Effects of Mineralocorticoid Receptor Antagonists on Atrial Fibrillation Occurrence: A Systematic Review, Meta‐Analysis, and Meta‐Regression to Identify Modifying Factors

Abstract

Background

Mineralocorticoid receptor antagonists (MRAs) have emerged as potential atrial fibrillation (AF) preventive therapy, but inconsistent results have been reported. We aimed to examine the effects of MRAs on AF occurrence and explore factors that could influence the magnitude of the effect size.

Methods and Results

PubMed, Embase, and Cochrane Central databases were used to search for randomized clinical trials and observational studies addressing the effect of MRAs on AF occurrence from database inception through April 03, 2018. We performed a systematic review and random effects meta‐analyses to compute odds ratios with 95% CIs. Meta‐regression was then applied to explore the sources of between‐study heterogeneity. We included 24 studies, 11 randomized clinical trials and 13 observational cohorts, representing a total number of 7914 patients (median age: 64.2 years; median left ventricular ejection fraction: 49.7%; median follow‐up: 12.0 months), 2843 (35.9%) of whom received MRA therapy. Meta‐analyses showed a significant overall reduction in AF occurrence in the MRA‐treated patients versus the control groups (15.0% versus 32.2%; odds ratio, 0.55; 95% CI, 0.44–0.70 [P<0.00001]), with the greatest benefit regarding recurrent AF episodes (odds ratio, 0.42; 95% CI, 0.31–0.59 [P<0.00001]) and with significant heterogeneity among the included studies (I ²=54%; P=0.0008). Meta‐regression analyses showed that effect size was significantly associated with older studies and higher AF occurrence rate in the control groups.

Conclusions

MRAs seem to be effective in AF prevention, especially regarding recurrent AF episodes.

Clinical Perspective

What Is New?

• Mineralocorticoid receptor antagonists may prevent atrial fibrillation (AF), but well‐sized randomized trials are lacking.

• This meta‐analysis involving 24 studies with 7914 patients, which represents the largest population studied to date, showed a significant overall reduction in AF occurrence in the mineralocorticoid receptor antagonists–treated patients versus the control groups (15.0% versus 32.2%, respectively; odds ratio, 0.55 [95% CI, 0.44–0.70]; P<0.00001).

• The greatest benefit was observed in cases of recurrent AF episodes (odds ratio, 0.42; 95% CI, 0.31–0.59 [P<0.00001]), especially in populations with high AF occurrence rate.

What Are the Clinical Implications?

• Our results suggest a clinical benefit of mineralocorticoid receptor antagonists in preventing AF and required well‐sized randomized trials to definitively answer the question.

Introduction

Atrial fibrillation (AF) is the most prevalent sustained arrhythmia, with an overall estimated prevalence of 1% to 2% among the general population. AF is associated with substantial morbidity, reduced functional status, impeded quality of life, and increased mortality1 The pathophysiological mechanisms underlying AF initiation and perpetuation are complex and not completely understood, but evidence indicates that atrial electrical, neurohormonal, and structural remodeling create the substrate for AF development1 There is evidence that aldosterone and the activation of its receptor, mineralocorticoid receptor, promote cardiac fibrosis and electrical disturbances2, 3 Mineralocorticoid receptor antagonists (MRAs) have been shown to reduce atrial fibrosis and prevent AF development.3, 4 Primary aldosteronism is strongly associated with the risk for developing AF in both clinical series (odds ratio [OR], 12.1; 95% CI, 3.2–45.2 [P<0.0001])5 Clinical data have suggested that MRAs could have positive effects on AF burden, but inconsistent results have been reported. Two previous meta‐analyses6, 7 investigated the impact of MRAs on AF occurrence but are affected by the noninclusion of nonrandomized clinical trials (RCTs) with the use of restricted search strategies and the absence of any analysis of heterogeneity to investigate modifying factors. Moreover, the benefit of MRAs on AF occurrence in patients who have heart failure (HF) with reduced ejection fraction (HFrEF)8 was not confirmed in patients without HFrEF or in those without any structural heart disease.

Therefore, we conducted a systematic review of the literature and meta‐analysis of both RCTs and observational studies to examine the potential effect of MRA use on AF occurrence using an appropriate strategy to avoid restrictive research (adapted to events considered as secondary end point in studies). We also performed subgroup and meta‐regression analyses to explore the source of heterogeneity and identify modifying factors.

Methods

This systematic review complied with the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses guidelines (Table S1)9 The protocol was prospectively registered in the International Prospective Register of Systematic Reviews (registration number: CRD42018096969). No ethics committee approval or informed consent was required since this was a retrospective analysis of previously published studies.

