Physiologic dead space and arterial carbon dioxide contributions to exercise ventilatory inefficiency in patients with reduced or preserved ejection heart failure

Abstract

Aims

Patients with heart failure (HF) with reduced (HFrEF) or preserved (HFpEF) ejection fraction demonstrate an increased ventilatory equivalent for carbon dioxide (V̇_E/V̇CO₂) slope. The physiologic correlates of V̇_E/V̇CO₂ slope remain unclear in the two HF phenotypes. We hypothesized that changes in physiologic dead space to tidal volume ratio (V_D/V_T) and arterial CO₂ tension (PaCO₂) differentially contribute to V̇_E/V̇CO₂ slope in HFrEF versus HFpEF.

Methods and Results

Adults with HFrEF (N=32) and HFpEF (N=27) (LVEF: 22±7 and 61±9%; BMI: 28±4 and 33±6 kg/m², P<0.01) performed cardiopulmonary exercise testing with breath-by-breath ventilation and gas-exchange measurements. PaCO₂ was measured via radial arterial catheterization. We calculated V̇_E/V̇CO₂ slope via linear regression, and V_D/V_T=1−[(863×V̇CO₂)/(V̇_E×PaCO₂)]. Resting V_D/V_T (0.48±0.08 vs. 0.41±0.11, P=0.04), but not PaCO₂ (38±5 vs. 40±3 mm Hg, P=0.21) differed in HFrEF versus HFpEF, respectively. Peak exercise V_D/V_T (0.39±0.08 vs. 0.32±0.12, P=0.02) and PaCO₂ (33±6 vs. 38±4 mm Hg, P<0.01) differed in HFrEF versus HFpEF, respectively. V̇_E/V̇CO₂ slope was higher in HFrEF compared to HFpEF (44±11 vs 35±8, P<0.01). Variance associated with V̇_E/V̇CO₂ slope in HFrEF and HFpEF was explained by peak exercise V_D/V_T (R²=0.30 vs. 0.50, respectively) and PaCO₂ (R²=0.64 vs. 0.28, respectively), but with different relative contributions from each (all P<0.01).

Conclusions

Relationships between V̇_E/V̇CO₂ slope with both V_D/V_T and PaCO₂ are robust, but differ in HFpEF compared to HFrEF. Increasing V̇_E/V̇CO₂ slope appears strongly explained by mechanisms influential to regulating PaCO₂ in HFrEF, which contrasts the strong role of increased V_D/V_T in HFpEF.

Introduction

A steep slope of the ventilatory equivalent for carbon dioxide (V̇_E/V̇CO₂) as quantified during cardiopulmonary exercise testing (CPET) demonstrates robust prognostic power in heart failure (HF) patients with reduced (HFrEF) or preserved (HFpEF) ejection fraction (1–4). Ventilatory inefficiency is suggested to accompany an excessive rate-mediated rise in V̇_E/V̇CO₂ slope, which is likely both a consequence and key contributor to exercise intolerance in HF (2–5). In this context, despite the broad clinical and research use of V̇_E/V̇CO₂ slope to phenotype the magnitude of exercise ventilatory inefficiency in HF, the understanding of this index has not been clearly defined in patients with HFrEF or HFpEF.

The V̇_E/V̇CO₂ slope is determined by dynamic coupling of gas exchange and hemodynamics reflecting pulmonary, cardiac, musculoskeletal, and nervous system responses to exercise (4, 6–8). ‘Ideal’ alveolar gas and air equations implicate the proportion of tidal volume distributed to physiologic dead space (V_D/V_T) accompanied by mechanisms underpinning regulation of arterial carbon dioxide tension (PaCO₂) as key contributors to V̇_E/V̇CO₂ slope (4, 6, 7, 9). While increased V̇_E/V̇CO₂ slope is likely to be the consequence of integrated factors including wasted alveolar ventilation (V̇_A), atelectasis, and/or impaired neural ventilatory control, there is controversy as to which event is the principal mediator in HFrEF (4, 7, 9, 10), with even less information available on this topic in HFpEF. Further complicating this discussion is the evolving knowledge of what unique characteristics unequivocally define the HFpEF phenotype (2, 3, 5, 11, 12).

Guazzi et al. (4) demonstrated that high V̇_E/V̇CO₂ slope from rest to peak exercise is predictive of survival in HFrEF. Similar to data from Woods et al. (7) during submaximal exercise, Guazzi et al. (4) attributed increased V̇_E/V̇CO₂ slope to influences of impaired mechanisms of ventilatory control (e.g., low PaCO₂ regulatory set-points) accompanied by contributions from elevated V_D/V_T. By contrast, without observations from specific tests of ventilatory control, Sullivan et al. (9) and others (10) suggest that ventilatory control related to regulation of PaCO₂ is intact in HFrEF, whereas abnormal physiologic hemodynamic factors are more likely present in this population.

As a necessary hypothesis-generating step towards advancing the understanding and interpretation of V̇_E/V̇CO₂ slope in HF, this study aimed to demonstrate how V̇_E/V̇CO₂ slope is related to changes in V_D/V_T and PaCO₂ during CPET in patients with HFrEF or HFpEF. We hypothesized that increased V_D/V_T accompanied by decreased PaCO₂ would differentially explain the rise in V̇_E/V̇CO₂ slope in HFrEF compared to HFpEF.

Methods

Participants

Patients with HFrEF or HFpEF (N=59) were referred by their primary cardiologist for CPET as part of a comprehensive HF evaluation at Mayo Clinic, Rochester, MN. Using the European Society of Cardiology criteria, we studied 32 HFrEF (left ventricular ejection fraction [LVEF] ≤40%) and 27 HFpEF (LVEF >50%) (Characteristics, Table 1) (13).

Table 1.

