Impact of Atrial Fibrillation on Exercise Capacity in Heart Failure with Preserved Ejection Fraction: A RELAX Trial Ancillary Study

Abstract

Background

Atrial fibrillation (AF) is common among patients with heart failure and preserved ejection fraction (HFpEF) but its clinical profile and impact on exercise capacity remains unclear. RELAX was a multicenter randomized trial testing the impact of sildenafil on peak VO₂ in stable outpatients with chronic HFpEF. We sought to compare clinical features and exercise capacity among HFpEF patients who were in sinus rhythm (SR) or AF.

Methods and Results

RELAX enrolled 216 HFpEF patients with 79 (37%) in AF, 124 (57%) in SR and 13 in other rhythms. Participants underwent baseline cardiopulmonary exercise testing (CPXT), echocardiogram, biomarker and rhythm status assessment prior to randomization. AF patients were older than those in SR but had similar symptom severity, co-morbidities and renal function. Betablocker use and chronotropic indices were also similar. Despite comparable LV size and mass, AF was associated with worse systolic (lower EF, stroke volume and cardiac index) and diastolic (shorter deceleration time and larger left atria) function compared to SR. Pulmonary artery systolic pressure was higher in AF. AF patients had higher NT-proBNP, aldosterone, endothelin-1, troponin I and CITP levels suggesting more severe neurohumoral activation, myocyte necrosis and fibrosis. Peak VO₂ was lower in AF, even after adjustment for age, sex, and chronotropic response, and V_E/VCO₂ was higher.

Conclusions

AF identifies an HFpEF cohort with more advanced disease and significantly reduced exercise capacity. These data suggest that evaluation of the impact of different rate or rhythm control strategies on exercise tolerance in HFpEF patients with AF is warranted.

Atrial fibrillation (AF) and heart failure (HF) commonly co-exist and the presence of each worsens the outcome of the other¹. The prevalence of AF in HF with preserved left ventricular ejection fraction (HFpEF; LVEF≥50%) is similar to that observed in patients with HF and reduced ejection fraction (HFrEF)². In HFrEF, patients with AF have worse exercise capacity than those in sinus rhythm (SR)³ and small studies suggest that rhythm control improves exercise capacity^4-6. However, HFpEF patients have been underrepresented in AF studies and it remains unclear whether the presence of AF further compromises exercise performance in HFpEF. Likewise, there is limited information regarding the profile of the HFpEF patient with AF, particularly in stringently defined HFpEF subjects with truly normal LVEF⁷.

The RELAX (Phosphodiesterase-5 Inhibition to Improve Clinical Status and Exercise Capacity in HFpEF) trial enrolled 216 HFpEF patients who met rigorous entry criteria designed to insure the presence of HF and cardiac limitation to exercise⁸. Rhythm status, echocardiography, biomarker assessment, and cardiopulmonary exercise testing (CPXT) were performed at baseline. We hypothesized that HFpEF patients with AF would display more advanced HF and greater impairment of exercise capacity when compared to HFpEF patients in SR. As HFpEF is associated with chronotropic incompetence⁹, while AF patients display an increased heart rate response during exercise¹⁰, we also evaluated the relationship between rhythm status and chronotropic response during exercise in HFpEF.

Methods

RELAX was a multicenter randomized (1:1) placebo-controlled trial testing the impact of chronic PDE5 inhibition (sildenafil) on exercise capacity in patients with HFpEF (ClinicalTrials.gov identifier NCT00763867)⁸. The trial was conducted by the Heart Failure Clinical Research Network (HFN) and funded by the National Heart, Lung, and Blood Institute. All patients provided written informed consent and the trial was approved by the institutional review board at each participating site. Notably, sildenafil did not improve exercise capacity in RELAX HFpEF patients and there was no evidence of an interaction between sildenafil and rhythm status⁸.

Study participants

RELAX enrolled patients with objective evidence of HF, LVEF≥50%, reduced exercise capacity, as evidenced by reduced (≤60% predicted) peak oxygen consumption (peak VO₂) at screening CPXT¹¹ and evidence of elevated filling pressures (elevated pulmonary capillary wedge pressure measured invasively or NT-proBNP≥400pg/ml). Additional entry criteria have been reported previously⁸. Rhythm status was classified as rhythm present on baseline electrocardiography (ECG). A history of AF was also recorded.