Data Sources and Search Strategy

An extensive, unrestricted, computerized MEDLINE, Embase, and Cochrane Library literature search of articles in English and French was independently conducted by 2 reviewers (J.A., L.D.) according to prespecified selection criteria from inception to April 3, 2018. We also considered studies selected from prior meta‐analyses related to the impact of MRAs on AF occurrence6, 7; trial protocols on trial registry platforms, including clinicaltrials.gov (https://clinicaltrials.gov/), the World Health Organization's International Clinical Trials Registry Platform (http://apps.who.int/trialsearch/), the UK Clinical Trials Gateway (https://www.ukctg.nihr.ac.uk/), and EudraCT (https://eudract.ema.europa.eu/); and data from scientific meeting abstracts and conferences. We used both controlled terms (ie, MeSH terms in MEDLINE) and free‐text terms related to MRAs as domain 1 (details of the search are provided in Figure 1). Regarding domain 2 (AF domain), as we did not expect that AF would be reported often in study titles or abstracts (because AF is often a secondary end point of MRA studies), we did not create a specific search domain using the free term “AF” to avoid restricting search strategy. Therefore, we computed a larger domain using the terms “cardiovascular disease” OR “heart disease” OR “atrial fibrillation” (domain 2). The final research was performed as follows: (domain 1) AND (domain 2). Second, a manual search was performed for relevant references from the selected articles.

Figure 1.

Flow diagram of study selection. Clinical trials (randomized, nonrandomized, parallel arm, and cluster designs) and clinical observational comparative studies (including retrospective or prospective cohorts, case‐control, or nested case‐control designs) were included. *Trials registry portals include clinicaltrials.gov (https://clinicaltrials.gov/), the World Health Organization international clinical trials registry platform (http://apps.who.int/trialsearch/), the UK clinical trials gateway (https://www.ukctg.nihr.ac.uk/), EudraCT (https://eudract.ema.europa.eu/).

Study Selection

Studies evaluating the effects of MRAs (study intervention) compared with non‐MRA drugs (placebo or other control drugs, study comparator) on AF occurrence in adult patients were included. Studies using comparators other than drugs were not included. Clinical trials (randomized or nonrandomized, parallel arm, and cluster designs) and clinical observational comparative studies (including retrospective or prospective cohorts and case‐control or nested case‐control designs) reporting any AF outcomes and the use of MRAs were included. We excluded cross‐sectional studies, case series, crossover studies, and case reports. Healthcare/health insurance database studies were also excluded because this type of database does not offer much valuable clinical information to allow the conduct of subgroup and meta‐regression analyses.

Data Extraction

Two review authors (J.A., L.D.) independently screened study titles and abstracts identified by the search against eligibility criteria. Full reports were obtained for all eligible articles/abstracts. The review authors independently extracted data from the selected studies in duplicate using a standardized data extraction form. Any disagreements were resolved by consensus with senior authors (J.J.P., P.M.). The κ statistic revealed excellent agreement between the 2 review authors (κ=0.86; 95% CI, 0.6–1.0 [P<0.0001]). Data extracted included patient demographic and baseline characteristics, patient selection, methodology and study design, inclusion and exclusion criteria, follow‐up duration, number of patients, type and dosing of MRAs (when available), and outcomes of interest reported at follow‐up. If studies lacked data, corresponding authors were contacted via email to provide the required information. The data that support the findings of this study are available from the corresponding author upon reasonable request.

Outcome

The primary outcome was the occurrence or recurrence of at least 1 symptomatic or asymptomatic AF, as defined in each study. All types of AF were studied, including postoperative AF (POAF).

Exploration of Heterogeneity of MRA Effect on AF Occurrence

To explore heterogeneity of MRA effects across trials, we planned to perform prespecified subgroup analyses and univariate meta‐regression analyses. The following parameters were considered for the subgroup analyses: study design (placebo and nonplacebo RCTs versus non‐RCTs), individual MRA agents used (spironolactone, eplerenone, canrenone, or unspecified MRA), type of AF (eg, new‐onset AF, recurrent AF, POAF, cardioversion, and catheter ablation), the presence of HFrEF (defined as patients with left ventricular ejection fraction [LVEF] ≤40% and New York Heart Association class ≥II), quality components including full‐text published studies versus scientific meeting abstracts/unpublished studies, and risk of bias (by omitting studies that were judged to be at least at a high or serious risk of bias and industry funding). The following parameters were considered in the meta‐regression analyses: clinical status (hypertension, considered as the proportion of patients with hypertension included in each study; patient age, considered as the mean age of the patients in each study; LVEF, considered as the mean LVEF in each study), AF incidence in the control and MRAs groups, and year of publication.

