Cardiac output response to exercise in relation to metabolic demand in heart failure with preserved ejection fraction

Study population

The total study population is compiled from three cohorts of patients with HFpEF and controls. Cohort 1 (n = 112) includes consecutive patients and controls who underwent invasive supine exercise ergometry studies with simultaneous expired gas analysis at the Mayo Clinic from 2002 to 2011. No data from the cohort 1 patients have been previously published. Cohort 2 (n = 50)¹⁵ and cohort 3 (n = 36)¹² are from previously published prospective studies examining upright exercise haemodynamics. Some clinical characteristics, exercise capacity, and ventricular–vascular function data from cohorts 2 and 3 have been published,^12,15 but the cardiovascular responses as they relate to VO₂, the primary aim of the current study, have not been reported. This study complies with the Declaration of Helsinki and has been approved by Institutional Review Boards of the Mayo Clinic and Johns Hopkins Hospital.

Heart failure with preserved EF was defined by LVEF ≥50% and cardiologist-adjudicated diagnosis of HF (Framingham criteria).^12,15 Exclusion criteria were significant valvular disease (moderate or greater left-sided regurgitation, any stenosis), cor pulmonale, significant pulmonary disease, unstable coronary disease or coronary spasm, primary renal or hepatic disease, constrictive pericarditis, or infiltrative, restrictive, or hypertrophic cardiomyopathies. Controls in cohort 1 were referred to the cath lab for assessment of exertional dyspnoea and were found to display no cardiac pathology after thorough invasive and non-invasive evaluation. Controls in the non-invasive cohorts were recruited from the community.

Study measurements performed at rest and during exercise in cohorts 1–3 are shown in Supplementary material, Table S1. All patients were studied in the post-absorptive state. Baseline ventricular morphology and function, diastolic filling characteristics, and left atrial volume were measured by transthoracic echocardiography. Echocardiographic LV stroke volume (SV) was determined from LV outflow Doppler, EF was determined by Simpson's biplane method, and LV end-diastolic volume was determined by SV/EF.^15,19 Heart rate (HR) was continuously recorded by electrocardiography. Breath-by-breath expired gas analysis was performed at rest and during exercise in all studies (MedGraphics, St Paul, MN, USA) to measure oxygen consumption (VO₂).

Invasive haemodynamic exercise assessment (cohort 1)

Right heart catheterization was performed in the supine position through the internal jugular vein.¹⁷ Right atrial (RAP), pulmonary artery (PAP), and pulmonary capillary wedge (PCWP) pressures were assessed at end-expiration at rest and after ≥2 min had elapsed at peak workload during supine cycle ergometry. Systemic blood pressure (BP) was measured by intra-arterial (radial) catheter (n = 55) or by cuff sphygmomanometry (n = 41).

Arterial–venous oxygen content difference (AVO₂diff) was measured directly as the difference between systemic and pulmonary arterial O₂ contents (=saturation ×haemoglobin ×1.34). CO was determined by the direct Fick method (=VO₂/AVO₂diff). The SV was determined by CO/HR. Pulmonary vascular resistance [PVR = (mean PAP – PCWP)/CO] and effective arterial elastance (Ea = 0.9 ×systolic BP/SV)^15,23 were determined as measures of pulmonary and systemic arterial afterload.

Non-invasive haemodynamic exercise assessment (cohorts 2 and 3)

Subjects in cohorts 2 and 3 underwent maximal-effort graded exercise testing on an upright cycle ergometer as previously described.^12,15 Cardiac volumes, SV, and EF were assessed at rest and during the final 2 min of exercise by echocardiography in cohort 2 and nuclear gated blood pool scan in cohort 3. CO was determined by SV ×HR. AVO₂diff was determined by the Fick method (=VO₂/CO).¹⁹ Ea was determined as in cohort 1.

Statistical analysis

Data are reported as mean ± standard deviation or median (25th, 75th interquartile range). Between-group differences were compared by t-test, Wilcoxon rank-sum test, or χ². Multivariable linear regression analysis was used to adjust for relevant baseline group differences and to examine differences with upright and supine exercise, in which the dependent variable was the normally distributed continuous outcome variable of interest, and factors entered into the model included age, gender, body mass, and history of hypertension, diabetes, creatinine, and medication use, or exercise posture. For non-normally distributed variables entered into regression models, the assumption of normally distributed residuals was verified by Quantile plots, and no violations were observed.

