Peak O2‐pulse predicts exercise training‐induced changes in peak V̇O2 in heart failure with preserved ejection fraction

Abstract

Aims

Exercise training (ET) has been consistently shown to increase peak oxygen consumption (V̇O₂) in patients with heart failure with preserved ejection fraction (HFpEF); however, inter‐individual responses vary significantly. Because it is unlikely that ET‐induced improvements in peak V̇O₂ are significantly mediated by an increase in peak heart rate (HR), we aimed to investigate whether baseline peak O₂‐pulse (V̇O₂ × HR⁻¹, reflecting the product of stroke volume and arteriovenous oxygen difference), not baseline peak V̇O₂, is inversely associated with the change in peak V̇O₂ (adjusted by body weight) following ET versus guideline control (CON) in patients with HFpEF.

Methods and results

This was a secondary analysis of the OptimEx‐Clin (Optimizing Exercise Training in Prevention and Treatment of Diastolic Heart Failure, NCT02078947) trial, including all 158 patients with complete baseline and 3 month cardiopulmonary exercise testing measurements (106 ET, 52 CON). Change in peak V̇O₂ (%) was analysed as a function of baseline peak V̇O₂ and its determinants (absolute peak V̇O₂, peak O₂‐pulse, peak HR, weight, haemoglobin) using robust linear regression analyses. Mediating effects on change in peak V̇O₂ through changes in peak O₂‐pulse, peak HR and weight were analysed by a causal mediation analysis with multiple correlated mediators. Change in submaximal exercise tolerance (V̇O₂ at the ventilatory threshold, VT1) was analysed as a secondary endpoint. Among 158 patients with HFpEF (66% female; mean age, 70 ± 8 years), changes in peak O₂‐pulse explained approximately 72% of the difference in changes in peak V̇O₂ between ET and CON [10.0% (95% CI, 4.1 to 15.9), P = 0.001]. There was a significant interaction between the groups for the influence of baseline peak O₂‐pulse on change in peak V̇O₂ (interaction P = 0.04). In the ET group, every 1 mL/beat higher baseline peak O₂‐pulse was associated with a decreased mean change in peak V̇O₂ of −1.45% (95% CI, −2.30 to −0.60, P = 0.001) compared with a mean change of −0.08% (95% CI, −1.11 to 0.96, P = 0.88) following CON. None of the other factors showed significant interactions with study groups for the change in peak V̇O₂ (P > 0.05). Change in V̇O₂ at VT1 was not associated with any of the investigated factors (P > 0.05).

Conclusions

In patients with HFpEF, the easily measurable peak O₂‐pulse seems to be a good indicator of the potential for improving peak V̇O₂ through exercise training. While changes in submaximal exercise tolerance were independent of baseline peak O₂‐pulse, patients with high O₂‐pulse may need to use additional therapies to significantly increase peak V̇O₂.

Introduction

Exercise training (ET) has a Class I recommendation for patients with heart failure with preserved ejection fraction (HFpEF).¹ While pharmacological trials, except for the recently published EMPEROR-HF trial,² have been overall unsuccessful,³ ET has been shown to reduce exercise intolerance—the hallmark symptom in HFpEF—as measured by increased peak oxygen consumption (V̇O₂).^4,5 However, as for any given treatment, ET is associated with a certain response heterogeneity with some patients showing better responses than others despite exercising to a similar extent. Heterogeneous responses to therapies led to the concept of personalized medicine which requires the identification of factors associated with treatment effects. This is especially important in HFpEF, as it is known to be a multifactorial and highly heterogeneous disease with several co-existing co-morbidities^6,7 and patients are almost exclusively suffering from multiple defects affecting the convective and diffusive oxygen delivery and utilization.⁸ In addition to abnormal active relaxation and increased passive stiffness (i.e. diastolic dysfunction) resulting in a blunted stroke volume (SV) response during exercise,⁹ a high prevalence of chronotropic incompetence, which is considered as the most relevant haemodynamic limitation in HFpEF,¹⁰ further limits the cardiac output response to incremental exercise. On the other hand, emerging data suggest that abnormalities in extracardiac factors such as haemoglobin concentration, alveolar ventilation, lung or muscle diffusion capacity, or mitochondrial respiration, which lead to a reduced arteriovenous oxygen content difference [C(a-v)O₂], play a significantly greater role in HFpEF compared with heart failure with reduced ejection fraction (HFrEF).^8,11,12

It is generally assumed that individuals with a lower baseline peak V̇O₂ have a higher potential to benefit from ET; however, this could not be confirmed in a recent meta-analysis in patients with heart failure.¹³ According to the Fick Principle, V̇O₂ is the product of heart rate (HR), SV, and C(a-v)O₂, and therefore, the increase in peak V̇O₂ following ET is mediated through an increase in any or a combination of these variables. Peak HR is known to be highly dependent upon age but not significantly different between trained and sedentary individuals.¹⁴ Accordingly, a meta-analysis from trials in healthy middle aged and older adults¹⁵ revealed that endurance ET did not significantly improve peak HR and that the improved peak V̇O₂ was related to significant changes in both peak SV and peak C(a-v)O₂. During cardiopulmonary exercise testing (CPET), the product of SV and C(a-v)O₂ can be indirectly obtained as O₂-pulse [V̇O₂ × HR⁻¹ = SV × C(a-v)O₂].

Even though peak HR, SV and C(a-v)O₂ can all be significantly reduced in patients with HFpEF, limited evidence suggests that the ET-induced improvements in peak V̇O₂ are mainly due to increases in C(a-v)O₂,^16,17 whereas most studies did not show increases in either peak HR (7/9 studies)^5,16,18–24 or peak SV (2/2 studies).^16,17 Based on these findings and the concept of a higher potential for improvement when starting with a lower baseline, the hypothesis of this study was that baseline peak O₂-pulse—not baseline peak V̇O₂—is inversely associated with the ET-induced change in peak V̇O₂ following 3 months of supervised ET compared with guideline control (CON).

Methods

Study setting

This study is a secondary analysis of the initial 3 month supervised period of the OptimEx-Clin (Optimizing Exercise Training in Prevention and Treatment of Diastolic Heart Failure) trial—a prospective, randomized, controlled, multicentre-trial investigating the effects of high-intensity interval training (HIIT), moderate continuous training (MCT) and CON in 180 sedentary patients with stable HFpEF (New York Heart Association Class II-III; left ventricular ejection fraction ≥50%; elevated estimated LV filling pressure [E/e′ medial ≥15] or E/e′ medial ≥8 with elevated natriuretic peptides [NT-proBNP ≥ 220 pg/mL or BNP ≥ 80 pg/mL]).²⁵ The study design²⁶ and the main results of the trial⁵ have been published before.

