Dynamic aspects of ventricular interaction during exercise in HFpEF and in pre‐capillary pulmonary hypertension

Abstract Aims The contribution of adverse ventricular interdependence remains undervalued in heart failure or pulmonary vascular disease, and not much is known about its dynamic nature during exercise and respiration. In this study, we evaluated ventricular interaction during exercise in patients with heart failure with preserved ejection fraction (HFpEF) and patients with chronic thromboembolic pulmonary hypertension (CTEPH) as compared with healthy controls. Methods and results Forty‐six subjects (10 controls, 19 CTEPH patients, and 17 HFpEF patients) underwent cardiac magnetic resonance imaging during exercise. Ventricular interaction was determined through analysis of the septal curvature (SC) of a mid‐ventricular short‐axis slice at end‐diastole, end‐systole, and early‐diastole, both in expiration and inspiration. Exercise amplified ventricular interaction in CTEPH patients and to a lesser extent in HFpEF patients (P < 0.05 for decrease in SC with exercise). Adverse interaction was most profound in early‐diastole and most pronounced in CTEPH patients (P < 0.05 interaction group * exercise) because of a disproportionate increase RV afterload (P < 0.05 to both controls and HFpEF) and diastolic pericardial restraint (P < 0.001 for interaction group * exercise) during exercise. Inspiration enhanced diastolic interdependence in CTEPH and HFpEF patients (P < 0.05 vs. expiration). Both at rest and during exercise, SC strongly correlated with RV volumes and pulmonary artery pressures (all P < 0.05). Conclusions Exercise amplifies adverse right–left ventricular interactions in CTEPH, while a more moderate effect is observed in isolated post‐capillary HFpEF. Given the strong link with RV function and pulmonary hemodynamic, assessing ventricular interaction with exCMR might be valuable from a diagnostic or therapeutic perspective.


Introduction
Ventricular interdependence, or the direct force transmission between both ventricles, originates directly from the cross-talk brought about by the interventricular septum, shared myocardial fibres, and the enclosing pericardium. 1 Ventricular interactions are dynamic, persist throughout the cardiac cycle, and are large enough to have meaningful impact on cardiac function. As the interventricular septum plays a crucial role in mediating ventricular interactions, ventricular interdependence can be easily examined through evaluation of septal motion by cardiac imaging. 2 Historically, ventricular interdependence has mainly been evaluated in the context of right ventricular (RV) pressure or volume overload. 3 Recently, however, also its role in patients with heart failure is increasingly recognized. Specifically, in patients with heart failure with preserved ejection fraction (HFpEF), recent evidence suggests that, in addition to left ventricular (LV) diastolic dysfunction, pericardial restraint, and the ensuing ventricular interdependence may contribute to the disproportionate increase in filling pressures during exercise. 4,5 Moreover, the upstream transfer of rising LV filling pressures during exercise contributes to RV afterload and could further exacerbate adverse ventricular interaction.
Exercise intolerance is a cardinal manifestation of both pulmonary vascular diseases and HFpEF, and numerous studies have already demonstrated the incremental value of exercise evaluation in both conditions. 6 Unfortunately, parameters of ventricular interaction, even despite their dynamic nature, are generally only assessed at rest, and the effect of respiration is often discarded. The goal of this study was therefore to gain further insight into the dynamic influence of exercise and respiration on ventricular interdependence and to explore differences between pre-capillary and post-capillary hypertension. Using exercise cardiac magnetic resonance (CMR) imaging with simultaneous haemodynamic measurements, we compared cardiac function and parameters of ventricular interaction of patients with pre-capillary pulmonary hypertension [due to chronic thromboembolic pulmonary hypertension (CTEPH)] with HFpEF patients and isolated post-capillary pulmonary hypertension. To provide a reference frame, we also included healthy controls.