Subject characteristics

Patients with HFpEF (n = 109) and controls (n = 73) were assembled from cohorts 1 (n = 71, 25), 2 (n = 21, 29), and 3 (n = 17, 19). The vast majority (n = 100) of patients met recently proposed diagnostic criteria for HFpEF.²⁴ In the remaining nine HFpEF subjects, there was incomplete echocardiographic data limiting classification, though each of these patients had been previously hospitalized for pulmonary oedema that improved with diuresis. Compared with controls, HFpEF patients were older, heavier, and more likely to display co-morbidities and receive treatment with antihypertensives and diuretics (Table 1). Levels of BNP were higher in HFpEF patients compared with controls, whereas haemoglobin and estimated glomerular filtration rate were lower. LV chamber size was similar in HFpEF patients and controls, while LV mass, left atrial volume, and E/e' ratio were greater in HFpEF patients.

Baseline haemodynamics

Resting HR, BP, CO, and Ea were similar in HFpEF patients and controls (Table 1). Pulmonary artery O₂ content was lower in those with HFpEF, suggesting relative inadequacy of O₂ delivery compared with controls. Pulse pressure was higher in HFpEF patients, consistent with greater systemic arterial stiffening. Right and left heart filling pressures, PAPs, and PVR were higher in HFpEF patients than in controls.

Exercise performance

Resting VO₂ was similar in HFpEF patients and controls, but when scaled to weight, resting VO₂ was lower in those with HFpEF (Table 2). Compared with controls, HFpEF patients achieved lower peak workload, reduced peak VO₂, and less increase in VO₂ during exercise. Each of these differences persisted after adjusting for age, body mass index (BMI), hypertension, diabetes, glomerular filtration rate, haemoglobin, vasodilator use, and beta-blocker use. Exercise systemic O₂ content was similar in HFpEF patients and controls, while pulmonary artery O₂ content was lower in those with HFpEF. No difference between the groups was observed in resting, peak exercise, or absolute exercise change in AVO₂diff.

Exercise haemodynamics

Compared with controls, HFpEF subjects demonstrated less increase in HR, BP, SV, EF, and CO with exercise (Table 2, Figure 1). Depressed EF reserve was due to lesser reduction in LV end-systolic volume in HFpEF patients, as changes in end-diastolic volume were similar in HFpEF patients and controls. Right and left heart filling pressures and PAPs with exercise were higher in those with HFpEF than in controls. Each of these differences persisted after adjusting for age, BMI, hypertension, diabetes, glomerular filtration rate, haemoglobin, vasodilator use, and beta-blocker use. Pulmonary hypertension in HFpEF patients was predominantly due to high PCWP, as PVR reductions with exercise were similar in HFpEF patients and controls (–0.5 ± 1.0 vs. –0.2 ± 0.7 mmHg/L/min, P = 0.3). Exercise Ea and changes in Ea were similar between HFpEF patients and controls.

Integrated responses

Depressed CO reserve in HFpEF patients could be caused by the lower absolute workload achieved or by cardiac limitation. To distinguish between these possibilities, the enhancement in CO relative to VO₂ was then analysed, identifying CO limitation as the primary culprit in HFpEF on average (Figure 2A and B). This is reflected by a significantly lower ΔCO/ΔVO₂ slope in HFpEF patients compared with controls (5.9 ± 2.5 vs. 7.4 ± 2.6, P = 0.0005; Table 2). Enhancement of CO with exercise was also reduced in HFpEF patients for any change in LV filling pressure (PCWP) or LV end-diastolic volume (Figure 3A and B). The ΔCO/ΔVO₂ relationships were similar in HFpEF subjects treated or not treated with beta-blockers (5.6 ± 2.4 vs. 6.3 ± 2.6, P = 0.2). In contrast to VO₂, scaling ΔCO to external work performed (in Watts) did not reveal a difference in HFpEF patients and controls overall or when substratified into lower and higher VO₂ categories (Supplementary material, Table S2).