In brief, participants were randomly assigned to HIIT (3 × 38 min/week with 4 × 4-min intervals at 80–90% of heart rate reserve), MCT (5 × 40 min/week at 35–50% heart rate reserve) and CON (one-time advice on physical activity). All patients were assessed at baseline and 3 months after randomization. CPET was performed on bicycle ergometers (starting at 20 watts, increasing by 10 watts per minute) and analysed at the study core lab in Munich, blinded to treatment arm assignment. Peak V̇O₂ was defined as the highest 30 s average within the last minute of exercise.²⁷ Peak HR and peak O₂-pulse were defined as the 30 s average derived from the same time span as peak V̇O₂. V̇O₂ at the first ventilatory threshold (VT1), a measure of submaximal exercise tolerance, was determined by the V-slope method.²⁸ The study was approved by the local ethic committees at all participating sites and conforms with the principles outlined in the Declaration of Helsinki. All participants provided written informed consent.

As the current research question did not aim at differences between HIIT and MCT, the results of the main analysis were not significantly different between both modes,⁵ and the required sample size to identify covariate–treatment interactions is substantially higher than for comparing group means,²⁹ the main analyses were performed with one ET group (combination of HIIT and MCT) versus CON. Only patients with complete paired baseline and 3 month follow-up CPET measurements were included.

Statistical analyses

The primary endpoint in the present analysis was the change in relative peak V̇O₂ (mL/kg/min). The change in relative V̇O₂ at VT1 (mL/kg/min) was analysed as a secondary endpoint. To ensure comparability in the evaluation of individual responses between single subjects with varying baseline values, all changes are expressed as %-change from baseline to 3 month follow-up. Next to baseline peak O₂-pulse, relative peak V̇O₂ and its other determinants (absolute peak V̇O₂, peak HR, weight, and haemoglobin as one of the determinants of C(a-v)O₂) were examined. Group means were compared with t-tests for independent means. For comparisons of ordinal data, the Mann–Whitney U-test was used. To determine the proportions of change in relative peak V̇O₂ that can be explained by the changes in peak O₂-pulse, peak HR and weight, a causal mediation analysis with multiple correlated mediators³⁰ was performed (R library ‘multimediate’³¹). Furthermore, the relationships between changes in relative peak V̇O₂, changes in peak HR, changes in peak O₂-pulse and changes in weight were analysed. The impact of baseline peak V̇O₂ and its determinants on the change in relative peak V̇O₂ was assessed using linear regression models with main effects of the independent variable and group as well as their interaction term (group × independent variable). These analyses were performed using robust linear regressions with MM-type estimators (function ‘rlm’ in R library ‘MASS’³² and function ‘f.robftest’ in R library ‘sfsmisc’³³). This method uses an iteratively reweighted least-squares procedure fitting bisquare estimators that is insensitive to influential data points and remains highly efficient (in comparison to ordinary least square estimates) in case of no outliers.³⁴ Analyses of the primary endpoint were performed in a complete data set (including all patients with valid assessments of the variables of interest) and a per-protocol set (excluding patients randomized to ET with adherence of <70% to the scheduled exercise sessions). Furthermore, we performed a sensitivity analysis within the original groups (HIIT vs. MCT vs. CON). Global interaction p-values were calculated using the function ‘lmrob’ with the recommended setting ‘KS2014’ in R library ‘robustbase’.³⁵ All statistical analyses were performed using R Statistical Software (Version 3.6.1, Foundation for Statistical Computing, Vienna, Austria)³⁶ with local significant levels of α = 0.05. As a secondary analysis, the results presented in this manuscript should be interpreted as exploratory.

Results

All 158 patients [106 (ET) vs. 52 (CON); 66% women; mean age of 70 ± 8 years; Table 1] with complete CPET data at follow-up were included in this sub-study (Figure 1). Due to indeterminable VT1, analyses including this endpoint were performed on all 154 patients with determinable VT1 at baseline and follow-up [104 (ET) vs. 50 (CON)]. The per-protocol set of ET patients included 87 individuals who performed at least 70% of the prescribed exercise sessions.

Table 1 Demographic and clinical characteristics at baseline

	Exercise (N = 106)	Guideline control (N = 52)
Sex
Female	70 (66%)	34 (65%)
Male	36 (34%)	18 (35%)
Age, years	70 (7)	69 (10)
Weight, kg	84.6 ± 18.0	78.7 ± 15.2
Body mass index, kg/m2	30.6 ± 6.1	29.0 ± 4.9
Resting heart rate, min	65 ± 11	65 ± 11
Blood pressure, mmHg
Systolic	128 ± 13	127 ± 15
Diastolic	74 ± 10	74 ± 10
Cardiovascular risk factors
Hypertension	90 (85%)	46 (88%)
Diabetes	30 (28%)	12 (23%)
Dyslipidaemia	74 (70%)	40 (77%)
Smoking
Never smoked	55 (52%)	30 (58%)
Ex-smoker	45 (42%)	21 (40%)
Current Smoker	6 (6%)	1 (2%)
Sleep apnoea	19 (18%)	9 (17%)
Severity of HFpEF
New York Heart Association class
II	80 (75%)	35 (67%)
III	26 (25%)	17 (33%)
E/e′ average, [no.]	13.4 ± 3.4 [103]	13.3 ± 4.7 [49]
E/A, [no.]	1.23 ± 0.65 [88]	1.20 ± 0.62 [46]
NT-proBNP, pg/mL, [no.]	321 (161–689) [102]	341 (175–622) [52]
Haemoglobin, mg/dL, [no.]	13.6 ± 1.6 [103]	13.2 ± 1.4 [52]
Other cardiac diagnoses
Coronary artery disease	29 (27%)	16 (31%)
Atrial fibrillation
Paroxysmal	13 (12%)	7 (13%)
Persistent	8 (8%)	3 (6%)
Permanent	10 (9%)	2 (4%)
Heart failure medication
Beta-blocker	68 (64%)	37 (71%)
Thiazide/loop diuretics	60 (57%)	31 (60%)
Angiotensin receptor blocker	44 (42%)	21 (40%)
Angiotensin-converting enzyme inhibitor	36 (34%)	16 (31%)
Aldosterone antagonists	12 (11%)	5 (10%)
Cardiopulmonary exercise testing parameters
Peak oxygen consumption, mL/kg/min	18.5 ± 5.1	19.5 ± 5.8
Peak oxygen pulse, mL/beat	12.7 ± 3.5	12.9 ± 4.0
Peak heart rate, b.p.m.	123 ± 26	121 ± 28
Peak respiratory exchange ratio	1.11 ± 0.09	1.10 ± 0.13

