Spironolactone in Atrial Fibrillation With Preserved Cardiac Fraction: The IMPRESS‐AF Trial

Abstract

Background

Patients with permanent atrial fibrillation have poor outcomes, exercise capacity, and quality of life even on optimal anticoagulation. Based on mechanistic and observational data, we tested whether the mineralocorticoid receptor antagonist spironolactone can improve exercise capacity, E/e' ratio, and quality of life in patients with permanent atrial fibrillation and preserved ejection fraction.

Methods and Results

The double‐masked, placebo‐controlled IMPRESS‐AF (Improved Exercise Tolerance in Heart Failure With Preserved Ejection Fraction by Spironolactone on Myocardial Fibrosis in Atrial Fibrillation) trial (NCT02673463) randomized 250 stable patients with permanent atrial fibrillation and preserved left ventricular ejection fraction to spironolactone 25 mg daily or placebo. Patients were followed for 2 years. The primary efficacy outcome was peak oxygen consumption on cardiopulmonary exercise testing at 2 years. Secondary end points included 6‐minute walk distance, E/e' ratio, quality of life, and hospital admissions. Spironolactone therapy did not improve peak oxygen consumption at 2 years (14.0 mL/min per kg [SD, 5.4]) compared with placebo (14.5 [5.1], adjusted treatment effect, −0.28; 95% CI, −1.27 to 0.71]; P=0.58). The findings were consistent across all sensitivity analyses. There were no differences in the 6‐minute walking distance (adjusted treatment effect, −8.47 m; −31.9 to 14.9; P=0.48), E/e’ ratio (adjusted treatment effect, −0.68; −1.52 to 0.17, P=0.12), or quality of life (P=0.74 for EuroQol‐5 Dimensions, 5‐level version quality of life questionnaire and P=0.84 for Minnesota Living with Heart Failure). At least 1 hospitalization occurred in 15% of patients in the spironolactone group and 23% in the placebo group (P=0.15). Estimated glomerular filtration rate was reduced by 6 mL/min in the spironolactone group with <1‐unit reduction in controls (P<0.001). Systolic blood pressure was reduced by 7.2 mm Hg (95% CI, 2.2–12.3) in the spironolactone group versus placebo (P=0.005).

Conclusions

Spironolactone therapy does not improve exercise capacity, E/e' ratio, or quality of life in patients with chronic atrial fibrillation and preserved ejection fraction.

Registration

UTL: https://www.clini​caltr​ial.gov; Unique identifier: NCT02673463. EudraCT number 2014‐003702‐33.

Nonstandard Abbreviations and Acronyms

AF

atrial fibrillation

BP

blood pressure

IMPRESS‐AF

Improved Exercise Tolerance in Heart Failure With Preserved Ejection Fraction by Spironolactone on Myocardial Fibrosis in Atrial Fibrillation

MLWHF

Minnesota Living with Heart Failure

NIHR

National Institute for Health Research

VO_2peak

peak oxygen consumption

Clinical Perspective

What Is New?

• The study presents a randomized placebo controlled clinical trial of spironolactone versus placebo in patients with chronic atrial fibrillation and preserved ejection fraction.

• Spironolactone therapy does not improve exercise capacity, endothelial function assessed as echocardiographic E/e' ratio, or quality of life in the tested population.

What Are the Clinical Implications?

• The trial does not support use of spironolactone therapy in patients with chronic atrial fibrillation and preserved ejection fraction.

Patients with permanent atrial fibrillation (AF) have poor outcomes and reduced quality of life even when they receive appropriate stroke prevention therapy and have preserved left ventricular function. ¹ The prognosis is worse when maintenance of the sinus rhythm cannot be achieved and the patients progress to persistent or permanent AF, even on optimal rate control. ²