Participant characteristics

	All	HFrEF	HFpEF	P-value
Variable	N = 59	N = 32	N = 27
Age (years)	62 ± 13	55 ± 10	71 ± 11	<0.001
Sex (male/female)	46/13	30/2	16/11	<0.001
LV ejection fraction (%)	40 ± 21	22 ± 7	61 ± 9	<0.001
NYHA class, n (%)
II	20 (34)	13 (41)	7 (26)	0.07
III	39 (66)	19 (59)	20 (74)	0.19
Height (cm)	172 ± 9	174 ± 8	170 ± 10	0.14
Weight (kg)	92 ± 20	86 ± 15	96 ± 23	0.05
Body mass index (kg/m2)	30 ± 6	28 ± 4	33 ± 6	<0.01
Body surface area (m2)	2.07 ± 0.26	2.03 ± 0.21	2.12 ± 0.31	0.19
Peak V̇O2 (% of pred.)	39 ± 14	31 ± 7	50 ± 14	<0.001
Peak V̇O2 (mL/kg/min)	8.6 ± 2.3	8.3 ± 2.2	8.8 ± 2.4	0.46
Weber-Janicki class [peak V̇O2], n (%)
C, 10–16 mL/kg/min	14 (24)	7 (22)	7 (26)	0.88
D, <10 mL/kg/min	45 (76)	25 (78)	20 (74)	0.75
Hemoglobin (g/dL)	13.3 ± 1.7	13.9 ± 1.6	12.5 ± 1.4	0.01
Creatinine (mg/dL)	1.33 ± 0.42	1.39 ± 0.40	1.26 ± 0.43	0.23
eGFR (mL/min per 1.73 m2)	56 ± 15	56 ± 15	55 ± 16	0.80
Drug Therapy, n (%)
ACE inhibitor	37 (63)	25 (78)	12 (44)	<0.001
ARBs	9 (15)	2 (6)	7 (26)	<0.001
Antiarrhythmic	15 (25)	10 (31)	5 (19)	0.09
β-blocker (β1 or non-select.)	45 (76)	28 (88)	17 (63)	0.04
Ca2+ channel blocker	7 (12)	1 (3)	6 (22)	<0.001
Digoxin	25 (42)	23 (72)	2 (7)	<0.001
Nitrate (oral, SL, or topical)	19 (32)	9 (28)	10 (33)	0.52
Aspirin	43 (73)	22 (69)	21 (78)	0.46
Diuretics	40 (68)	29 (91)	11 (41)	<0.001
Echocardiography
LA volume (mL)	104 ± 43	115 ± 42	84 ± 40	0.03
LA volume index (mL/m2)	51 ± 20	56 ± 20	43 ± 20	0.04
Mitral E-wave VEL (cm/s)	89.3 ± 30.1	89.3 ± 40.1	89.3 ± 22.2	0.58
Mitral A-wave VEL (cm/s)	68.9 ± 29.7	59.0 ± 32.8	77.5 ± 24.2	0.01
Mitral E/A ratio	1.4 ± 0.8	1.7 ± 0.9	1.2 ± 0.6	0.11
Mitral septal tissue Doppler VEL (e′) (cm/s)	5.3 ± 1.8	4.5 ± 1.7	6.1 ± 1.7	<0.001
Mitral E/e′ ratio	19.1 ± 9.8	22.1 ± 10.9	16.2 ± 7.7	0.02
IV septum thickness (mm)	9.9 ± 1.6	9.5 ± 1.6	10.2 ± 1.6	0.06
Posterior wall thickness (mm)	9.8 ± 1.7	9.7 ± 1.5	9.9 ± 1.9	0.96

Data are mean ± SD, n, or percentage (%).

Patients were excluded based on the following criteria: significant coronary artery disease (stenosis ≥50%), cor pulmonale, diagnosed pulmonary disease secondary to HF, primary renal or hepatic disease, valvular heart disease (any stenosis, >mild regurgitation, etc.), hypertrophic or infiltrative cardiomyopathy, constrictive pericarditis, or deep vein thrombosis. All aspects of this study were approved by the Mayo Clinic Institutional Review Board while conforming to principles outlined in the Declaration of Helsinki. All authors have read and agreed to the manuscript as written.

Echocardiography

Resting two-dimensional and tissue Doppler echocardiography according to guidelines of the American Society of Echocardiography were used to assess LVEF, morphology, and function; as well as early transmitral flow velocity (E), late transmitral flow velocity (A), E/A ratio, early diastolic mitral annular velocity (e′), peak E to e′ ratio, and left atrial volumes (Table 1) (13).

CPET protocol

Patients performed CPET in the morning in a fasted state while remaining on standard pharmacologic therapy. Patients exercised in a semi-recumbent position at an initial workload of 20-watts (W) while pedaling at 60–65 rpm, thereafter increasing by 10-W every 3 min until volitional fatigue. Heart rate and rhythm were continuously monitored using a 12-lead electrocardiogram.

Ventilation and gas exchange

Breath-by-breath open-circuit spirometry (CPX/D, MGC Diagnostics, St.Paul, MN) was used to continuously collect ventilation and gas exchange data throughout CPET. The final 30-s of rest and the final completed stage (i.e., peak pulmonary oxygen uptake, V̇O_2peak) were used for data analyses. Data were acquired for variables: V̇O₂, V̇CO₂, respiratory exchange ratio (RER), respiratory rate (f_B), V_T, V̇_E, end-tidal CO₂ tension (P_ETCO₂), and end-tidal O₂ tension (P_ETO₂). Percent (%) of predicted V̇O_2peak was calculated from Hansen et al. (14).

Arterial gas measurement and parameters

Continuous monitoring and periodic sampling of arterial blood gases occurred using standard technique via percutaneous insertion of a 20-gauge catheter at the left radial artery. Arterialized blood drawn into heparinized 3-mL glass syringes occurred so as to approximately align with breath-by-breath periods (i.e., final ~30-s of each CPET stage), with the final draw of the last completed stage used to describe peak exercise. Arterialized samples were chilled on ice and immediately sent to our institutional core laboratory where standard processing techniques occurred for: PaCO₂, oxygen tension (PaO₂), pH (pHa), and oxygen saturation (SaO₂). From these measurements, we used standard physiologic and blood biochemical equations (described in Appendix) to derive bicarbonate (HCO₃⁻) and base excess (BE) of arterialized blood.