Subjects
HFpEF and CTEPH patients were recruited from the heart failure clinic and the centre for pulmonary vascular diseases at our institution. HFpEF was defined by symptoms of heart failure, normal LV ejection fraction (LVEF ≥50%), and elevated left heart filling pressures on RHC [pulmonary arterial wedge pressure (PAWP), >15 mmHg at rest or ≥25 mmHg with exercise], although CTEPH patients were diagnosed according to current guidelines. 7,8 Patients with significant valvular heart disease, coronary artery disease, hypertrophic or infiltrative cardiomyopathy, primary renal or hepatic disease, or significant ventilatory disease (FEV 1 < 50%, TLC or VC < 70%) were excluded. Control subjects without prior cardiopulmonary disease volunteered to participate after responding to local advertisements. All were asymptomatic and had a normal ECG, transthoracic echocardiogram, and right heart catheterization at rest. Finally, to be included, all participants had to able to perform at least 50 W on an upright bicycle stress test. The study conformed to the Declaration of Helsinki and was approved by the local Ethics Committee. All participants provided written informed consent.

Study design
Firstly, cardiopulmonary exercise testing with continuous monitoring of expiratory gases was performed on an upright cycle ergometer (ER900 and Oxycon Alpha, Jaeger, Germany). Through breath-by-breath analysis minute ventilation, oxygen consumption and carbon dioxide production were assessed. Additional measures included peak heart rate, peak power, and the ventilatory equivalent for carbon dioxide. Secondly, within 24 h, all subjects underwent exercise CMR imaging with simultaneous invasive pressure measurements through a 7-Fr MRI-compatible pulmonary artery catheter (Edwards Lifesciences, CA, USA) and a 20-G radial arterial catheter. Pressures were recorded through MRI-compatible transducers connected to a PowerLab recording system (ADInstruments, Oxford, UK) and analysed offline using LabChart V6.1.1 (ADInstruments, Oxford, UK). Images were acquired during free breathing at rest, and during exercise at 25%, 50% of peak power achieved during CPET. Workloads were imposed nearly instantaneously and were maintained for ≈3 min at each stage, 30 s to achieve a physiological steady state, and then 2-3 min for image acquisition and will hereinafter be referred to as rest, low, and moderate intensity.

CMR equipment, image acquisition, and analysis
Cardiac function was assessed using a free-breathing real-time exercise CMR method that we previously validated against invasive standards and that has been described in depth elsewhere. 9 Detailed information can be found in the Supporting Information. To assess ventricular interdependence, the LV and RV endocardial and epicardial borders of a mid-ventricular short-axis slice were traced during expiration and inspiration at end-diastole (ED), at end-systole (ES), and at early-diastole [defined as mitral valve opening (MVO)] both at rest and during low-intensity and moderate-intensity exercise. Using custom software, septal curvature (SC) was calculated as the reciprocal of the mean radius of curvature of the midline of the mid-septal segment (60% of total septal length to avoid tethering effects of RV insertion points) as illustrated in Figure 1. A negative SC denotes septal bowing into the LV. In HFpEF patients, right atrial pressure was used a surrogate for pericardial pressure. 10 LV transmural pressure was calculated as pulmonary capillary wedge pressure minus right atrial pressure. Pulmonary vascular reserve (P/Q slope) was evaluated through the relationship between mean pulmonary artery pressure and cardiac output.
(Shapiro-Wilk test), and variables are presented as means (± standard deviation) or as medians (with 25% and 75% percentiles) accordingly. Categorical data were compared using a Chi 2 test and continuous variables with either a Kruskal-Wallis H test or a one-way analysis of variance (ANOVA) with the Bonferroni post hoc correction. The cardiac response to exercise and the evolution of ventricular interaction were assessed using repeated measures ANOVA with exercise intensity and respiration as within-subject effect and subject group as between-subject effect. In a separate analysis, age was added as a covariate to the model. The relationship between SC, volumes, and pressures was determined using Pearson correlation coefficients. Intraobserver and interobserver reproducibility (see Table  S1) was assessed at rest and during exercise in a sample of 10 subjects using the intra-class correlation coefficient (two-way mixed and absolute agreement quoted). A P-value of <0.05 was considered statistically significant. Sample size calculation is provided in the supplements.

