Increased myocardial stiffness more than impaired relaxation function limits cardiac performance during exercise in heart failure with preserved ejection fraction: a virtual patient study

Abstract

Aims

The relative impact of left ventricular (LV) diastolic dysfunction (LVDD) and impaired left atrial (LA) function on cardiovascular haemodynamics in heart failure with preserved ejection fraction (HFpEF) is largely unknown. We performed virtual patient simulations to elucidate the relative effects of these factors on haemodynamics at rest and during exercise.

Methods and results

The CircAdapt cardiovascular system model was used to simulate cardiac haemodynamics in wide ranges of impaired LV relaxation function, increased LV passive stiffness, and impaired LA function. Simulations showed that LV ejection fraction (LVEF) was preserved (>50%), despite these changes in LV and LA function. Impairment of LV relaxation function decreased E/A ratio and mildly increased LV filling pressure at rest. Increased LV passive stiffness resulted in increased E/A ratio, LA dilation and markedly elevated LV filling pressure. Impairment of LA function increased E/A ratio and LV filling pressure, explaining inconsistent grading of LVDD using echocardiographic indices. Exercise simulations showed that increased LV passive stiffness exerts a stronger exercise-limiting effect than impaired LV relaxation function does, especially with impaired LA function.

Conclusion

The CircAdapt model enabled realistic simulation of virtual HFpEF patients, covering a wide spectrum of LVDD and related limitations of cardiac exercise performance, all with preserved resting LVEF. Simulations suggest that increased LV passive stiffness, more than impaired relaxation function, reduces exercise tolerance, especially when LA function is impaired. In future studies, the CircAdapt model can serve as a valuable platform for patient-specific simulations to identify the disease substrate(s) underlying the individual HFpEF patient’s cardiovascular phenotype.

Graphical Abstract

Introduction

Heart failure with preserved ejection fraction (HFpEF) is a complex clinical syndrome characterized by signs and symptoms of heart failure (HF), a left ventricular (LV) ejection fraction (LVEF) greater than 50%, LV diastolic dysfunction (LVDD), and elevated mean left atrial (LA) pressure (mLAP) during rest and/or exercise.¹ HFpEF accounts for half of the total HF burden and, in sharp contrast to HF with reduced ejection fraction, no evidence-based treatments are available to this date.² Key reasons behind this lack of available treatments stem from the diversity in phenotypes and an incomplete understanding of the underlying pathophysiology.^3,4 HFpEF is a heterogeneous disease for which the standard one-size-fits-all treatment does not fit, and a targeted phenotype-specific treatment seems to be more promising.⁵

The diagnosis of HFpEF can be challenging since symptoms are often non-specific or difficult to elucidate, especially in an early disease stage.⁶ In addition, indicators of HFpEF or LVDD, i.e. impaired LV relaxation function and/or increased LV passive stiffness, are difficult to measure non-invasively. While echocardiography is the recommended non-invasive modality, its diagnostic value can be limited, particularly in the presence of impaired LA function.⁷ Hence, current clinical guidelines recommend exercise testing to determine whether LV diastolic pressures remain normal, especially in patients with inconclusive resting echocardiogram.¹ In addition, studying HFpEF in a controlled manner in clinical setting can be cumbersome, as recognition of individual factors can be confounded by comorbid conditions that impair exercise intolerance. Therefore, a computational model of HFpEF may be promising in recognizing specific HFpEF phenotypes and their pathophysiology, potentially leading to predicting the response to treatment.

