Reducing Long‐Term Mortality Post Transcatheter Aortic Valve Replacement Requires Systemic Differentiation of Patient‐Specific Coronary Hemodynamics

Among a cohort of 201 patients undergoing TAVR during the study period, 31 patients with severe AS who received TAVR (Table: patients' characteristics) and underwent a follow‐up echo and computed tomography (CT) post‐TAVR were retrospectively selected from anonymized databases between 2020 and 2022 at the Hamilton General Hospital in Hamilton, Ontario, Canada. Patients' selections for post‐TAVR CT were done by the heart team blinded to the study objectives and was based on choosing patients with higher risks of future adverse events, excluding those with either medical (eg, substantial possibility of contrast nephropathy) or social (eg, unable to go or refused to take part) impediments to executing the follow‐up CT. In addition, those treated for surgical aortic bioprostheses degeneration (ie, valve‐in‐valve) and those with unsuccessful devices as per Valve Academic Research Consortium criteria were excluded.¹⁶ Other exclusion criteria were any patient data with missing required inputs for the computational modeling. We did not exclude any data based on image quality. All 31 patients had normal or mild CAD (<50% obstruction) interpreted with a >50% threshold to define obstructive lesions, per guidelines.¹⁷ All discrepancies were sorted out by communicating with the local investigators. The protocols were reviewed and approved by the Hamilton Integrated Research Ethics Board, and informed consents were collected from all participants. Measurements were obtained according to guidelines including American Heart Association, American College of Cardiology, and American Society of Echocardiography. Data were collected at both pre‐procedure and post‐procedure time points. Results were expressed as mean±SD.

Doppler echocardiography data included the measurements and reports that were collected at baseline and post‐procedure at different time points. Transthoracic echocardiography or transesophageal echocardiography was performed at 2 time points to collect the stenotic valve and transcatheter valve hemodynamics data: baseline (pre‐TAVR) and 1 year post‐TAVR. This research made use of the same date follow‐up echocardiogram data from the 31 patients who had undergone post‐procedure CT scans at the same time. Echocardiograms and reports were reviewed and analyzed in a blinded fashion by senior cardiologists using OsiriX imaging software (OsiriX version 8.0.2; Pixmeo, Bernex, Switzerland). The following metrics were collected for clinical assessment and computational model: aortic valve area, mean pressure gradient, maximum aortic valve velocity, ejection fraction, Doppler velocity index, forward left ventricular outflow tract stroke volume, left ventricle (LV) outflow tract velocity time integral, LV outflow tract diameter, cardiac cycle time, ejection time, aortic valve effective orifice area, mitral valve effective orifice area, ascending aorta cross‐sectional area, LV outflow tract area, grading of aortic and mitral valves regurgitation severity measured by Doppler echocardiography, as well as brachial systolic and diastolic pressures measured by a sphygmomanometer.

CT imaging was performed on a GE 320‐detector Revolution scanner (General Electric, Milwaukee, WI) or a Siemens 64‐detector dual‐energy scanner (Somatom Definition Flash; Siemens, Erlangen, Germany) with a retrospective gated acquisition protocol, with 0.625 mm slice thickness, 0.35‐second tube rotation, 0.24:1 pitch, 100 or 120 kV tube voltage based on patient size and iterative reconstruction at 30%. The image acquisitions were chosen at the best diastolic phase of the cardiac cycle (70%–75%). CT images of the patients at the baseline and follow‐up post‐TAVR were used to segment and reconstruct the 3‐dimensional (3D) geometries. Detailed information regarding geometry reconstruction is outlined in Data S1. The CT was performed at 2 time points at the same time of echo data collection: baseline (pre‐TAVR) and around 1 year post‐TAVR. The following parameters were collected from the CT images and reconstructed geometries: left main coronary artery (LMCA) average diameter, left anterior descending (LAD) coronary artery average diameter, left circumflex coronary artery (LCX) coronary artery average diameter, right coronary artery (RCA) average diameter, bifurcation angle, aorta angle, commissural misalignment, aortic root diameter, ascending aorta diameter, coronary height, and valve to coronary distance.¹⁸

The TAVR procedure was performed through the transfemoral access as per the common practice with Edwards SAPIEN valve system.¹⁹ In certain cases, the native aortic valve was predilated according to the operator's discretion. The choice of transcatheter heart valve size was based on the pre‐TAVR CT measurements of the effective aortic annular diameter and area as well as leaflet calcium volume.^{20, 21} The present research includes patients who underwent TAVR with balloon expandable Sapien 3 (Edwards Lifesciences Inc., Irvine, CA). The used valve sizes were 23, 26, and 29. Depending on the operator's discretion, balloon post‐dilatation was conducted in chosen circumstances.

