How myocardial work could be relevant in patients with an aortic valve stenosis?

Abstract

Graphical Abstract

Open in new tabDownload slide

Introduction

Aortic stenosis (AS) is the most common primary valvular heart disease leading to an intervention with growing prevalence due to the ageing population.¹ Current recommendations state that aortic valve replacement (AVR) is a class I indication in cases of symptoms or reduced left ventricular ejection fraction (LVEF, <50%). Whatever, LVEF is preserved in many patients with AS even when symptoms develop. Stratification of pre-operative and post-operative risk of each patient is currently challenging. Unfortunately, valvular parameters such as aortic valve area (AVA) and transvalvular gradient did not permit an ideal risk stratification.² Several studies suggest the additional value of global longitudinal strain (GLS) to best stratify this population. Magne et al.³ demonstrated in a meta-analysis that GLS <14, 7% with preserved LVEF increased with an OR of 2.6 risk of death. Despite these results, GLS is not widely used in clinical routine. A possible explanation is the after-load dependence of GLS.⁴ Indeed, GLS decreases with the increasing LV after-load that is why an after-load independent feature to better describe LV function would be necessary.

Myocardial work (MW) is a very promised new tool to assess more precisely LV function⁵ taking into account LV after-load. Its efficiency in patient’s stratification has already been suggested in cardiac resynchronization therapy.^6,7 hypertrophic cardiomyopathies⁸, and mitral regurgitation.⁹ However, in order to calculate the MW, an accurate estimation of the pressure curve is needed. Russell et al.^5,10 have proposed a non-invasive method for the estimation LV pressure based on a black-box non-linear method that fits a reference waveform to the duration of the isovolumic and ejection phases of a given patient, as measured by echocardiographic timing of aortic and mitral valve events. Peak LV pressure was estimated from a non-invasive cuff-based measurement of the brachial artery pressure.¹¹ Thanks to this pressure curve estimation, a MW computation tool was developed for these patients with normal or subnormal afterload.⁵ However, this pressure estimation method could not be applied in the case of AS, where high pressure gradients could be observed between LV and the aorta. Fortuni et al.¹² have adapted the pressure estimation method by calculating peak LV pressure as the sum of mean aortic transvalvular gradient and aortic systolic pressure to calculate MW for this type of patients. On the other side, our team recently proposed a novel model-based approach to assess non-invasively LV pressure and MW in AS patients.¹³ Both methods have shown excellent correlations with MW indices calculated with the invasive LV pressure. The objective of this article is to compare each method performance to evaluate MW on a prospective database of AS patients. As the essential part of the MW determination is the estimation of LV pressure, pressure curves calculated with each method were compared in severe and moderate AS patients.

Methods

Population

Sixty-seven adults (>18 years old) with severe (AVA < 1 cm², n = 62) and moderate (n = 5) AS, who underwent a coronary angiography with left heart catheterization (LHC), were prospectively included. Ten patients were excluded from the final analysis because of atrial fibrillation, concomitant significant aortic regurgitation, or incomplete set of images for getting robust GLS measurements.

The study was carried out in accordance with the principles outlined in the Declaration of Helsinki on research in human subjects and received specific ethical approval from the local Medical Ethics Committee. All patients were informed, and a consent was obtained.

Echocardiography

All patients underwent a standard transthoracic echocardiography (TTE) using a Vivid S70 or E95 ultrasound system (General Electric Healthcare, Horten, Norway). Images were recorded on a remote station for off-line analysis by dedicated software (EchoPAC PC, version BT 202, General Electric Healthcare, Horten, Norway). Aortic and mitral valve events were manually evaluated in apical long-axis view: mitral valve closure (MVC), aortic valve opening (AVO), aortic valve closure (AVC), and mitral valve opening (MVO). Standard speckle tracking strain analysis was applied in order to extract regional myocardial strain curves. The AVA (cm²) and mean pressure gradient were also quantified according to current recommendations.

Invasive ventricular pressure

The LHC was performed via a retrograde access from the radial artery with a 5 French Judkin R4 catheter (ICU Medical, San Clemente, CA, USA) placed at the mid LV cavity using fluoroscopic screening. It has been performed with cautious to optimize the quality of the recording but using the catheter people are used to. Before coronary angiography, transducers were calibrated, with a 0-level set at the mid-axillary line. In a second time, catheter was placed in the thoracic ascendant aorta to measure aortic pressure. The experimental invasive data set includes the measured ventricular pressure $Plvexp$⁠, the systolic and diastolic arterial pressures.