Baseline characteristics
Forty-six subjects (10 controls, 19 CTEPH patients, and 17 HFpEF patients) were included in the study. The baseline characteristics and the results of the cardiopulmonary exercise test are summarized in Table 1. HFpEF patients were older and had a higher BMI compared with both other groups. Parameters of peak exercise capacity were significantly lower and pulmonary artery pressures higher in CTEPH and HFpEF patients compared with controls. Finally, HFpEF patients had higher right atrial and pulmonary capillary wedge pressures compared with both other groups.

Cardiac volumes and function during exercise
The cardiac response to exercise is depicted in Table 2 and Figure S2. In control subjects, stroke volume and ejection fraction increased significantly during exercise, which was primarily driven by a significant reduction in end-systolic volume. In CTEPH patients and to a lesser extent in HFpEF patients, RV end-systolic volumes increased more than end-diastolic volumes because of the exercise-induced afterload increase (higher P/Q slope, P < 0.001 between groups). Hence, RV ejection fraction declined in both groups during exercise. LV volumes, on the other hand, either remained unchanged (HFpEF) or declined in tandem (CTEPH). LV stroke volume and ejection fraction therefore remained stable during exercise. Global biventricular volume increased minimally in all groups ( Figure 2A, P = 0.926 between groups), but there were marked differences in the behaviour of LV and RV volumes during exercise. RV/LV volume ratio increased significantly in CTEPH patients (P < 0.001), increased slightly in HFpEF patients (P = 0.015), and remained unchanged in controls ( Figure 2, P < 0.001 for interaction group * exercise). Right atrial and pulmonary arterial wedge pressures increased significantly during exercise in HFpEF patients (both P < 0.001 for rest-to-peak), whereas LV transmural pressure only increased minimally (P = 0.048 for rest-to-peak).

Dynamics of septal curvature across the cardiac cycle
The dynamics of average (expiration and inspiration combined) SC at end-diastole, end-systole, and mitral valve opening (i.e. end relaxation/early diastole) can be appreciated Figure 3 and are also represented in Figure S3. At rest, a similar change was seen in the three groups ( Figure 3A, P = 0.064 for interaction group * cardiac cycle) with slightly lower SC at ED compared with ES in controls and HFpEF patients and a significantly lower SC at MVO compared with ES in CTEPH patients. At moderate exercise intensity, the dynamics of SC differed across the groups ( Figure 3B, P < 0.001 for interaction group * cardiac cycle) with a more pronounced decrease in SC at ED and especially at MVO in HFpEF and CTEPH patients.

Effect of exercise on ventricular interaction
The effect of exercise on average (expiration and inspiration combined) SC is outlined in Figure 3 and Table S2. Throughout the cardiac cycle, there were significant differences between the three groups (between-group difference, all P < 0.01), but SC was only differentially affected by exercise at MVO (P = 0.024 for interaction group * exercise). In controls, SC did not change with exercise either at ED, ES, or MVO (main effect of exercise both P > 0.05). Conversely, in CTEPH and to a lesser extent in HFpEF patients, SC decreased significantly during exercise at ED and MVO (main effect of exercise all P < 0.01), although no effect was observed at ES (main effect of exercise P > 0.05). CTEPH patients had consistently lower SC compared with both other groups, whereas in HFpEF patients, SC was only significantly lower compared with controls at MVO during exercise. Other parameters of ventricular interaction, such as eccentricity index or the ratio of RV to LV diameter displayed similar changes during exercise ( Figure S4).

Effect of respiration on ventricular interaction
The effect of respiration did not differ between groups (interaction group * respiration all P > 0.05; Figure 4). Inspiration enhanced end-diastolic ventricular interaction at rest in CTEPH and HFpEF patients and during exercise in HFpEF patients ( Figure 4A). At end systole on the other hand, respiration did not influence SC, neither at rest nor during exercise ( Figure 4B). Finally, at MVO, a significant effect was noted in CTEPH patients at rest and in all groups during exercise ( Figure 4C). Detailed description of average, expiratory, and inspiratory SC values are available in Tables S2-S4.

