Heart 98:1693-1700 doi:10.1136/heartjnl-2012-302178
  • Review

Aldosterone antagonists improve ejection fraction and functional capacity independently of functional class: a meta-analysis of randomised controlled trials

  1. Thomas H Marwick
  1. Section of Cardiovascular Imaging, Department of Cardiovascular Medicine, Cleveland Clinic, Cleveland, Ohio, USA
  1. Correspondence to Dr Thomas H Marwick, Section of Cardiovascular Imaging, Department of Cardiovascular Medicine, Heart and Vascular Institute, Cleveland Clinic, 9500 Euclid Ave, Cleveland, OH 44195, USA; marwict{at}
  1. Contributors DP: conception and study design, study search and data collection, analysis and drafting of manuscript. PT: study search and data collection, drafting of manuscript. PC: analysis and revising the manuscript. TM: conception and study design, analysis and final approval of manuscript.

  • Received 4 April 2012
  • Accepted 18 June 2012
  • Published Online First 12 July 2012


Context Current guidelines recommend the use of aldosterone antagonists (AA) in patients with moderately severe to severe symptoms [New York Heart Association (NYHA) class III to IV] and systolic heart failure.

Objective To determine the efficacy of AA in improving ejection fraction (EF) and functional capacity and to assess whether this effect was influenced by baseline NYHA classification.

Study design Meta-analysis of randomized controlled trials. Data extraction performed independently by two researchers.

Data Sources MEDLINE and the Cochrane Library.

Study Selection Prospective randomized controlled trials using AA were included if there was a clear description of the baseline NYHA classification and change in EF in patients from study initiation to completion.

Results Data from 1,575 patients enrolled in fourteen studies were included. Overall, there was a weighted mean improvement in EF of 3.2% and in NYHA classification of 0.13 in subjects treated with AA when compared to controls (p<0.001). A mixed effects meta-regression analysis revealed that baseline NYHA was not predictive of improvement in EF (p=0.67) nor NYHA status (p=0.18).

Conclusions The results of this meta-analysis suggest that AA is associated with significant improvements in EF and functional class independent of baseline functional capacity. This supports and expands on the recently published EMPHASIS-HF trial and suggests that the current restriction of AA use to patients with NYHA class III-IV symptoms should be reconsidered.


Aldosterone is a mineralocorticoid hormone which has many deleterious effects on the cardiovascular system.1–4 Aldosterone antagonists (AA) prevent ventricular remodelling, myocardial fibrosis and vascular inflammation, reduce oxidative stress and improve endothelial function.5–7 A number of large randomised control trials (RCT) investigating the benefit of AA in patients with systolic heart failure (HF) have demonstrated a clear reduction in death and cardiovascular morbidity.8–10 Despite these clear benefits, AA remain underprescribed.11 Current guidelines recommend the use of AA in selected patients with reduced left ventricular ejection fraction (LVEF) and at least moderately severe symptoms, evidenced by New York Heart Association (NYHA) class III–IV.12 ,13 This recommendation was mainly based on results from the Randomized Aldosterone Evaluation Study (RALES)8 and the Eplerenone Post-Acute Myocardial Infarction Heart Failure Efficacy and Survival Study.9 However, the recently published Eplerenone in Mild Patients Hospitalization And Survival Study in Heart Failure (EMPHASIS-HF)10 demonstrated a significant reduction in the primary end-points of death and hospitalisation with the addition of AA to other recommended HF therapy in patients with reduced LVEF (ejection fraction (EF) ≤35%) and mild symptoms (NYHA II): a cohort of patients for whom treatment with AA is currently not recommended.