Risk of Bias (Quality) Assessment

Regarding clinical trials, 2 authors (J.A., L.D.) evaluated risk of bias in individual studies using Cochrane Collaboration's risk of bias tool9 Any disagreements were resolved by consensus with senior authors (J.J.P., P.M.). Regarding observational studies, we used the Risk of Bias Tool in Nonrandomized Studies of Interventions (ROBINS‐I). After a careful risk of bias assessment for each study, the 2 authors (J.A., L.D.) qualified the studies as “high” or “medium/low” risk of bias. The potential for reporting/publication bias will be further visually explored by funnel plots if ≥10 studies are available for the comparison and with Egger test. We planned to use the Grading of Recommendations Assessment, Development and Evaluation (GRADE) working group methodology to assess the quality of evidence for all outcomes.

Statistical Analysis

Statistical analyses were performed with Review Manager version 5.3 (RevMan 5.3, Cochrane Collaboration) and R software for Windows version 3.4.4 (R Foundation for Statistical Computing). Heterogeneity was estimated using I ² statistics and P values for Cochrane heterogeneity tests. Substantial between‐study heterogeneity was defined as I ² >50%, and significant heterogeneity was defined if P<0.10. We used Mantel‐Haenszel summary OR with random effect. Continuous variables were analyzed as the mean difference. For categorical variables, we calculated the OR with 95% CI using the total number of events and patients extracted from the individual studies, with an OR <1 signifying a reduced occurrence of AF in the MRA group. Robustness of the main result was assessed by several sensitivity analyses by excluding: (1) each study sequentially; (2) asymmetric studies on the funnel plot; (3) most influential trials (defined as studies with a weight ≥5.0%); and (4) less influential trials (defined as studies with a sample size <100 patients). Regarding meta‐regression analyses, each trial was weighted using inverse variance, and each parameter significantly associated with treatment effect (MRAs versus controls) on AF occurrence was then studied with linear regression analysis between the OR logarithms and quantitative variables. Unweighted logistic regression analysis between the positive status of the trial and quantitative variables was performed. A P<0.05 was considered statistically significant.

Results

The flow chart is presented in Figure 1. The inverted funnel plot for the overall mortality end point did not suggest any substantial publication bias (Figure S1) and the Egger test did not show any significant asymmetry (P=0.25).

Descriptions of Included Studies

Details of the study characteristics are presented in Table. Twenty‐four studies enrolled a total of 7914 adult patients, with 2843 patients in the MRAs arms (35.9%) and 5071 patients in the control arms (64.1%). The median age for the entire population was 64.2 (interquartile range, 51.6–68.0) years. The administered MRAs were spironolactone in 62.5%,10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25 eplerenone in 12.5%8, 21, 26, 27 canrenone in 8.3%28, 29 and nonspecified in 16.7% of the studies (G. Marchetti, et al, unpublished data, 2012)30, 31 Among the 7914 patients, 4831 (61.1%) were included in new‐onset AF studies (including 1397 patients in POAF studies), and the remaining 3083 (38.9%) were included in AF recurrence studies (including 408 and 233 patients in electrical cardioversion and catheter ablation studies, respectively). The median LVEF reported in the 24 studies was 49.7% (interquartile range, 26.0–58.5%). Of the 7914 patients, 2839 were patients with HFrEF (35.9%). The median proportion of patients with hypertension was 58.4% (14.6–80%). The median follow‐up was 12.0 (interquartile range, 3.0–36.1) months (range: 0.2–49.8 months) in non‐POAF trials and 8.0 (interquartile range, 5.5–21) days (range: 5–30 days) in POAF studies.

Table 1.

