Hemodynamic Correlates and Diagnostic Role of Cardiopulmonary Exercise Testing in Heart Failure with Preserved Ejection Fraction

Abstract

Background

Peak oxygen consumption (VO₂) is depressed in patients with heart failure with preserved ejection fraction (HFpEF). The hemodynamic correlates of reduced peak VO₂ and its role in the clinical evaluation of HFpEF are unclear.

Objectives

To define the invasive hemodynamic correlates of peak VO₂ in both supine and upright exercise in HFpEF, and evaluate its diagnostic role as a method to discriminate HFpEF from non-cardiac etiologies of dyspnea (NCD).

Methods

Consecutive patients with dyspnea and normal EF (n=206) undergoing both noninvasive upright and invasive supine cardiopulmonary exercise testing were examined. Patients with invasively-verified HFpEF were compared to those with NCD.

Results

Compared to NCD (n=72), HFpEF patients (n=134) displayed lower peak VO₂ during upright and supine exercise. Left heart filling pressures during exercise were inversely correlated with peak VO₂ in HFpEF, even after accounting for known determinants of O₂ transport according to the Fick principle. Very low upright peak VO₂ (<14 ml/kg/min) discriminated HFpEF from NCD with excellent specificity (91%) but poor sensitivity (50%). Preserved peak VO₂ (>20 ml/kg/min) excluded HFpEF with high sensitivity (90%) but had poor specificity (49%). Intermediate peak VO₂ cutoff points were associated with substantial overlap between cases and NCD.

Conclusions

Elevated cardiac filling pressure during exercise is independently correlated with reduced exercise capacity in HFpEF, irrespective of body position, emphasizing its importance as a novel therapeutic target. Noninvasive cardiopulmonary testing discriminates HFpEF and NCD at high and low values, but additional testing is required for patients with intermediate peak VO₂.

INTRODUCTION

Heart failure (HF) can be defined as an inability of the heart to pump blood to the body at a rate commensurate with its needs, or to do so only at the cost of elevated filling pressures.(1) About half of patients with HF have a preserved ejection fraction (HFpEF).(2) High filling pressures during exercise are pathognomonic of HFpEF(1–5), but the relationships between hemodynamics and exercise capacity remain unclear. Indeed, while exertional dyspnea is often considered to be caused by left atrial hypertension in HF, studies in HF with reduced EF (HFrEF) have failed to detect any association between exercise filling pressures and aerobic capacity.(6–8)

Peak oxygen consumption (VO₂) as measured by cardiopulmonary exercise testing (CPET) is the gold standard assessment for aerobic capacity.(9–11) Reduced peak VO₂ in HFpEF is used as an endpoint for clinical trials,(12,13) and is known to be prognostic.(14–16) However, CPET is not currently incorporated into HFpEF diagnostic guidelines.(17,18) Reductions in peak VO₂ are often considered to reflect poor cardiac output reserve, but peripheral abnormalities also contribute in many patients with HFpEF.(19,20) Invasive CPET, which combines expired gas analysis with direct measures of hemodynamics, has emerged as the gold standard test to identify or exclude HFpEF in patients with unexplained dyspnea.(4,9,21,22)

Because little data is available relating filling pressures to aerobic capacity in people with HFpEF, and because no study has directly evaluated the potential utility of noninvasive CPET in diagnosis, we performed a detailed evaluation of both noninvasive and invasive CPET in a large, consecutive series of patients with and without HFpEF.

METHODS

All consecutive patients undergoing invasive hemodynamic exercise testing for the evaluation of unexplained dyspnea over a 15-year period from January 2000 to January 2014 at the Mayo Clinic in Rochester, MN were identified. From this cohort, patients who had underwent both transthoracic echocardiography and CPET within 1 year of the date of invasive CPET were included. All patients were evaluated by a cardiologist at Mayo Clinic before and after testing. The study was reviewed and approved by the Mayo Clinic Institutional Review Board.