While absolute changes in oxygen extraction (AVO₂diff) at peak exercise were similar in HFpEF patients and controls (Table 2, Figure 2C), HFpEF subjects relied on a greater increase in AVO₂diff for any absolute change in VO₂ (Figure 2D). The increase in O₂ extraction relative to O₂ consumption (ΔAVO₂diff/ΔVO₂ slope) was thus greater in HFpEF patients compared with controls (8.9 ± 3.4 vs, 5.5 ± 2.0 min/dL, P < 0.0001). Diabetics had higher resting AVO₂diff (P = 0.04), but changes in AVO₂diff with exercise were not different from those of non-diabetics. There were no differences in AVO₂diff at rest or with exercise in the hypertensive and obese subgroups.

Importantly, these differences in ΔCO/ΔVO₂ and ΔAVO₂diff/ΔVO₂ in HFpEF patients and controls were consistently observed when cohorts 1–3 were analysed separately rather than together (Table 3). Thus, regardless of body position during exercise, or method used to measure CO, the increase in CO was on average consistently impaired in those eith HFpEF, relative to metabolic demand. These differences persisted after adjusting for age, BMI, hypertension, diabetes, glomerular filtration rate, haemoglobin, vasodilator use, and beta-blocker use.

Impact of body position

Compared with supine ergometry, upright exercise was associated with greater increases in VO₂, CO, AVO₂diff, HR, and BP (Supplementary material, Table S3). However, in multivariable regression analysis including exercise position in the model, HFpEF subjects continued to display highly significant impairments in the exercise augmentation in HR, SV, BP, CO, VO₂, and ΔCO/ΔVO₂, with greater ΔAVO₂diff/ΔVO₂ (Supplementary material, Table S3). Significant group ×position interactions were observed for exercise increases in CO and BP, where HFpEF patients displayed the smallest increases compared with controls during upright exercise.

Cardiac output and peak VO₂ in heart failure with preserved ejection fraction

Functional capacity (peak VO₂) is similarly impaired in HFpEF and HFrEF.⁹ Peak VO₂ is an integrated measure of cardiac reserve that is being used to diagnose HFpEF²⁵ and as an endpoint in clinical trials.^26–28 Thus, better characterization of the fundamental mechanisms underlying VO₂ limitation in HFpEF is essential for improved understanding of pathophysiological mechanisms and to better inform future trial design and identify therapeutic _targets.

Most studies have reported that exercise VO₂ and CO are individually depressed in HFpEF.^{10,12,15–20,23,29,30} However, VO₂ and CO at any time during exercise are largely determined by the intensity of work being performed.⁴ Thus, group differences in CO might be caused by cardiac limitations or non-cardiovascular factors including patient motivation, peripheral limitations, fitness, or orthopaedic issues. Elevation in cardiac filling pressures (at rest or with stress) is the most conspicuous and consistently observed haemodynamic feature in HFpEF.¹⁷ Non-diastolic limitations have been reported,^{12,15,16,19,23,29,30} but their roles have been questioned.²² Impaired exercise reserve responses in non-diastolic parameters (e.g. HR or contractile response) may not be causal of exercise intolerance, but rather consequence of premature cessation of exercise in response to dyspnoea from high filling pressures,¹⁸ abnormal metabolic–neural signalling,^31,32 or non-cardiac factors such as deconditioning or obesity.^11,14 Indeed, textbooks of exercise physiology describe how patients with HFrEF are limited by inadequate CO reserve with exercise, whereas patients with HFpEF have normal CO responses but elevated filling pressures.³³ It has even been questioned whether HFpEF truly represents a form of cardiac failure, since patients are frequently elderly with co-morbidities that might in themselves produce symptoms of effort intolerance that are not directly attributable to cardiac dysfunction.²¹

The current data provide compelling evidence that the reduction in exercise capacity in HFpEF is determined largely by inadequate CO reserve, which, when combined with stress-induced elevations in cardiac filling pressures,^{10,13,16,17,20} markedly limits exercise capacity. The strength of this experimental approach lies in the simultaneous assessment of both whole-body O₂ delivery (CO) and O₂ consumption (VO₂). Because increases in CO during exercise are ultimately driven by increases in VO₂,^5,6 this analysis allows for direct comparisons of exercise responses between patients with HFpEF and controls, without the need to adjust for measures of effort adequacy (such as the respiratory exchange ratio) that are necessary to gauge metabolic status when CO is not directly measured. Intriguingly, scaling CO reserve to external work failed to reveal the cardiac limitation in HFpEF, suggesting that this index is less sensitive to haemodynamic impairments in HFpEF.