Characteristics of the Clinical Trials Included in the Meta‐Analysis

Study	Type of Publication	Sites, Location	Study Design	Patients, No. (% Treated With MRA)	Population	Primary or Secondary AF Prevention	Mean Age, SD	LVEF at Inclusion, %	HTA, No. (%)	AF Occurrence in the Control Group, %	Comparison	Main Outcome	Follow‐Up Duration
Paziaud et al 200310	Full article	France	Retrospective observational study	96 (21.9)	Consecutive patients referred for electric cardioversion	Recurrence of AF	64.3	58	14.6	16.0	Spironolactone+conventional therapy vs conventional therapy	Successful electric cardioversion	NA
Gao et al 200711	Full article	China	Randomized placebo‐controlled study	116 (50.0)	Cardiac patients with HF without any AF history	New‐onset AF	55	42	58	41.4	Spironolactone 20 mg/d vs placebo	New‐onset cardiac arrhythmia occurrence	6 mo
Boldt et al 200831	Full article	Germany	Prospective observational study	148 (27.0)	Consecutive patients with HFrEF referred for electric cardioversion	Recurrence of AF	67	32	83	35.2	MRA+conventional therapy vs conventional therapy	Successful electric cardioversion	12 d
Kim et al 200912	Full article	Korea	Prospective observational study	74 (6.8)	Patients referred for electric cardioversion	Recurrence of AF	59	44.7	29	69.6	Spironolactone+conventional therapy vs conventional therapy	AF recurrence	13 mo
Letsas et al 200913	Full article	Germany	Prospective observational study	72 (8.3)	Patients with paroxysmal and persistent AF who underwent successful PVI	Recurrence of AF	54.9	54.8	19	39.4	Spironolactone+conventional therapy vs conventional therapy	AF recurrence	12.5 mo
Brinkley et al 201014	Abstract	United States	Retrospective observational study	171 (41.5)	Consecutive patients with dual‐chamber ICD (primary prevention) without any AF history	New‐onset AF	NA	NA	NA	58.0	Spironolactone+conventional therapy vs conventional therapy	New‐onset AF detected on ICDs	21 mo
Dabrowski et al 2010 (SPIR‐AF)15	Full article	Poland	Randomized controlled study	164 (50.0)	Consecutive patients with recurrent AF episodes	Recurrence of AF	66	69	73.2	80.5	Spironolactone 25+β‐blocker±enalapril vs β‐blocker±enalapril	Incidence of symptomatic AF recurrence	12 mo
Disertori et al 2010 (GISSI‐AF)30	Full article	Italy	Prospective observational study	1442 (6.4)	Patients with symptomatic paroxysmal or persistent AF	Recurrence of AF	68	55	85.4	52.4	Valsartan±MRA and conventional therapy vs valsartan±conventional therapy	AF recurrence	12 mo
Lopes et al 201016	Abstract	Portugal	Prospective observational study	156 (29.5)	Cardiac patients with HF without any AF history	New‐onset AF	63	NA	55.8	13.6	Spironolactone+conventional therapy vs conventional therapy	New‐Onset AF occurrence	43.2 mo
Özaydin et al 201017	Full article	Turkey	Prospective observational study	269 (25.6)	Patients with HF and LVEF ≤50% referred for CABG and/or valve surgery without any AF history	New‐onset AF	59	43	53.2	23.0	Spironolactone+conventional therapy vs conventional therapy	New‐onset POAF incidence	NA
Williams et al 201124	Full article	United States	Retrospective observational study	83 (27.7)	Patients with ICDs with concomitant AF	Recurrence of AF	71	33.1	80.4	53.3	Spironolactone+conventional therapy vs conventional therapy	Hospitalization for AF or need for electric cardioversion	49.8 mo
Billota et al 201228	Full article	Italy	Randomized controlled study	56 (50.0)	Patients with neurocritical care with cerebral edema	New‐onset AF	57	NA	NA	7.1	Canrenone 200 mg/d+mannitol+furosemide vs mannitol+furosemide	Incidence of new cardiac arrhythmia	8 d
(G. Marchetti, et al, unpublished data, 2012)	Abstract	Italy	Randomized controlled study	90 (50.0)	Patients with HFrEF with an LVEF ≤35% and first documented AF treated by electric cardioversion	Recurrence of AF	73	NA	57.8	44.4	MRA+conventional therapy vs conventional therapy	Progression to permanent AF	3 mo
Pretorius et al 201218	Full article	United States	Randomized placebo‐controlled study	294 (50.0)	Patients with LVEF ≥30% referred for CABG and/or valve surgery without any AF history	New‐onset AF	59	57	64.5	27.2	Spironolactone 25 mg/d vs placebo	New‐onset POAF incidence	6 d
Swedberg et al 2012 (EMPHASIS‐HF)8	Full article	Worldwide	Randomized placebo‐controlled study, post hoc analysis	1794 (50.8)	Patients with HFrEF with LVEF ≤35% and NYHA class II without AF history	New‐onset AF	67.9	26	64.5	4.5	Eplerenone 25 to 50 mg/d vs placebo	Composite of death from cardiovascular causes or hospitalization for HF	21 mo
Tumasyan et al 201219	Abstract	Armenia	Randomized controlled study	135 (25.2)	Patients with chronic HF with NYHA class III and AF	Recurrence of AF	61.1	NA	NA	67.3	Spironolactone+conventional therapy vs ramipril or valsartan or aliskiren+conventional therapy	AF rhythm control and adequate control of ventricular response	12 mo
Ito et al 201326	Full article	Japan	Retrospective observational study	161 (34.2)	Consecutive patients referred for catheter ablation of long‐standing persistent AF	Recurrence of AF	60.5	64	50	60.4	Eplerenone+conventional therapy vs conventional therapy	AF recurrence after catheter ablation	24 mo
Grigoryan et al 201520	Abstract	Armenia	Randomized placebo‐controlled study	42 (50.0)	Patients with paroxysmal AF with LVEF ≥40%	Recurrence of AF	51.6	NA	NA	28.6	Spironolactone 25 to 50 mg/d vs placebo	AF recurrence	8 mo
Simopoulos et al 201521	Full article	Greece	Retrospective observational study	332 (39.8)	Patients with HF and LVEF ≤40% referred for CABG and/or valve surgery without any AF history	New‐onset AF	64.2	36	NA	45.0	Spironolactone or eplerenone 25 to 50 mg/d+conventional therapy vs conventional therapy	New‐onset POAF incidence	1 mo
Vukicevic et al 201622	Abstract	Serbia	Retrospective observational study	226 (15.0)	Consecutive patients referred for CABG without any AF history	New‐onset AF	63.9	NA	NA	22.4	Spironolactone+conventional therapy vs conventional therapy	New‐onset POAF incidence	5 d
Bosone et al 201729	Full article	Italy	Randomized controlled study	289 (33.9)	Patients with hypertension and type 2 diabetes mellitus with AF history	Recurrence of AF	68	61	100	40.8	Canrenone 10 to 100 mg/d+ramipril 5 mg/d vs amlodipine 5 mg/d or ramipril 5 mg/d+hydrochlorothiazide	AF recurrence	12 mo
Cikes et al 2018 (TOPCAT)25	Full article	Worldwide	Randomized placebo‐controlled study	1207 (50.9)	Patients with symptomatic HF and LVEF ≥45%	New‐onset AF (n=1005) and recurrence of AF (n=314)	71	60	NA	9.3	Spironolactone 15 to 45 mg/d vs placebo	Composite of cardiovascular mortality, aborted cardiac arrest, or HF hospitalization	34.8 mo
Tsutsui et al 2018 (J‐EMPHASIS‐HF)27	Full article	Japan	Randomized placebo‐controlled study	221 (50.2)	Patients with HFrEF with LVEF ≤35% and NYHA class II without AF history	New‐onset AF	68.7	26.1	58.4	1.8	Eplerenone 25 to 50 mg/d vs placebo	Composite of death from cardiovascular causes or hospitalization for HF	29 mo
Shavit et al 201823	Data not published in the original study but collected during the investigation23	Israel	Prospective observational study	276 (35.9)	Consecutive patients referred for cardiac surgery without any AF history	New‐onset AF	69	NA	80	28.8	Spironolactone 25 mg/d+conventional therapy vs conventional therapy	New‐onset POAF incidence	30 d

