reduction in cardiovascular hospitalizations but no other adverse cardiac events [7]. Finally, our conclusions supporting MRA usage in HFrEF align with current American Heart Association guidelines which recommend MRAs for patients with HFrEF and NYHA class II-IV symptoms or following acute MI complicated by HF and EF ≤ 40% [1].

Implications

Current guidelines suggest MRAs are useful in treating HFrEF and acute MI complicated by HF [1, 34]. We demonstrate that treatment of HFpEF with MRAs does not reduce adverse cardiac events. However, MRAs do cause harm from hyperkalemia (NNH 26 [20–34]) and gynecomastia (e.g., nonselective MRA, NNH 33 [19–63]). Selective MRAs offer a slight advantage in terms of no significant gynecomastia while having equivalent reductions in adverse cardiac outcomes. We suggest continued usage of MRAs in HFrEF, where there is a significant reduction in adverse cardiac outcomes, e.g., cardiovascular death (NNT 34 [26–50]), or all-cause mortality (NNT 32 [24–45]). We suggest that MRAs be avoided in HFpEF as they do not reduce adverse cardiovascular outcomes.

Conclusions

Our systematic review provides evidence that MRAs should not be used in HFpEF. MRA usage in HFpEF provides a risk of hyperkalemia and/or gynecomastia without reducing adverse cardiac events. In contrast, MRA usage in HFrEF significantly reduces adverse cardiac events.