Computational models of the human cardiovascular system allow for well-controlled variations in cardiac and vascular functional properties and, hence, enable to study relations between the structural and functional cardiovascular abnormalities and the distinctive pathophysiological features of HFpEF patients. Therefore, we simulated common pathophysiological features of the HFpEF disease, including impaired LV relaxation, increased LV passive stiffness, and impaired LA function, covering a wide variety of disease severity. These simulations were used to determine the relative effects of these abnormalities on cardiovascular system dynamics at rest and during exercise. In order to quantify these relative effects, simulated cardiovascular function at rest and exercise were assessed using current clinical guideline recommendations for the diagnosis of HFpEF.^1,8

Methods

Model description

The multi-scale CircAdapt model of the human heart and circulation^9–11 enables realistic beat-to-beat simulation of cardiovascular mechanics and haemodynamics at rest and during exercise under a wide variety of (patho-)physiological circumstances (www.circadapt.org). CircAdapt is configured as a closed-loop system consisting of several module types, including cardiac walls, cardiac valves, large blood vessels, systemic and pulmonary peripheral vasculature, the pericardium and local cardiac tissue mechanics (Figure 1A). Furthermore, CircAdapt allows structural tissue adaptation to changes in haemodynamic loading by varying the structure of the cardiovascular system (i.e. mass, size, and passive stiffness of cardiac and vascular walls) through physiological adaptation rules, based on mechano-sensing feedback control.¹²

Figure 1.

Schematic representation of the methods used in this study. (A) Structure of the CircAdapt cardiovascular system model. (B) Flow chart showing how the two different reference simulations were obtained. (C) Following from these two control simulations, LV diastolic dysfunction is simulated and the resulting simulation output were assessed at both rest and exercise haemodynamics. AV, aortic valve; Eed, end-diastolic elastance; LA, left atrium; LV, left ventricle; MAP, mean arterial pressure; MV, mitral valve; PV, pulmonary valve; RA, right atrium; RV, right ventricle; tau, isovolumic relaxation duration; TV, tricuspid valve.

Reference HFpEF simulations

For this study, resting conditions were taken to be a cardiac output (CO) of 5.1 L/min with a heart rate (HR) of 70 beats per minute, being the typical average values measured in HFpEF patients.¹³ The CO of 5.1 L/min is the effective systemic blood flow, which equals the output of the LV, because the aortic and mitral valves are assumed competent. Furthermore, using the abovementioned structural adaptation to haemodynamic loading, LV wall mass was allowed to increase by 25% from the healthy value with an LV end-diastolic volume of 120 mL, thereby resembling LV concentric hypertrophy similar to HFpEF patients.¹⁴ Large clinical studies observed systemic arterial hypertension in patients with HFpEF with the mean arterial pressure (MAP) typically at 115 mmHg.¹⁵ Correspondingly, using homeostatic pressure-flow regulation model, MAP and CO were kept constant at their predefined reference values in all simulations through changes in both systemic vascular resistance and total blood volume. Two reference simulations were used in this study, one with normal LA function and one with impaired LA function (Figure 1B). The latter was created by changing LA myocardial contractility and LA passive stiffness to 15% and 250% of their respective healthy values, so that LA ejection fraction and LA stiffness was in the range of measured LA function in patients with HFpEF (i.e. 39 ± 17% and 0.79 ± 0.75 mmHg/mL, respectively).¹⁶ The resulting reference simulations with LV concentric hypertrophy, systemic arterial hypertension, and with or without impaired LA function were used as starting point for all subsequent simulations.

Simulating the virtual HFpEF cohort

The following two LV diastolic abnormalities associated with HFpEF and LVDD were imposed on the septal- and LV free wall to evaluate their effect on cardiovascular system dynamics:

1. Impaired LV relaxation function: the time constant of LV isovolumic relaxation (tau), as determined from the LV pressure trace using the Weiss method¹⁷ as shown in Supplementary material online, Methods A, was increased from 35 ms in the control simulation to 70 ms, in steps of 5 ms. This range of tau was chosen because it falls well within the range of healthy controls and HFpEF patients as reported by Zile et al.¹⁸

2. Increased LV passive stiffness: the end-diastolic elastance (Eed), as determined from the tangent of the end-diastolic pressure-volume relationship as described by Burkhoff et al.¹⁹ as shown in Supplementary material online, Methods B, was increased from 0.2 mmHg/mL (normal passive stiffness) to 2.6 mmHg/mL, in steps 0.1 mmHg/mL. This maximum Eed was chosen because it is significantly in excess of the 1.28 mmHg/mL reported by Borlaug et al.²⁰