AF indicates atrial fibrillation; CABG, coronary artery bypass graft; HF, heart failure; GISSI‐AF, Gruppo Italiano per lo Studio della Sopravvivenza nell'Infarto Miocardico–Atrial Fibrillation; HFrEF, heart failure with reduced ejection fraction; HTA, hypertension; ICD, implantable cardioverter‐defibrillator; LVEF, left ventricular ejection fraction; MRA, mineralocorticoid receptor antagonist; NA, not available; NYHA, New York Heart Association; POAF, postoperative atrial fibrillation; PVI, pulmonary vein isolation; SPIR‐AF, Effect of combined spironolactone‐β‐blocker ± enalapril treatment on occurrence of symptomatic atrial fibrillation episodes in patients with a history of paroxysmal atrial fibrillation; TOPCAT, Treatment of Preserved Cardiac Function Heart Failure With an Aldosterone Antagonist trial.

AF Occurrence

As shown in Figure 2, compared with the control, MRAs reduced the risk for AF occurrence (15.0% versus 32.2%; OR, 0.55 [95% CI, 0.44–0.70]; P<0.00001), with a significant heterogeneity between the included studies (I ²=54%; P=0.0008). Prespecified sensitivity analyses were not significantly different from those of the primary analysis (Tables S2 and S3).

Figure 2.

Atrial fibrillation occurrence comparing mineralocorticoid receptor antagonist (MRAs) therapy vs controls.

Subgroup and Meta‐Regression Analyses

Subgroup analyses based on prespecified parameters were performed (reported in Figures 2 and 3, Figures S2 through S9). The benefit of MRAs for reducing the risks for AF occurrence was consistent considering individual MRA agents, quality components, and presence or not of HFrEF. There was a significant interaction between MRA effect and type of AF (higher effect for AF recurrence versus new‐onset AF, P=0.01).