HFpEF patients were defined by typical symptoms of HF (dyspnea), normal EF (≥50%), and elevated pulmonary capillary wedge pressure (PCWP) at rest or exercise (rest ≥15 mmHg, exercise ≥25 mmHg).(1–4) Patients with significant valvular heart disease (>mild stenosis, >moderate regurgitation), pulmonary arterial hypertension, constrictive pericarditis, high output failure, unstable coronary disease, primary cardiomyopathies, history of low EF (<50%), significant pulmonary disease and heart transplant were excluded.

Control subjects with non-cardiac causes of dyspnea (NCD) were required to display no evidence of cardiac etiology of dyspnea after exhaustive clinical evaluation, including a normal EF, normal pulmonary artery (PA) pressures (rest <25 mm Hg, exercise <40 mm Hg) and normal rest and exercise PCWP (criteria above).

Invasive CPET

Right heart catheterization was performed via the right internal jugular vein in the fasted state and supine position after minimal sedation as previously described to measure hemodynamics at rest and peak exercise.(3–5) Details of the invasive hemodynamic assessment are provided in the Online Supplement.

Noninvasive CPET

On a separate day from the invasive CPET, standard noninvasive upright treadmill or upright cycle ergometry exercise was performed using the same expired gas analysis technique as the invasive studies (MedGraphics, St. Paul, MN) to measure breath-by-breath VO₂ and CO₂ production (VCO₂) as well as respiratory exchange ratio (RER=VCO₂/VO₂), ventilatory efficiency [Minute ventilation (V_E)/VCO₂ nadir], and O₂ pulse (VO₂/HR).(9–11) Additional details are available in the Online Supplement.

Because peak VO₂ varies with age, sex, and muscle mass, absolute measured values were converted to relative values normalized to body weight (mL/kg/min); as well as percent predicted peak VO₂ using both the Wasserman-Hansen equation (current recommendation), (11) as well as the Fletcher et al nomogram.(23,24)

Statistical analysis

Data are reported as mean and standard deviation or median (25^th–75^th interquartile range). Chi square, Wilcoxon rank sum, or T test were used as appropriate to compare HFpEF and controls. Logistic regression was used to evaluate whether CPET variables could discriminate patients with HFpEF from NCD. Receiver Operating Curves (ROC) were constructed to evaluate the diagnostic performance of each test by the Youden index and c statistic. To apply clinically, noninvasive CPET variables were dichotomously classified as normal or abnormal based upon standard partition values identifying low and higher risk patients according to Lewis and colleagues (>20 and <14 ml/kg/min).(9) Linear regression was used to explore relationships between peak VO₂ at the time of CPET and invasive central hemodynamics in univariate analysis and in multivariable analysis after adjusting (a priori) for known determinants of peak VO₂ according the Fick Principle (stroke volume, heart rate, AVO₂ difference), as well as other previously established correlates including age and sex. Collinearity between exercise variables was assessed by Variance Inflation Factors (VIF), with values>5 indicating significant collinearity in the model.(25) Correction for multiple hypothesis testing was not performed. All tests were 2-sided, with a P value <0.05 considered significant. Analyses were performed using JMP 13.0.0 (SAS Institute, Cary, NC).

RESULTS

Patients with HFpEF (n=134) were older and heavier than NCD (n=72), with greater prevalence of comorbid conditions including diabetes, hypertension, kidney disease, anemia and atrial fibrillation (Table 1). As expected, HFpEF patients displayed more evidence of congestion, with higher NT-proBNP levels, E/e’, and estimated PA systolic pressure.

Table 1.