Derived ventilation and gas exchange parameters

By using arterialized gas measurements combined with acquired basic ventilatory and gas exchange responses, ‘ideal’ alveolar gas and air equations with associated parameters were used to calculate: V̇_A, alveolar volume (V_A), alveolar CO₂ tension (PACO₂), alveolar O₂ tension (PAO₂), alveolar-to-arterial O₂ difference (PA-aO₂), indirect simplified estimate of whole lung diffusing capacity for O₂ (DLO₂) via Fick’s law of diffusion, and V_D ventilation (V̇_D) (described in Appendix), whereas V_D/V_T equaled (4, 6–10),

$$\frac{{\mathrm{V}}_{\mathrm{D}}}{{\mathrm{V}}_{\mathrm{T}}}=(1-\frac{863\times {\stackrel{.}{\mathrm{V}}\text{CO}}_2}{{\stackrel{.}{\mathrm{V}}}_{\mathrm{E}}\times {\text{PaCO}}_2})$$

where 863 converts gas concentration to partial pressure while correcting for the fact that V̇CO₂ is expressed as a dry gas volume at standard temperature (0°C), pressure (760 mm Hg), and dry (0% water vapor) (STPD), whereas ventilation is expressed as a wet gas volume at body temperature, pressure, and saturation (BTPS).

Mixed-expired CO₂ (PECO₂) equaled the quotient of 863 and V̇_E/V̇CO₂ (15). All data from rest to peak exercise was used to calculate V̇_E/V̇CO₂ slope via linear regression, which provided optimal data interpretability and temporal consistency across participants (16).

Statistical Analyses

Parametric data are presented as mean±SD. Results of Shapiro-Wilk and Levene’s tests did not suggest data were significantly skewed or variance was heterogeneous. Between group baseline differences were compared using independent Student’s t-tests for parametric data or χ²-tests for categorical variables. Repeated measures ANOVA models with group-by-time interaction terms were used to test between-within group differences at rest and peak exercise. When F-tests from group-by-time interaction terms were significant, post-hoc Tukey-Kramer tests were used to assess between-within pairwise differences.

Ordinary least squares univariate linear regressions were used to test relationships between dependent (ordinate, V̇_E/V̇CO₂ slope) and independent (abscissa, V_D/V_T, PaCO₂, etc.) variables. Coefficient of determination (R²) and regression equations, y_est= a+bx, were computed from these models. Interpretation of R² were based on standards of Cohen (17): modest=0.02; moderate=0.15; or strong≥0.25. Two-tailed significance was determined using an alpha level set at 0.05. Statistical analyses were performed using SAS statistical software v.9.4 (SAS Institute, Cary, NC).

Results

Participant characteristics

Both HFrEF and HFpEF demonstrated age, sex distribution, and LVEF consistent with currently recognized clinical phenotypes (Table 1). Despite greater %predicted V̇O_2peak in HFpEF versus HFrEF, there were no differences in V̇O_2peak (mL/kg/min). New York Heart Association and Weber-Janicki functional classifications aligned in HFpEF, but to a lesser extent in HFrEF. Compared to HFpEF, Hb was higher in HFrEF whereas creatinine and eGFR were similar. Frequency of therapies including ACE inhibitors, β-blockers (HFrEF, 6 and 22 on non-selective and β₁-selective, respectively; HFpEF, 17/17 on β₁-selective), digoxin, and diuretics were higher in HFrEF versus HFpEF. Resting echocardiography indicated there were increased LV filling pressures in HFrEF and HFpEF. Participants completed all aspects of this study in the absence of demonstrating adverse events (e.g., no signs/symptoms of pulmonary embolism, which could contribute to impaired ventilatory function).

Workload and basic ventilation and gas exchange

Resting V̇O₂, V̇CO₂, RER, V̇_E, RR, V_T, P_ETO₂, and PECO₂/P_ETCO₂ were similar between groups, whereas P_ETCO₂ and PECO₂ were lower in HFrEF versus HFpEF (Table 2). Despite no differences in peak W and RER, HFrEF demonstrated lower V̇O₂ and V̇CO₂ versus HFpEF, whereas differences in V̇_E/V̇CO₂, P_ETCO₂, and PECO₂ persisted from rest. Except for V̇_E/V̇CO₂ (↓HFpEF), P_ETCO₂ (↓HFrEF), and PECO₂ (↑HFpEF), changes in other metrics from rest to peak exercise within HFrEF and HFpEF were similar (Table 2).

Table 2.

Basic ventilatory and gas exchange responses at rest and peak exercise

	All	HFrEF	HFpEF	P-value
Rest	N = 59	N = 32	N = 27
V̇O2 (L/min)	0.24 ± 0.06	0.23 ± 0.05	0.25 ± 0.07	0.97
V̇CO2 (L/min)	0.21 ± 0.06	0.20 ± 0.05	0.21 ± 0.07	0.99
RER	0.87 ± 0.10	0.88 ± 0.10	0.85 ± 0.09	0.63
V̇E (L/min)	8.6 ± 2.7	9.0 ± 2.2	8.1 ± 3.3	0.95
fB (breaths/min)	17 ± 4	17 ± 4	16 ± 5	0.91
VT (L)	0.54 ± 0.18	0.54 ± 0.16	0.53 ± 0.20	0.99
V̇E/V̇CO2	42 ± 8	45 ± 7	38 ± 6	<0.001
PETO2 (mm Hg)	103 ± 6	103 ± 6	102 ± 5	0.98
PETCO2 (mm Hg)	35 ± 6	33 ± 5	39 ± 5	<0.001
PECO2 (mm Hg)	21 ± 4	20 ± 3	23 ± 4	<0.01
PECO2/PETCO2	0.60 ± 0.06	0.61 ± 0.06	0.60 ± 0.06	0.98
Peak exercise
Workload, W	38 ± 11	39 ± 11	36 ± 12	0.39
V̇O2 (L/min)	0.76 ± 0.22*	0.70 ± 0.18*	0.82 ± 0.25*	0.04
V̇CO2 (L/min)	0.82 ± 0.23*	0.77 ± 0.19*	0.89 ± 0.25*	0.04
RER	1.10 ± 0.08*	1.10 ± 0.09*	1.10 ± 0.06*	0.92
V̇E (L/min)	32 ± 9*	33 ± 9*	35 ± 9*	0.43
fB (breaths/min)	31 ± 8*	30 ± 7*	31 ± 10*	0.97
VT (L)	1.08 ± 0.33*	1.13 ± 0.35*	1.03 ± 0.31*	0.43
V̇E/V̇CO2	40 ± 10	44 ± 10	35 ± 7*	<0.001
PETO2 (mm Hg)	112 ± 7*	113 ± 7*	111 ± 6*	0.62
PETCO2 (mm Hg)	34 ± 8	30 ± 6*	39 ± 7	<0.001
PECO2 (mm Hg)	23 ± 5	20 ± 4	25 ± 5*	<0.01
PECO2/PETCO2	0.68 ± 0.05*	0.69 ± 0.04*	0.66 ± 0.06*	0.14