Figure 3.

Mineralocorticoid receptor antagonist (MRA) benefit in reducing the risk for atrial fibrillation (AF) occurrence in subgroups analyses regarding the type of AF (A) (new‐onset AF vs AF recurrence), the presence or not of heart failure with reduced ejection fraction (HFrEF) (B) (HFrEF, defined as patients with left ventricular ejection fraction [LVEF] ≤40% and New York Heart Association [NYHA] class ≥II), the study status (C) (full‐text and published studies vs meetings abstracts or unpublished studies), and individual MRA used (D). Circle sizes are proportional to trial sample sizes. CR indicates canrenone; EP, eplerenone; NS, nonspecified MRA; OR, odds ratio; SP, spironolactone.

On the prespecified univariate meta‐regression analyses, 2 variables were found to be statistically associated with the effects of MRAs on AF occurrence (Figure 4). First, the reduction in AF occurrence when receiving MRAs was significantly higher (P=0.045) in the “oldest” studies. Second, studies with a higher AF occurrence rate in the control groups were significantly more likely to report a beneficial effect of MRAs on AF occurrence than those with a lower AF occurrence rate (P=0.023). The probability of a significant positive MRA impact was associated with a higher AF occurrence rate in the control group (P=0.03, Figure S10). These 2 predictors (year of publication and AF occurrence rate in the control group) explained 17% and 21% of the variance, respectively. The association between the year of publication with positive outcome for MRAs in AF was mainly driven by the published full text; when excluding the 7 conference abstracts, the association became nonsignificant (P=0.18). The association between AF rate in the control group with positive outcome for MRAs in AF remained consistent with or without the exclusion of conference abstracts. Hypertension, patient age, and LVEF did not significantly influence the effects of MRAs on AF occurrence (P=0.82, P=0.51, and P=0.68, respectively).

Figure 4.

Treatment effects (both mineralocorticoid receptor antagonists [MRAs] and controls) on atrial fibrillation (AF) occurrence were associated with a high AF occurrence rate in the control group (A) but not with the AF occurrence rate in the MRA group (B). Treatment effects (both in the MRA group and controls) on AF occurrence were associated with the year of publication of the study (C). Circle sizes are log‐proportional to trial sample sizes. Blue (AF occurrence panels) and red (year of publication panel) circles indicate trials with a positive primary outcome effect (AF occurrence, as defined by the trial authors).

Study Quality and Publication Bias

The quality of included studies is presented in Tables S4 and S5., 32, 33 According to the GRADE methodology, our primary outcome had a fair consistency, with moderate to low risk of bias studies, good precision, and no evident publication bias. However, it had substantial heterogeneity. Hence, the quality of evidence was judged to be moderate.

Discussion

Two previous meta‐analyses6, 7 investigated the impact of MRAs on AF occurrence but presented a lack of power caused by insufficient studies included in the meta‐analyses, restricted search strategies (with the use of “AF” as a search domain, whereas AF is often not the primary end point of MRA studies and therefore is rarely present in the title and abstract of studies), restricted MRA search strategies (canrenone was not included), and absence of any analysis of heterogeneity to investigate modifying factors. One originality of our search strategy was to include studies considering AF occurrence as a secondary end point to avoid restrictive research. In fact, the main end points of MRA drug studies are generally HF and hypertension. AF is therefore rarely reported in these studies and rarely mentioned in the study title or abstract. Therefore, in our meta‐analysis, using “AF” as a search domain would have inevitably caused us to miss some studies that perfectly met our inclusion criteria.