Baseline Characteristics

	Control (n=72)	HFpEF (n=134)	p value
Anthropometrics
Age, years	54±16	67±11	<0.0001
Female, n (%)	58	60	0.8
Body mass index, kg/m2	28.4±5.8	32.7±6.8	<0.0001
Comorbidities
Diabetes, %	10	25	0.007
Angiographic CAD, % (n=46/88)	28	31	0.8
Hypertension, %	84	96	0.007
Atrial fibrillation, %	6	20	0.004
Laboratories
Hemoglobin, g/dl	12.7±1.5	12.2±1.4	0.005
eGFR, ml/min	92±24	80±25	0.0009
NT proBNP, pg/ml	78 [29, 187]	435 [107, 1134]	<0.0001
Medications
ACE inhibitor/ARB, %	31	50	0.007
Beta blocker, %	35	65	<0.0001
Diuretic, %	32	55	0.002
Echocardiography
LVEDD, mm	48±5	49±6	0.5
EF, %	63±5	62±6	0.4
LAVI, ml/m2	30±12	40±15	<0.0001
E/e’ ratio	10±4	13±6	<0.0001
RVSP, mmHg	29±5	37±11	<0.0001

Values are expressed as mean ± standard deviation or number (percentage), CAD- Coronary Artery Disease, ACE-Angiotensin Converting Enzyme, ARB-Angiotensin Receptor Blocker, eGFR- estimated glomerular filtration rate, NT pro BNP- N terminal pro Brain Natriuretic Peptide, LVEDD- Left ventricular end diastolic dimension, LAVI- Left Atrial Volume Index, E- Early diastolic mitral inflow velocity, e’-early diastolic septal tissue Doppler velocity, EF- Ejection Fraction, RVSP-Right Ventricular Systolic Pressure

Invasive CPET

Right and left sided filling pressures and PA pressures were higher at rest and with peak exercise in the HFpEF group (Table 2). Peak VO₂ and peak workload performed at invasive CPET were reduced in HFpEF as compared to NCD. This was explained principally by lower CO reserve, as AVO₂ difference reserve was similar in HFpEF and NCD. All group differences remained significant after adjusting for age and BMI.

Table 2.

Invasive Hemodynamic and Cardiopulmonary Exercise Data

	Rest			Peak Exercise
	Control (n=72)	HFpEF (n=134)	p value	Control (n=72)	HFpEF (n=134)	p value
Vital Signs
HR	67±13	63±13	0.02	109±26	100±22	0.005
SBP	139±26	148±30	0.1	168±38	179±35	0.2
MBP	95±17	100±17	0.1	106±20	116±20	0.02
Central Hemodynamics
Right atrial pressure mmHg	5±2	9±3	<0.0001	7±5	18±6	<0.0001
PA mean pressure, mmHg	16±4	25±8	<0.0001	25±8	45±10	<0.0001
PASP, mmHg	28±7	39±11	<0.0001	41±13	63±14	<0.0001
PCWP, mmHg	9±4	16±6	<0.0001	14±5	31±6	<0.0001
PCWP/Watts, mmHg/W	-	-	-	0.2 ± 0.2	0.9 ± 0.4	<0.0001
O2 Delivery & Metabolism
Peak Watts achieved, W	-	-	-	67 ± 29	40 ± 18	<0.0001
O2 consumption, ml/min	222±60	225±61	0.8	979±350	812±270	0.002
O2 consumption, ml/kg/min	2.7±0.6	2.5±0.6	0.01	12.5±4.7	9.1±2.8	<0.0001
Cardiac output, L/min	5.4±1.6	4.9±1.4	0.04	10.4±3.2	8.4±3.0	<0.0001
Cardiac index, L/min/m2	2.80±0.72	2.49±0.67	0.003	5.42±1.54	4.21±1.40	<0.0001
Stroke volume index, ml/m2	43±10	41±12	0.3	50±16	43±17	0.02
AVDO2, ml/dl	4.2±0.8	4.7±0.9	0.0001	9.6±2.1	10.0±2.2	0.2

Values are mean ± SD, %. Adjustment of multiple hypothesis testing was not performed. HR-heart rate; SBP-systolic blood pressure; MBP-mean BP; PA-Pulmonary artery; PASP-PA systolic pressure; PCWP- pulmonary capillary wedge pressure; W-Watts; AVDO₂-arteriovenous O₂ content difference.