The mechanisms leading to reduced exercise capacity, related morbidity, and mortality in anticoagulated patients with permanent AF are likely related to disturbed diastolic ventricular function, myocardial fibrosis, and stiffening. ³ , ⁴ In fact, activators of fibroblast signaling such as fibroblast growth factor 23 are elevated in patients with chronic AF. ⁵ These changes can lead to ventricular filling abnormality, reducing cardiac output, and decreasing exercise capacity. ³ , ⁴ Aldosterone increases cardiac collagen deposition and left ventricular fibrosis. ⁶ This involves direct stimulation of cardiac fibroblasts by aldosterone to produce collagen with chronification of oxidative stress and inflammation in the heart. ⁷ , ⁸ Cardiac expression of mineralocorticoid receptors is increased in AF, thus augmenting the genomic effects of aldosterone. ⁹ According to a substudy of the RALES (Randomized Aldactone Evaluation Study) trial, the improved survival in patients with heart failure with impaired cardiac contractility treated by spironolactone was linked to its ability to reduce serum markers of ongoing fibrosis. ¹⁰ The recently published RACE‐3 (Routine Versus Aggressive Upstream Rhythm Control for Prevention of Early Persistent Atrial Fibrillation in Heart Failure Study), reporting a complex intervention that included spironolactone (>75% difference in use between groups) suggested beneficial effects compared with usual care in patients with recent onset AF. ¹¹

The IMPRESS‐AF (Improved Exercise Tolerance in Heart Failure With Preserved Ejection Fraction by Spironolactone on Myocardial Fibrosis in Atrial Fibrillation) trial therefore tested the effects of mineralocorticoid receptor inhibition with spironolactone in patients with permanent AF with preserved left ventricular ejection fraction compared with placebo on exercise capacity (primary outcome); and its effect on quality of life, diastolic function, all‐cause hospital admissions, and spontaneous cardioversion to sinus rhythm (secondary outcomes).

Methods

The data that support the findings of this study will be made available, provided the request is scientifically sound and from an appropriately qualified research group. All requests for data should be addressed to the corresponding authors.

The IMPRESS‐AF trial is a double‐masked randomized placebo‐controlled multicenter enrollment single‐site management trial conducted in Birmingham, United Kingdom. The trial randomized 250 patients with permanent AF and preserved left ventricular function 1:1 to either spironolactone or placebo. Permanent AF was defined as at least 1‐year history of AF on all prescreening assessments with a previous decision to the accept rate control strategy and with further confirmation of AF on ECG at the screening. The previously published trial protocol was developed following the SPIRIT (Standard Protocol Items: Recommendations for Interventional Trials) statement and the latest PRO‐specific guidance. ¹² , ¹³ The study was approved by the National Research and Ethics Committee West Midlands—Coventry and Warwickshire (REC Reference 14/WM/1211). All patients provided signed informed consent. The study is registered at EudraCT number 2014‐003702‐33 and Clini​caltr​ial.gov (NCT02673463). Because of the sensitive nature of the data collected for this study, requests to access the data set from qualified researchers trained in human subject confidentiality protocols may be sent to the University of Birmingham.

Patients were recruited from primary care general/family practices and outpatient AF clinics in Sandwell and West Birmingham Hospitals Trust, Birmingham. The patients were ≥50 years with permanent AF, left ventricular ejection fraction ≥55%, and controlled blood pressure (BP). ¹⁴ An interpreter and translated materials were available if English was not their preferred language. The exclusion criteria (full list published previously ¹³ ) included life expectancy of <2 years, severe chronic obstructive pulmonary disease, severe mitral or aortal valve stenosis or regurgitation, rapidly progressing or severe renal impairment, potassium ≥5mmol/L, recent coronary artery bypass graft surgery (within 3 months), use of aldosterone antagonist or potassium sparing diuretic within 14 days. All patients received optimized treatment following established clinical guidelines on management of AF, heart failure, and hypertension. ¹⁵

All patients were screened, investigated, and managed in the Research Clinic in the Institute of Cardiovascular Sciences, City Hospital, Birmingham. During the baseline visit, eligible patients underwent cardiopulmonary exercise testing using a cycling ergometer to measure peak oxygen consumption (VO_2peak), 6‐minute walk test, and completed paper‐based quality of life questionnaires (validated Minnesota Living with Heart Failure ¹⁶ and EuroQol‐5 Dimensions, 5‐level version quality of life ¹⁷ questionnaires). A block randomization (block size of 4) was performed after the assessments using a secure web‐based randomization system. Patients were stratified by their baseline VO_2peak of ≤16 and >16 mL/min per kg. Patients, the trial team, care providers, outcome assessors, and data analysts were masked to the treatment allocation. Patients attended the research clinic every 3 months for follow‐up (Table S1). The day‐to‐day management of the trial was coordinated by the Birmingham Clinical Trials Unit, registered by the National Institute for Health Research (NIHR) as a trials unit. A Trial Steering Committee was responsible for overseeing the progress of the trial. An independent Data Monitoring and Ethics Committee was responsible for the regular monitoring of trial data and adverse events.