Association between ventricular interaction, RV end-diastolic volume, and pulmonary artery pressures
Both at rest and during moderate exercise, strong correlations were noted between SC and RVEDV (r between À0.535 and À0.702, all P < 0.001) and sPAP (r between À0.672 and À0.788, all P < 0.001) throughout the cardiac cycle ( Figure 5 and Table S6). Likewise moderate to strong correlations were noted with total pulmonary vascular resistance (r between À0.369 to À0.700, all P < 0.05, Table S6). In addition, SC at moderate exercise intensity also correlated with P/Q slope at end-diastole and mitral valve opening (r À 0.402 and r À 0.409, respectively, both P < 0.01) but not at end systole (r À 0.160). Finally, at end systole, SC correlated (r 0.679 at rest, r 0.641 at moderate exercise, both P < 0.001; Figure S5) with a surrogate of the interventricular pressure gradient (the delta of systolic blood pressure and systolic pulmonary artery pressures, i.e. sBP-PAP). Because we did not measure PAWP during the XMR protocol, similar analyses could not be obtained with other potential surrogates at end-diastole (pulmonary arterial wedge pressure, RAP) or at mitral valve opening (sPAP, pulmonary arterial wedge pressure).

Discussion
Using exercise CMR imaging, we evaluated the effects of exercise and breathing on ventricular interdependence in healthy controls and HFpEF and CTEPH patients. We demonstrate that ventricular interaction differs across the three groups and is influenced by exercise and respiration. Specifically, exercise enhanced adverse right-left ventricular interactions in CTEPH patients and had a more limited effect in HFpEF patients with isolated post-capillary PH and no effect in controls subjects. Inspiration enhanced diastolic right-left interactions, particularly in CTEPH and HFpEF patients. Finally, associations were noted between SC and RV volumes or pulmonary artery pressures, both at rest and during exercise, suggesting that non-invasive evaluation of ventricular interdependence with exCMR might be attractive in the evaluation of treatments lowering right ventricular afterload or those specifically targeting adverse ventricular interdependence.