Baseline studies

The RELAX study design has been reported previously^{8, 12}. Pertinent to this analysis, patients underwent baseline transthoracic echocardiography performed according to a standard protocol¹³ with measurements performed at the HFN echocardiography core laboratory (Mayo Clinic, Rochester, MN). Values reported in AF represent an average of 3 to 5 beats.

Plasma biomarker measurement included markers of neurohumoral activation (aldosterone, N-terminal pro-B-type natriuretic peptide [NT-proBNP], endothelin-1), cardiac injury or inflammation (troponin I, c-reactive peptide [CRP]), renal function (cystatin C, uric acid), and fibrosis (procollagen III n-terminal peptide [NT-procollagen III], galectin 3, c-telopeptide for type I collagen [CITP]). Assays were performed at the HFN biomarker core laboratory (University of Vermont, Burlington, VT)^{8, 12}. Details of the RELAX CPXT protocol and HFN CPXT core laboratory (Massachusetts General Hospital, Boston, MA) methodologies have been reported¹².

Briefly, symptom-limited CPXT with simultaneous expired ventilatory gas analysis was performed by treadmill or bicycle ergometry. Percent predicted peak VO₂ was calculated according to published age and sex-normalized values¹¹. Age, sex, body weight and mode-specific predicted peak VO₂ was also calculated using the Wasserman equation^{14, 15}. Entry criteria specified maximal effort as evidenced by a peak respiratory exchange ratio (RER)>1.0. The ventilatory anaerobic threshold (VAT) was determined by the V-slope method¹⁶. Peak oxygen pulse, reflecting oxygen consumption per heart beat during exercise, was obtained from the ratio of peak VO₂ to peak exercise heart rate (HR). Peak circulatory power was defined as the product of peak VO₂ and peak systolic blood pressure (SBP), while circulatory stroke work (SW) was derived from the ratio of circulatory power to peak exercise HR¹⁷. The V_E/VCO₂ slope for the entire duration of exercise was calculated based on 10s averaged V_E (L/min) and VCO₂ (L/min) data¹⁸.

Resting HR was documented after 5 minutes in a stationary position prior to CPXT. Peak HR was defined as HR at peak VO₂. Chronotropic response was expressed as the change in HR from rest to peak exercise. Age-predicted maximal HR (APMHR) was defined with the Astrand (220-age)¹⁹, and Brawner (164-[0.7*age])²⁰ formulae, and each used to calculate a chronotropic index (CI) reflecting the percentage of HR reserve used (change in HR from rest to peak exercise/ APMHR minus resting HR). Clinically significant chronotropic incompetence was defined as a CI<0.8 for patients not taking betablockers, and <0.62 in patients reporting active betablocker use (Astrand formula). No betablocker correction was employed for chronotropic incompetence defined by the Brawner formula. HR recovery (HRR) was taken as the absolute difference in HR between peak exercise and at 1 minute during exercise unloading or cessation. A HRR≤18bpm was considered abnormal²¹.

Statistical analysis

Continuous data are presented as mean ±standard deviation or median (25^th, 75^th percentiles) as appropriate; categorical data are presented as frequency (%) within each group (AF vs. SR). Baseline characteristics and CPXT parameters for the RELAX study population stratified by rhythm status were compared using the t test or Wilcoxon rank sum test for continuous variables, and chi-squared test for categorical variables. Univariable and multivariable linear regression analyses for pre-specified pertinent variables were performed to define the association between rhythm status and peak VO₂. To adjust for the pathophysiological role of chronotropic response to exercise, a linear regression model was used to examine the relationship between CI and peak VO₂ or peak workload with an interaction term included for rhythm status, thereby comparing the slope of the VO₂-CI or workload-CI relationship between patients in AF and SR. Normality of model residuals was tested using the Kolmogorov-Smirnov test and visually assessed for symmetry. Analyses were performed using SAS version 9.2.; p<0.05 (2-sided) was considered statistically significant.