The mechanism by which the use of AA translates into an improvement in mortality has not been clearly defined. It is now well accepted that in patients with systolic HF a therapeutic improvement in LVEF translates into longer term improvements in mortality.14 Therefore, the mortality benefits seen with AA may be related to reverse remodelling of the LV and an improvement in EF. A 2008 meta-analysis of RCTs using AA in LV dysfunction confirmed both a mortality benefit and an improvement in LVEF; however, this study only included trials where the primary end-point of mortality was described and they did not investigate whether this benefit was predicated on significantly reduced functional capacity.15 Furthermore, since this work several RCTs have been published which have investigated the clinical effects of the addition of AA to recommended therapy in patients with mild HF symptoms.16–18

The aims of this meta-analysis were (1) to provide an up-to-date quantification of the improvement in EF after the addition of AA in systolic HF; (2) to clarify whether this effect is dependent on baseline NYHA class; and (3) as a secondary end-point, to assess the impact of AA on functional class.


Search strategy

Prospective randomised controlled trials using AA were included if there was a clear description of the baseline NYHA classification and change in EF in patients from study initiation to completion. A MEDLINE search and a search of the Cochrane Library were supplemented with careful examination of reference lists from identified articles for any pertinent trials missed in the initial search. Independent searches were performed by two experienced reviewers (DP and PT) and were then correlated to ensure that all relevant articles were identified. The electronic search combined terms describing all derivatives of AA (spironolactone, aldactone, eplerenone, inspra, canrenoate, canrenoic acid, mineralocorticoid receptor antagonist(s)/blocker(s), aldosterone antagonist(s), aldosterone receptor antagonist(s)/blocker(s)) with pathologies of interest (heart failure, cardiomyopathy, hypertensive heart disease, ventricular dysfunction) and outcomes of interest (ejection fraction, NYHA, left ventricular function, exercise capacity, exercise tolerance).

Study selection

Trials were excluded where data pertinent to the analysis were not available including: (1) no data on EF at baseline and completion of the trial, (2) no description of the functional status of the study cohort at baseline, (3) studies of HF with preserved ejection and (4) non-randomised trials. The search was restricted to English language literature and studies in humans.

Data extraction

Data including study characteristics, population demographics, aetiology of cardiac disease, inclusion/exclusion criteria, echocardiographic parameters, and functional capacity at baseline and at completion were extracted and entered into an electronic database. For multiple papers published from a single dataset, the largest study with findings relevant to this analysis was used.

Outcome measures

The primary end-point was the change in EF from initiation of AA to the end of the study. The δ EF in the control group was subtracted from the δ EF in the treatment group to account for any improvement not attributable to AA. In order to assess whether there was a difference in the impact of AA on EF based on functional capacity classification, studies were divided into two groups: Group 1 consisted of studies which enrolled subjects with an average NYHA (rounded to the nearest whole number) of II or less; Group 2 consisted of those with an average NYHA classification of III or more (current guidelines only recommend the use of AA in the latter group). When average NYHA was not provided, it was calculated by: ((n with NHYA class I × 1) + (n with NYHA class II × 2) + (n with NYHA class III × 3) + (n with class IV × 4))/n. The secondary end-point was to assess the impact of AA on functional classification in trials where the primary outcome was measurable.

Statistical analysis

Continuous data were expressed as mean ± SD or median (25th, 75th centile) where appropriate. Categorical data were presented as absolute values and percentages. Jadad's scoring method was used to assess the quality of included trials.19 Begg's funnel plot of SE versus difference in mean EF was used to assess the presence of publication bias. Symmetry of distribution was tested using Begg and Mazumdar rank correlation (Kendall's τ with continuity correction, two tailed p) and a fail-safe N calculation was also performed. I 2 statistic was calculated to estimate heterogeneity across studies.20 ,21 The ORs and corresponding 95% CI were computed using random effects models. Forest plots were constructed to graphically describe the overall effects of intervention versus control. As important baseline variables differed between studies, meta-regression was used to assess the influence of potential covariates (baseline NYHA, duration of therapy, age, use of ACE inhibitor (ACEi)/angiotensin receptor blocker (ARB), use of β-blockers and baseline EF were all prespecified variables) on the outcomes measured (EF and NYHA). Bonferroni correction was applied to multiple meta-regression analyses. All analyses were performed using Comprehensive Meta-analysis V.2 software (Biostat, Englewood, New Jersey, USA). A p value of <0.05 was considered statistically significant.