Data are mean ± SD. Oxygen consumption (V̇O₂); carbon dioxide output (V̇CO₂); respiratory exchange ratio (RER); minute ventilation (V̇_E); respiratory rate (f_B); tidal volume (V_T); end-tidal oxygen (P_ETO₂); end-tidal carbon dioxide (P_ETCO₂); mixed expired carbon dioxide (PECO₂).

^*Within group, rest vs. peak exercise, P<0.05. Table P-values and symbols represent significance after Tukey-Kramer post-hoc testing from repeated measures ANOVA models.

Alveolar equation and blood biochemistry parameters

Excluding lower resting V_D/V_T in HFpEF versus HFrEF (Figure 1), parameters associated with alveolar equations or blood biochemistry were similar between groups (Table 3). By contrast, consistent with higher V̇_E/V̇CO₂ slope in HFrEF (Figure 1A), HFrEF demonstrated higher V_D, V̇_D, and V_D/V_T (Figure 1B) as well as lower PaCO₂ (Figure 1C) and PACO₂ accompanied by higher pHa versus HFpEF at peak exercise (Table 3). Absence of between group differences for other metrics at rest persisted to peak exercise. For within group differences from rest to peak exercise, HFrEF and HFpEF demonstrated similar directional changes for variables in Table 3 except for PaO₂, which increased within HFrEF but not within HFpEF.

Figure 1.

Measurements or univariate linear regressions involving the ventilatory equivalent for carbon dioxide (V̇_E/V̇CO₂) slope, physiologic dead space to tidal volume ratio (V_D/V_T), or arterial carbon dioxide tension (PaCO₂). Data presented in panels A to C are interquartile range with mean represented as (+). Variables presented on the abscissa in panels D to F are at rest, peak exercise, and peak exercise, respectively. For panels D to F representative of all participants: filled circle is mean ± SD of V̇_E/V̇CO₂ slope at the mean of the variable set on the abscissa, solid line is goodness of fit line of the model fit equation, dotted lines are 95% prediction bands of the model fit equation, and grey bands are isopleths representing linked changes in observed V̇_E/V̇CO₂ slope and PaCO₂ or V_D/V_T when either PaCO₂ or V_D/V_T are theoretically constrained values. Interpretation of R²: modest=0.02; moderate=0.15; or strong≥0.25. The Y-intercept in panels D to F differed from 0.0 (P<0.01). Differences in R²: panel D vs. E or F, P<0.001; panel E vs. F, P=0.16. Heart failure with reduced (HFrEF, N=32) or preserved (HFpEF, N=27) ejection fraction. *Following post-hoc Tukey-Kramer testing, different within group for both HFrEF and HFpEF or all HF, rest to peak exercise, P<0.05.

Table 3.

Alveolar equation, related derivatives, and blood biochemistry parameters

	All	HFrEF	HFpEF	P-value
Rest	N = 59	N = 32	N = 27
VD (L)	0.24 ± 0.10	0.26 ± 0.08	0.22 ± 0.13	0.64
V̇D (L/min)	4.0 ± 1.7	4.3 ± 1.3	3.5 ± 2.2	0.79
VA (L)	0.29 ± 0.11	0.28 ± 0.11	0.30 ± 0.11	0.98
V̇A (L/min)	4.64 ± 1.43	4.67 ± 1.32	4.61 ± 1.57	0.99
PaO2 (mm Hg)	71 ± 12	71 ± 14	72 ± 11	0.99
PAO2 (mm Hg)	96 ± 7	98 ± 8	94 ± 6	0.11
PA-aO2 (mm Hg)	25 ± 12	26 ± 13	22 ± 11	0.54
PA-aO2/V̇O2 (mm Hg/mL/kg/min)	9.6 ± 5.0	10.5 ± 5.7	8.5 ± 3.8	0.41
DLO2 (mL/min/mm Hg)	12 ± 7	12 ± 8	13 ± 7	0.98
PACO2 (mm Hg)	39 ± 4	38 ± 5	40 ± 4	0.45
SaO2 (%)	94 ± 3	94 ± 3	95 ± 3	0.88
pHa	7.41 ± 0.04	7.42 ± 0.04	7.39 ± 0.05	0.19
HCO3− (mEq/L)	24.33 ± 2.52	24.23 ± 2.78	24.46 ± 2.22	0.99
Base excess (mEq/L)	−0.50 ± 2.60	−0.44 ± 2.71	−0.56 ± 2.50	0.98
Peak exercise
VD (L)	0.38 ± 0.15*	0.43 ± 0.15*	0.31 ± 0.12*	<0.01
V̇D (L/min)	11.4 ± 4.8*	12.9 ± 4.3*	9.7 ± 4.9*	<0.01
VA (L)	0.71 ± 0.28*	0.70 ± 0.26*	0.72 ± 0.30*	0.98
V̇A (L/min)	21 ± 7*	20 ± 6*	21 ± 7*	0.97
PaO2 (mm Hg)	76 ± 15	78 ± 14*	74 ± 15	0.59
PAO2 (mm Hg)	109 ± 6*	111 ± 7*	107 ± 5*	0.10
PA-aO2 (mm Hg)	33 ± 14*	33 ± 14*	33 ± 14*	0.99
PA-aO2/V̇O2 (mm Hg/mL/kg/min)	4.3 ± 2.7*	4.3 ± 2.6*	4.3 ± 2.9*	0.99
DLO2 (mL/min/mm Hg)	29 ± 21*	27 ± 18*	31 ± 25*	0.81
PACO2 (mm Hg)	35 ± 6*	34 ± 6*	37 ± 5*	0.03
SaO2 (%)	95 ± 4	95 ± 3	94 ± 5	0.63
pHa	7.40 ± 0.05	7.42 ± 0.04	7.38 ± 0.05	0.02
HCO3− (mEq/L)	21.62 ± 3.23*	21.19 ± 3.75*	22.13 ± 2.47*	0.61
Base excess (mEq/L)	−3.04 ± 3.21*	−3.21 ± 3.62*	−2.83 ± 2.70*	0.96