[IMAGE OMITTED. SEE PDF]

Comparison of mean changes

Relative peak V̇O₂ increased by 8.0 ± 15.7% in the ET group compared with a reduction of −2.0 ± 18.3% in the CON group. Mean changes were significantly different between ET and CON for relative peak V̇O₂, absolute peak V̇O₂, peak O₂-pulse and weight (P < 0.05), while no significant differences have been observed for the change in peak HR, haemoglobin (Figure 2, Table 2) or the levels of exhaustion as measured by peak respiratory exchange ratio (RER) (Table 2). Mean change in relative V̇O₂ at VT1 was significantly different between groups (P = 0.03; Table 2). The difference in change in relative peak V̇O₂ between groups was primary mediated by changes in peak O₂-pulse (~72%), while changes in peak HR and weight accounted for approximately 18% and 10%, respectively. From baseline to follow-up, the beta-blocker dosage was changed in 14 patients with an increase in two patients randomized to ET (none in CON) and a decrease in 10 ET and 2 CON patients (P = 0.13). When these patients were excluded, the difference in mean change in peak HR between groups diminished to 0.3% (95% CI, −3.5 to 4.1, P = 0.89), whereas the mean changes in peak V̇O₂, peak O₂-pulse, and weight remained significantly different between the groups (data not shown). In this subset, change in peak O₂-pulse accounted for approximately 88% of the difference in change in relative peak V̇O₂ between groups.

[IMAGE OMITTED. SEE PDF]

Table 2 Mean changes following exercise training and guideline control

	Mean ± SD
Exercise training [N = 106]*	Guideline control [N = 52]*
Baseline	3 months	Change	Baseline	3 months	Change	Difference (95% CI), P-value
Relative peak V̇O2, mL/kg/min	18.5 ± 5.1	19.8 ± 5.7	8.0 ± 15.7%	19.5 ± 5.8	18.9 ± 5.7	−2.0 ± 18.3%	10.0% (4.1 to 15.9), P = 0.001
Absolute peak V̇O2, mL/min	1,528 ± 447	1,617 ± 465	6.8 ± 15.2%	1,519 ± 488	1,475 ± 485	−2.1 ± 17.3%	8.9% (3.3 to 14.5), P = 0.002
Peak O2-pulse, mL/beat	12.7 ± 3.5	13.2 ± 3.1	5.1 ± 13.9%	12.9 ± 4.0	12.5 ± 4.3	−2.5 ± 18.5%	7.7% (1.9 to 13.4), P = 0.01
Peak heart rate, b.p.m.	123 ± 26	124 ± 25	2.3 ± 13.5%	121 ± 28	121 ± 30	0.3 ± 13.1%	2.6% (−2.5 to 6.4), P = 0.39
Weight, kg	84.6 ± 18.0	83.8 ± 18.1	−1.0 ± 2.6%	78.7 ± 15.2	78.8 ± 15.6	0.1 ± 2.6%	−1.0% (−1.9 to −0.2), P = 0.02
Haemoglobin, g/dL*	13.6 ± 1.6	13.6 ± 1.4	0.4 ± 7.2%	13.2 ± 1.4	13.1 ± 1.4	−0.6 ± 5.5%	0.9% (−1.1 to 3.0), P = 0.38
Relative V̇O2 at VT1, mL/kg/min*	11.1 ± 3.1	11.9 ± 3.1	9.4 ± 16.5%	11.5 ± 2.9	11.4 ± 2.6	2.0 ± 21.3%	7.4% (2.0 to 14.2), P = 0.03
Peak Respiratory Exchange Ratio	1.11 ± 0.09	1.10 ± 0.13	−1.0 ± 5.9%	1.10 ± 0.09	1.11 ± 0.12	0.9 ± 11.0%	−0.1% (−5.2 to 1.3), P = 0.24

Covariate-treatment interactions

The influence of baseline peak O₂-pulse (in mL/beat) on change in relative peak V̇O₂ was significantly higher following ET compared with CON (interaction P = 0.04; Figure 3A). In the ET group, every 1 mL/beat higher baseline peak O₂-pulse was associated with a decreased mean change in relative peak V̇O₂ of −1.45% (95% CI, −2.30 to −0.60, P = 0.001). In contrast, the change in relative peak V̇O₂ following CON was not dependent on baseline peak O₂-pulse [β-coefficient: −0.08% (95% CI, −1.11 to 0.96), P = 0.88]. After adjustment for sex, age, and baseline weight, this difference remained significant [ET: −1.89% (95% CI, −2.84 to −0.94); CON: −0.42% (95% CI, −1.77 to 0.62); interaction P = 0.049]. The interaction between baseline peak O₂-pulse and group on change in peak V̇O₂ may also depend on baseline peak RER with a higher difference between groups in patients with peak RER > 1.10 (Supporting Information, Figure S1). No significant interactions on change in relative peak V̇O₂ were found between groups and relative peak V̇O₂, absolute peak V̇O₂, peak HR, weight, and haemoglobin (interaction P > 0.05; Figure 3, Table 3). None of these factors was significantly associated with the change in relative V̇O₂ at VT1 in either group (Supporting Information, Table S1; Figure S2). The influence of baseline peak O₂-pulse on change in relative peak V̇O₂ was similar in HIIT and MCT; however, neither peak O₂-pulse (interaction P = 0.15) nor any of the other factors showed a significant interaction with the original study groups (HIIT, MCT, and CON) on the change in relative peak V̇O₂ (Supporting Information, Table S2; Figure S3). Results of the per-protocol analysis were similar to the main results (Table 3; Supporting Information, Figure S4). Following ET, every 1 mL/beat higher baseline peak O₂-pulse was associated with a decreased mean change in relative peak V̇O₂ of −1.88% (95% CI, −2.79 to −0.97, P < 0.001; interaction P = 0.01). Accordingly, the mean difference in change in relative peak V̇O₂ between a patient who attended at least 70% of ET sessions and a patient randomized to CON was 20.5% for a baseline peak O₂-pulse of 8.6 mL/beat (10th percentile), and 3.7% for a baseline peak O₂-pulse of 17.9 mL/beat (90th percentile).