Patients randomized to spironolactone received 25 mg once daily for 2 years. In the case of an increase in potassium level to 5.1 to 5.5 mmol/L or non‐life‐threatening side effects (eg, gynecomastia) the dose was down‐titrated to 25 mg each second day with an attempt to re‐up‐titrate the dose if the reason for down‐titration has resolved. When potassium levels exceeded 5.5 mmol/L or if toxicity was suspected, the trial medication was stopped for the duration of the trial, but the patients were requested to attend the remaining follow‐up visits and their outcomes were included in the intention‐to‐treat analysis. Patient compliance to treatment was defined adequate when ≥80% of allocated capsules were taken.

The primary hypothesis was assessed by the difference between trial groups in VO_2peak on cardiopulmonary exercise testing ¹⁸ , ¹⁹ at 2 years, adjusted for the baseline values. The Secondary efficacy end points were 6‐minute walk test distance, quality of life using Minnesota Living with Heart Failure, and EuroQol‐5 Dimensions questionnaires, diastolic function assessed by E/E’ ratio ²⁰ , ²¹ on echocardiography, rates of all‐cause hospital admissions and spontaneous return to sinus rhythm on ECG. Prespecified safety outcomes were occurrence of breast pain or swelling, allergic reaction, raised serum creatinine (>220 μmol/L), low estimated glomerular filtration (<30 mL/min per 1.73m²), and hyperkalemia (≥5.1 and ≥6.0 mmol/L).

Statistical Analysis

All primary analyses followed modified intention to treat principles including all patients reporting outcomes regardless of their compliance with the medication. For the primary outcome patients who died before the 2‐year follow‐up assessment were additionally included with VO_2peak scores at 2 years imputed as 0 regardless of cause; 3 sensitivity analyses were undertaken for the primary outcome (further details are available in Data S1). Further analyses included additional adjustments for age, sex, and body mass index at baseline. Pre‐defined subgroup analyses compared the difference in primary outcome VO_2peak between spironolactone and placebo according to VO_2peak; age, sex, body mass index, systolic and diastolic BP. Details of the statistical methodology for the secondary outcome measures are available in Data S1. Analyses of the study outcomes were defined in a statistical analysis plan, signed off before unmasking. Results are presented in accordance with CONSORT and CONSORT PRO. ²² STATA version 12 software was used for all analyses.

Sample Size

Sample size was determined to be able to show a clinically important difference in the primary outcome of VO_2peak. Published values of VO_2peak in subjects with heart failure give baseline values (mean 16 [SD 5] mL/min per kg). ²³ A difference of 2 mL/min per kg was judged to be clinically relevant. A sample size of 100 patients in each arm would give the power of at least 80% to detect differences in primary and secondary end points of a magnitude consistent with published results from similar studies. We increased the sample size to 125 per arm for provision for a 20% dropout rate. Statistical power would be higher with the benefits of adjusting for baseline values.

Results

A total of 250 patients were randomized to spironolactone or placebo (125 per group). Patients were elderly (mean age, 72.3 [SD, 7.4] years, further details in Table 1). Random groups were well balanced (Table 1). The final study visit was attended by 101 (81%) patients randomized to spironolactone and 106 (85%) randomized to placebo (Figure 1). Spironolactone had the expected effect on BP and kidney function. Systolic BP was reduced by 7.2 mm Hg (95% CI, 2.2–12.3) in the spironolactone group, almost no change in the placebo group (Table S2). There was no significant treatment effect for diastolic BP. Spironolactone increased creatinine (mmol/L) by 6.9 (95% CI, 3.4–10.5) and lowered estimated glomerular filtration rate (mL/min per 1.73) by 6.0 (95% CI, −9.3 to −2.8) at 2 years.

Table 1.