Data are mean ± SD. Physiologic dead space (V_D); V_D ventilation (V̇_D); alveolar volume (V_A); alveolar ventilation (V̇_A); arterial oxygen tension (PaO₂); alveolar oxygen tension (PAO₂); alveolar-to-arterial O₂ gradient (PA-aO₂); pulmonary oxygen uptake (V̇O₂); estimated pulmonary diffusing capacity for oxygen (DLO₂); alveolar carbon dioxide tension (PACO₂); arterial oxygen saturation (SaO₂); arterial pH (pHa); bicarbonate (HCO₃⁻);.

^*Within group, rest vs. peak exercise, P<0.05. Table P-values and symbols represent significance after Tukey-Kramer post-hoc testing from repeated measures ANOVA models.

In sub-analyses including only HFrEF (N=22) and HFpEF (N=17) on β₁-blockers, between-within differences for V̇_E/V̇CO₂ slope (F=13.7, P<0.01), V_D/V_T (F=20.7, P<0.01), and PaCO₂ (F=5.12, P<0.01) mirrored outcomes illustrated in Figure 1(A–C). Within HFpEF, there were no differences in V̇_E/V̇CO₂ slope (P=0.96), peak exercise V_D/V_T (P=0.69), and peak exercise PaCO₂ (P=0.90) between patients on versus not on β₁-blockers.

Predictors of V̇_E/V̇CO₂ slope

Resting LVEF was unrelated to V̇_E/V̇CO₂ slope when stratified by HF group (Figure 1D), whereas combining all HF resulted in a modest relationship (y=−0.21x+48, R²=0.16, P=0.01). In comparison, peak exercise V_D/V_T and PaCO₂ strongly explained the variance in V̇_E/V̇CO₂ slope across all HF (Figure 1E and 1F).

Consistent with plots of HFrEF versus HFpEF in Figures 1E and 1F, unique contributions from each group to pooled regressions are illustrated in detail in Figure 2. While decreasing PaCO₂ strongly explained the variance associated with increasing V̇_E/V̇CO₂ slope in HFrEF (Figure 2B), V_D/V_T was more strongly related to elevated V̇_E/V̇CO₂ slope in HFpEF (Figure 2C). Although there was significant linearity between increasing V_D/V_T and V̇_E/V̇CO₂ slope in HFrEF (Figure 2A) as well as PaCO₂ and V̇_E/V̇CO₂ slope (Figure 2D, inverse relationship) in HFpEF, variance explained in Figures 2A and 2D was approximately half of Figures 2B and 2C in HFrEF and HFpEF, respectively. In the absence of V_D/V_T (Y-intercept not different from 0.0, P=0.14; Figure 2A), and not until V_D/V_T >0.2, was V̇_E/V̇CO₂ slope elevated except for the influence of PaCO₂ in HFrEF (Figure 2B). In contrast, V_D/V_T accompanied by lesser effects from PaCO₂ contributed to increased V̇_E/V̇CO₂ slope at V_D/V_T ≥1.0 in HFpEF (Figure 2C; and Figure 2D isopleths).

Figure 2.

Univariate linear regressions involving the ventilatory equivalent for carbon dioxide (V̇_E/V̇CO₂) slope (dependent, ordinate), peak exercise physiologic dead space to tidal volume ratio (V_D/V_T) (independent, abscissa), or peak exercise arterial CO₂ tension (PaCO₂) (independent, abscissa). For all panels: filled circle is mean ± SD of V̇_E/V̇CO₂ slope at the mean of the variable set on the abscissa, solid line is goodness of fit line of the model fit equation, dotted lines are 95% prediction bands of the model fit equation, and grey bands are isopleths representing linked changes in observed V̇_E/V̇CO₂ slope and PaCO₂ or V_D/V_T when either PaCO₂ or V_D/V_T are theoretically constrained values. Interpretation of R²: modest=0.02; moderate=0.15; or strong≥0.25. The Y-intercept in panel A did not differ from 0.0 (P=0.14), whereas the Y-intercept in panels B to D differed from 0.0 (P<0.01). Differences in R²: panel A vs. B, P=0.06; panel A vs. C, P=0.33; panel B vs. D, P=0.06; panel C vs. D, P=0.30. Heart failure with reduced (HFrEF, N=32) or preserved (HFpEF, N=27) ejection fraction.

Finally, while increasing V̇_D and PAO₂ accompanied by decreasing PACO₂ demonstrated significant linearity with increasing V̇_E/V̇CO₂ slope in HFrEF and HFpEF, in an inverse manner, HCO₃⁻ and BE explained the variance in V̇_E/V̇CO₂ slope in HFrEF, whereas this did not occur in HFpEF (Table 4).

Table 4.