Noninvasive CPET

Noninvasive CPET was obtained a median of 3 days (IQR 1–20) prior to invasive CPET. As expected by the greater muscle mass recruited in the upright position, peak VO₂ attained during upright noninvasive CPET was ~60% higher than supine invasive CPET (16.6 vs. 10.2 ml/kg/min, p<0.0001). However, the two tests were well correlated with one another (r=0.64, p<0.0001, Figure 1).

Figure 1.

Peak O₂ consumption (VO₂) in HFpEF (red) and NCD (black) indexed to body weight [A] and after converting to percent predicted peak VO₂ by the Wasserman nomogram [B,C]. [D] Correlation between peak VO₂ measured during invasive hemodynamic testing in the supine and upright position.

As with invasive CPET, HFpEF patients displayed lower peak VO₂ compared to NCD during noninvasive upright CPET, with similar peak RER (Table 3, Figure 1). The VO₂ at VT was lower in HFpEF patients, but O₂ pulse and V_E/VCO₂ nadir were similar in cases and controls. Residual breathing reserve and peak arterial saturations were normal and similar between HFpEF and NCD at peak exercise, indicating that there was not a significant pulmonary limitation to exercise in either group.

Table 3.

Noninvasive Cardiopulmonary Exercise Data

	Control (n=72)	HFpEF (n=134)	p value
Vital Signs
Baseline HR, min−1	77±14	73±13	0.02
Peak HR, min−1	142±27	118±20	<0.0001
HR recovery, min−1	16±11	10±11	0.0003
Baseline systolic BP, mmHg	119±19	126±21	0.02
Peak systolic BP, mmHg	153±31	149±29	0.3
Peak Exercise capacity
Peak RER	1.14±0.11	1.14±0.13	0.7
Peak VO2, ml/min	1659±547	1334±479	<0.0001
Peak VO2, ml/kg/min	20.5±5.6	14.5±4.3	<0.0001
O2 sat, %	96±12	95±11	0.8
Breathing reserve, %	46±15	45±15	0.6
Percent Predicted peak VO2 by Wasserman formula
Peak VO2, % predicted	89±24	81±20	0.03
Peak VO2, <80% predicted, %	41	48	0.3
Peak VO2, <70% predicted, %	20	30	0.1
Peak VO2, <60% predicted, %	9	16	0.1
Peak VO2, <50% predicted, %	4	5	0.9
Percent Predicted peak VO2 by Fletcher formula
Peak VO2, % predicted	64±18	49±13	<0.0001
Peak VO2, <80% predicted, %	80	97	<0.0001
Peak VO2, <70% predicted, %	71	92	0.0001
Peak VO2, <60% predicted, %	44	78	<0.0001
Peak VO2, <50% predicted, %	23	54	<0.0001
Other Expired Gas Variables
Peak O2 pulse, ml/min/beat	12±3	11±3	0.4
VO2 at VT, ml/min	1205±383	985±361	<0.0001
VO2 at VT, ml/kg/min	14.9±4.0	10.8±3.3	<0.0001
VT percent of peak VO2, %	74±9	75±13	0.4
VE/VCO2 nadir	30±5	31±5	0.2

Values represent mean ± standard deviation. Adjustment of multiple hypothesis testing was not performed. HR- Heart Rate, FVC-Forced Vital Capacity, FEV1-Fraction of Forced Vital Capacity Expired in the first second, MVV-Maximum Voluntary Ventilation, VT-Tidal Volume, VE-Minute Ventilation, ETCO2-End Tidal Carbon Dioxide, O2-Oxygen, RR-Respiratory Rate, RER- Respiratory Exchange Ratio, VD/VT –Ratio of physiological dead space to tidal volume, RER- Respiratory Exchange Ratio. VO2 –Oxygen Consumption, VCO2 –Carbon Dioxide Production, BP-Blood Pressure, AT-Anaerobic Threshold

Relationships between Hemodynamics and Exercise Capacity

Peak VO₂ measured at invasive CPET and noninvasive CPET decreased with increasing exercise PCWP as measured during peak supine exercise and also varied inversely with exercise PA pressure and PCWP indexed to workload (Figure 2). These relationships were stronger for upright exercise as compared to supine exercise for PCWP and PA pressure (interaction p≤0.003). Peak VO₂ correlated directly with exercise cardiac output (Figure 2).