Baseline Characteristics

	Spironolactone (n=125)	Placebo (n=125)
Demographic characteristics
Age, y	73 (68–77)	72 (67–78)
Sex
Women	28 (22%)	31 (25%)
Men	97 (78%)	94 (75%)
Ethnicity
White	118 (94%)	118 (94%)
Black	3 (2%)	3 (2%)
Asian	3 (2%)	2 (2%)
Ethnicity other than White, Black, or Asian or mixed ethnicity of White, Black and/or Asian background	1 (1%)	2 (2%)
Smoker
Current smoker	6 (5%)	8 (6%)
Ex‐smoker	66 (53%)	68 (54%)
Never smoked	53 (42%)	49 (39%)
Alcohol use, units/wk	3 (0–12)	6 (0–14)
Characteristics of the study outcomes
VO2peak, mL/kg per min	14 (11–18)	14 (11–18)
VO2peak of ≤16 mL/kg per min	77 (62%)	78 (62%)
6‐min walk test, m	266 (196–316)	271 (200–330)
E/E’ ratio	9.8 (8.0–12.0)	9.7 (7.5–13.0)
E/E’ ratio categories
<10	66 (52.8%)	64 (51.2%)
≥10 to <14	41 (32.8%)	39 (31.2%)
≥14	18 (14.4%)	22 (17.6%)
EQ‐5D‐5L
Score	0.84 (0.74–0.94)	0.88 (0.74–0.94)
Missing data	4 (3%)	5 (4%)
MLWHF score
Score	7.0 (6.3–35.8)	14.0 (5.8–30.0)
Missing data	8 (6%)	4 (3%)
Clinical characteristics
Left ventricular ejection fraction (%)	58 (57–62)	58 (56–63)
Brain natriuretic peptide, pg/mL	122 (73–230)	136 (82–241)
Brain natriuretic peptide >130 ng/L	56 (44.8%)	66 (52.8%)
Body mass index, kg/m2	29 (26–33)	30 (26–34)
Systolic blood pressure, mm Hg	130 (117–140)	129 (118–142)
Diastolic blood pressure, mm Hg	75 (67–83)	74 (68–82)
Resting heart rate, bpm	85 (74–99)	83 (74–97)
Diabetes mellitus	24 (19%)	21 (17%)
Medications
Non‐vitamin K oral anticoagulants	60 (48%)	57 (46%)
Vitamin K antagonists	47 (38%)	47 (38%)
Aspirin	10 (8%)	9 (7%)
Clopidogrel	6 (5%)	4 (3%)
Loop diuretic	25 (20%)	24 (19%)
Thiazide diuretic	14 (11%)	20 (16%)
Angiotensin‐converting enzyme inhibitor or angiotensin receptor blocker	67 (54%)	80 (64%)
Calcium channel blocker	43 (34%)	37 (30%)
Beta blocker	66 (53%)	70 (56%)
Digoxin	26 (21%)	23 (18%)
Amiodarone	0 (0%)	1 (1%)
Statin	88 (70%)	69 (55%)
Inhalers of asthma or COPD	20 (16%)	17 (14%)

Continuous data presented as median (interquartile range). To score Minnesota Living with Heart Failure questionnaire, it was allowed that at most 20% of 21 responses were missing which was equivalent to 4 data items. If there were ≤4 data items missing then we used mean substitution to impute the missing responses and then scored the questionnaire by summating the responses to all 21 questions; otherwise, the person's score was left missing. EQ‐5D‐5L indicates EuroQol‐5 Dimensions, 5‐level version quality of life questionnaire; COPD, chronic obstructive pulmonary disease; MLWHF indicates Minnesota Living with Heart Failure; and VO_2peak, peak oxygen consumption.

Figure 1. CONSORT flow diagram.

EQ‐5D indicates EuroQol‐5 Dimensions, 5‐level version quality of life questionnaire; FU, follow up; IMP, investigational medicinal product; and MLWHF, Minnesota Living with Heart Failure.

Data on the primary outcome, VO_2peak at the end of the trial were available for 106 patients in placebo group and 103 patients in spironolactone group (Table 2), which included 3 deaths in placebo group and 5 deaths in spironolactone group. In both groups, 3 patients were not able to perform cardiopulmonary exercise testing because of frailty. In the primary intention to treat analysis, VO_2peak (mL/min per kg) changed from mean 14.5 (SD 4.6) to mean 14.0 (SD 5.4) in the spironolactone group and from mean 14.6 (SD, 5.1) to mean 14.5 (SD, 5.1) in placebo group. The treatment effect showed no difference between the groups (differences in mean −0.28, 95% CI, −1.27 to 0.71; P=0.58). The findings were consistent across the performed sensitivity analyses (Table 2, Table S3).

Table 2.