The relationship between CO and VO₂ is typically depressed in patients with HFrEF,² in keeping with the definition of HF as an inability to pump blood adequately to the body at normal filling pressures.¹ However, only one previous study has examined the relationship between CO and VO₂ in HFpEF.¹⁸ Bhella and colleagues found that peak VO₂ and CO were reduced in HFpEF, similar to the current data, but, when plotting CO relative to VO₂, the authors surprisingly found that the enhancement in CO was elevated in HFpEF. The authors speculated that abnormalities in skeletal muscle might generate metabolic signals that drive excessive increases in CO, leading to increased ventricular filling pressures during exercise in HFpEF.¹⁸

The current data argue against this hypothesis, showing that on average the increase in CO relative to VO₂ was impaired in HFpEF. The reasons for the discordant findings are not obvious, although it is notable that the SV enhancement during exercise in HFpEF patients noted by Bhella et al. (+74% increase) was remarkably high, and well in excess of the +16% increase noted in the current study and the –7 to +10% changes with exercise reported by other groups.^16,23,29,30 Secondly, Bhella and colleagues found that the resting VO₂ was elevated in the HFpEF patients—suggesting a hypermetabolic state. In contrast, in the current study, the resting O₂ consumption (scaled to body mass) was lower in HFpEF. These differences may relate to changes in metabolism reflecting variability in HF severity or chronicity between the two study populations.

Determinants of cardiac output limitation

Cardiac output reserve limitation in HFpEF was related to impaired SV and HR, similar to previous studies,^{10,12,15–17,19,23,29,30} Inadequate SV (and EF) reserve was due to an inability to reduce LV end-systolic volume with exercise, since end-diastolic volume increased similarly in HFpEF patients and controls. Impaired reduction in end-systolic volume could be caused by inadequate enhancement in contractility, blunted afterload reduction, or both. Previous studies have reported attenuated reductions in systemic vascular resistance (SVR) during exercise in HFpEF patients.^12,29,30 In this study, effective arterial elastance (Ea), which characterizes total (resistive and pulsatile) arterial afterload, changed similarly in HFpEF patients and controls. This finding may appear at odds with previous studies showing increased arterial stiffness and impaired flow-mediated dilation in HFpEF.^12,15,29 However, it is also known that Ea varies directly with HR, in addition to SVR.³ Because HR was ∼30% higher at peak exercise in controls, this would inflate the exercise Ea value in this group, even if other components of afterload were lower. The similar change in Ea observed during exercise in the current study suggests that the blunted SV and EF responses with exercise in HFpEF patients were caused primarily by limitations in contractile reserve.

Enhancement in LV end-diastolic volume was similar in HFpEF patients and controls, though diastolic reserve was clearly impaired, as evidenced by the three-fold greater elevation in PCWP in HFpEF patients. It is currently unknown if mitigation of PCWP elevation in HFpEF would directly improve aerobic capacity, though a recent trial found that exercise training was associated with a reduced resting E/e' ratio (a marker of PCWP), and the extent of resting E/e' improvement was correlated with the improvement in peak VO₂.²⁷ Elevation in LV diastolic pressure is often considered to limit exercise capacity by provoking dyspnoea, yet it is notable that PCWP increases during exercise in patients with HFpEF were associated with dramatic elevations in PAPs, increasing right ventricular afterload. Given the well-described impact of right ventricular dysfunction on exercise capacity in HFrEF,³⁵ the enhanced load sensitivity of the right ventricle,³⁶ and the deleterious impact of increased PCWP on pulsatile right ventricular load,³⁷ it is likely that PCWP elevation from diastolic dysfunction in HFpEF has additional implications for right ventricular reserve that may also limit CO responses to exercise.