Results

Patient characteristics

RELAX enrolled 216 patients with HFpEF (mean age 69±10 years, 48% female) of whom 79 (37%) had AF, 124 (57%) were in SR, and 13 (6%) were in other rhythms (excluded from this analysis). Patients in AF were older than those in SR, but had similar reported symptom severity (NYHA class, MLWHFQ score), distribution of co-morbidities, hemoglobin and renal function (Table 1). Loop diuretic and digoxin therapy were more frequent, angiotensin converting-enzyme inhibitor (ACEI) or angiotensin receptor blocker (ARB) use less frequent, and betablocker use similar among AF patients compared to those in SR.

Table 1. Baseline characteristics by rhythm status at enrollment.

Variable	Atrial Fibrillation n=79	Sinus rhythm n=124	p-value
Age, y	72.7±9.17	65.7±10.3	<0.0001
Female sex	32 (41)	67 (54)	0.06
BMI, kg/m2	33.2±6.4	34.7±8.2	0.14
Rest Systolic BP, mmHg	126±15	129±18	0.17
Rest Diastolic BP, mmHg	71±10	69±10	0.18
Rest HR, bpm	72±13	68±11	0.04
NYHA III/IV	42 (53)	65 (52)	0.92
MLWHFQ score	46±21	45±24	0.70
Ischemic etiology	33 (42)	44 (36)	0.37
Comorbidities
Hypertension	65 (82)	108 (87)	0.35
Diabetes	28 (35)	56 (45)	0.17
COPD	15 (19)	24 (19)	0.95
History of AF	79 (100)	28 (23)	<0.0001
GFR, mL/min/1.73m2	65.3 (51.6,77.2)	65.6 (43.5,86.0)	0.77
Cystatin C, mg/L	1.4 (1.1,1.7)	1.3 (1.0,1.7)	0.07
Hb, mg/dL	12.6 (12.1,13.8)	13.0 (11.9,13.9)	0.58
NT-procollagen III, μg/L	8.1 (6.4,10.7)	7.5 (5.5,9.5)	0.10
CITP, μg/L	6.7 (5.5,10.2)	5.7 (4.2,9.0)	0.003
Galectin 3, ng/mL	14.3 (11.7,19.7)	13.6 (10.8,16.9)	0.15
Medication at enrollment
ACEI/ARB	49 (62)	97 (78)	0.01
Beta blocker	63 (80)	91 (73)	0.30
Aldosterone antagonist	10 (13)	11 (9)	0.39
Loop diuretic	77 (98)	78 (63)	<0.0001
Digoxin	18 (23)	3 (2)	<0.0001
Calcium channel blocker	28 (35)	36 (29)	0.34
Amiodarone	1 (1)	7 (6)	0.15
Other antiarrhythmic	3 (4)	6 (5)	1.00
Aspirin/Thienopyridine	31 (39)	90 (73)	<0.0001
Warfarin	69 (87)	22 (18)	<0.0001
Echocardiography
LVEF, %	59±8	63±7	0.002
Rest cardiac index, L/min/m2	2.4±0.6	2.6±0.7	0.04
Stroke volume, mL	70±20	84±24	0.0001
eFS, %	36±7	40±6	0.0002
mFS, %	19±4	21±3	0.0009
LV mass index, g/m2	84±33	82±33	0.77
LV diastolic dimension, cm	4.6±0.7	4.7±0.6	0.71
Deceleration time, ms	172±48	203±44	<0.0001
LAVI, mL/m2	62±25	41±13	<0.0001
PASP, mmHg	45±12	41±12	0.045
RA pressure, mmHg	11±5	7±4	<0.0001
E/e’ (medial)	18.6±9.2	18.2±9.7	0.74

Continuous variables shown as mean ±SD (median [25^th,75^th percentiles] for biomarkers); categorical variables shown as count (percentage).

ACEI, angiotensin converting enzyme inhibitor; ARB, angiotensin receptor blocker; BMI, body mass index; BP, blood pressure; CITP, c-telopeptide for type I collagen; COPD, chronic obstructive pulmonary disease; eFS, endocardial fractional shortening; GFR, glomerular filtration rate; HR, heart rate; LAVI, left atrial volume index; LV, left ventricular; LVEF, left ventricular ejection fraction; mFS, midwall fractional shortening; MLWHFQ, Minnesota living with heart failure questionnaire total score; NT-procollagen III, procollagen III n-terminal peptide; NYHA, New York Heart Association class; PASP, pulmonary arterial systolic pressure; RA, right atrial.