Varying combinations of both impaired LV relaxation function and increased passive stiffness were imposed on the abovementioned reference simulations, yielding two sets of 200 virtual patient simulations (Figure 1C). For each simulation, LV diastolic function was assessed using the echocardiographic indices of LVEF, mitral flow velocities (E, A, and E/A ratio) and maximum LA volume (LAV), and compared to current diagnostic criteria for grading LVDD.⁸ LVEF was quantified as the change in LV volume during systole divided by end-diastole volume, mitral E and A velocities as the maximum transmitral flow velocities during early- and late diastole, respectively, and LAV as the maximum LA volume over a cardiac cycle. Furthermore, to provide definitive evidence of LV diastolic pressures being normal or elevated at rest, mLAP is calculated directly from the simulated LA pressure trace over a cardiac cycle. Lastly, CO, HR, and MAP were maintained at 5.1 L/min, 70 beats per minute and 115 mmHg, respectively, in all simulations.

Modelling exercise haemodynamics

Exercise can be simulated in CircAdapt by simultaneously increasing CO and HR, using a fixed relationship from literature²¹ as shown in Supplementary material online, Methods C. Exercise intensities are achieved while increasing the total blood volume in the cardiovascular system, with MAP maintained at 115 mmHg through changes in systemic vascular resistance. We hypothesized that mLAP is a limiting factor for the cardiovascular system’s ability to increase exercise intensity, because of backward transmission of elevated left-sided filling pressures into the pulmonary circulation. To allow comparison between simulations, we assumed that it is impossible to increase exercise intensity when mLAP exceeds 35 mmHg. This pressure threshold was chosen as it resembles the measured pulmonary capillary wedge pressure (i.e. LA pressure) reported by Borlaug et al.²² at peak exercise. For each of the virtual HFpEF patient simulations described above, CO and HR were gradually increased from rest until the predefined mLAP threshold was reached, with the threshold-reaching CO being defined as the cardiac exercise performance (CEP). In addition to CEP, the average increase of mLAP per increase in CO is calculated from the mLAP-CO relationship to quantify the effect of abnormal LV and/or LA function to the increase of LV filling pressure with exercise.

Results

Relation of LV diastolic dysfunction with impaired LV relaxation function and increased passive stiffness

Simulations of the two LV diastolic abnormalities introduced in the model with normal LA function demonstrate marked changes in LV diastolic function (Figure 2). The following changes were identified:

Figure 2.

Effect of gradually impaired left ventricular (LV) relaxation function (top), increasing LV passive stiffness (middle), and the combination of both LV abnormalities (bottom) on LV and left atrial (LA) pressure, transmitral flow velocity, and LV pressure–volume relationship. For comparison, the pressure and volume traces of the control simulation are displayed (transparent lines). A, late diastolic inflow velocity; E, early diastolic inflow velocity; Eed, end-diastolic elastancetau; tau, isovolumic relaxation duration.

1. Impaired LV relaxation function increased the duration of isovolumic relaxation and reduced peak E-wave velocity resulting in a decreased E/A ratio, with a prolonged E-wave deceleration duration, and normal LA pressures. The LV pressure–volume relationship showed minimal changes in LV end-diastolic pressures.

2. Increased LV passive stiffness decreased the duration of isovolumic relaxation, increased peak E-wave velocity and decreased A-wave velocity, resulting in a sharp increase in E/A ratio. LA and LV diastolic pressures elevated severely, especially at end-diastole.

3. The combination of both LV functional abnormalities increased peak E-wave velocity and slightly reduced peak A-wave velocity, however E/A ratio is normal in all simulations (i.e. between 1.0 and 2.0). Yet, mLAP and LV diastolic pressures elevated severely, similar to simulations with isolated increased LV passive stiffness. Hence, the sum of both LV functional abnormalities leads to a pseudo-normalization of the LV filling pattern, thereby demonstrating the limited predictive value of E/A ratio to elevated LV filling pressures.