[IMAGE OMITTED. SEE PDF]

Table 3 Results of the predictor analyses for the inter-individual response variability in peak V̇O₂ for the main and per protocol analysis

	Mean change in relative peak V̇O2 [β-coefficient (95% CI), P-value]	Interaction P for exercise training vs. guideline control
Exercise training	Guideline control [N = 52]*	Main analysis	Per protocol analysis
Complete set [N = 106]*	Per protocol set [N = 87]*
Peak O2-pulse (per 1 mL/beat)	−1.45% (−2.30 to −0.60), P = 0.001	−1.88% (−2.79 to −0.97), P < 0.001	−0.08% (−1.11 to 0.96), P = 0.88	0.04	0.01
Peak V̇O2 (per 1 mL/kg/min)	−0.38% (−0.99 to 0.22), P = 0.17	−0.26% (−0.98 to 0.45), P = 0.47	−0.39% (−1.10 to 0.32), P = 0.28	0.97	0.78
Peak V̇O2 (per 100 mL/min)	−0.76% (−1.45 to 0.07), P = 0.03	−0.74% (−1.51 to 0.02), P = 0.06	−0.17% (−1.03 to 0.69), P = 0.70	0.31	0.34
Peak heart rate (per 10 b.p.m.)	0.42% (−0.79 to 1.63), P = 0.50	0.74% (−0.57 to 2.06), P = 0.27	−0.48% (−1.96 to 0.99), P = 0.52	0.35	0.22
Weight (per 10 kg)	−1.11% (−2.84 to 0.63), P = 0.21	−2.04% (−4.15 to 0.08), P = 0.06	1.35% (−1.35 to 4.05), P = 0.32	0.14	0.054
Haemoglobin (per 1 mg/dL)*	0.10% (−1.81 to 2.00), P = 0.92	0.21% (−1.91 to 2.33), P = 0.85	1.60% (−1.46 to 4.65), P = 0.30	0.44	0.49

Associations between changes in peak V̇O₂ and its determinants

In the overall sample, changes in relative peak V̇O₂ were positively correlated with changes in peak O₂-pulse and peak HR and negatively correlated with weight (P < 0.001). Furthermore, the changes in peak O₂-pulse were negatively correlated with the changes in peak HR (P < 0.001). Changes in weight were not significantly correlated with either changes in peak HR nor changes in peak O₂-pulse (P > 0.45). A correlation matrix is shown in Supporting Information, Figure S5. None of these associations were significantly different between ET and CON (interaction P > 0.05; Supporting Information, Table S3).

Discussion

This is the first study to examine potential predictors of inter-individual response variability in relative peak V̇O₂ following ET versus CON in HFpEF and, to the best of our knowledge, the first study to examine the effects of baseline peak O₂-pulse on the change in peak V̇O₂ following ET in any population. The main finding of this study (confirming the primary hypothesis) is that in patients randomized to ET, lower baseline peak O₂-pulse was associated with a larger improvement in relative peak V̇O₂, whereas no such association was found in CON patients. This difference between ET and CON remained significant after adjusting for sex, age and baseline weight, and increased after excluding ET-patients with an adherence lower than 70%. Furthermore, the predictive effect of baseline O₂-pulse seemed to be independent of the exercise mode (HIIT and MCT).

O₂-pulse is the fraction of V̇O₂ and HR; therefore, low peak O₂-pulse can be caused by either low peak V̇O₂, high peak HR or a combination of both. Nevertheless, the association of baseline peak V̇O₂ (absolute or adjusted to body weight) with the change in relative peak V̇O₂ was not significantly different between groups, confirming the results of a recent meta-analysis in heart failure.¹³ When adjusted to body weight, we observed an almost parallel decline in change in peak V̇O₂ with increasing baseline peak V̇O₂ in both groups. Accordingly, the baseline level of relative peak V̇O₂ may only be a prognostic marker for its change in patients with HFpEF, with differences between ET and CON being similar for different base levels. In a cohort study including 120 patients with heart failure with reduced ejection fraction,³⁷ it has already been shown that the severity of chronotropic incompetence may be associated with an impaired response to ET. However, in the present trial, peak HR alone was not a significant predictor of the ET-induced changes in relative peak V̇O₂. Alternatively, a high O₂-pulse can also be defined as a high SV and/or C(a-v)O₂. Accordingly, a higher peak O₂-pulse at baseline is accompanied with a lower reserve to increase peak SV and/or C(a-v)O₂—the components of peak V̇O₂ that are most likely to improve following ET.¹⁵ This interpretation is supported by the results after splitting the sample based on baseline peak RER. While a peak RER ≥1.10 is considered excellent effort and maximal exhaustion,³⁸ patients with a lower peak RER could have stopped for other reasons (e.g. low motivation or musculoskeletal complaints), which seems to reduce the predictive power of baseline peak O₂-pulse for ET-induced changes in peak V̇O₂ in this subgroup. On the other hand, patients with a low peak O₂-pulse despite high volitional effort are very likely to be truly limited by their ability to increase SV and/or C(a-v)O₂. Therefore, in addition to the results of the per-protocol analysis, the higher difference between groups in patients with peak RER ≥ 1.10 strengthens the assumption of a true association between baseline peak O₂-pulse and change in peak V̇O₂ through ET. Consequently, the change in peak O₂-pulse was also the primary mediator for the change in relative peak V̇O₂ following ET in this study. Importantly, the predictive power of baseline peak O₂-pulse did not apply to the change of V̇O₂ at VT1. Instead, patients were able to equally improve their functional capacity through exercise training irrespective of baseline peak O₂-pulse. This may be explained by the fact that peak O₂-pulse is not a determinant of V̇O₂ at VT1. Furthermore, the extent to which ET-induced changes in V̇O₂ at VT1 are mediated by changes in HR, SV and C(a-v)O₂ is less well investigated.