Study Outcomes

	Spironolactone		Placebo		Treatment Effect*	P Value*
	Mean (SD)	n	Mean (SD)	n	(95% CI)†
Primary outcome†
VO2peak, mL/kg per min	14.0 (5.4)	103	14.5 (5.1)	106	−0.28 (−1.27 to 0.71)	0.58
Primary outcome sensitivity analysis
Per protocol analysis:‡ VO2peak, mL/kg per min	14.8 (4.3)	57	14.9 (4.9)	77	0.21 (−0.78 to 1.21)	0.67
Complete case analysis: VO2peak, mL/kg per min	14.8 (4.5)	98	14.9 (4.6)	103	−0.09 (−0.86 to 0.68)	0.81
Multiple imputation method:§ VO2peak, mL/kg per min	13.4 (6.0)‖	125	14.0 (5.5) ‖	125	−0.53 (−1.57 to 0.51)	0.32
Adjusted for stratification variable and BNP level: VO2peak, mL/kg per min	14.0 (5.4)	103	14.5 (5.1)	106	−0.32 (−1.31 to 0.68)	0.53
Analysis removing patients who died and could not perform CPET (assigned to a peak VO2 of 0) at follow‐up: VO2peak, mL/kg per min¶	14.8 (4.6)	98	14.9 (4.6)	103	−0.09 (−0.86, 0.68)	0.81
Secondary outcomes
6‐min walk test distance, m†	313 (108)	105	330 (112)	107	−8.47 (−31.9 to 14.9)	0.48
E/Eʹ ratio	9.00 (3.05)	101	9.72 (3.57)	106	−0.68 (−1.52 to 0.17)	0.12
Brain natriuretic peptide, pg/mL	179 (171)	101	186 (110)	105	4.95 (−28.3 to 38.2)	0.77
EQ‐5D‐5L score†	0.81 (0.26)	98	0.84 (0.21)	104	−0.008 (−0.06 to 0.04)	0.74
MLWHF score#	17.4 (22.7)	96	15.3 (20.4)	104	0.49 (−4.32 to 5.29)	0.84

BNP, brain natriuretic peptide; CPET indicates cardiopulmonary exercise testing; EQ‐5D‐5L, EuroQol‐5 Dimensions, 5‐level version quality of life questionnaire; MLWHF, Minnesota Living with Heart Failure; VO_2peak, peak oxygen consumption.

^*The mean differences between the spironolactone group and placebo group, 95% CIs, and the corresponding P values were estimated from linear regression models adjusting for the baseline continuous peak oxygen consumption score. In the sensitivity analyses additional adjustments were made for age, sex, and body mass index.

^†A value of 0 was assigned to peak oxygen consumption, 6‐minute walk test, and EuroQol‐5 Dimensions, 5‐level version quality of life questionnaire scores for those who died.

^‡Per‐protocol population was defined as ≥80% of capsules taken.

^§Predictive Mean Matching imputation method was used to generate 20 imputed data sets. Data for participants in spironolactone group and placebo group were imputed separately. peak oxygen consumption_, age, body mass index, systolic/diastolic blood pressure, 6‐minute walk test, brain natriuretic peptide level, E/E’ ratio, EuroQol‐5 Dimensions, 5‐level version quality of life, Minnesota Living with Heart Failure scores at baseline and sex were included in the imputation model.

^‖SD estimates were obtained by multiplying the standard error by the square root of 125.

^¶Total of 8 exclusions; 5 from spironolactone, and 3 from placebo.

^#For Minnesota Living with Heart Failure questionnaire, score ranges from 0 to 105 with a higher score reflecting poorer quality of life; the highest value across the whole participants was assigned to those who died. All patients choose to use the English version of the questionnaires with only 1 proxy completion case recorded (placebo group is the reference group).

The subgroup analyses showed no significant interaction of treatment with baseline VO_2peak values (mL/min per kg, ≤16 versus >16, P=0.54), body mass index (kg/m², <25 versus 25–30 versus ≥30, P=0.13), sex (P=0.91), median BP (P=0.36 for systolic BP and P=0.93 for diastolic BP), and E/E’ ratio (<10 versus 10–14 versus ≥14, P=0.73) (Table 3). There was a significant interaction between treatment and age: higher VO_2peak values were observed in older patients randomized to spironolactone, but in younger patients in placebo group (P=0.03 for interaction). The magnitude of the differences was small, with the point estimates for the treatment effect in each subgroup being smaller than the pre‐stated clinically important treatment effect.