Arterial–venous oxygen extraction reserve

The Fick equation dictates that VO₂ is equal to the product of CO and AVO₂diff, and patients with HF may display reduced peak VO₂ that is related to abnormalities in the latter, the former, or both.^18,19,38 We found that AVO₂diff at peak exercise was not different between HFpEF patients and controls, though the increase in AVO₂diff as a function of VO₂ (ΔAVO₂ difference/ΔVO₂ slope) was enhanced in those with HFpEF. We speculate that this functions to compensate partly for inadequate O₂ delivery from CO reserve impairment at submaximal workloads in HFpEF.

The similar increase in peak exercise AVO₂diff in cases and controls is similar to recent findings from Maeder and colleagues,¹⁶ but differs from two recent reports showing reduced AVO₂diff at peak exercise in HFpEF patients.^18,19 These discrepancies may relate to the populations studied and different analytical approaches. The latter studies enrolled disease-free controls, as opposed to the current study that included controls with co-morbidities that might influence O₂ extraction. Haykowsky et al. performed comparisons of AVO₂diff relative to workload (Watts) as opposed to VO₂. They also found that the oxygen uptake/work slope was attenuated in HFpEF, meaning that VO₂ was lower for any workload in HFpEF. Thus, the ΔAVO₂diff/ΔVO₂ slope in their HFpEF group might be expected to have been steeper if their data were examined according to VO₂.

When CO is reduced, circulation time is increased, which may provide greater time for gas diffusion at the capillaries and greater O₂ extraction.³⁹ Thus, our findings should not be taken to indicate that patients with HFpEF are more ‘adept’ at peripheral O₂ extraction, or that abnormalities in the periphery do not play key roles in limiting exercise capacity in HFpEF, as indicated in multiple recent studies.^2,18,19,38 Indeed, even with limited CO reserve, improvements in peripheral function may be attainable with interventions such as exercise training, probably related to the greater plasticity in the vasculature and skeletal muscle.^26,27,40 Future research may clarify whether clinical evaluation of both VO₂ and CO reserve will allow for more refined insight as to whether limitations in a specific patient are due predominantly to cardiac or peripheral factors, possibly to better tailor therapy.⁴

Limitations

Patients with HFpEF and control subjects were drawn from three cohorts, two prospectively enrolled and one retrospective. Cohort 1 was a cath lab referral population, introducing potential bias. The protocols and methods to measure CO were different. However, all are well established, and the finding of impaired CO reserve relative to VO₂ was uniformly and consistently observed in both upright and supine exercise when analysed separately and taking into account body position (Table 3; Supplementary material, Table S3). Cardiac pressure and volume were not measured simultaneously during exercise and in the same patient, but pressure and volume data in the three cohorts were included to provide insight into the mechanisms for CO limitation in HFpEF. Importantly, the primary endpoints (CO and VO₂) were directly measured at rest and during exercise in all subjects. Baseline differences were present, including greater age, adiposity, beta-blocker use, and creatinine in those with HFpEF. However, all group differences remained highly significant after adjusting for each of these baseline differences. Subjective symptoms (dyspnoea, fatigue) were not quantified, and many subjects did not exercise to their ideal maximum capacity, particularly during supine ergometry, where peak HR and VO₂ were lower. However, attainment of true maximal objective exercise workload is not necessary in this analysis, because adequacy of CO reserve was evaluated by scaling it to the physiological variable that drives it (VO₂). This study did not assess for the development of mitral regurgitation during exercise, which may also contribute to impaired CO reserve and exertional pulmonary hypertension in HFpEF.²⁹ Not all HFpEF subjects were taking chronic diuretics (69%), but this prevalence is similar to other studies of compensated HFpEF outpatients.¹⁹

Funding

The Mayo Clinic Center for Translational Science Activities, the National Institutes of Health (UL RR024150); the Marie Ingalls Career Development Award in Cardiovascular Research (to B.A.B.).

Conflict of interest: none declared.