To date, only two studies have examined the effects of ET versus CON on SV and C(a-v)O₂ in HFpEF.^16,17 In a non-randomized study, Fu et al.¹⁶ found a significant improvement in peak V̇O₂ after 12 weeks of HIIT compared with CON (n = 60) along with significant improvements in peak C(a-v)O₂ and without significant changes in either peak SV or peak HR. In a secondary analysis of the randomized controlled PARIS study (n = 40), Haykowsky et al.¹⁷ showed that the improvement in peak V̇O₂ following ET was due to significant increases in peak HR and peak C(a-v)O₂. However, despite the significant increase in peak HR, only 16% of the training related improvement in peak V̇O₂ was attributed to an improved cardiac output. While changes in peak HR were not significantly different between the groups in most ET trials in HFpEF,^5,16,19–22 significant improvements as observed in two trials^23,24 could also be influenced by factors not directly related to ET, that is, changes in levels of exhaustion or changes in HR-affecting medications (e.g. beta-blockers). For instance, patients randomized to the ET group of the PARIS study²³ had a significantly higher change in peak HR compared with CON (+4 vs. −7 b.p.m.); however, the results for peak systolic blood pressure (+1 mmHg vs. −10 mmHg, P = 0.04), and peak RER (+0.03 vs. −0.02, P = 0.07) indicate that different levels of exhaustion between groups may have contributed to the significant difference in peak HR. In the present trial, changes in beta-blocker dosage were more common in the ET group (11.3%) compared with CON (3.8%) and when these patients were excluded, the difference in change in peak HR between groups diminished from 1.9% to 0.3% (both P > 0.05).

The evidence that ET-related improvements in peak V̇O₂ are most likely mediated through increases in peak C(a-v)O₂ may also explain the overall positive effects of ET in patients with HFpEF as C(a-v)O₂ has been shown to be reduced in 75% and being the leading cause of exercise intolerance in 40% of patients with HFpEF.¹² Furthermore, it has been shown that a normalization of impaired muscle oxygen diffusion would result in a significantly larger improvement in peak V̇O₂ than a normalization of convective oxygen delivery.^8,12 Due to interactions between the components of the Fick equation, Houstis et al.⁸ demonstrated that doubling a patient's cardiac output would lead to a decrease in C(a-v)O₂ of 45%, and thus, peak V̇O₂ would increase by only 10%. On the other hand, normalization of the 36% deficit in skeletal muscle oxygen diffusion that was observed in their study led to a predicted improvement in peak V̇O₂ of 27%. Similarly, the results of the present trial show that a change in peak HR (independent of groups assignment) was not associated with an equivalent change in relative peak V̇O₂ (a 10% increase in peak HR was associated with ~6.4% increase in peak V̇O₂), which is likely to be explained by a reduced SV (shortening the time of the diastole) and a reduced C(a-v)O₂ (reducing the contact time in the muscle).

Future implications

The results of this exploratory analysis implicate that patients with HFpEF and high O₂-pulse may not be able to significantly improve their peak V̇O₂ by performing regular ET. Although we still highly recommend regular ET for patients with HFpEF and high O₂-pulse to reduce the decline in peak V̇O₂ with ageing and disease progression and to improve parameters beyond peak V̇O₂ (e.g. V̇O₂ at VT1), it should probably be supplemented by additional therapies if the intention is to increase maximal exercise tolerance.

Despite lacking evidence for its benefits in HFpEF, most patients were treated with beta-blockers (66% in the present trial). This proportion is likely to be decreasing, as the effects of a long-term administration of beta-blockers in hypertension, stable coronary artery disease or atrial fibrillation (common co-morbidities in HFpEF) are questioned and some studies led to the concern that beta-blockers may be even deleterious in HFpEF.³⁹ By increasing the duration of diastole, a reduced HR may allow a better left ventricular filling; however, it may also impair the cardiac output response to exercise⁴⁰ and according to the results of the present analysis may contribute to a blunted ET response by its effect on peak O₂-pulse. Indeed, a recently published trial investigating the effects of beta-blocker withdrawal in HFpEF⁴¹ has shown a significant short-term increase in peak HR (~31%) and peak V̇O₂ (~17%). However, further research is necessary to show whether interventions to increase peak HR in patients with HFpEF (e.g. by reducing beta-blockers or rate-adaptive pacing⁴²) have a positive long-term effect on clinical outcomes and exercise capacity.

Interestingly, we also found a trend towards lower ET-induced changes in relative peak V̇O₂ for patients with higher body weight at baseline (P = 0.14 in the full analysis, P = 0.054 in the per-protocol analysis). Whether patients with HFpEF and higher baseline weight benefit less from exercise training or especially HIIT (see Supporting Information, Figure S3E) should be investigated in future studies. Nevertheless, the results of the present study underscore the need for additional trials that specifically target weight loss in HFpEF. As most patients with HFpEF are overweight or obese (85% in the present trial), losing weight, which has a disproportionate impact on relative peak V̇O₂ (a weight loss of 20% leads to an increase in relative peak V̇O₂ of 25%), is another important option to increase exercise tolerance in HFpEF that has been largely ignored so far. To date, in the only lifestyle intervention trial targeting weight loss in patients with HFpEF (N = 100; mean BMI, 39.3 kg/m²),²¹ ET and caloric restriction resulted in similar and additive changes in relative peak V̇O₂ (main effect of ET: 1.2 mL/kg/min vs. diet: 1.3 mL/kg/min; joint effect: 2.5 mL/kg/min) by significantly improving absolute peak V̇O₂ (ET) and reducing weight (ET and caloric restriction).

A combination of treatments targeting several deficits may overcome the interaction effects between the determinants of peak V̇O₂ and will likely have a disproportionate impact compared with the correction of a single deficit.⁸ For example, based on the interaction with SV and C(a-v)O₂, that is, O₂-pulse, increasing peak HR will possibly not only have a direct effect on peak V̇O_2,⁴¹ but also enhance the potential for improving peak V̇O₂ following ET in patients with high O₂-pulse.

Methodological aspects, strengths, and limitations

This study has several strengths and limitations. The individual pre-post change in peak V̇O₂ can be divided into a ‘true’ change depending on the intervention, a ‘true’ change which is independent of the intervention (e.g. ageing), and a change due to random errors which is also independent of group assignment (e.g. measurement errors, day-to-day variability, different levels of exhaustion during the CPETs at baseline and follow-up).²⁹ Therefore, to examine predictors of the ET-induced response variability (instead of prognostic factors that are independent of the intervention and possibly influenced by regression to the mean) it is mandatory to include a comparator arm, which, however, has not been performed in many previous studies. Furthermore, the dependent and all independent parameters were analysed as continuous variables, which has several advantages over arbitrary categorization (e.g. retaining higher power and avoiding misclassification due to random errors).