Estimation of LV pressure from computational model

The model of cardiovascular system integrates four main sub-models, based on the previous works of our team^14–19 (Figure 1): (i) cardiac electrical system, (ii) elastance-based cardiac cavities, (iii) systemic and pulmonary circulations, and (iv) heart valves. The proposed model and the equations were described in detail in Owashi et al.¹³

Figure 1

Model of the cardiovascular system. P, V, E, and R, respectively, stand for the pressure, volume, elastance, and resistance. The indexes ao, sa, sys, sv, vc, pa, pul, and pu, respectively, stand for aorta, systemic arteries, systemic, systemic veins, vena cava, pulmonary arteries, pulmonary, and pulmonary veins.

Open in new tabDownload slide

To sum up, the cardiac electrical system is composed of a set of interconnected cellular automata, adapted from,^14,16 which represent different cardiac regions [the sinoatrial node (SAN), right and left atria (RA and LA), the atrioventricular node (NAV), upper bundle of His (UH), left and right bundle branch (RBB and LBB), and both ventricles (RV and LV)] and cycle between four electrical activation state. All these automata electrical activities are synthesized in an electrocardiogram (ECG). The LA and RA and ventricle are simulated by four elastance-based cardiac cavities. The cavities are connected by a complete circulatory model that integrates the pulmonary and systemic arteries, capillaries, and veins. Lastly four detailed model of heart valves (mitral, aortic, tricuspid, and pulmonary) were integrated.²⁰

Patient-specific ventricular pressure simulation

The set of model parameters, that was identified, was selected from the sensitivity analysis similar to the one performed in Owashi et al.¹³ Other model parameter values were defined from previous publications: ventricular and circulatory parameters were taken from Refs.,^13,19,22,23 heart valve parameters were adapted from Ref.20 and cardiac electrical conduction system fromRefs.^14,18 For each patient LV pressure $Plvmodel$ was simulated using the best set of parameters found during the parameter identification process (Supplementary data online, Appendix S1).

Template-based approach

As suggested by Russel et al.,⁵ valvular timings (MVC, AVO, AVC, and MVO) obtained from TTE may be used to estimate a normalized, patient-specific LV pressure curve. A predefined LV pressure curve template, calculated from the average of observed data in previous works of the group, is temporally adjusted and scaled in amplitude so as to fit the observed valvular timings and non-invasive systolic pressure value of a given patient. Mean aortic valve pressure gradient, estimated with echocardiography, was added to the instantaneous systolic pressure value to scale the normalized AS patient-specific LV pressure curve.¹² This method leads to a template-based estimate of a patient-specific LV pressure curve $Plvtemplate$⁠, which was directly extracted from the echocardiography workstation (EchoPAC version 202, General Electric Healthcare, Horten, Norway). The method is summarized in the top right part of Figure 2.

Figure 2

Myocardial work evaluation from model-based approach and experimental invasive measure.

Open in new tabDownload slide

Estimation of MW

The MW indices were calculated from experimental and simulated LV pressure, in order to obtain:

• model-based indices: GCW^model, GWW^model, GWE^model, GPW^model, GNW^model, and GWI^model;

• template-based indices: GCW^template, GWW^template, GWE^template, GPW^template, GNW^template, and GWI^template;

• experimental indices: GCW^exp, GWW^exp, GWE^exp, GPW^exp, GNW^exp, and GWI^exp.

Evaluation of the difference between estimated and measured data

Comparison of the pressure waveforms

Where $Plvexpk$ and $Plvestimatedk$ are, respectively, the experimental and the model or template-based estimated pressures at the sample k. Then a regression line was performed to access the mean correlation coefficient (r²) slope and intercept between the experimental and the two estimated methods.

Comparison of the MWs indices

Work indices (GCW, GWW, GWE, GPW, GNW, and GWI) were evaluated for each patient with the invasive pressure curve and the two estimated methods. Model- and template-based indices were compared with invasive indices using linear regression method and Bland–Altman (BA) analysis.^24,25 BA plots represents the average of the invasive and estimated variables in x-axis and the difference in the y-axes for all the patients. The mean difference and the 1.96 times the standard deviation are used to assess visually the good agreement.

Results

Patient characteristics

The baseline characteristics of the population are depicted in Table 1. The continuous variables are presented as mean ± standard deviation in the case of normal distribution, as median (inter-quartile range) in the non-normal distribution case, categorical variable as absolute frequencies and percentage. The population had a mean age of 82 years. The majority of patients was males (57%), with 58% of NYHA class I–II and 42% of NYHA class III–IV. All the patients suffering from severe (93%) or moderate AS with a mean AVA equal to 0.77 cm². The LV pressure and work indices extracted from invasive measurement are summarized in Table 2. The overall population presents a mean GWW higher than normal²⁶ (459 mmHg.%) and a mean GWE reduce (83%).