Figure 2.

Correlation between peak VO₂ in both upright (black) and supine (red) positions and exercise pulmonary capillary wedge pressure (PCWP), pulmonary artery (PA) pressure, cardiac output (CO) and PCWP indexed to work performed in Watts (W).

To determine whether PCWP was an independent predictor of peak VO₂ , we performed multivariable linear regression modeling including known determinants of peak VO₂: age, sex and the components based upon the Fick principle (HR, stroke volume, and AVO₂ difference). After adjusting for these covariates, PCWP remained an independent correlate of peak VO₂ in patients with HFpEF (p=0.0004), but not in patients with non-cardiac causes of dyspnea (p=0.5, Table 4).

Table 4.

Independent Correlates of Peak VO₂

Non-cardiac Dyspnea	Variance Inflation Factor	Standardized β estimate	p value
Age, years	2.46	−0.15	0.4
Female sex	1.22	−0.07	0.6
Peak stroke volume, ml	1.57	0.35	0.03
Peak heart rate, min−1	1.85	0.37	0.04
Peak AVDO2,ml/dl	1.48	0.31	0.05
Peak PCWP, mmHg	1.49	−0.10	0.5
HFpEF
Age, years	1.29	−0.20	0.04
Female sex	1.15	−0.19	0.04
Peak stroke volume, ml	1.70	0.26	0.02
Peak heart rate, min−1	1.42	0.27	0.007
Peak AVDO2,ml/dl	1.22	0.15	0.10
Peak PCWP, mmHg	1.01	−0.29	0.0006

One half (50%) of HFpEF patients displayed peak VO₂ in the range considered to reflect higher risk in HF (<14 ml/kg/min).(9) Compared to patients with peak VO₂≥14 ml/kg/min, this group displayed more severe hemodynamic derangements during exercise, with poorer cardiac output reserve and higher PCWP (Figure 3).

Figure 3.

As compared to HFpEF patients with peak VO₂≥14 ml/kg/min (black), patients with severely decreased peak VO₂ (<14 ml/kg/min, red) display [A] worse CO response relative to metabolic demand (ΔCO/ΔVO₂), [B] higher PCWP, [C] higher PCWP relative to cardiac output reserve, and [D] higher PCWP relative to ergometric work performed.

Diagnosis of HFpEF using Noninvasive CPET

In addition to older age, multiple CPET variables were predictive of the presence of HFpEF, including chronotropic incompetence, abnormal HR recovery, reduced VO₂ at ventilatory threshold, and low O₂ pulse (Table 5). However, none of these variables clearly discriminated cases and controls (all AUC<0.73). Peak VO₂<14 ml/kg/min displayed very high specificity for HFpEF (91%), but poor sensitivity (50%), while peak VO₂>20 ml/kg/min conversely had high sensitivity (90%) but poor specificity (49%). Peak VO₂<17 ml/kg/min displayed the highest Youden index to discriminate groups from our data, with an AUC of 0.78, indicating fair to good discrimination. V_E/VCO₂ nadir, which is roughly equivalent to V_E/VCO₂ slope, was not effective to discriminate HFpEF from NCD (Table 5).

Table 5.