Despite these strengths, the original trial was designed to detect differences between group means and therefore, methods to further reduce the bias of random errors (e.g. repeated pre-measurements and post-measurements)²⁹ have not been applied. The wide scatter of individual changes underlines the importance of conducting predictor analyses; however, this heterogeneity might have been amplified by the multimorbid condition of the patients with a high number of adverse events in both groups⁵ and the fact that, strictly speaking, this study was not a predictor analysis for ‘ET versus control’ as it compared the offer for supervised ET (HIIT and MCT) with a recommendation to perform regular physical activity (CON). To account for different levels in adherence, we performed a per-protocol analysis excluding ET patients with an adherence of <70%; however, it is unclear if and how many patients assigned to CON started exercising between baseline and follow-up. Nevertheless, as mean peak V̇O₂ slightly decreased following CON, it is unlikely that many patients performed regular exercise training in this group. On average, patients included in the present trial had a relatively preserved exercise capacity at baseline. However, the wide range of baseline values and their linear relationships with the change in peak V̇O₂ suggest that the results of the regression analyses are likely to be generalizable. Lastly, the measurement of O₂-pulse does not allow to distinguish between SV and C(a-v)O₂; however, it can be more easily obtained in routine care. While both peak SV and peak C(a-v)O₂ are significantly reduced in HFpEF,¹⁰ further research is necessary to show whether both factors play a relevant role for the improvements in peak V̇O₂ following ET.

Conclusions

In patients with HFpEF, lower baseline peak O₂-pulse is associated with higher ET-induced changes in relative peak V̇O₂. This is an important finding towards a deficit-oriented personalized medicine in HFpEF, provides an easily measurable indicator of the potential for improving maximal exercise tolerance through exercise training and underlines the value of CPET to guide therapy. While changes in submaximal exercise tolerance were independent of baseline peak O₂-pulse, patients with HFpEF and high O₂-pulse may need to use additional therapies (e.g. reduction of negative chronotropic drugs, rate-adaptive pacing, and/or weight loss) to significantly increase maximal exercise tolerance.

Acknowledgements

The authors thank the staff of the University Hospital ‘Klinikum rechts der Isar’ of the Technical University of Munich, Antwerp University Hospital, Heart Centre Leipzig, Charité Universitätsmedizin Berlin and Norwegian University of Science and Technology Trondheim who were involved in recruitment, evaluations, administration, and support during the conduct of the OptimEx-Clin study. Open Access funding enabled and organized by Projekt DEAL.

Conflict of interest

Stephan Mueller reported receiving grants from Deutsche Forschungsgemeinschaft (DFG) through the TUM International Graduate School of Science and Engineering and the German Centre for Cardiovascular Research (DZHK) during the conduct of the study. Ephraim B. Winzer reported receiving personal fees from Novartis (honoraria for lectures and advisory board activities), Boehringer Ingelheim (honoraria for advisory board activities), and CVRX (honoraria for lectures) outside the submitted work. Luciene F. Azevedo reported receiving grants from Brazilian National Council for Scientific and Technological Development and Technical University of Munich—Laura Bassi Award. André Duvinage reported receiving grants from Novartis outside the submitted work. Frank Edelmann reported receiving grants from DFG, BMBF, Servier, and personal fees from Bayer Healthcare, Merck, Novartis, Servier, Berlin Chemie, Boehringer Ingelheim, Vifor Pharma, AstraZeneca, PharmaCosmos outside the submitted work. Axel Linke reported receiving speaker fees from Abbott, Medtronic, Edwards Lifesciences, AstraZeneca, Boston Scientific, and Novartis; grants from Edwards Lifesciences and Novartis; advisory board fees from Transverse Medical, Picardia, Edwards Lifesciences, and Heart Leaflet Technology; and stock options from Claret Medical and Transverse Medical, and being a co-owner of Dresden Cardiovascular Research Institute and Core Laboratories outside the submitted work. Emeline M. Van Craenenbroeck reported receiving grants from the Flemish Research Funds (FWO) as a senior clinical investigator during the conduct of the study. Burkert Pieske reported receiving personal fees from Bayer Healthcare (steering committee, lectures), Merck (steering committee, lectures), Novartis (steering committee, lectures), Servier, AstraZeneca (lectures), Bristol-Myers Squibb (lectures), and Medscape (lectures) outside the submitted work. Martin Halle reported receiving grants from the TUM International Graduate School of Science and Engineering and the German Centre for Cardiovascular Research (DZHK) during the conduct of the study and grants from Novartis (principal investigator of the Activity Study in HFrEF) and personal fees from Bristol-Myers Squibb, Berlin Chemie-Menarini, Novartis, Daiichi-Sankyo, AstraZeneca, Roche, Abbott (advisory board on exercise and diabetes), Sanofi, Pfizer, Boehringer Ingelheim, and Bayer, and serves as an advisor for Medical Park SE, Germany outside the submitted work. No other disclosures were reported. Bernhard Haller, Anna Feuerstein, Paul Beckers, Mark J. Haykowsky, Andreas B. Gevaert, Jennifer Hommel, Katrin Esefeld, Isabel Fegers-Wustrow, Jeffrey W. Christle, Elisabeth Pieske-Kraigher, Evgeny Belyavskiy, Daniel A. Morris, Martin Kropf, and Radhakrishnan Aravind-Kumar declared that they have no conflict of interest.

Funding

This work was supported by the European Commission, Framework Program 7 [grant number: EU 602405-2]; the German Centre for Cardiovascular Research (DZHK) [grant number: 81Z0600603] [S.M. and M.H.]; and the Deutsche Forschungsgemeinschaft (DFG) through the TUM International Graduate School of Science and Engineering (IGSSE; Garching, Germany) [S.M., A.D., and M.H.]. The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

References

Pelliccia A, Sharma S, Gati S, Bäck M, Börjesson M, Caselli S, Collet JP, Corrado D, Drezner JA, Halle M, Hansen D, Heidbuchel H, Myers J, Niebauer J, Papadakis M, Piepoli MF, Prescott E, Roos‐Hesselink JW, Graham Stuart A, Taylor RS, Thompson PD, Tiberi M, Vanhees L, Wilhelm M. ESC guidelines on sports cardiology and exercise in patients with cardiovascular disease. Eur Heart J. 2020; 2020: 17–96.