LV dimensions and LV mass index (LVMI) were comparable between AF and SR; however AF was associated with worse systolic function at rest (lower LVEF, stroke volume [SV], endocardial [eFS] and midwall fractional shortening [mFS]). Although E/e’ was similar between groups, other parameters of LV diastolic function were significantly worse in AF (shorter deceleration time, higher right atrial pressure [RAP], larger left atrial volume index [LAVI]). Pulmonary artery systolic pressures [PASP] were also higher in AF. Neurohumoral activation was more severe in AF relative to SR (elevated plasma NT-proBNP, aldosterone, endothelin-1; Figure 1). Troponin I levels were higher in AF than SR, consistent with greater myocardial necrosis (Figure 1). Plasma markers of fibrosis (NT-procollagen III, CITP, galectin 3) were higher in AF than SR, however only CITP reached statistical significance (Table 1).

Figure 1. Biomarkers of neurohumoral activity in HFpEF patients in atrial fibrillation and sinus rhythm.

(a) Plasma NT-proBNP, (b) aldosterone, (c) endothelin-1, (d) troponin I, (e) uric acid, (f) c-reactive protein. White bars (atrial fibrillation), black bars (sinus rhythm). Median (75^th percentile) shown.

Exercise performance

Fewer patients in AF performed bicycle ergometry (52% AF vs. 68% SR, p=0.02). Both groups performed a maximal or near-maximal CPXT, inferable from subjective (Borg score) and objective (RER) measures of exertion at peak exercise (Table 2). The most common reason for exercise cessation was dyspnea in AF (49% AF vs. 37% SR) and fatigue among patients in SR (31% AF vs. 52% SR). Exercise duration was shorter for AF than SR (mean 9.0 vs. 10.1min, p=0.02) but not after age-sex adjustment (p=0.14).

Table 2. Cardiopulmonary exercise test data.

Variable	Atrial Fibrillation n=79	Sinus rhythm n=124	p-value
Exercise duration, min	9.0±3.0	10.1±3.0	0.02
6MWD, m	283±107	313±109	0.055
Rest VO2, mL/kg/min	3.1±0.6	2.9±0.8	0.049
Rest pulse pressure, mmHg	51±15	59±19	0.001
Peak RER	1.1±0.1	1.1±0.1	0.62
Peak Borg score	6.8±2.5	6.8±2.3	0.91
Peak SBP, mmHg	138±30	163±29	<0.0001
Peak DBP, mmHg	69±14	74±15	0.02
Peak pulse pressure, mmHg	69±26	89±24	<0.0001
Peak VO2, mL/kg/min	11.7±2.7	12.8±3.2	0.008
VO2 at VAT, mL/kg/min	7.2±1.8	7.7±1.9	0.07
VO2 at VAT (% of peak VO2)	62.6±9.2	60.6±8.6	0.12
% age-sex predicted VO2
Wasserman, %	63.6±14.2	68.8±16.0	0.02
Standard, %	40.4±8.6	42.8±9.6	0.11
Peak O2 pulse, mL/kg/bpm	0.11±0.03	0.12±0.03	0.07
Peak O2 pulse, mL/bpm	10.5±3.2	11.6±4.3	0.04
Peak circulatory power, mmHg*mL/kg/min	1644±588	2109±751	<0.0001
Circulatory stroke work, mmHg*mL/kg/bpm	15.5±5.7	19.4±6.1	<0.0001
Peak workload, Watts	67±29	77±32	0.03
Peak VCO2, mL/kg/min	12.8±3.3	14.1±3.8	0.01
Peak VE, L/min	43.6±12.7	47.0±16.0	0.12
VE/VCO2 slope	35.1±7.2	32.6±8.1	0.03
Chronotropic indices
Rest HR (pre-exercise)	72±12	69±14	0.14
Peak HR, bpm	109±27	111±24	0.69
Chronotropic response, bpm	37±23	42±20	0.17
Age-predicted maximal HR, bpm
Astrand formula†	147±9	154±10	<0.0001
Brawner formula‡	113±6	118±7	<0.0001
Chronotropic index
Astrand formula†	0.51±0.34	0.50±0.25	0.73
Brawner formula‡	1.10±1.08	0.92±0.61	0.19
HRR, bpm	11±15	10±9	0.51
HRR≤18bpm, n (%)	55 (73)	94 (83)	0.14

Continuous variables shown as mean ±SD; categorical variables shown as count (percentage).