Noninvasive Cardiopulmonary Exercise Test predictors of HFpEF

Univariate	OR [95% CI]	AUC	Sensitivity	Specificity	p value
Age>60 years	7.5 [4.0–14.4]	0.7256	80	35	<0.0001
Chronotropic incompetence	2.2 [1.2–4.0]	0.5935	69	49	0.009
HR recovery <6 beats/min	4.2 [1.9–10.8]	0.6081	31	90	0.001
VO2 at AT<9 ml/kg/min	6.1 [2.3–21.2]	0.6091	28	94	0.001
VO2 at AT<11 ml/kg/min	6.8 [3.3–15.1]	0.6957	54	85	<0.0001
O2 pulse < 0.14 ml/beat/kg	3.2 [1.8–6.0]	0.6371	73	54	0.0001
VE/VCO2>30	1.1 [0.6–2.0]	0.5112	48	50	0.8
VE/VCO2>36	1.7 [0.7–4.4]	0.5284	16	90	0.3
Peak VO2/kg
Peak VO2<14 ml/kg/min	10.7 [4.6–29.1]	0.7071	50	91	<0.0001
Peak VO2<20 ml/kg/min	8.1 [4.0–17.2]	0.6906	90	49	<0.0001
Peak VO2<17 ml/kg/min	12.4 [6.2–24.6]	0.7778	80	76	<0.0001
% Pred (Wasserman)
Peak VO2<50% predicted	1.1 [0.3–5.1]	0.5012	5	96	0.9
Peak VO2<60% predicted	2.0 [0.8–5.7]	0.5367	16	91	0.1
Peak VO2<70% predicted	1.7 [0.8–3.4]	0.5474	30	80	0.1
Peak VO2<80% predicted	1.4 [0.8–2.5]	0.5376	48	59	0.3
% Pred (Fletcher)
Peak VO2<50% predicted	3.9 [2.1–7.7]	0.6544	54	77	<0.0001
Peak VO2<60% predicted	4.6 [2.5–8.6]	0.6704	78	56	<0.0001
Peak VO2<70% predicted	4.5 [2.0–10.3]	0.6018	92	29	0.0002
Peak VO2<80% predicted	8.1 [2.8–29.7]	0.5851	97	20	<0.0001

OR-Odds Ratio, AUC-Area Under the Curve, HR-heart rate, AT-Anaerobic Threshold, VO2-Oxygen Consumption, V_E/VCO₂-ventilatory efficiency

Conversion of relative peak VO₂ (in ml/kg/min) to percent predicted peak VO₂ using the Wasserman-Hansen equation attenuated the separation between cases and controls, with a substantial worsening of overlap (Figure 1, Table 3). Indeed, 52% of patients with invasively-verified HFpEF displayed peak VO₂≥80% predicted using the Wasserman equation. In contrast, 97% of patients with invasively-verified HFpEF displayed peak VO₂<80% predicted according to the Fletcher scheme (Table 3). The Fletcher-based cutoffs for percent predicted peak VO₂ modestly discriminated cases from controls, though they were not superior to absolute peak VO₂ indexed to body weight (Table 5).

DISCUSSION

This study examined the invasive hemodynamic correlates of exercise capacity in HFpEF and the utility of non-invasive CPET as a potential test to discriminate HFpEF from non-cardiac causes of exertional dyspnea. There are two key findings that have broad-reaching implications. First, we observed that elevation in cardiac filling pressures during supine exercise is an independent correlate of reduced exercise capacity in patients with HFpEF, regardless of body position, and even after accounting for other known determinants related to oxygen delivery, diffusion, utilization and extraction according to the Fick principle. Second, we demonstrate that there is good discrimination of HFpEF from NCD using peak VO₂ values below and above currently suggested cutoffs (peak VO₂ values of <14 and >20 ml/min/kg),(9) but many patients lie within the intermediate gray zone between these boundaries where CPET may be less robust to distinguish HFpEF from non-cardiac etiologies of dyspnea. In these patients, more definitive evaluation using exercise hemodynamic testing is required to establish the diagnosis (Figure 4).

Figure 4.

Proposed diagnostic algorithm for Cardiopulmonary exercise testing (CPET) for unexplained dyspnea. When obvious causes have been excluded, measurement of peak VO₂ by CPET allows for discrimination of HFpEF from non-cardiac dyspnea at very low (<14 ml/kg/min) or relatively peak VO₂ (>20 ml/kg/min), respectively. However, given the substantial overlap in the intermediate range, additional testing, as with hemodynamic exercise testing, is required in patients with peak VO₂ 14–20 ml/kg/min.