Anker SD, Butler J, Filippatos G, Ferreira JP, Bocchi E, Böhm M, Brunner‐La Rocca H‐P, Choi D‐J, Chopra V, Chuquiure‐Valenzuela E, Giannetti N, Gomez‐Mesa JE, Janssens S, Januzzi JL, Gonzalez‐Juanatey JR, Merkely B, Nicholls SJ, Perrone SV, Piña IL, Ponikowski P, Senni M, Sim D, Spinar J, Squire I, Taddei S, Tsutsui H, Verma S, Vinereanu D, Zhang J, Carson P, Lam CSP, Marx N, Zeller C, Sattar N, Jamal W, Schnaidt S, Schnee JM, Brueckmann M, Pocock SJ, Zannad F, Packer M. Empagliflozin in heart failure with a preserved ejection fraction. N Engl J Med. 2021; 385: 1451–1461.

Bonsu KO, Arunmanakul P, Chaiyakunapruk N. Pharmacological treatments for heart failure with preserved ejection fraction‐a systematic review and indirect comparison. Heart Fail Rev. 2018; 23: 147–156.

Fukuta H, Goto T, Wakami K, Kamiya T, Ohte N. Effects of exercise training on cardiac function, exercise capacity, and quality of life in heart failure with preserved ejection fraction: A meta‐analysis of randomized controlled trials. Heart Fail Rev. 2019; 24: 535–547.

Mueller S, Winzer EB, Duvinage A, Gevaert AB, Edelmann F, Haller B, Pieske‐Kraigher E, Beckers P, Bobenko A, Hommel J, Van de Heyning CM, Esefeld K, von Korn P, Christle JW, Haykowsky MJ, Linke A, Wisløff U, Adams V, Pieske B, van Craenenbroeck EM, Halle M. Effect of high‐intensity interval training, moderate continuous training, or guideline‐based physical activity advice on peak oxygen consumption in patients with heart failure with preserved ejection fraction: A randomized clinical trial. JAMA. 2021; 325: 542–551.

Shah SJ, Kitzman DW, Borlaug BA, van Heerebeek L, Zile MR, Kass DA, Paulus WJ. Phenotype‐specific treatment of heart failure with preserved ejection fraction: A multiorgan roadmap. Circulation. 2016; 134: 73–90.

Pandey A, Vaduganathan M, Arora S, Qamar A, Mentz RJ, Shah SJ, Chang PP, Russell SD, Rosamond WD, Caughey MC. Temporal trends in prevalence and prognostic implications of comorbidities among patients with acute decompensated heart failure: The ARIC study community surveillance. Circulation. 2020; 142: 230–243.

Houstis NE, Eisman AS, Pappagianopoulos PP, Wooster L, Bailey CS, Wagner PD, Lewis GD. Exercise intolerance in heart failure with preserved ejection fraction: Diagnosing and ranking its causes using personalized O(2) pathway analysis. Circulation. 2018; 137: 148–161.

Zile MR, Baicu CF, Gaasch WH. Diastolic heart failure‐‐abnormalities in active relaxation and passive stiffness of the left ventricle. N Engl J Med. 2004; 350: 1953–1959.

Pandey A, Khera R, Park B, Haykowsky M, Borlaug BA, Lewis GD, Kitzman DW, Butler J, Berry JD. Relative impairments in hemodynamic exercise reserve parameters in heart failure with preserved ejection fraction: A study‐level pooled analysis. JACC Heart failure. 2018; 6: 117–126.

Haykowsky MJ, Brubaker PH, John JM, Stewart KP, Morgan TM, Kitzman DW. Determinants of exercise intolerance in elderly heart failure patients with preserved ejection fraction. J Am Coll Cardiol. 2011; 58: 265–274.

Dhakal BP, Malhotra R, Murphy RM, Pappagianopoulos PP, Baggish AL, Weiner RB, Houstis NE, Eisman AS, Hough SS, Lewis GD. Mechanisms of exercise intolerance in heart failure with preserved ejection fraction: The role of abnormal peripheral oxygen extraction. Circ Heart Fail. 2015; 8: 286–294.

Taylor RS, Walker S, Smart NA, Piepoli MF, Warren FC, Ciani O, Whellan D, O'Connor C, Keteyian SJ, Coats A, Davos CH, Dalal HM, Dracup K, Evangelista LS, Jolly K, Myers J, Nilsson BB, Passino C, Witham MD, Yeh GY. Impact of exercise rehabilitation on exercise capacity and quality‐of‐life in heart failure: Individual participant meta‐analysis. J Am Coll Cardiol. 2019; 73: 1430–1443.

Nes BM, Janszky I, Wisløff U, Støylen A, Karlsen T. Age‐predicted maximal heart rate in healthy subjects: The HUNT fitness study. Scand J Med Sci Sports. 2013; 23: 697–704.

Montero D, Díaz‐Cañestro C. Endurance training and maximal oxygen consumption with ageing: Role of maximal cardiac output and oxygen extraction. Eur J Prev Cardiol. 2016; 23: 733–743.

Fu TC, Yang NI, Wang CH, Cherng WJ, Chou SL, Pan TL, Wang JS. Aerobic interval training elicits different hemodynamic adaptations between heart failure patients with preserved and reduced ejection fraction. Am J Phys Med Rehabil. 2016; 95: 15–27.

Haykowsky MJ, Brubaker PH, Stewart KP, Morgan TM, Eggebeen J, Kitzman DW. Effect of endurance training on the determinants of peak exercise oxygen consumption in elderly patients with stable compensated heart failure and preserved ejection fraction. J Am Coll Cardiol. 2012; 60: 120–128.

Angadi SS, Mookadam F, Lee CD, Tucker WJ, Haykowsky MJ, Gaesser GA. High‐intensity interval training vs. moderate‐intensity continuous exercise training in heart failure with preserved ejection fraction: A pilot study. J Appl Physiol. 2015; 119: 753–758.

Donelli da Silveira A, Beust de Lima J, da Silva PD, dos Santos MD, Zanini M, Nery R, Laukkanen JA, Stein R. High‐intensity interval training is effective and superior to moderate continuous training in patients with heart failure with preserved ejection fraction: A randomized clinical trial. European journal of. Prev Cardiol. 2020; 27: 1733–1743.