Table 3.

Subgroup Analysis of the Primary Outcome

Analyses	Spironolactone		Placebo		Treatment Effect (95% CI)*	Estimate of Difference (95% CI)* , †	P Value for Interaction*
	Mean (SD)	n	Mean (SD)	n
Pre‐specified subgroup analyses
Peak VO2, mL/min per kg
VO2 ≤16	11.2 (4.4)	60	11.9 (3.7)	63	−0.56 (−1.85 to 0.73)	0.64 (−1.38 to 2.66)	0.54
VO2 >16	18.1 (3.9)	43	18.2 (4.6)	43	0.07 (−1.47 to 1.62)
Age, y
Age ≤ median‡	14.4 (6.3)	54	16.6 (4.8)	53	−1.40 (−2.76 to −0.05)	2.24 (0.28 to 4.20)	0.03
Age > median	13.7 (4.3)	49	12.3 (4.6)	53	0.83 (−0.55 to 2.22)
BMI, kg/m2
BMI <25	14.7 (3.9)	14	15.2 (4.9)	14	0.30 (−2.40 to 2.99)	…	0.13
25 to <30	14.9 (6.5)	43	16.4 (5.0)	36	−1.59 (−3.21 to 0.02)	−1.89 (−5.05 to 1.27)
BMI ≥30	13.0 (4.5)	46	13.0 (5.0)	55	0.58 (−0.85 to 2.00)	0.28 (−2.79 to 3.35)
Sex
Women	11.0 (3.7)	20	12.1 (2.9)	26	−0.41 (−2.54 to 1.72)	0.14 (−2.28 to 2.57)	0.91
Men	14.8 (5.5)	83	15.2 (5.5)	80	−0.27 (−1.39 to 0.86)
Systolic blood pressure, mm Hg
SBP ≤ median‡	13.5 (6.2)	52	14.8 (5.2)	54	−0.71 (−2.10 to 0.68)	0.93 (−1.06 to 2.93)	0.36
SBP > median	14.6 (4.4)	51	14.0 (5.1)	51	0.23 (−1.19 to 1.64)
Diastolic blood pressure, mm Hg
DBP ≤ median‡	13.5 (5.6)	54	13.7 (5.2)	58	−0.24 (−1.58 to 1.11)	−0.09 (−2.08 to 1.90)	0.93
DBP > median	14.7 (5.1)	49	15.3 (5.0)	47	−0.33 (−1.78 to 1.12)
Post‐hoc subgroup analysis
E/E’
E/E’ <10	15.3 (5.2)	56	15.1 (5.1)	58	−0.26 (−1.60 to 1.10)	…	0.73
10 <14	13.0 (4.9)	34	13.9 (5.5)	30	0.06 (−1.75 to 1.85)	0.31 (−1.95 to 2.58)
E/E’ ≥14	11.2 (6.2)	13	13.4 (4.8)	18	−1.23 (−3.86 to 1.40)	−0.97 (−3.93 to 1.99)

BMI indicates body mass index; DBP, diastolic blood pressure; and SBP, systolic blood pressure; and VO_2peak, peak oxygen consumption; mean is unadjusted.

^aThe mean differences between the spironolactone group and the placebo group, 95% CIs, and the corresponding P values were estimated from linear regression models adjusting for the baseline continuous peak oxygen consumption score.

^bThe lower level was always treated as the reference group for the estimates of treatment difference apart from sex for which women were the reference group.

^cThe median age is 72.58 years, median systolic blood pressure is 129 mm Hg, and median diastolic blood pressure is 74 mm Hg (placebo group is the reference group).

There was no difference between random groups in any of the secondary outcomes (Table 2, Table S4). The findings remained consistent after adjustment of age, sex, and body mass index for all outcomes (Tables S3 and S4). Spontaneous return to sinus rhythm on ECG performed at 2 years was uncommon in both study groups (4 [4%] in placebo group and 8 [8%] in spironolactone group, P=0.21) (Table 4, Table S5). At least 1 hospitalization was observed in 15% of patients in spironolactone group and 23% in placebo group (hazard ratio, 0.65; 95% CI, 0.36–1.17) (Table 4, Figure 2, Table S5). There was no significant difference in overall mortality, death from cardiac causes, hospitalizations because of cardiac causes, and rates of stroke and systemic thromboembolism between the study arms (Tables S6 through S8). Patients in the spironolactone group had a higher occurrence of breast pain (17 versus 5 in controls), breast swelling (11 versus 4 in controls), and hyperkalaemia (≥5.1 mmol/L, 46 versus 17 in controls) (Table S6).