Comparison of LV pressure waveforms

Figure 3 presents the comparison between model-based (⁠$Plvmodel$⁠) and invasive (⁠$Plvexp$⁠) pressures obtained for the 67 AS patients. The mean correlation coefficient (r²) was equal to 0.81 (min: 0.23; max: 0.99). Mean slope and intercept of the regression line between the simulated and the measured pressure data were 0.94 (min: 0.49, max: 1.27) and −8.30 mmHg (min:−42.4, max: 21.9), respectively. RMSE was equal to 33.9 mmHg (min: 9.15, max: 90.4). Similarly, a comparison was performed between template-based estimation (⁠$Plvtemplate$⁠) and $Plvexp$ (Figure 4), mean RMSE was equal to 40.4 mmHg (min: 14.0, max: 89.2), mean r² is 0.72 (min: 0.25, max: 0.99), mean slope and mean intercept to 0.84 (min: 0.45, max: 1.21) and 23.8 (min: 5.87, max: 64.1), respectively. Despite results are slightly better for the model-based LV pressure estimation, the difference is not significant to conclude for a superior method.

Figure 3

Model-based LV pressure of the 67 AS patients: (i) experimental (black) and (ii) simulated (blue) curves.

Open in new tabDownload slide

Figure 4

Template-based LV pressure of the 67 AS patients: (i) experimental (black) and (ii) estimated (red) curves.

Open in new tabDownload slide

Comparison of MWs indices

Model-based MW

Scatter and BA plots for GCW, GWW, GWE, GPW, GNW, and GWI indices are presented in Figure 5. Concerning constructive work, slope and intercept of the regression line between estimations and measures were 0.79 and 251 mmHg.%, and r² = 0.81. In BA analysis, the mean bias of estimation is –251 mmHg.%. For wasted work, slope and intercept of the regression line between estimations and measures are 0.84 and –39.3 mmHg.% and r² = 0.91. In BA analysis, the mean bias of estimation is −32.0 mmHg.%. For work efficiency, slope and intercept of the regression line between estimations and measures are 1.00 and –0.003 and r² = 0.92. In BA analysis, the mean bias of estimation was −0.007. For GCW, GWW, and GWI the slope and intercept were 0.74 and 327 mmHg.%, 0.83 and 59.6 mmHg.%, 0.77 and 148 mmHg.%, r² were 0.76, 0.80, and 0.77 and the mean bias were –214 mmHg.%, –70.0 mmHg.%, and –144 mmHg.%, respectively. The negative mean bias observed on all the BA could be explained by an under-estimation of MW indices due to a slight advance observed in LV estimated pressure curves in most of the patients with this method.

Figure 5

Result of global work indices comparison, on all patients for model-based method. Scatter plots and Bland–Altman analysis of: (A) Global Constructive Work (GCW), (B) Global Wasted Work (GWW), (C) Global Work Efficiency (GWE), (D) Global Positive Work (GPW), (E) Global Negative Work (GNW), and (F) Global Work Index (GWI).

Open in new tabDownload slide

Template-based MW

Figure 6 presents the comparison between template-based and invasive indices. Despite an overestimation of all the indices, the quality of the result is similar with a good correlation coefficient. For GCW, GWW, GWE, GPW, GNW, and GWI, slope and intercept of the regression line between estimations and measures were 0.86 and 413 mmHg.%, 0.90 and 103 mmHg.%, 0.89 and 0.08, 0.71 and 576 mmHg.%, 0.88 and 251 mmHg.%, 0.69 and 216 mmHg.% with r² = 0.66, r² = 0.93, r² = 0.93, r² = 0.60, r² = 0.82, and r² = 0.72, respectively. In BA analyses the mean bias were 76.8 mmHg.%, 57.4 mmHg.%, –0.013, –19.2 mmHg.%, 156 mmHg.%, and –175 mmHg.%, respectively, for the six indices. The bias, here, could be explained by larger pattern of the LV pressure curve in some patients.

Figure 6

Result of global work indices comparison, on all patients for template-based method. Scatter plots and Bland–Altman analysis of: (A) Global Constructive Work (GCW), (B) Global Wasted Work (GWW), (C) Global Work Efficiency (GWE), (D) Global Positive Work (GPW), (E) Global Negative Work (GNG), and (F) Global Work Index (GWI).

Open in new tabDownload slide

In order to propose another error computation and better understand the results we also calculate for each patient and each MW indices the relative error (⁠$Xexp-XestimatedXexpfor X∈$ {GCW, GWW, GWE, GPW, GNW, GW1}). These results are gathered with the regression line summary in Table 3 for the two methods. We can notice that GCW, GWW, and GWE, where the bias is lower, have reasonable relative error (in %) with 14.77%, 16.51%, and 3.10% for the model-based method and 18.38%, 26.70%, and 2.97% for the template-base method, respectively, for these three indices.