Edelmann F, Gelbrich G, Düngen HD, Fröhling S, Wachter R, Stahrenberg R, Binder L, Töpper A, Lashki DJ, Schwarz S, Herrmann‐Lingen C, Löffler M, Hasenfuss G, Halle M, Pieske B. Exercise training improves exercise capacity and diastolic function in patients with heart failure with preserved ejection fraction: Results of the ex‐DHF (exercise training in diastolic heart failure) pilot study. J Am Coll Cardiol. 2011; 58: 1780–1791.

Kitzman DW, Brubaker P, Morgan T, Haykowsky M, Hundley G, Kraus WE, Eggebeen J, Nicklas BJ. Effect of caloric restriction or aerobic exercise training on peak oxygen consumption and quality of life in obese older patients with heart failure with preserved ejection fraction: A randomized clinical trial. JAMA. 2016; 315: 36–46.

Smart NA, Haluska B, Jeffriess L, Leung D. Exercise training in heart failure with preserved systolic function: A randomized controlled trial of the effects on cardiac function and functional capacity. Congest Heart Fail. 2012; 18: 295–301.

Kitzman DW, Brubaker PH, Morgan TM, Stewart KP, Little WC. Exercise training in older patients with heart failure and preserved ejection fraction: A randomized, controlled, single‐blind trial. Circ Heart Fail. 2010; 3: 659–667.

Kitzman DW, Brubaker PH, Herrington DM, Morgan TM, Stewart KP, Hundley WG, Abdelhamed A, Haykowsky MJ. Effect of endurance exercise training on endothelial function and arterial stiffness in older patients with heart failure and preserved ejection fraction: A randomized, controlled, single‐blind trial. J Am Coll Cardiol. 2013; 62: 584–592.

Paulus WJ, Tschöpe C, Sanderson JE, Rusconi C, Flachskampf FA, Rademakers FE, Marino P, Smiseth OA, De Keulenaer G, Leite‐Moreira AF, Borbély A, Édes I, Handoko ML, Heymans S, Pezzali N, Pieske B, Dickstein K, Fraser AG, Brutsaert DL. How to diagnose diastolic heart failure: A consensus statement on the diagnosis of heart failure with normal left ventricular ejection fraction by the heart failure and echocardiography associations of the European Society of Cardiology. Eur Heart J. 2007; 28: 2539–2550.

Suchy C, Massen L, Rognmo O, Van Craenenbroeck EM, Beckers P, Kraigher‐Krainer E, Linke A, Adams V, Wisloff U, Pieske B, Halle M. Optimising exercise training in prevention and treatment of diastolic heart failure (OptimEx‐CLIN): Rationale and design of a prospective, randomised, controlled trial. Eur J Prev Cardiol. 2014; 21: 18–25.

Mezzani A, Agostoni P, Cohen‐Solal A, Corrà U, Jegier A, Kouidi E, Mazic S, Meurin P, Piepoli M, Simon A, Laethem CV, Vanhees L. Standards for the use of cardiopulmonary exercise testing for the functional evaluation of cardiac patients: A report from the exercise physiology section of the European Association for Cardiovascular Prevention and Rehabilitation. Eur J Cardiovasc Prev Rehabil. 2009; 16: 249–267.

Beaver WL, Wasserman K, Whipp BJ. A new method for detecting anaerobic threshold by gas exchange. J Appl Physiol (Bethesda, Md: 1985). 1986; 60: 2020–2027.

Ross R, Goodpaster BH, Koch LG, Sarzynski MA, Kohrt WM, Johannsen NM, Skinner JS, Castro A, Irving BA, Noland RC, Sparks LM, Spielmann G, Day AG, Pitsch W, Hopkins WG, Bouchard C. Precision exercise medicine: Understanding exercise response variability. Br J Sports Med. 2019; 53: 1141–1153.

Jérolon A, Baglietto L, Birmelé E, Alarcon F, Perduca V. Causal mediation analysis in presence of multiple mediators uncausally related. Int J Biostat. 2020; 17: 191–221.

Jérolon A. multimediate: Causal mediation analysis in presence of multiple mediators uncausally related. 2021. R package version 0.0.0.9000. Available from: https://github.com/AllanJe/multimediate

Venables WN, Ripley BD. Modern Applied Statistics with S, Fourth ed. New York: Springer; 2002.

Maechler M. sfsmisc: Utilities from 'Seminar fuer Statistik' ETH Zurich. 2020. R package version 1.1–7. Available from: https://CRAN.R‐project.org/package=sfsmisc

Yu C, Yao W. Robust linear regression: A review and comparison. Commun Stat Simul Comput. 2017; 46: 6261–6282.

Maechler M, Rousseeuw P, Croux C, Todorov V, Ruckstuhl A, Salibian‐Barrera M, Verbeke T, Koller M, Conceicao E, Anna di Palma M. robustbase: Basic Robust Statistics. 2021. R package version 0.93–9. Available from: http://robustbase.r‐forge.r‐project.org/

R Core Team. R: A Language and Environment for Statistical Computing. 2021. Available from: https://www.R‐project.org/

Schmid JP, Zurek M, Saner H. Chronotropic incompetence predicts impaired response to exercise training in heart failure patients with sinus rhythm. Eur J Prev Cardiol. 2013; 20: 585–592.

Balady GJ, Arena R, Sietsema K, Myers J, Coke L, Fletcher GF, Forman D, Franklin B, Guazzi M, Gulati M, Keteyian SJ, Lavie CJ, Macko R, Mancini D, Milani RV. Clinician's guide to cardiopulmonary exercise testing in adults: A scientific statement from the American Heart Association. Circulation. 2010; 122: 191–225.

Meyer M, LeWinter MM. Heart rate and heart failure with preserved ejection fraction: Time to slow β‐blocker use? Circ Heart Fail. 2019; 12: [eLocator: e006213].

Kalogeropoulos AP, Butler J. Heart rate in heart failure with preserved ejection fraction: Target or marker? Eur J Heart Fail. 2017; 19: 1504–1506.

Palau P, Seller J, Domínguez E, Sastre C, Ramón JM, de La Espriella R, Santas E, Miñana G, Bodí V, Sanchis J, Valle A, Chorro FJ, Llácer P, Bayés‐Genís A, Núñez J. Effect of β‐blocker withdrawal on functional capacity in heart failure and preserved ejection fraction. J Am Coll Cardiol. 2021; 78: 2042–2056.

Zweerink A, van der Lingen ACJ, Handoko ML, van Rossum AC, Allaart CP. Chronotropic incompetence in chronic heart failure. Circ Heart Fail. 2018; 11: [eLocator: e004969].