Table 4.

Secondary Outcomes: Return to Sinus Rhythm and Hospitalization for All Causes

Analyses at 2 y	Spironolactone		Placebo		Odds Ratio (95% CI)	P Value
Sinus rhythm, n (%)*	n=101	8 (8%)	n=106	4 (4%)	2.19 (0.64–7.52)†	0.21†
Hospitalization for all causes
Participants with at least 1 event, n (%)	n=118	18 (15%)‡	n=123	28 (23%)	0.65 (0.36–1.17)§	0.15§
Incidence rate (no. per 10 000 person‐days)	n=118	2.46	n=123	3.78

^aSpontaneous return to sinus rhythm on ECG.

^bThe odds ratios, 95% CIs, and the corresponding P values were estimated from a logistic regression model, after adjustment for the continuous baseline peak oxygen consumption.

^cOne first hospitalized event had no date and was excluded from the time‐to‐event analysis.

^dThe adjusted hazard ratio, 95% CIs, and the corresponding P values were estimated from a Cox regression model adjusting for the baseline continuous peak oxygen consumption score for the primary analysis (placebo group is the reference group).

Figure 2. Kaplan–Meier plot of time to first hospitalization.

 

Discussion

Mechanistic studies have demonstrated that elevation in cardiac filling pressures (estimated by E/e' in this trial) contributes to pulmonary limitations and impaired VO_2peak. ²⁴ , ²⁵ Treatment with spironolactone in the IMPRESS‐AF study does not improve either exercise capacity or quality of life in this cohort of stable patients with permanent AF with preserved ejection fraction despite lowering BP. CIs excluded clinically important effects. On the contrary, kidney function, assessed by estimated glomerular filtration rate, worsened in patients randomized to spironolactone.

Clinical trials of the aldosterone antagonists spironolactone and eplerenone (RALES, EPHESUS [Eplerenone Post‐AMI Heart Failure Efficacy and Survival Study], EMPHASIS‐HF [Eplerenone Post–Acute Myocardial Infarction Heart Failure Efficacy and Survival Study]) uniformly showed their clinical benefits in systolic heart failure. The IMPRESS‐AF study conclusively demonstrates that use of an aldosterone antagonist, spironolactone does not improve aerobic capacity, estimated filling pressures, or quality of life in patients with AF without systolic impairment, and generally consistent with the overall lack of effect in patients with heart failure and preserved ejection fraction found in the ALDO‐DHF (Aldosterone Receptor Blockade in Diastolic Heart Failure) ²⁶ and TOPCAT (Treatment of Preserved Cardiac Function Heart Failure with an Aldosterone Antagonist) ²⁷ trials. Of interest, in the ALDO‐DHF trial of predominantly hypertension‐related heart failure and preserved ejection fraction aldosterone inhibition reduced E/e' but did not increase in VO_2peak, and in another trial spironolactone improved VO_2peak and reduced exercise‐induced increase in E/e' in a selected population of patients with heart failure and preserved ejection fraction that excluded patients with AF. ²⁶ , ²⁸ Of note, although the TOPCAT trial showed overall neutral results this may be confounded by the quality issues in Eastern Europe. ²⁹ In the IMPRESS‐AF trial, the hospitalization rates were numerically higher in the placebo arm. A larger study that is powered for hospitalizations and cardiovascular death might produce other results, although such a study would be difficult to justify based on the overall results of spironolactone trials in patients with preserved ejection fraction.

Although patients receiving spironolactone had numerically more cases of spontaneous return to sinus rhythm such cases were few in both study arms, the difference was not significant statistically, and could therefore be a chance finding. The recently published RACE‐3 trial suggested that a complex intervention including spironolactone was associated with a higher rate of sinus rhythm maintenance at 1 year (P=0.042). ¹¹ The RACE‐3 intervention consisted of mineralocorticoid receptor antagonists, statins, angiotensin‐converting enzyme inhibitors and/or angiotensin receptor blockers, and cardiac rehabilitation. ¹¹ Among the targeted trial medicines, the mineralocorticoid receptor antagonists showed the most prominent contrast in use (85% in the treatment group and 4% in the control group). Whilst the RACE‐3 trial indicated the mineralocorticoid receptor antagonists could help maintenance of the sinus rhythm in the recent onset AF, spontaneous return in sinus rhythm was infrequent in both arms of the IMPRESS‐AF trial. The actual baseline VO_2peak of 14.25 mL/kg per minute was lower than the projected 16 mL/kg per minute, possibly reflecting the background chronic AF, but this is unlikely to affect the study conclusions.