HR, heart rate; HRR, absolute heart rate recovery at 1 minute post exercise; RER, respiratory exchange ratio; VAT, ventilatory anerobic threshold; VE, minute ventilation; VCO₂, carbon dioxide output; VE/VCO₂, slope of the relationship between minute ventilation and carbon dioxide output; VO₂, oxygen consumption.

^†220-age¹⁹

^‡164-[0.7*age]²⁰

Resting VO₂ was higher in AF patients compared to SR. However peak VO₂, scaled to body mass (standard), was significantly reduced in AF, and confirmed by a lower percent-predicted peak VO₂ (Wasserman formula; Table 2). VO₂ at VAT tended to be lower in AF, though as a proportion of peak VO₂ was similar between groups. Peak exercise workload was also lower in AF relative to SR. On multivariable analysis, AF was associated with a reduced peak VO₂ even after adjustment for age, sex, LVEF, and exercise modality (Table 3).

Table 3. Relationship between atrial fibrillation and exercise capacity measured by peak VO₂.

Model		Linear regression analysis (peak VO2)
	Sample size	Estimated difference between AF and SR (mL/kg/min)	95%CI	p-value
Unadjusted	202	−1.2	−2.0 to −0.3	0.008
Adjusted for age/sex	202	−0.9	−1.7 to −0.1	0.03
Adjusted for	199	−0.9	−1.8 to −0.1	0.03
age/sex/EF
Adjusted for	199	−1.0	−1.9 to −0.2	0.02
age/sex/EF and
exercise modality*
Adjusted for age/ sex/
chronotropic index
Astrand formula	200	−1.2	−2.2 to −0.3	0.01
Brawner formula	200	−1.2	−2.2 to −0.2	0.02

CI, confidence interval; EF, ejection traction.

^*Bicycle vs. treadmill

Minute ventilation was similar in AF and SR but the V_E/VCO₂ slope was elevated in AF. Even after age-sex adjustment, patients with AF demonstrated a significantly reduced peak exercise SBP (p<0.0001; Table 3). AF patients had lower pulse pressure, lower circulatory power and stroke work; O₂ pulse was also or tended to be lower (Table 3). Six-minute walk distance, representing submaximal exercise capacity, also displayed a lower trend in AF versus SR (Table 2).

Heart rates at rest and at peak exercise were similar between patients in AF and SR (Table 2, Figure 2A). Likewise, despite a lower mean APMHR (Astrand/Brawner formula) for AF patients, the chronotropic response as reflected by the CI (Table 2) and the prevalence of chronotropic incompetence (Figure 2B) were equal between groups. Albeit a majority of patients reported betablocker use (Table 1), the median CI was similar among AF patients with and without betablockers (0.85 vs. 0.9, p=0.90), and higher in SR patients on betablockers compared to those without (0.74 vs. 0.42, p<0.0001). The relationship between peak VO₂ or peak workload and CI did not differ between HFpEF patients in AF or SR (Figure 2C-D) and AF remained associated with a lower peak VO₂ even after adjustment for CI (Table 3).

Figure 2. Chronotropic response to exercise in HFpEF patients in atrial fibrillation and sinus rhythm.

(a) Heart rates at rest and at peak exercise, (b) prevalence of chronotropic incompetence during exercise (calculated using standard [Astrand] or Brawner formulae; p-values AF vs. SR), (c) relationship between chronotropic index (Brawner formula) and peak VO₂, (d) relationship between chronotropic index and peak workload. Results for unadjusted linear regression shown for patients in atrial fibrillation (red line) and sinus rhythm (black line). p-values (c) and (d) refer to interaction terms between rhythm status and chronotropic index.