Lastly, the pressure–volume relationships demonstrate that the imposed LV diastolic abnormalities exert no effect on LVEF, which is preserved in all simulations.

Simulation cohort of virtual HFpEF patients

Figure 3 shows LV systolic- and diastolic function as assessed by LVEF, E/A ratio, maximum LA volume, and mLAP in the virtual patient simulation cohort, in absence and presence of impaired LA function (Figure 3A and B, respectively). In general, it is shown that LVEF is preserved in all virtual patients that E/A ratio decreases with impairing LV relaxation function, and increases with increasing LV passive stiffness. Furthermore, maximum LA volume and mLAP increase predominantly in response to increased LV passive stiffness. Impaired LA function increased E/A ratio, due to diminished A-wave velocity, and increased the sensitivity of mLAP to increase with LV passive stiffness. These modulating effects of impaired LV relaxation function and increased LV passive stiffness on LV systolic- and diastolic function were observed regardless of the predefined MAP (Supplementary material online, Results A).

Figure 3.

Contour maps indicating the relationship between the changes in LV relaxation function and LV passive stiffness on LV systolic- and diastolic function indices considered: LVEF, E/A ratio, maximum LA volume, and mLAP (from left to right), in the absence (A) and presence (B) of impaired LA function. Each grid point corresponds to a single simulation. The highlighted white isoline represent the clinically used cut-off criteria^1,8 for each respective index. Four representative cases with various degrees of LV diastolic function are marked by coloured dots as indicated. Eed, end-diastolic elastancetau; tau, isovolumic relaxation duration.

Following these results, the various severity degrees of LVDD as commonly observed in HFpEF could be reproduced using the abovementioned LV functional abnormalities. These representative virtual patients are highlighted by the coloured dots in Figure 3A and B, representing virtual patients with normal LV diastolic function (grey), impaired relaxation (grade I LVDD, yellow), pseudo-normal filling (grade II LVDD, orange), and restrictive filling (grade III LVDD, red).

Figures 4 and 5 demonstrate the cardiovascular haemodynamics of each respective virtual patient simulation in comparison to clinical measurements, in the absence and presence of impaired LA function. In general, the isolated effects of impaired LA function are shown to be an increased E/A ratio as well as mLAP, regardless of LV function. Simulations of moderately impaired LA function, included in Supplementary material online, Results B, show qualitatively similar but less pronounced effects on resting haemodynamics. As a result, grading of LVDD in the presence of impaired LA function becomes inconsistent, as E/A ratio in each virtual patient is greater than 2.0 (Figure 5), yet mLAP at rest is not necessarily elevated in all simulations.

Figure 4.

Transmitral inflow velocity patterns, E/A ratio, and mean left atrial (LA) pressure (mLAP) with increasing left ventricular (LV) diastolic dysfunction (DD) severity. For comparison, example continuous-wave Doppler recordings adapted from Sohn et al.²³ with clinical cut-off criteria of each respective severity are shown (top). The four representative simulations were qualitatively similar in mitral flow morphology compared with the example measurements (bottom). For comparison, the pressure and volume traces of the control simulation are displayed (transparent lines). A, late diastolic inflow velocity; Ao, aortic; E, early diastolic inflow velocity.

Figure 5.

Simulated transmitral inflow velocity patterns, E/A ratio, and mean left atrial (LA) pressure (mLAP) in the four representative simulations with LA dysfunction. For comparison, the pressure and volume traces of the simulation with normal LA function with the annotated LV functional abnormality are displayed (transparent lines). A, late diastolic inflow velocity; Ao, aortic; E, early diastolic inflow velocity.