Univariate predictors of V̇_E/V̇CO₂ slope

	b (95% CL)	β (95% CL)	Intercept (95% CL)	R2 (95% CL)	P-value
	HFrEF
V̇D	33 (8, 59)	0.44 (0.11, 0.68)	29 (18, 41)	0.20 (−0.03, 0.43)	0.01
V̇A	0.26 (−0.40, 0.92)	0.14 (−0.22, 0.47)	38 (24, 52)	0.02 (−0.07, 0.11)	0.43
PaO2	0.11 (−0.18, 0.40)	0.14 (−0.22, 0.47)	35 (12, 58)	0.02 (−0.07, 0.11)	0.45
PAO2	1.29 (0.89, 1.69)	0.77 (0.58, 0.88)	−100 (−144, −55)	0.59 (0.39, 0.79)	<0.001
PA-aO2	0.19 (−0.10, 0.49)	0.24 (−0.12, 0.54)	37 (26, 48)	0.06 (−0.09, 0.21)	0.19
DLO2	1.41 (−0.16, 2.99)	0.32 (−0.03, 0.60)	37 (29, 45)	0.10 (−0.08, 0.28)	0.08
PACO2	−1.37 (−1.87, −0.87)	−0.72 (−0.85, −0.50)	90 (73, 107)	0.51 (0.28, 0.74)	<0.001
HCO3−	−1.70 (−2.64, −0.76)	−0.56 (−0.76, −0.26)	79 (59, 100)	0.31 (0.06, 0.56)	<0.01
Base excess	−1.44 (−2.48, −0.40)	−0.46 (−0.70, −0.13)	39 (34, 44)	0.21 (−0.02, 0.44)	<0.01
HFpEF
V̇D	31 (8, 53)	0.49 (0.14, 0.73)	25 (17, 32)	0.24 (−0.02, 0.50)	<0.01
V̇A	0.00 (−0.42, 0.44)	0.01 (−0.37, 0.39)	34 (25, 44)	0.00 (0.0, 0.0)	0.98
PaO2	−0.03 (−0.24, 0.18)	−0.06 (−0.43, 0.33)	36 (21, 52)	0.00 (−0.04, 0.04)	0.77
PAO2	0.92 (0.38, 1.47)	0.57 (0.24, 0.78)	−65 (−124, −7)	0.33 (0.06, 0.60)	<0.01
PA-aO2	0.13 (−0.08, 0.35)	0.25 (−0.14, 0.58)	30 (22, 37)	0.06 (−0.10, 0.22)	0.21
DLO2	0.95 (−0.09, 1.98)	0.35 (−0.03, 0.64)	30 (25, 35)	0.13 (−0.09, 0.35)	0.07
PACO2	−0.80 (−1.40, −0.20)	−0.48 (−0.73, −0.12)	64 (41, 86)	0.23 (−0.03, 0.49)	0.01
HCO3−	−0.40 (−1.66, 0.87)	−0.13 (−0.49, 0.26)	43 (15, 71)	0.02 (−0.08, 0.12)	0.53
Base excess	−0.04 (−1.21, 1.12)	−0.02 (−0.40, 0.36)	34 (30, 39)	0.00 (−0.01, 0.01)	0.94

Data are model parameters from univariate linear regressions predicting V̇_E/V̇CO₂ slope from rest to peak exercise in heart failure with reduced (HFrEF, N=32) or preserved (HFpEF, N=27) ejection fraction. Standardized beta (β). See definition of variables in Table 3 caption.

Discussion

These data suggest increased V̇_E/V̇CO₂ slope in both HFrEF and HFpEF patients is fundamentally linked to elevated V_D/V_T accompanied by altered regulation of PaCO₂ during CPET. Both HFrEF and HFpEF demonstrated V̇_E/V̇CO₂ slope at levels consistent with worsened prognosis (1–4). However, moderate-to-strong relationships between increasing V_D/V_T and decreasing PaCO₂ with rising V̇_E/V̇CO₂ slope differed in HFrEF compared to HFpEF. Decreasing PaCO₂ at peak exercise in HFrEF explained >2× the variance in V̇_E/V̇CO₂ slope compared to the same relationship tested in HFpEF. By contrast, despite demonstrating lower peak exercise V_D/V_T compared to HFrEF, variance explained between increasing V_D/V_T and V̇_E/V̇CO₂ slope in HFpEF was >1.5× that of HFrEF. These data are consistent with the hypothesis that increased V̇_E/V̇CO₂ slope is not exclusively a function of central hemodynamic limitations in patients with HFrEF or HFpEF (4, 5, 18–20).

While abnormal V_D/V_T and PaCO₂ contribute to increased V̇_E/V̇CO₂ (slope/ratio) in HFrEF (4, 7, 9, 10), these relationships have not been tested in HFpEF. These data in HFrEF are consistent with work from Guazzi et al. (4) and Woods et al. (7) who during peak and submaximal CPET, demonstrated low PaCO₂ and high V_D/V_T paralleled increased V̇_E/V̇CO₂ in HFrEF. By comparison, these observations are in contrast to interpretations of others (9, 10) suggesting PaCO₂ coupled to ventilatory control mechanisms are not altered in HFrEF; instead, suggesting increased V̇_E/V̇CO₂ slope is related to high V_D/V_T provoked by impaired pulmonary perfusion and, hence, V̇_A/Q̇ >1.0. Although this study did not include specific tests of ventilatory control related to chemoreflexes and/or group III/IV muscle afferents, these data align with several related studies in HF suggesting neurophysiologic control pathways abnormally activated during increased energy demand may play an appreciable role in excessive ventilatory responses to exercise in HF (4, 7, 18–21).

Energy demand in-excess of oxidative capabilities leads to a marked rise in CO₂ production followed by an acidotic blood milieu if accumulated CO₂ blood content is not adequately compensated (6, 8). In healthy adults, metabolic-driven production of CO₂ leads to intrinsic activation of pathways relating to mechanisms of HCO₃⁻ formation accompanied by increased ventilation. Timely coordination of these events during exercise functions well as a buffering mechanism (Hb binds H⁺) to help maintain blood acid-base balance while facilitating decreased blood CO₂ content as Na⁺ binds HCO₃⁻ (and to a lesser extent Hb binds CO₂) to be transported to the pulmonary system for expiration. However, it remains unclear how these interconnected compensatory mechanisms of breathing and acid-base buffering equilibrate in conditions where both the rate and absolute level of CO₂ accumulation in blood may become accelerated, such as might occur in HF.

Patients with HF demonstrate greater reliance on nonoxidative energy demand leading to increased CO₂ production (20). Both HFrEF and HFpEF may demonstrate skeletal muscle fiber shifts leading to an increased proportion of type II (glycolytic) relative type I fibers, decreased mitochondrial density and enzymatic activity, and exacerbated central chemoreceptor and/or group III/IV afferent activation via elevated circulating H⁺ (18–21). Thus, these and other pathologic events proposed in HF are compatible with augmented blood CO₂ content and low pHa, which could be expected to provoke augmented ventilatory compensation.