While a majority of HFpEF patients had abnormal HRR consistent with autonomic dysfunction, the severity of impaired HRR was similar between groups.

Discussion

This analysis of the RELAX HFpEF cohort demonstrates that important phenotypic differences exist between HFpEF patients with and without AF and is the first comprehensive analysis of the impact of AF on exercise capacity in HFpEF. Consistent with our hypothesis, HFpEF patients in AF were older and exhibited worse LV systolic and diastolic function at rest, more severe neurohumoral activation, and greater impairment of exercise capacity compared to HFpEF patients in SR. Peak VO₂ was lower in AF despite adjustment for pertinent variables, and was not accompanied by a higher chronotropic response as has been seen in HFrEF patients with AF²². Ventilatory efficiency was lower (steeper V_E/VCO₂ relationship) in AF, suggesting greater impairment of pulmonary perfusion during exercise. Peak exercise SBP, circulatory power, and circulatory stroke work were lower and peak O₂ pulse tended to be lower in AF suggesting impaired systolic reserve function. These findings demonstrate that mechanisms beyond altered chronotropic response mediate the more impaired exercise tolerance in HFpEF patients with AF and suggest that AF is a marker of a more advanced HFpEF phenotype.

In patients with HFrEF, AF is associated with more severe symptoms and LV systolic impairment²³. In HFrEF, a rhythm control strategy does not result in better outcomes than rate control²⁴, and betablockers may have reduced pharmacological efficacy contrary to established benefits for patients in SR²⁵. Importantly, characteristics associated with AF in HFpEF are less well documented, despite the high prevalence of AF in HFpEF and evidence of equivalent²⁶ or perhaps stronger association with morbidity and mortality in HFpEF than HFrEF^{2, 27}. Existing studies are highly discrepant, have limited phenotypic characterization and enrolled patients during a hospitalization for HF which may not accurately reflect the HFpEF population at large^27-29. Although the CHARM-Preserved trial studied stable HFpEF outpatients with and without AF², the prevalence of coronary artery disease was higher than typically observed in HFpEF and HFpEF was defined by a LVEF>40% thereby including patients with potential HFrEF pathophysiology.

RELAX enrolled stable chronic HFpEF patients in SR and AF, at least 3 months out from any HF hospitalization, exhibiting a LVEF≥50% consistent with current diagnostic criteria⁷. In this setting AF was associated with older age and worse LV systolic and diastolic function, mirroring observations in HFrEF²³ but contrary to sparse available data in HFpEF and AF, where systolic function (LVEF) has been reported as similar to SR^27-29. Furthermore, in contrast to the findings of Linssen et al.²⁷, and perhaps due to the exclusion of patients with LVEF40-49%, we found NT-proBNP to be significantly elevated in AF relative to HFpEF-SR patients. This corroborates baseline findings in the I-Preserve population³⁰, and may either pertain to AF and/or more severe HF in HFpEF-AF patients. More severe HF is also suggested by elevation in other heretofore unstudied biomarkers of HF severity in HFpEF-AF patients (aldosterone, ET-1, troponin I), greater diuretic use and a lower cardiac index. A higher resting VO₂ was observed in AF which, as in HFrEF, may signify increased resting energy demands or hypermetabolism corresponding to increased HF severity ^{31, 32}. These findings support the rationale for aggressive upstream therapy addressing the atrial substrate in HF patients with AF, which is currently under investigation³³.

AF has been associated with reduced functional capacity in HFrEF²² and in patients without structural heart disease³⁴, however prior studies of exercise tolerance in HFpEF have mostly excluded patients in AF^{9, 35, 36}. Similarly, HFpEF has been underrepresented in AF intervention trials^{4, 24, 37}, thus the impact of AF on exercise capacity and cardiopulmonary function during exercise in HFpEF remains unclear. Fung et al. reported the mean 6MWD to be lower in hospitalized HFpEF patients with AF (279±66m, n=42) compared to SR (338±86m, n=104)²⁸, however formal cardiopulmonary stress testing was not performed. In the present analysis we confirm and extend this observation in several respects.