Effect on cardiac exercise haemodynamics

Figure 6 demonstrates that increased LV passive stiffness limits CEP more than impaired LV relaxation function, regardless of LA function. In the simulation with normal LV and LA function (grey line, Figure 6A) mLAP increases on average 1.8 mmHg per 1 L/min increase in CO, leading to a CEP of 20.1 L/min. Impaired LV relaxation function (yellow line, Figure 6A) led to a slight, disproportional increase in diastolic pressures with exercise, with mLAP increasing on average 2.1 mmHg per 1 L/min increase in CO. As a result, CEP is reached at a 16% lower CO than the simulation with normal LV and LA function. Increased LV passive stiffness in addition to impaired LV relaxation function (orange line, Figure 6A) results in an almost linear increase of mLAP with CO, with mLAP increasing on average 3.4 mmHg per 1 L/min increase in CO, which results in a reduction in CEP of 49% compared to the simulation with normal LV and LA function. In the simulation with a more severely stiffened LV (red line, Figure 6A), mLAP increased on average 8.3 mmHg per 1 L/min increase in CO and CEP was reduced by 75% compared to the simulation with normal LV and LA function. Impaired LA function has a marginal exercise-limiting effect in simulation with normal LV function as mLAP increases on average 1.9 per 1 L/min increase in CO and CEP lowered by 10% as compared to the simulation with normal LV and LA function (grey lines in Figure 6A vs. Figure 6B). Similarly, impaired LV relaxation with impaired LA function led to an average increase in mLAP of 2.2 mmHg per 1 L/min increase in CO with CEP lowered an additional 12% as compared to simulation with normal LA function (yellow lines in Figure 6A vs. Figure 6B). However, in the simulations with increased LV passive stiffness a more pronounced exercise-limiting effect of impaired LA function is observed with mLAP increasing on average 5.1 and 10.3 mmHg per 1 L/min increase in CO and CEP being an additional 33% and 18% lower (orange and red lines in Figure 6A vs. Figure 6B). Lastly, a moderately impaired LA function produced qualitatively similar relationships between exercise capacity and the two LV diastolic abnormalities (see Supplementary material online, Results B), which emphasizes the importance of a normal LA function to maintain exercise performance.

Figure 6.

Cardiac exercise performance (CEP) as a function of left ventricular (LV) relaxation function and LV passive stiffness with normal left atrial (LA) function (A) and impaired LA function (B). The contour maps on the left side of the figure show the continuous effects of LV relaxation function and LV passive stiffness on exercise capacity. Each grid point corresponds to a single simulation. Four representative cases with various degrees of LV diastolic function are marked by coloured dots as indicated. The right side of the figure show how mean LA pressure (mLAP) increases with cardiac output (CO) in these four representative cases. CEP is defined as the CO at which the exercise-limiting mLAP threshold is exceeded. Eed, end-diastolic elastancetau; tau, isovolumic relaxation duration.

Discussion

In the present study, we demonstrated the pathophysiological contributions and associations of a wide range of (patho-)physiological LV and LA function on LV diastolic (dys)function and exercise haemodynamics in virtual patients with HFpEF. Given the qualitative and quantitative resemblance of clinical resting echocardiographic observations in HFpEF patient and model simulations and the direct mechanistic translation to exercise haemodynamics, this modelling approach could in the future be a valuable tool in recognizing specific phenotype in the individual patient and provide a better insight into the underlying pathophysiology. While there are some computational modelling studies of HFpEF and LVDD,^24–26 the existing studies, to the best of our knowledge, have not linked the relative contributions of LV and/or LA functional abnormalities as found in HFpEF patients to both rest and exercise haemodynamics.

In general, our virtual patient simulations suggest that increased LV passive stiffness reduced exercise capacity more substantially than impaired LV relaxation function did, and even more so when LA function is impaired. Furthermore, simulations revealed that impairment of LV relaxation function and increase of LV myocardial stiffness exert the opposite effect on the pattern of LV filling dynamics in terms of transmitral flow velocity, giving rise to inconsistent LVDD grading (i.e. pseudo-normal filling pattern).