The rate of increase in V̇_E relative to metabolic demand was exaggerated across our HF patients. Likewise, a formidable drop in HCO₃⁻ from rest to peak exercise accompanied by negative BE is consistent with the presence of metabolic acidosis and base deficit (6, 8, 22, 23). The concurrent decrease in PaCO₂ linked to patterned responses in HCO₃⁻, BE, V̇_E, and pHa is consistent with the notion that ventilatory control pathways may function at an abnormally high gain coupled with a low PaCO₂ (apneic) set-point in HF (4). In this context, our group and others have demonstrated that metabolically activated muscle afferents during exercise contribute to increased ventilation and V̇_E/V̇CO₂ slope in HFrEF (18, 19, 21). While this study did not include tests of muscle afferent activation linked to ventilatory function, assuming skeletal muscle pathology occurs secondary to HF (20), this musculoskeletal neural pathway connecting systemic blood biochemical changes to brainstem centers of ventilatory control is a candidate mechanism contributing to increased V̇_E/V̇CO₂ slope in HF.

Although PaCO₂ and V_D/V_T are interrelated, the shift in the relationship between increased V̇_E/V̇CO₂ slope from PaCO₂ in HFrEF to V_D/V_T in HFpEF suggests there may have been an influence of sub-clinical pathology of mechanical and/or structural pulmonary features on increased V̇_E/V̇CO₂ slope in HFpEF (4, 9, 15, 24). However, if sub-clinical pulmonary disease occurring independently or secondary to HF were to have been present in HFpEF, we hypothesize that while inhomogeneous distribution of V̇_A/Q̇ >1.0 lung units may have contributed to increased V_D/V_T and V̇_E/V̇CO₂ slope (9, 10), the stronger influence of V̇_A/Q̇ on high V_D/V_T would have derived from effects of V̇_A/Q̇ <1.0 (suggested by PECO₂/P_ETCO₂ ≤0.70 (15)). Even with the potential for moderately reduced (~20% drop from normal) lung diffusing capacity in HFpEF (24), because of distinct properties of CO₂ (e.g., high blood solubility and diffusion rate) it could be suggested that any perfusion (V̇_A/Q̇ >0.0) of normally ventilated alveoli will result in little interference in CO₂ transfer (6, 8). In contrast, it is likely that with inhomogeneous distribution of V̇_A/Q̇ <1.0, V̇_A, and DLO₂ throughout lungs that effects of venous admixture and moderately reduced gas transfer could be expected to nominally influence PaCO₂ while effectively diluting arterial O₂ content and increase PA-aO₂. Inasmuch as HFpEF demonstrated alignment between PaCO₂ and PACO₂ suggesting no impairment in CO₂ diffusing capacity, PA-aO₂ simultaneously extended beyond normal limits (6, 8). These combined observations are consistent with the suggestion that V̇_A/Q̇ <1.0 lung areas demonstrate stronger effects on PaO₂ compared to effects of V̇_A/Q̇ >1.0 lung sectors on impaired CO₂ exchange and, hence, PaCO₂ (6, 8). Thus, although peak exercise PaCO₂ inversely related to V̇_E/V̇CO₂ slope in HFpEF, it does not appear that effects of abnormal pulmonary CO₂ transfer and whole lung V̇_A/Q̇ >1.0 played an appreciable role in this relationship.

Limitations

Information regarding the long-term prognostic translation of the present observations is not yet available in this limited sample size of HF patients, and causality cannot be deduced from these data. Increased V̇_E/V̇CO₂ slope in HFpEF of this study represents the higher limit in this population. A possible explanation for this observation is a secondary effect of aging as these data are consistent with reports of Haykowski et al. (5) and Borlaug et al. (12) in HFpEF of similar age, whereas younger HFpEF in Shafiq et al. (2) and Guazzi et al. (1) demonstrated lesser V̇_E/V̇CO₂ slope. Additional effects of obesity might also be considered impactful to increased V̇_E/V̇CO₂ slope secondary to attenuated pulmonary perfusion as others have demonstrated direct correlation between BMI and degree of diastolic dysfunction independent of comorbidities commonly recognized in HFpEF (25). While this study did not include age and body sized matched controls with respect to each HF group, elucidating the independent influences of aging and obesity on ventilatory inefficiency is important in this line of study as both factors contribute to the HFpEF phenotype.

Discussions regarding the influence of pulmonary perfusion on V̇_E/V̇CO₂ slope and V_D/V_T are hypothesis generating interpretations without direct measurements of central hemodynamics. Despite this limitation, the study design based upon ‘ideal’ alveolar equations accounts for cardiopulmonary hemodynamic interactions in addition to influences of peripheral perfusion physiology on the derivation of V̇_E/V̇CO₂ slope (4, 6–9). Although PaCO₂ is taken as a cross-sectional measurement, whereas V̇_E/V̇CO₂ slope is quantified as a continuous variable, we suggest breath-to-breath variability in PaCO₂ (i.e., negligible compared to inter-breath variability in P_ETCO₂) would not have an appreciable effect on relationships demonstrated in this study.

We acknowledge that β-blocker selectivity may potentially impact the interpretation of V̇_E/V̇CO₂ slope in HFrEF (26). This study was not powered to test the effect of β-blocker selectivity on our outcomes, and therefore additional studies are needed to clarify the potential role of β-blocker selectivity on exercise ventilation in HF. The immediate translation of these data to a specific therapeutic strategy to improve ventilatory efficiency in HF is also not yet available. However, based on recent observations in HFpEF suggesting exercise training attenuates pathologic processes of skeletal muscle remodeling (leading to improved oxidative metabolism (27)) and exercise accompanied by inorganic nitrite supplementation improves central—peripheral hemodynamics (e.g., V̇_A/Q̇ ~1.0) (12, 28), we hypothesize that questions tested in this study applied to those paradigms can be used to advance the understanding of how to interpret V̇_E/V̇CO₂ slope while also lending an ability to properly manage ventilatory function in HF. Lastly, our suggestion of a possible role of impaired skeletal muscle afferents on abnormal ventilatory control that contributes to increased V̇_E/V̇CO₂ slope in HFpEF has not been formally tested in the setting of the present aims. However, because of skeletal muscle pathophysiology (e.g., shift from type I to II fibers (20)), HFpEF may be predisposed to abnormal skeletal muscle afferent function similar to that recognized in HFrEF (18, 19, 21).