Influence of exercise on ventricular interdependence
Ventricular interdependence is an undervalued feature of both pulmonary vascular disease and heart failure and generally only assessed at rest. 3,4,[11][12][13] In the current study, we therefore examined ventricular interdependence during exercise in CTEPH and HFpEF patients while comparing with healthy control subjects. In CTEPH patients, we observed a strong increase in ventricular interdependence during exercise, whereas a more modest effect was observed in HFpEF patients. In contrast, SC remained largely unchanged in controls. The disparity between the groups is most likely explained by the differences in RV afterload during exercise. Pulmonary artery pressures are flow dependent but increase disproportionally in those with pulmonary vascular disease or heart failure. 14 In pulmonary hypertension, the heightened RV afterload prolongs RV contraction and results in an interventricular relaxation dyssynchrony causing a rapid left-to-right septal motion in early LV diastole. [15][16][17] The magnitude of the septal shift correlates with invasive haemodynamics and has shown to be marker of disease severity. 18,19 Hence, the increased early diastolic septal shift observed in our CTEPH cohort during exercise can be interpreted as a sign of worsening relaxation dyssynchrony. Similarly, also in HFpEF patients, the most profound differences with controls were observed at early diastole and potentially hint to the presence of some form of relaxation dyssynchrony during exercise in HFpEF subjects as well. Although still fairly limited in our cohort, relaxation dyssynchrony could become more important in those with combined pre-capillary and post-capillary pulmonary hypertension (CpC-PH) and thus might contribute, along with heightened right-sided congestion, to the increased ventricular interdependence and impaired RV reserve observed in this subtype. 20 In addition to prolonged RV contraction, also diastolic pericardial constraint appears to have contributed to the observed differences in CTEPH patients. But the pericardium is capable of adapting chronically, acute ventricular volume changes, such as occur during exercise or acute volume overload, leading to a swift increase in pericardial restraint once the critical inflection point of the pericardial stress-strain relationship is surpassed. 21,22 In both CTEPH and HFpEF patients, RV volumes increased with exercise, but only in CTEPH patients, LV volumes declined. As RV stroke volume remained unchanged with exercise, the decline in LV volumes is a manifestation of increased diastolic ventricular interdependence, which corresponds with the significantly lower SC observed in CTEPH patients at ED. In contrast, in our HFpEF cohort, the small increase in RV volumes during exercise was not accompanied by a decline in LV volumes and resulted in only small changes in the RV/LV volume ratio and the end-diastolic SC with exercise. This is in keeping with a study by Parasuraman et al. who also did not observe a decline in LV volumes in HFpEF patients during exercise and corroborates earlier reports that pericardial restraint likely plays only a minor role in non-obese HFpEF patients with isolated Figure 2 Global ventricular volume and RV/LV volume ratio. Relative change in biventricular volume from rest to moderate exercise intensity and evolution of RV/LV volume ratio during exercise in controls and CTEPH and HFpEF patients. Data presented as Tukey boxplots or as mean ± SEM. Interaction P-value between exercise and subject group. Coloured P-value denotes main effect of exercise. *, † P < 0.05 Bonferroni post hoc to controls and HFpEF, respectively.  20,24 Thus, our results confirm that diastolic pericardial restraint is not universal in HFpEF, exemplifying the substantial pathophysiological heterogeneity of the disease.

Inspiration enhances diastolic interaction during exercise
During exercise, venous return to the heart is augmented by venoconstriction and activation of the skeletal muscle and the abdominothoracic pumps. During inspiration, the decrease in intrathoracic pressure is transmitted to the heart and is known to augment right heart filling. Previously, we have demonstrated in healthy subjects that the reciprocal effects of respiration on LV and RV volumes persist during exercise. 25 In the current study, we expand upon these findings and examine the dynamic nature of respiration on ventricular interdependence during exercise in CTEPH and HFpEF patients. In general, inspiration had a similar effect on SC in the different groups and the associated increased venous return to the right heart enhanced early and to a lesser extent also late diastolic interaction during exercise. The observed effect was small and most pronounced in CTEPH and HFpEF patients, again signalling increased ventricular interdependence compared with controls.

Ventricular interdependence as therapeutic target?
Although contemporary PH treatment mainly targets RV afterload, it has been speculated that strategies aimed at mitigating the deleterious consequences of increased afterload on cardiac function (e.g. adverse ventricular interdependence) might confer additional benefit. For instance, small proof-of-concept studies have suggested that cardiac pacing might counter adverse ventricular interdependence in patients with pulmonary hypertension. [26][27][28][29] Whether this then also improves patient outcome remains to be proven, but the substantial increase in ventricular interdependence during exercise suggests, at least in theory, that this may merit further consideration as an ancillary treatment. Likewise, also in heart failure, ventricular interdependence is increasingly considered as a potential therapeutic target. Borlaug et al. demonstrated that limited pericardiotomy attenuates the increase in LV filling pressure that develops during acute volume loading. 5,30 From our results, it would appear unlikely that the limited increase in ventricular interaction we observed in our cohort of HFpEF patients with isolated post-capillary pulmonary hypertension would be sufficient for a substantial haemodynamic benefit during exercise. However, in other HFpEF subtypes with more pronounced interdependence such as obesity-related HFpEF and certainly those with associated pre-capillary pulmonary vascular disease (i.e. combined post-capillary and pre-capillary pulmonary hypertension), the potential benefit might be larger than suggested by our data. 20,24 The strong link between SC and exercise haemodynamics certainly provides a rationale for patient selection based on exercise imaging. Given its gold-standard accuracy and feasibility, exercise Effect of respiration on ventricular interaction. Tukey boxplots showing the effect of respiration on SC both at rest (left) and during moderate exercise intensity (right) at (A) end-diastole, (B) end-systole, and (C) early-diastole in controls and CTEPH and HFpEF patients. Interaction p-value denotes the interaction between respiration and subject group. *P < 0.05, **P <0.01, ***P <0.001 Bonferroni post hoc for comparison between expiration (exp) and inspiration (insp).