Study characteristics

Our search strategy retrieved 203 potentially relevant articles. After exclusion of studies that did not meet inclusion criteria, 48 studies were evaluated in detail (figure 1). For our analysis, 14 articles reporting on 1575 patients were included.16–18 22–32 Six studies were performed in Europe, six in Asia (predominantly Japan), one in the USA and one in Iran (table 1). The average Jadad score for the included studies was 3. The duration of treatment with AA ranged from 2 to 12 months, with the mean duration of treatment being 7.3±3.5 months. Spironolactone was the predominant AA used (12 trials); canrenone24 and epleronone16 were each only used in one trial.

Figure 1

Study flow diagram. AA, aldosterone antagonists; EF, ejection fraction; NYHA, New York Heart Association.

Table 1

Characteristics of studies included

Table 2 describes patient characteristics in included studies. The weighted mean age of included patients was 59 years and the majority were male subjects (74%). Background medical therapy varied between studies. Overall, 73% of patients were taking β-blockers on enrolment (73.7% in the treatment arms vs 71.6% in the control arms) while 93% were either taking an ACEi or an ARB (93% in each arm). Eight studies enrolled patients with an average NYHA of II (1216 patients) and six studies enrolled patients with an average NYHA of III (359 patients). EF was measured using echocardiography alone in 11 trials (one used cardiac magnetic resonance, one used CT and echocardiography and one used radionuclide ventriculography). Only two trials used a core lab to measure the EF: the others either did not specify or were analysed by blinded local readers.

Table 2

Clinical characteristics of patients in included studies

Effect of aldosterone antagonism on EF

Mean EF improved in 13 of the 14 studies when compared with baseline levels. In one study the EF fell by 1% in the treatment arm.23 The treatment period in this study was only 2 months. However, after controlling for the δ EF in the control group, all studies revealed an improvement in EF. The EF changes seen in each study are represented using a forest plot (figure 2). Overall, the weighted mean difference in EF was 3.2% (OR 1.9, 95% CI 1.4 to 2.7, p<0.001) between treatment and control groups. Funnel plot analysis illustrated no significant publication bias (Kendall's τ = −11, p=0.58) with a fail-safe N value of 136 studies.

Figure 2

Meta-analysis of the effect of aldosterone antagonists (AA) on change of ejection fraction.

There remained a significant improvement in EF when a meta-analysis was performed in the eight studies containing patients with an average NYHA of II (weighted mean improvement in EF of 3.0%) (OR 2.1, 95% CI 1.3 to 3.3, p=0.001) (figure 3). A mixed effects meta-regression analysis of all 14 studies revealed that improvement in EF was independent of baseline NYHA (p=0.67, figure 4). In addition, the change in AA-induced EF improvement was independent of the duration of therapy (p=0.81), age (p=0.94), use of ACEi/ARB (p=0.71), use of β-blockers (p=0.16) and baseline EF (p=0.68).

Figure 3

Meta-analysis of the effect of aldosterone antagonists (AA) on change of ejection fraction in studies with an average New York Heart Association of <III.

Figure 4

Mixed effects regression analysis demonstrating that improvement in ejection fraction (EF) is independent of baseline New York Heart Association (NYHA) (p=0.67).

Effect of aldosterone antagonism on functional capacity

There was incomplete reporting of change in NYHA class from baseline to study completion with only five trials providing these data; however, all studies describe an improvement in symptoms after treatment with AA (figure 5). There was a significant improvement in average NYHA with AA treatment (WMD 0.13, OR 1.7, 95% CI 1.2 to 2.2, p=0.001). Funnel plot analysis revealed no significant publication bias (Kendall's τ =8, p=0.09) with a fail-safe N value of seven studies. Using random effects meta-regression analysis, the change in NYHA was independent of baseline NYHA (p=0.18). Similarly, change in NYHA status was independent of duration of therapy (p=0.46), mean age (p=0.18), use of ACEi/ARB (p=0.37), use of β-blockers (p=0.47) and baseline EF (p=0.69).