Our novel principal finding is that, compared to SR, HFpEF-AF patients have a lower peak VO₂ both in absolute terms and relative to age, sex, body-size and exercise mode-adjusted predicted values. Further, we demonstrate that the impaired exercise capacity in AF is not explained by differences in resting LVEF or altered chronotropic response, suggesting a specific effect of AF on cardiac reserve. Rhythm irregularity³⁸ and loss of atrial contribution to LV filling³⁹ are recognized to impair cardiac output in AF. Indeed, the lower pulse pressure and circulatory stroke work observed in HFpEF-AF patients suggest diminished SV at peak exercise. Additionally, since peak VO₂ represents the product of cardiac output and arterial-venous oxygen content difference, a greater impairment of peripheral oxygen availability or utilization in AF may also be important. Endothelial dysfunction⁴⁰ and peripheral muscle blood flow abnormalities have been correlated with exercise performance in patients with chronic AF, although in the absence of structural heart disease⁴¹.

The mean V_E/VCO₂ slope was significantly steeper in AF-HFpEF patients compared to SR, reflecting impaired ventilatory efficiency. As the prevalence of lung disease was similar between groups, this likely reflects impaired augmentation of pulmonary perfusion during exercise, a factor which may contribute to earlier exercise cessation and reduced functional capacity in AF⁴². An increased V_E/VCO₂ slope has been observed in patients with permanent AF compared to control subjects in SR⁴³ and accompanying lone AF before cardioversion to SR^{34, 44}. However, Ariansen et al.⁴³ reported no difference in ventilatory efficiency between AF and SR when HF patients were excluded from their analysis. Further, Agostoni et al.²² did not observe an association between AF and V_E/VCO₂ slope in chronic HFrEF. Therefore it remains unclear whether the steeper V_E/VCO₂ slope can be ascribed to a specific effect of AF on exercise hemodynamics. More likely AF is a marker of reduced pulmonary vascular reserve as well as more impaired diastolic reserve function in HFpEF patients, as suggested by the differences in baseline characteristics.

Interestingly HFpEF-AF patients did not display an exaggerated chronotropic response to exercise compared to SR patients, contrary to previous findings in lone AF³⁴ and AF with HFrEF^{3, 22}. Peak HR and CI were similar between AF and SR, while peak VO₂ and workload were lower in AF patients; however, the relationships between CI and peak VO₂ or workload were not significantly different for AF than SR. Chronotropic incompetence is a common feature in HFpEF patients in SR⁹ and our data demonstrate that HFpEF patients in AF have a similar prevalence of chronotropic incompetence when their prevalence of BB use is equivalent to SR. Notably, no symptomatic or prognostic benefit has been demonstrated with strict versus lenient HR control in the general AF population⁴⁵ including a post-hoc analysis of patients with HF⁴⁶ although the latter did not examine the impact of rate control strategy on events stratified by HFpEF and HFrEF. Therefore, although current ACC/AHA guidelines recommend HR control in HFpEF patients with AF (class IC)⁴⁷, the optimal level of HR control in HFpEF, particularly with respect to exercise capacity, has yet to be established.

Study Limitations

Rhythm was classified by the presence or absence of AF on enrollment ECG. Consequently HFpEF patients with previous or paroxysmal AF may have been classified as SR, though this would likely serve to underestimate the observed differences. Greater frequency of treadmill exercise versus bicycle ergometry in AF would also minimize any differences in peak VO₂ from SR as bicycle ergometry is generally associated with a lower peak VO₂. Information on right heart structure and function, along with duration of AF and decisions regarding rate or rhythm control were unavailable, though antiarrhythmic agents were used in <10% of the overall population. Therefore, these data pertain to patients with and without persistent AF in the context of heterogeneous AF and HF therapeutic strategies.

Conclusions

In patients with HFpEF, AF is associated with important phenotypic and functional differences: older age, impaired LV systolic and diastolic function and functional reserve, more severe neurohumoral activation, and impaired exercise tolerance, supporting AF as a marker of more advanced HF phenotype in HFpEF. Furthermore, in the context of betablocker therapy, chronotropic incompetence is equally common in HFpEF patients with AF or SR. Further study is required to determine whether different rate control strategies or indeed, rhythm control in HFpEF patients with AF may favorably impact exercise tolerance.