Discussion

A model-based and template-based method were evaluated against invasive haemodynamic assessment of LV-pressure in a prospective cohort and results shown the validity of the estimations made in patients with an AS, combining the mean pressure gradient to the software currently commercially available. MW indices can thus be easily applied in routine clinical practice.

Estimation of LV pressure and MW indices

Concerning the evaluation of LV pressure, both methods show a good agreement between estimated and measured pressure waveforms. To our knowledge, our study is the first to provide a quantitative comparison between two estimated LV pressures and invasively measured curves in the context of AS on such a database. Moreover, myocardial indices calculated with the two estimation methods were compared with indices calculated with invasive pressures.

Model-based method allows for the in silico assessment of MW indices, while integrating physiological knowledge. This method has the advantage of requiring only AVA, pressure gradient evaluated in echocardiography, systolic and diastolic pressure values. The computational model directly integrates a representation of the pathophysiology of the aortic valves and takes into account characteristics associated with the subject and pathology.

Compared with model-based approach, template-based estimations require additional information related to aortic and MVO and closure, which should be manually identified on apical 3-chamber view and pulsed wave Doppler recordings. As a consequence, evaluations of valve timings could be cumbersome. Despite the manual evaluation of valvular events, the template-based method appears to be more appropriate in a clinical context. In fact, LV pressure and work indices could be directly extracted from the echocardiography workstation, whereas the model-based method implies an off-line procedure associated with a computational cost. Although template-based method could be privileged in clinical practice, model-based approach could be interesting for the evaluation of retrospective databases that do not integrate valve timings.

Work estimation

Despite the imprecise evaluation of LV pressure in both cases, the estimation of LV work indices strongly correlates with invasive measurements.¹¹ Model-based and template-based approaches appear as accurate methods for the estimation of MW in AS. This good correlation of all the works indices despite the imperfect estimation of LV pressure curves could be explained by different points. First, the temporal integration during the work computation induces a smoothing of the error between experimental and estimated work in both methods. Moreover, the computation of the indices uses only the pressure curve between AVO and AVC. This issue of using LV pressure estimation in order to analyse the MW could be avoid by using other indices based only on strain curves.²⁷

Myocardial function for AS patients

Current guidelines recommend surgical AVR in patients with Severe AS who have symptoms, or those who have reduced LVEF. The LVEF considered up to now was 50% but recent papers clearly showed that already for LVEF reaching 55–60%, patient prognosis is already dismal.^28,29 The severity of AS is not assessed merely by gradient and valve area but also resides in the interplay between increased LV-after-load of a stenotic valve and its deleterious effects on the myocardium. In a subpopulation of patients with long-standing AS that does not improve after intervention, with increased morbidity and mortality, adverse and irreversible LV-remodelling has often been implicated.³⁰ Prior meta-analysis revealed that asymptomatic severe AS patients who were treated with a watchful-waiting strategy had a 3.5-fold higher rate of all-cause mortality at 4 years, compared with those who underwent early AVR.³ Also, Taniguchi et al.³¹ demonstrated in a propensity score-matched analysis that patients treated with the initial AVR strategy had a lower risk of all-cause death and heart failure requiring hospitalization, than patients treated with a conservative strategy. Several studies underscore the relevance of a precise assessment of the myocardial consequences of the severe AS. Load is a key factor that impacts parameters quantifying LV systolic function. MW provides a unique opportunity to assess with much less load dependant, LV systolic function in AS patients.^32,33 The classic ‘pressure–volume’ loop, from invasive haemodynamics, has formed the basis of our understanding of the contributions of preload, after-load, and contractility to LV systolic function. The ‘area’ within this loop is referred to as LV stroke work and was the first way to conceptualized MW. It was followed by ‘pressure–strain’ loop and the MW indices that offer a complementary picture of LV systolic function. Also, Jain et al.³⁴ underline that LV function do not fully recover in days and months following transcatheter aortic valve replacement (TAVR). By comparing this index pre- and post-TAVR, they demonstrated that GLS improved as MW reduced in patients treated with TAVR for severe AS. Strain indices and MW appear particularly promising, providing a sensitive evaluation of LV function that could guide for potential earlier-TAVR for pauci-symptomatic patients. One limitation of this paper is to treat almost only patient with severe AS. Still large randomized trials are needed for confirming the value of echo-parameter and for demonstrate that currently, we might propose valve replacement at a late timing according to the heart consequences of the chronic increase in after-load remated to the AS.

Conclusion

The two non-invasive methods of LV pressure estimation and the work indices computation correlate with invasive measurements and computations for AS patients. Although the model-based approach requires less information and is associated with slightly better performances, the implementation of template-based method is easier and seems more appropriate in a clinical practice.

Supplementary data

Supplementary data are available at European Heart Journal - Cardiovascular Imaging online.

References

Author notes