Figure 5

Meta-analysis of the effect of aldosterone antagonists (AA) on change in New York Heart Association class.


The present analysis of 1575 patients with systolic HF in 14 prospective randomised controlled trials demonstrated an average improvement in EF of 3.2% following AA treatment, despite appropriate use of background medical therapy. This outcome was independent of baseline functional class. The improvement in EF was associated with an additional improvement in functional capacity in all studies that reported NYHA as an end-point.

Aldosterone is a mineralocorticoid hormone principally known for its action on sodium reabsorption in the distal nephron of the kidney.33 Beyond this action, aldosterone, via non-genomic pathways, has many deleterious effects on the cardiovascular system including superoxide formation, vascular inflammation, endothelial dysfunction and myocardial fibrosis.1–3 ,34 ,35 Experimental data have revealed that treatment with mineralocorticoid receptor antagonism can abrogate these injurious effects.36–39 In patients with chronic HF, increased plasma concentration of aldosterone is associated with a greater 6 month mortality.40 Importantly, despite complete blockade of vascular ACE via treatment with other renin-angiotensin-aldosterone system inhibitors, aldosterone levels typically remain elevated in HF.41 Furthermore, mineralocorticoid receptors are overexpressed in the failing heart.42

Two large randomised controlled clinical trials have shown incremental morbidity and mortality benefits with the addition of AA to the medical regimen of patients with reduced LVEF.8 ,9 Only patients with NYHA classification III–IV were included in the RALES while inclusion into the Eplerenone Post-Acute Myocardial Infarction Heart Failure Efficacy and Survival Study was restricted to patients with reduced EF postmyocardial infarction that had clinical evidence of HF or diabetes. The results of these trials are reflected in current guidelines which restrict the use of AA in systolic HF to patients who have persistent NYHA III–IV symptoms despite optimal treatment with ACEi or ARB, β-blockers and diuretics.12 ,13

The mechanisms behind the cardioprotective effect of AA are likely multiple and in this regard a subgroup analysis of the RALES trial provided some particularly interesting insights.43 In this retrospective analysis, the beneficial effects of spironolactone on the primary end-points were seen exclusively in those patients who expressed the highest levels of markers of collagen turnover (suggesting accelerated interstitial myocardial fibrosis). Moreover, treatment with spironolactone significantly decreased these markers.43 AA-induced reduction in markers of collagen turnover has subsequently been associated with positive effects on LV remodelling including reductions in myocardial fibrosis and cyclic variation in integrated backscatter; these changes have been associated with improvements in both LV diastolic and systolic function.25 ,32 ,44 ,45 From a mechanistic viewpoint, however, there is no reason to believe that AA would only be effective in patients with advanced symptomatology. This issue was addressed in the recently published EMPHASIS-HF study which described a significant reduction in the rate of death from a cardiovascular cause or hospitalisation for HF, in this case, in patients with NYHA class II.10 Our analysis expands on these data by describing associated positive LV remodelling in a similar mildly symptomatic cohort.

These data may be particularly relevant with regard to prophylactic implantable cardioverter-defibrillator (ICD) therapy where an absolute numerical cut-off for EF forms a vital component of the decision making algorithm. A 2010 meta-analysis assessing the efficacy of ICD therapy in primary prevention included eight RCTs.46 These trials have informed current guidelines which mandate a period of ‘optimal medical therapy’ prior to device implantation. However, it is important to note that AA were not included as part of background medical therapy, with the exception of one study (which did not show a benefit to ICD therapy). Interestingly, this meta-analysis of ICD therapy in primary prevention demonstrated an RR reduction of 27% for all cause death, while the RR reduction using AA in EMPHASIS-HF was 19.2% in a less symptomatic population. While caution must be exercised in extrapolating data across different trials, this observation in combination with the detected improvement in EF in our meta-analysis raises the question as to whether the first step should be to add AA to the armamentarium of HF drugs in suitable patients and reassess LV function prior to ICD implantation.