Conclusions

These data suggest patients with HF demonstrate abnormally increased V̇_E/V̇CO₂ slope, which can be explained by both V_D/V_T and PaCO₂ during CPET. However, variance in V̇_E/V̇CO₂ slope is more strongly explained by V_D/V_T than PaCO₂ in HFpEF as compared to HFrEF patients, in whom variance in V̇_E/V̇CO₂ slope can be primarily accounted for by decreased PaCO₂. These hypothesis-generating data emphasize the need to refine the interpretation and clinical utility of V̇_E/V̇CO₂ slope in HF. Advanced studies focusing on the integrated pathophysiology of abnormal V̇_E/V̇CO₂ slope, V_D/V_T, and PaCO₂ are warranted in HF, particularly those examining the unique relevance of peripheral musculoskeletal pathologies in contributing to the variable HF phenotype which has come to the forefront in discussions of HFpEF (1–3, 5, 11, 12, 20).

Appendix: Calculation methods for derived blood biochemistry and ‘ideal’ alveolar (gas/air) equation parameters

In using arterialized blood draws as described in the above text (methods), the Henderson-Hasselbalch equation was used to derive bicarbonate (HCO₃⁻) in blood as,

$${\text{HCO}}_3^{-}=s\times {\text{PaCO}}_2\times {10}^{\text{pHa}-\mathrm{p}{\mathrm{K}}^{\prime }};$$

where at standard body temperature of 37°C and pHa at 7.4, s equals 0.0307 mmol/L/mm Hg, which is the plasma solubility coefficient of CO₂, whereas pK′ equals 6.0907, which is the acid-dissociation constant of carbonic acid (29, 30). Both s and pK′ were corrected for temperature (T) and pHa as follows (29, 30):

$$\begin{array}{c}s=0.0307+0.00057\times (37-\mathrm{T})+0.0002\times {(37-\mathrm{T})}^2,\text{and}\\ \mathrm{p}{K}^{\prime }=6.086+0.042\times (7.4-\text{pHa})+(38-\mathrm{T})\times [0.00472+0.00139\times (7.4-\text{pHa})]\end{array}$$

Base excess of arterialized blood (BE) was calculated using the Van Slyke equation derived by Siggaard-Andersen from BE and buffer base curves at body temperature and corrected for SaO₂ (22, 23),

$$\text{BE}=\left[\mathrm{Z}\times \left(({\text{HCO}}_3^{-}-24.4)+\mathrm{\beta }\times (\text{pHa}-7.4)\right)\right]+[0.3\times \text{Hb}\times \frac{100-{\text{SaO}}_2}{100}];$$

where Z describes the distribution of HCO₃⁻ between red blood cells and plasma depending on hemoglobin (Hb) concentration equal to, 1–0.0143×Hb; 24.4 is the plasma HCO₃⁻ concentration at a reference pHa of 7.4; β is the slope constant of the HCO₃⁻ vs. pHa plot equal to, 7.7+1.43×Hb; and the final bracketed term is to correct for differences in SaO₂ from ideal oxygenation at 1.0.

Derived alveolar gas and air equation parameters

By using arterial blood gas measurements and acquired basic ventilation and gas exchange responses, ‘ideal’ alveolar gas and air equations and associated parameters could be used to calculate the following in addition to V_D/V_T (as explained above in methods) (4, 6, 7, 9, 10): alveolar ventilation and volume,

$${\stackrel{.}{\mathrm{V}}}_{\mathrm{A}}={\stackrel{.}{\mathrm{V}}}_{\mathrm{E}}\times \left(1-\frac{{\mathrm{V}}_{\mathrm{D}}}{{\mathrm{V}}_{\mathrm{T}}}\right)\text{and}{\mathrm{V}}_{\mathrm{A}}=\frac{{\stackrel{.}{\mathrm{V}}}_{\mathrm{A}}}{f_{\mathrm{B}}},\text{respectively};$$

alveolar CO₂ tension,

$${\text{PACO}}_2=\frac{863\times {\stackrel{.}{\mathrm{V}}\text{CO}}_2}{{\stackrel{.}{\mathrm{V}}}_{\mathrm{A}}};$$

alveolar O₂ tension,

$${\text{PAO}}_2={\mathrm{P}}_{\mathrm{I}}{\mathrm{O}}_2+\frac{{\text{PACO}}_2\times 0.21\times (1-\text{RER})}{100\times \text{RER}}-\frac{{\text{PACO}}_2}{\text{RER}};$$

ventilation of V_D (V̇_D),

$${\stackrel{.}{\mathrm{V}}}_{\mathrm{D}}={\stackrel{.}{\mathrm{V}}}_{\mathrm{E}}-{\stackrel{.}{\mathrm{V}}}_{\mathrm{A}};$$

and alveolar-to-arterial O₂ difference (PA-aO₂) was calculated as the difference between PAO₂ and PaO₂. To account for potential effects of differences in absolute V̇O₂ on the interpretability of PA-aO₂, we also provide PA-aO₂ standardized to V̇O₂ (PA-aO₂/V̇O₂).

Because of complexities involved in properly deriving end-pulmonary capillary O₂ tension, we provide an indirect simplified estimate of whole lung diffusion capacity for O₂ via Fick’s law of diffusion,

$${\text{DLO}}_2=\frac{{\stackrel{.}{\mathrm{V}}\mathrm{O}}_2}{{\text{PAO}}_2-{\text{PaO}}_2}.$$

For above parameters related to ‘ideal’ alveolar gas and air equations: V̇_E is minute ventilation, V_D is physiologic dead space, V_T is tidal volume, f_B is respiratory rate, V̇CO₂ is carbon dioxide output, 863 is the correction factor needed when computing partial pressure from fractional concentration involving both volumes/gas (STPD) and volumes/flows (BTPS, body temperature and pressure, saturated) standards of measurement, P_IO₂ is the inspired tension of O₂, 0.21 is inspired O₂ fraction (room air), RER is the respiratory exchange ratio at the lung, V̇O₂ is pulmonary O₂ uptake, and PaO₂ is arterial O₂ tension.