Exercise Hemodynamics and Exercise Capacity

Elevation in LV filling pressures increases the hydrostatic pressure in the pulmonary capillaries, altering Starling forces to favor fluid movement out of the vascular space and into the interstitium, which can alter lung compliance and promote dyspnea.(26) Elevation in filling pressures during exercise in HFpEF is associated with increased risk of death, even when resting pressures are normal.(27) Conversely, reduction in filling pressures improves dyspnea as well as morbidity and mortality, at least in patients with HFrEF.(28–30) While it would seem self-evident that increases in PCWP during exercise should cause dyspnea and thus worsen exercise capacity, studies performed to date conducted in patients with HFrEF have failed to identify any correlation between cardiac filling pressures and peak VO₂.(6–8)

In contrast, we observed a significant inverse relationship between exercise PCWP and peak VO₂. Intracardiac pressures are lower in the upright position, leading some to question the relevance of hemodynamics measured in the supine position to activities of daily life, which are more likely to be performed in the upright position. In this regard, it is notable that we observed that the inverse relationships between elevated LV filling pressures during supine exercise and peak VO₂ were even more pronounced for upright exercise (Figure 2).

Importantly, these data show for the first time that elevated exercise PCWP is an independent correlate of reduced peak VO₂. According to the Fick principle, VO₂ is equal to the product of cardiac output and arterial venous O₂ content difference. Remarkably, exercise PCWP remained a significant correlate of peak VO₂ even after accounting for these determinants, as well as other known correlates including age, sex and body mass. These data emphasize the important role of elevated PCWP in the pathophysiology of exercise intolerance in HFpEF, and support ongoing research into pharmacologic and device-based interventions to reduce exercise PCWP in HFpEF.(13,31–36)

The reasons for the variance between the current data in patients with HFpEF and prior studies performed in patients with HFrEF are unclear.(6–8) While HFpEF patients are limited in part by cardiac output reserve,(1,5) this impairment may not be as profound as is observed in patients with HFrEF.(20,37) It may be that in HFrEF, exercise capacity is more constrained by poor output, whereas HFpEF patients may be more likely to cease exercising because of dyspnea caused by high filling pressures, or other sequelae of high PCWP, such as inadequate RV ejection in the setting of exercise-induced pulmonary hypertension.(5)

Noninvasive CPET in Diagnosis

Studies to date have uniformly reported that peak VO₂ is depressed in HFpEF.(1,4,5,19,20,38–46) However, surprisingly little data exists on the diagnostic utility of CPET for HFpEF, with most prior studies focusing on its prognostic role.(14–16) Current HF guidelines acknowledge the difficulty in establishing the diagnosis of HFpEF,(17) but do not directly address the diagnostic potential of CPET for HFpEF.(10,11,47) The 2016 HF guidelines from the ESC do give a grade IIa recommendation for CPET in the evaluation of HF among patients with unexplained dyspnea but there is no evidence supporting this recommendation (level of evidence C).(18)

The current data fills this knowledge gap, providing the first direct evidence on the ability of noninvasive CPET to distinguish HFpEF from NCD. Importantly, the diagnosis was determined in all participants using the gold standard of invasive CPET testing, allowing for the ability to definitively establish or refute the presence of HFpEF, rather than reaching a “diagnosis of exclusion”, which is often used in clinical practice,(17) or relying on echocardiography and natriuretic peptide testing, which are poorly sensitive.(3,4) The inclusion of a comparator group of older patients with non-cardiac dyspnea and a normal EF is a strength, since this reflects the sort of patients that are encountered in everyday practice.

Peak VO₂ relative to body weight emerged as the most useful discriminatory noninvasive CPET variable to diagnose HFpEF among these patients with unexplained dyspnea, with very high sensitivity and specificity at extreme values (>20 or <14 ml/kg/min). However, there was a large intermediate zone (14–20 ml/kg/min) with substantial overlap. This appears to be related to the fact that many patients with non-cardiac dyspnea display mild impairments in aerobic capacity beyond what would be seen in totally healthy volunteers. Because a large proportion of patients fall within this intermediate range, the current data suggest that CPET alone may not be best suited to serve as a key for noninvasive diagnosis of HFpEF, and invasive hemodynamic exercise testing, ideally with simultaneous CPET, should remain as the reference standard in the evaluation of possible HFpEF.