Although a reduction in mortality is obviously of chief concern in the management of HF, improvement in a patient's exercise capacity may be as important. It is well documented that modulation of neuro-hormonal activation can increase exercise capacity in patients with HF.47 ,48 Despite the use of adequate medical therapy, it is encouraging to observe that AA do appear to have an adjunctive effect on functional capacity even in patients who have mild symptoms at the outset. However, given the relatively small change in functional class observed, the number of trials which reported functional status at study completion (five studies with 703 patients) and the potential problems with bias, we feel that, although significant in our analysis, these findings should only be viewed as exploratory and hypothesis generating.

A comparison between our study and a 2008 meta-analysis investigating the effects of AA in patients with LV dysfunction is warranted. Our data are consistent with the reported improvement in EF noted in that study with the addition of seven RCTs and 651 patients in our analysis. As our study was specifically related to EF and functional capacity, we attempted to provide a more detailed review of each trial with additional relevant information on background medical therapy, baseline and follow-up EF and NYHA, and how and where the EF was calculated. However, the main difference between these two studies is that we wished to investigate a very specific question which is of particular relevance since the publication of EMPHASIS-HF, namely, is improvement in LVEF determined by baseline functional classification? To address this, we performed a repeat meta-analysis in studies which enrolled mostly patients with mild symptoms and a meta-regression analysis investigating the influence of baseline NYHA on EF change, neither of which was performed in the earlier study.

Our study has several limitations which necessitate consideration. First, while corresponding authors for each trial were contacted, we were unable to obtain complete individual patient data which limits our ability to identify subgroups, further describe differential effects of AA therapy and calculate mean paired SD or SE of the mean. We therefore calculated these values using unpaired t tests, which is a conservative approach that is likely to overestimate these values. Despite this, both δ EF and NYHA both changed significantly. Second, as with all meta-analyses, our results may be influenced by publication bias, although this was not apparent statistically. Third, the authors appreciate the difficulty with describing NYHA—a categorical variable—as an average. However, this does effectively express the concept that treatment with AA resulted in more patients improving their NYHA classification than controls; furthermore, this is not without precedent.26 Fourth, there were five studies which did not specify whether EF was assessed by an investigator blinded to patient treatment (table 1), and none of the five studies which assessed NYHA at completion specified whether the investigator assigning NYHA was blinded to patient treatment, which are potential sources of bias in these study results. Fifth, the period of follow-up limits our capability to draw definitive conclusions about the permanence of the benefits seen with AA treatment. However, as recently described by Kramer et al, short-term therapeutic effects on LV remodelling predict long-term therapeutic effects on mortality.14 Finally, we did not procure details on complications related to AA therapy as these have been well described in the past including in a recent meta-analysis.15 ,49


The results of this meta-analysis suggest that AA is associated with a significant improvement in EF and NYHA class which are not dependent on baseline functional capacity. These findings expand upon and support the recently published EMPHASIS-HF trial and suggest that the current restrictions on the use of AA in patients with NYHA class III–IV symptoms need to be revisited.


  • Competing interests None.

  • Patient consent This is a meta-analysis of published RCT.

  • Provenance and peer review Not commissioned; internally peer reviewed.

  • Data sharing statement Dr Kimura verbally provided details on NYHA classification of patients enrolled in his study to a mutual colleague working at the Cleveland Clinic Foundation on the understanding that this information would be used specifically in this meta-analysis which he understood we were publishing. We had previously written to Dr Kimura who understood the nature of our work. No additional data.


Free sample
This recent issue is free to all users to allow everyone the opportunity to see the full scope and typical content of Heart.
View free sample issue >>

Don't forget to sign up for content alerts so you keep up to date with all the articles as they are published.