Using this methodology, MRAs were associated with a significantly lower AF risk compared with no MRA treatment (OR, 0.55; 95% CI, 0.44–0.70 [P<0.00001]). This effect remained consistent across subgroups with respect to sensitivity and meta‐regression analyses. The effect seems to be larger regarding AF recurrence compared with new‐onset AF. This may be explained by the antifibrotic effects of MRAs, since fibrosis is present in patients with AF to a greater extent compared with those without AF34 Unfortunately, when restricting meta‐analysis only to RCT versus placebo subgroup, the efficacy of MRAs did not reach statistical significance (OR, 0.76; 95% CI, 0.55–1.06 [P=0.11]). This may be explained by the low AF rate in the control group (9.2%) compared with other types of studies (44.9% when considering RCT, prospective, and retrospective observational studies). Interestingly, the MRA efficacy does not seem confined to patients with HF, as initially suggested by previous meta‐analyses6, 7 and in the post hoc analysis of EMPHASIS‐HF (Eplerenone in Mild Patients Hospitalization and Survival Study in Heart Failure)8 This trial was the first large and randomized placebo‐controlled study to test the hypothesis that MRAs could decrease AF occurrence. In this study, 1794 patients with HFrEF who had LVEF ≤35% and New York Heart Association class II without AF history were enrolled. The median LVEF was 26%. Patients were randomized to receive either eplerenone 25 to 50 mg/d or placebo during a 21‐month follow‐up period. The primary end point was a composite of death from cardiovascular causes or hospitalization for HF. New‐onset AF occurred in 25 of 911 (2.7%) patients in the eplerenone group and 40 of 883 (4.5%) patients in the placebo group (hazard ratio, 0.58; 95% CI, 0.35–0.96 [P=0.034]). The beneficial effects of MRAs, independent of the presence of HF, were also recently highlighted in another meta‐analysis35 This meta‐analysis studied the effects of MRAs in patients with ST‐segment–elevation myocardial infarction without HF or with a reduced LVEF of <40% (10 RCTs, 4147 patients). MRA treatment decreased mortality (2.4% versus 3.9%; OR, 0.62 [95% CI, 0.42–0.91]; P=0.01) compared with the control group. A possible mechanism for MRA impact on AF may pass through the prevention of electrical remodeling and fibrosis.2, 3, 4, 36

MRAs did not significantly reduce the risk of new‐onset POAF, but only 1 of the 5 studies included in this analysis was an RCT and we observed a significant heterogeneity across studies included18 POAF is a multifactorial phenomenon, and aldosterone might play an important role in POAF development. Experimental studies have shown that aldosterone promotes myocardial inflammation and fibrosis, modulates ionic currents, induces oxidative stress, and enhances cardiac damage during ischemia‐reperfusion, particularly by increasing cardiomyocyte apoptosis2, 4, 36, 37 All of these phenomena constitute a potential substrate for POAF occurrence. Preliminary findings support this hypothesis, with higher preoperative aldosterone plasma levels in patients with POAF than in those without POAF37 The ALDOCURE (Spironolactone and Perioperative Atrial Fibrillation Occurrence in Cardiac Surgery Patients) multicenter double‐blind RCT from our group (NCT03551548), specifically designed to test the impact of spironolactone on POAF occurrence after elective coronary artery bypass graft±aortic valve replacement in patients with preserved LVEF, may resolve this issue.

In our meta‐analysis, we observed large variations in AF occurrence rates in the control group (Table). We explored the influence of AF rate variations between trials on AF occurrence using prespecified meta‐regression analyses (Figure 4). For the 24 trials included, AF occurrence rate ranged between 1.8%27 and 80.5%15 (mean: 36.1%). “Positive” MRA trials had a higher occurrence of AF in the control group than “negative” trials. The probability of a significant positive MRA impact was associated with a higher AF occurrence rate in the control group (P=0.03, Figure S10). This finding may indicate that the results of MRA trials in the field of AF are influenced by excessively high AF occurrence rates in control groups. Moreover, we can suppose that MRAs are more effective for patients presenting with frequent AF recurrences at baseline. This hypothesis is supported by the results of the study by Dabrowski et al15 in which a spectacular effect of spironolactone on the number of AF episodes during a 12‐month follow‐up period was reported. In this study, the included patients exhibited at baseline (before randomization) ≈4 episodes during 3 months and a long history of AF (for 4 years). A long history of highly recurrent AF might indicate larger cardiac fibrosis and atrial remodeling and could therefore select patients who can benefit most from MRA therapy. Large trials dedicated to assessing this hypothesis are warranted.

Finally, the year of publication for a study was significantly associated with the study results, with a higher probability of having a positive effect of MRAs in “old” studies compared with “recent” ones. This result may be explained by the constant improvement of therapeutics every year, making it more difficult to demonstrate efficiency. This may be likely in the setting of AF with the advent of catheter ablation and the common use of antiarrhythmic drugs such as amiodarone.

Study Limitations

A potential limitation of our meta‐analysis is the inclusion of nonrandomized studies and meeting abstracts. However, this methodology allows us to perform a systematic review and to limit the risk of publication bias. The studies acquired different cohorts and included different MRA agents, different types of AF, and different clinical contexts (POAF studies with short follow‐up versus no‐POAF studies with longer follow‐up), which led to a moderate heterogeneity according to the GRADE score (I²=54% regarding the principal analysis). We explored most of these factors with subgroup and meta‐regression analyses, but the absence of individual data clearly limited our ability to address within‐study heterogeneity. We decided a priori to use a random effect model because we had concerns of heterogeneity, and because the choice between the 2 models should not be based solely on the observed significant test for heterogeneity. Figure S11 showed AF occurrence comparing MRA therapy versus controls using a fixed effect model and did not exhibit any significant difference with the random one. The absence of individual data prevented us from highlighting, for example, potential differences in the efficacy among MRA agents, especially in patients with diabetes mellitus where studies suggested that spironolactone increased glycated hemoglobin and cortisol levels and did not improve endothelial function, whereas eplerenone did. Furthermore, the methods for detecting AF during follow‐up are heterogeneous across studies. This is inherent to AF detection and may lead to an underestimation of the AF risk.