Converting Peak VO₂ to Percent Predicted

Conversion of measured peak VO₂ to percent predicted peak VO₂ is currently recommended when evaluating patients with unexplained dyspnea using CPET.(10,11) The Hansen-Wasserman formula,(23) derived using data from 77 male shipyard workers, is the recommended method to make this determination. We observed that percent predicted peak VO₂ using this method only obscured differences between HFpEF and NCD (Figure 1). Importantly, a large proportion of patients with invasively-proven HFpEF displayed percent predicted peak VO₂ that was minimally impaired according to the Wasserman formula: 52% of HFpEF patients displayed peak VO₂ ≥ 80% by the Wasserman equation, and 84% of HFpEF patients displayed peak VO₂ ≥ 60% by this method. These data raise serious questions with the use of this formula in the evaluation of patients with HFpEF.

A different nomogram from Fletcher et al.(24) performed better to discriminate the groups, but this was not superior to using measured peak VO₂ values adjusted for body weight, which is much easier to apply clinically and appears to discriminate HFpEF from NCD. Further study is required to identify the optimal metrics to normalize aerobic capacity for diagnostic purposes.

Limitations

There is selection bias in that patients were referred for an invasive procedure for unexplained dyspnea. However, invasive CPET testing is necessary to clearly establish or refute the presence of HFpEF, which is not possible with non-invasive studies, and this study would not have been possible without the gold standard assessment.(3,4) Patients with significant pulmonary disease were excluded, because this cohort is more readily identifiable using examination, imaging and pulmonary function testing, and we can therefore make no conclusions regarding the ability of CPET testing to discriminate this group from patients with isolated HFpEF or NCD. Functional class was not ascertained in this study, and the specific causes of non-cardiac dyspnea were rarely identified, since the focus was on identifying or excluding HFpEF specifically. Noninvasive CPET testing, echocardiogram and invasive CPET were not performed simultaneously, but since the median time between noninvasive CPET and invasive CPET was only 3 days, there is minimal risk of medication or other physiologic changes might have influence test results. Exercise was performed in different positions, but the observation that upright exercise performance was even more tightly correlated with hemodynamics measured in the supine position is a strength of this study, particularly since cardiac catheterization is most commonly performed in the supine position. CPET with or without stress imaging is also valuable to detect ischemic disease and other non-cardiac causes of dyspnea (Figure 4), and these were not evaluated in this study as patients with significant pulmonary disease were excluded. Furthermore, among patients with HFpEF, CPET can provide valuable information about pulmonary and chronotropic reserve, and may also be used to longitudinally follow clinical course or gauge the efficacy of therapeutic interventions.

Conclusion

Elevation in left ventricular filling pressure with exercise is independently correlated with depressed exercise capacity, supporting its central role in the pathophysiology of HFpEF. Very low or relatively preserved peak VO₂ measured noninvasively by CPET is useful to discriminate HFpEF from non-cardiac dyspnea, but people with exertional dyspnea and only mildly depressed peak VO₂ require additional testing to clarify the diagnosis, since non-cardiac etiologies of dyspnea may also reduce peak VO₂.

Clinical Perspectives.

Although exercise capacity (peak oxygen consumption, VO₂) is depressed in HFpEF, its hemodynamic determinants and role in diagnosis remains unclear. We demonstrate that 1) peak VO₂ correlates with invasive cardiac output and biventricular filling pressures regardless of body position during exercise, 2) peak PCWP is an independent correlate of peak VO₂ only in HFpEF, and 3) peak VO₂ <14 ml/kg/min or >20 ml/kg/min was useful to rule in and rule out HFpEF, respectively, whereas intermediate values and currently used complex nomograms were less discriminatory.

Translational Outlook.

Therapies targeting PCWP with exercise and its impact on exercise capacity require further study. Future research should seek to clarify role of additional testing to improve the ability of noninvasive CPET to discriminate HFpEF from non-cardiac causes of dyspnea.