Overall, spironolactone was well tolerated. There were comparable rates of withdrawal from the study in the treatment and control groups. As expected, spironolactone reduced BP, thus demonstrating adequate overall compliance with the drug as confirmed by the expected effect. However, we found a safety signal as there was a reduction of 6.0 mL/kg in the estimated glomerular filtration rate over 2 years. These data indicate potential harm to the kidney caused by treating patients with AF with spironolactone. However, most patients with pharmacological inhibition of renin‐angiotensin‐aldosterone system, including spironolactone have hemodynamically mediated reductions in glomerular filtration rates. ³⁰ These changes can be renal protective by decreasing chronic glomerular hyperfiltration in patients with chronic kidney disease and they may have contributed the decrease in HF hospitalization that was observed in the TOPCAT trial. Not unexpectedly, a higher occurrence of breast pain/swelling and hyperkalemia was noted in the spironolactone group.

Limitations

The study outcomes were assessed by tests of physical capacity, but these tests could be inherently affected by various musculoskeletal problems despite every effort to perform the tests until the limits of the cardiac reserve are reached. Although recognized questionnaires were used to assess quality of life, specific validation of the tests in the study population has not been done.

There was a relatively high drop‐out rate in this study, and overall 16% of patients did not complete the primary outcome tests. However, the study power was estimated to allow 20% loss of the patients during follow‐up, and the validity of the findings was maintained. The study did not have power to reliably define effects of spironolactone on hard outcomes, such as hospitalizations or return to sinus rhythm. However, given the detrimental effects of the drug on kidney function in this trial population further testing of spironolactone will need careful consideration.

Conclusions

Treatment with the aldosterone antagonist, spironolactone in patients with permanent AF and preserved ejection fraction does not improve exercise tolerance, quality of life, and diastolic function. Furthermore, spironolactone leads to worsening of renal function which should be considered in this patient population and use of mineralocorticoid receptor antagonists may mandate closer monitoring of renal function.

Sources of Funding

The trial was funded by the NIHR, NIHR‐EME Programme (grant number 12/10/19) with support from NIHR Clinical Research Network West Midlands, United Kingdom. The University of Birmingham was the sponsor of this trial.

Disclosures

Prof Lip reports acting as a Consultant for Bayer/Janssen, BMS/Pfizer, Medtronic, Boehringer Ingelheim, Novartis, Verseon, and Daiichi‐Sankyo; Speaker for Bayer, BMS/Pfizer, Medtronic, Boehringer Ingelheim, and Daiichi‐Sankyo. No fees were directly received personally. Prof Calvert reports grants from NIHR HTA, during the conduct of the study; grants from NIHR Birmingham Biomedical Research Centre, grants from NIHR Surgical Reconstruction and Microbiology Research Centre, grants from Health Data Research UK, grants from Innovate UK, grants from Macmillan Cancer Support, personal fees from PCORI, from Astellas, from Takeda, from Glaukos, from Merck, outside the submitted work. Prof Fisher reports grants from NIHR, during the conduct of the study; grants from BMS/Pfizer, outside the submitted work. Prof Kirchhof is a board member of the European Society of Cardiology and has received travel support from the European Society of Cardiology, including for meetings pertinent to this work, during the conduct of the study; Prof Kirchhof has received research support from European Union, British Heart Foundation, Leducq Foundation, Medical Research Council (UK), and German Centre for Heart Research, from several drug and device companies active in atrial fibrillation, and has received honoraria from several such companies, outside the submitted work; In addition, Prof Kirchhof is listed as an inventor on 2 patents held by University of Birmingham (Atrial Fibrillation Therapy WO 2015140571, Markers for Atrial Fibrillation WO 2016012783), pending. The remaining authors have no disclosures to report.

Supporting information

Data S1

Tables S1–S8

References 31 , 32

(J Am Heart Assoc. 2020;9:e016239 DOI: 10.1161/JAHA.119.016239.)