Effects of impaired LV relaxation function and increased passive stiffness on LV filling dynamics during exercise

Previous studies have demonstrated that LV filling pressures increase significantly on exertion in patients with HFpEF as compared to healthy controls.²² In addition, invasive exercise haemodynamics testing was shown to enhance diagnosis of HFpEF, even in patients with normal echocardiography and normal LV filling pressures at rest. These clinical observations are concordant with the results of the current study, in which simulations with isolated impaired LV relaxation function are associated with normal LV filling pressures at rest and only have significant effects on mLAP at higher exercise intensities. Furthermore, simulations suggest that increased LV passive stiffness leads to a significant reduction in exercise capacity, with an increase in filling pressures on exertion similar to measurements in HFpEF patients.²⁰ Moreover, our results are in concordance with the clinical observation that elevated resting LV filling pressures are associated with worse exercise tolerance as compared to patients with normal resting haemodynamics.^1,8

LA dysfunction can confound grading of LV diastolic dysfunction

Assessment of LV diastolic function using Doppler echocardiography is an integral part of the diagnosis HFpEF, with increased LVDD disease severity being associated with decreased exercise capacity.²⁷ However, diagnosing LVDD becomes increasingly difficult in the presence of impaired LA function (e.g. atrial fibrillation).²⁸  Figure 5 and the Supplementary material online, Results B show that impaired LA function leads to a reduced A-wave, resulting in an overall increase in E/A ratio. As a result, even the simulation with no LV functional abnormality has an E/A ratio exceeding 2.0, which would suggest grade III LVDD with elevated mLAP and poor exercise capacity.⁸ However, a comparison of Figure 6A and B (grey lines) contradicts this LVDD gradation, as the mLAP–CO relationship and CEP are only marginally reduced by impaired LA function alone. Hence, our simulations demonstrate that the different grades of LVDD are more a reflection of LA-LV functional interaction than of LV function alone. In future studies, our computational modelling and simulation approach can be used to identify diagnostic indices specific to LA function, which may help to differentiate between LA and LV components of diastolic function.

LA dysfunction can beget HFpEF and vice versa

Clinical trials have shown that abnormal LA function is rather common in HFpEF patients and that patients with an impaired LA function have worse functional class and poorer exercise capacity than those without.^29,30 Whether impaired LA function precedes the incident of HFpEF or vice versa remains a topic of debate.³¹ In a study by Sanders et al.³² it was demonstrated that chronic haemodynamic overload of the LA as a result of HF triggers heterogeneous structural remodelling of the LA wall, e.g. atrial fibrosis and conduction abnormalities. Similarly, our simulations corroborate these clinical findings by showing that increased LV passive stiffness results in severe LA dilation and elevated filling pressures (Figure 3A). Hence, our virtual patient simulations suggest that impaired LA function in HFpEF can be a consequence of increased LV passive stiffness. However, simulations show that impaired LA function on its own increases mLAP at rest and decreases exercise tolerance, regardless of LV function (Figures 3 and 6, respectively). These detrimental effects of impaired LA function on exercise haemodynamics are even more pronounced in the presence of increased LV passive stiffness. In addition, both mLAP at rest and CEP appear to progress towards detrimental values more severely in response to increased LV passive stiffness. These results suggest that to patients with LA dysfunction are more prone to develop symptoms of HF at a lower LV myocardial stiffness, compared to those without LA dysfunction. In addition, Ling et al.³³ observed that long-standing LA dysfunction could mediate adverse LV remodelling by driving LV fibrosis, thereby leading to signs and symptoms of HFpEF. Following these results, our cohort of virtual patient simulations suggests that impaired LA function can serve as both a consequence of and a risk factor for the development of HFpEF.