Conclusions

Results from our meta‐analysis suggest a substantial efficacy of MRAs in reducing the risk of AF in patients with or without HF, especially in the setting of AF recurrence prevention. These findings support the hypothesis of mineralocorticoid receptor inhibition as an emerging treatment option for the prevention of AF, particularly in patients with “active” AF with frequent episodes. Future adequately powered randomized studies are required to assess such a hypothesis.

Sources of Funding

This systematic review was supported by Caen Normandy University Hospital (CHU Caen Normandie, France) and Normandy University (Université de Caen Normandie, France).

Disclosures

None.

Supporting information

Table S1. Preferred Reporting Items for Systematic Reviews and Meta‐Analyses Checklist for the Meta‐Analysis

Table S2. Sensitivity Analyses to Evaluate the Contribution of Each Study to the Pooled Estimation by Excluding Each of the Studies One After the Others

Table S3. Sensitivity Analyses to Evaluate the Contribution of Asymmetric Studies on the Funnel Plot of Biggest Trials (Weight Percentage ≥5.0%) and of Smaller Trials (Sample Size <100 Patients) to the Pooled Estimation

Table S4. Risk of Bias in Randomized Studies, Based on the Cochrane Risk of Bias Tool for Randomized Controlled Trials

Table S5. Risk of Bias in Observational Studies Based on the Risk of Bias Tool in Nonrandomized Studies of Interventions (ROBINS‐I) Assessment Tool (Version 19, September 2016 for Cohort‐Type Studies)

Figure S1. Funnel plot of standard error (log odds ratio) by odds ratio to evaluate publication bias for effect of mineralocorticoid receptor antagonists (MRAs) on reducing atrial fibrillation (AF) occurrence.

Figure S2. Impact of mineralocorticoid receptor antagonists (MRAs) vs control in new‐onset atrial fibrillation (AF) vs AF recurrence.

Figure S3. Impact of mineralocorticoid receptor antagonists (MRAs) on atrial fibrillation (AF) occurrence vs that of controls in the presence of heart failure with reduced ejection fraction (HFrEF) or not (defined as patients with left ventricular ejection fraction [LVEF] ≤40% and New York Heart Association [NYHA] class ≥II).

Figure S4. Impact of mineralocorticoid receptor antagonists (MRAs) on atrial fibrillation (AF) occurrence vs that of controls in full‐text published vs meetings abstracts or unpublished studies.

Figure S5. Impact of mineralocorticoid receptor antagonists (MRAs) on atrial fibrillation (AF) occurrence vs that of controls regarding the risk of bias of studies (evaluated by omitting studies judged to be at least at a high or serious risk of bias).

Figure S6. Impact of mineralocorticoid receptor antagonists (MRAs) on atrial fibrillation (AF) occurrence vs that of controls regarding the funding sources.

Figure S7. Impact of mineralocorticoid receptor antagonists (MRAs) on atrial fibrillation (AF) occurrence vs that of controls among the MRAs used (spironolactone, eplerenone, canrenone, or unspecified MRA).

Figure S8. Impact of mineralocorticoid receptor antagonists (MRAs) on atrial fibrillation (AF) occurrence vs that of controls in the following subgroups: new‐onset postoperative atrial fibrillation (POAF), AF recurrence after electrical cardioversion, and AF recurrence after catheter ablation.

Figure S9. Impact of mineralocorticoid receptor antagonists (MRAs) on atrial fibrillation (AF) occurrence vs that of controls in postoperative atrial fibrillation (POAF) and no‐POAF studies.

Figure S10. Atrial fibrillation (AF) occurrence rate in the control group was significantly calibrated to predict the positive effect of mineralocorticoid receptor antagonist (MRA) therapy on AF occurrence (A).

Figure S11. Atrial fibrillation (AF) occurrence comparing mineralocorticoid receptor antagonist (MRA) therapy vs control using a fixed effect model.

(J Am Heart Assoc. 2019;8:e013267 DOI: 10.1161/JAHA.119.013267.)