Limitations
Firstly, the small sample size increases the probability for Type II statistical errors and multiple comparisons increase the likelihood of Type I errors. Nevertheless, the accuracy of our exercise CMR method enables evaluation of relevant differences with high statistical significance despite the modest cohort size. Secondly, HFpEF patients were older than both other groups, which might have impacted the exercise response. However, there was no difference in SC between the youngest and oldest HFpEF tertile and integration of age as a covariate in the model did not alter the relation between the predictor and outcome variables (see Table S5). Thirdly, because we retained maximal clinical feasibility, we opted to analyse ventricular interaction from a single mid-ventricular slice, instead of considering the full septal shape. This could have lowered sensitivity for mild or more complex septal motion patterns. Moreover, because the intraventricular septum is not necessarily a perfect circular arc with a single radius of curvature, SC may vary depending on the local geometry and the measured length of arc, and thus, its relation to ventricular interdependence is inevitably confounded by a certain amount of error, bias, or variability. Fourthly, the limitation in temporal resolution did not allow evaluation of ventricular interdependence at higher heart rate and workloads. Nonetheless, as pulmonary pressures increase nearly linearly with cardiac output, the observed effects would be even more pronounced at peak exercise. Finally, as we did not perform biventricular pressure measurements, we cannot relate our findings to the timing and magnitude of the interventricular pressure gradient.

Conclusions
Exercise enhances adverse right-left ventricular interactions in pre-capillary pulmonary hypertension, although only a modest effect is observed in isolated post-capillary HFpEF. Inspiration further enhances diastolic ventricular interdependence, particularly in HFpEF and CTEPH patients. Given the strong link between SC and RV volumes or pulmonary haemodynamics, assessing ventricular interaction with exercise imaging might be valuable from a diagnostic or therapeutic perspective.

Conflict of interest
MC: none; TP: none; JB: none; ALG has received grants from the Fund for Scientific Research Flanders (FWO) and from the National Health and Medical Research Council (NHMRC) of Australia; JL: none; MD received fees as speaker, investigator, consultant, or steering committee member for Actelion, Bayer, Bellarophon, Eli Lilly, GlaxoSmithKline, MSD, Pfizer, Figure 5 Association between ventricular interaction, RV end-diastolic volume, and haemodynamics at rest and at moderate exercise intensity in all subjects. Pearson correlation of average (expiration and inspiration combined) SC at rest and during moderate exercise intensity with RV end-diastolic volume (left), systolic pulmonary artery pressure (middle), and right atrial pressure (right) at (A) end-diastole, (B) end-systole, and (C) early-diastole. *P < 0.05, **P < 0.01, and ***P < 0.001.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Table S1. Observer variability for echocardiographic measures Table S2. Average Septal Curvature (expiration + inspiration) Table S3. Expiratory Septal Curvature Table S4. Inspiratory Septal Curvature Table S5. Average septal curvature in HFpEF patients according to age tertile and effect of age as covariate on average septal curvature during exercise. Table S6. Pearson correlations of SC with volumes and hemodynamics. Figure S1. exCMR setup and ventricular volume analysis. Figure S2. Ventricular volumes and cardiac function at rest and during exercise. Figure S3. Evolution of septal curvature across the cardiac cycle. Figure S4. Effect of exercise on Eccentricity Index and RVLV ratio Figure S5. Association between ventricular interaction and the approximated end-systolic interventricular pressure gradient at rest and at moderate exercise intensity in all subjects.