Clinical implications

In clinical practice, diagnosis of HFpEF is difficult and often missed.⁶ Consequently, selecting an optimal treatment strategy is difficult, and the currently used one-size-fits-all approach does not yield the desired results in patients. A more personalized phenotype-specific treatment seems to be more promising.⁵ The mechanistic nature of our modelling approach enables studying the cause-consequence relationship between HFpEF disease substrates and their relative pathological contribution to rest and exercise haemodynamics. As a result, our model may be able to bridge the knowledge-gap between a patient’s clinical manifestation and their underlying cardiovascular pathology. Using clinical (non-invasive) diagnostic data, the CircAdapt model may be personalized to obtain a patient-specific cardiovascular model. Such a ‘Digital Twin’ of the patient’s heart may then be used to aid in selecting therapy strategy, as the effects of therapeutic interventions could be evaluated by simulation first.³⁴ Future work will focus on patient-specific modelling for diagnostic and potentially also therapeutic stratification of HFpEF patients.

Study limitations

Central to the clinical assessment of LVDD is the ratio between peak E-wave velocity to early diastolic mitral annular velocity (e′; E/e′ ratio), which has been shown to reasonably agree with LV filling pressures.^1,8 In CircAdapt, longitudinal septal/LV lateral wall motion with respect to LA wall motion is inadequately represented as the cardiac cavities are approximated by spherical cavities.^9,10 However, in the model LV filling pressure can be obtained directly from the LA pressure traces, hence usage of the E/e′ ratio could be omitted.

Although the impaired LA function simulations and Supplementary material online, Results B show qualitative similar effects on LV filling haemodynamics, the appropriate degree of LA contractile dysfunction could not be verified from clinical data,¹⁶ as it is difficult to assess what portion of the observed changes in LA structure and function is caused by abnormal LV diastolic function and what portion by primary LA dysfunction. Nevertheless, the mechanistic concept of LA failure mediating HFpEF and confounding LVDD grading are still supported by the model.

In this virtual patient study, we assumed that resting values of MAP (115 mmHg) and CO (5.1 L/min) were sustainable through homeostatic pressure-flow regulation. This assumption may not be the case in all patients with severe LVDD and comorbidities such as valvular disease.

We acknowledge that only clinical studies can comprehend the clinical complexity of an HFpEF cohort. However, computational models like the one used in this study allow for improving our understanding of basic pathophysiological mechanisms under well-controlled conditions that are difficult to achieve in clinical setting. Highly prevalent comorbidities that could result in exercise intolerance in patients with HFpEF, such as valvular diseases, chronotropic incompetence, and pulmonary hypertension¹⁵ can be implemented in the CircAdapt model³⁵ and thus be investigated in future studies.

Conclusion

We present a unique in silico research platform for realistic simulations of cardiovascular haemodynamics in virtual HFpEF patients. Through well-controlled variations of active and passive myocardial tissue properties, such as LV relaxation function, LV passive stiffness, and LA function, we were able to simulate HFpEF patients with a wide spectrum of LVDD and related limitations of CEP, all with preserved resting LVEF. Our virtual patient simulations revealed that increased LV passive stiffness, more than impaired LV relaxation function, reduces CEP. Furthermore, our findings emphasize the importance of assessing LA function in HFpEF, because impaired LA function can confound the diagnostic assessment of LVDD and exert an additional exercise-limiting effect, especially in the presence of increased LV passive stiffness. In future studies, the CircAdapt model can serve as a valuable platform for patient-specific simulations to identify the disease substrate(s) underlying the individual HFpEF patient’s cardiovascular phenotype.

Funding

We acknowledge the support from the Dutch Heart Foundation (Dr Dekker grant 2015T082 and ERA-CVD JTC2018 grant 2018T094, both to J.L.) and the Netherlands Organization for Scientific Research (NWO-ZonMw, VIDI grant 016.176.340 to J.L.).

Data availabilty

The simulation algorithms used during the current study are available from the corresponding author on reasonable request. The CircAdapt source-code is open access and the url (www.circadapt.org) is provided in text.

Conflicts of interest: none declared.