Appropriate statistical tests were performed using Jamovi (v.1.8.1.0). Normality of each continuous variable was assessed using the Shapiro–Wilk test (Figure 7; Table). Continuous variables were expressed as mean ± SD, categorical variables were expressed as counts. Correlations, and comparisons between the variables was performed using Pearson's r, Spearman's rho. Paired samples t tests were performed on pre‐ and post‐TAVR variables using Mann–Whitney U or Wilcoxon W depending on normality. Statistical significance was considered when the P value was <0.05. Detailed information regarding normalization of relevant parameters is outlined in Data S1.

A computational fluid dynamics and lumped parameter modeling framework was developed to calculate 3D blood flow dynamics in all main coronary artery branches (LMCA, LAD, LCX, and RCA), aortic root, ascending aorta, sinus, and neo‐sinus regions for both pre‐ and post‐TAVR (Figure 1). We used computed tomography images to reconstruct the geometries. This framework is based on lumped parameter modeling and 3D FSI modeling as implemented in open‐source foam‐extend library and validated against in vivo Doppler echocardiography data as explained in Khodaei et al^{15, 40, 41} and Keshavarz‐Motamed et al.²⁴ Please refer to Figure S1 and Table S1 for all details about our numerical study.

As shown in Figure 4, on average, the peak systole flow increased 12.8% (1.25–1.41 mL/s) (W=129; P=0.242), 4.39% (1.14–1.19 mL/s) (W=159.5; P=0.694) and 25.67% (1.48–1.86 mL/s) (t=2.258; P<0.05) for LAD, LCX, and RCA, respectively, whereas the peak diastole flow decreased by 8.98% (1.67–1.52 mL/s) (W=244; P=0.084), 16.83% (1.01–0.84 mL/s) (t=2.193; P<0.05), and 22.73% (0.88–0.68 mL/s) (W=250; P=0.059) for LAD, LCX, and RCA, respectively. Looking at each patient individually (Figures 3^, 4), we found distinctive patterns among patients in terms of coronary flow improvements after TAVR.

We calculated the TAWSS over the diastole for all patients in pre‐ and post‐TAVR states. On average, TAWSS decreased for 84% of the patients for different coronary branches. As shown in Figures 5 and 7I, maximum TAWSS decreased 9.47% (0.95–0.86 pa) (t=2.3; P<0.05), 7.75% (1.16–1.07 pa) (W=343; P=0.06), 6.94% (2.16–2.01 pa) (t=1.7; P=0.1), 8.07% (1.61–1.48 pa) (t=1.7; P=0.1), and 6.28% (1.75–1.64 pa) (t=1.5; P=0.2) for bifurcation, LMCA, LAD, LCX, and RCA branches, respectively.

We also observed that the distribution of TAWSS and its maximum is very different for each patient and is highly dependent on the geometrical characterization of the coronary walls. For patient #26 as a sample (Figure 2B), the TAWSS ranges from 0.12 pa to 1.5 pa across the coronary arteries with considerable reduction (to a range of 0.1–1.2 pa) followed by TAVR.

Despite the significant decrease in aortic valve pressure gradient for all patients after TAVR (Figure 6), the LV workload did not reduce for 54.84% of the patients. As shown in Figure 6, although on average, the aortic valve mean pressure gradient dropped by 64% (46.1–16.6 mm Hg) (t=12.4; P<0.001) after TAVR, the LV workload burden was not removed (2.52% increase [1.59–1.63 J] [W=213; P=0.221] in LV workload) after TAVR. In addition, in 24 patients (77.42%) the LV workload remained above 1 J after TAVR, which is an indicator of overloaded ventricle.¹⁴

We examined the correlation between coronary flow and other hemodynamics and geometrical parameters (please see Data S1 for the comprehensive statistical analysis and correlation matrix). As shown in Figure 7A and 7B, we observed a strong correlation between LV workload and mean coronary flow for both pre‐ and post‐TAVR cases (r=0.75; P<0.001 pre‐TAVR and r=0.52; P<0.003 post‐TAVR). Moreover, LV workload was strongly correlated with pulse pressure (r=0.59; P<0.001) and systole blood pressure (r=0.65; P<0.001), as shown in Figures 7G and 7H. This is an interesting finding, as LV workload could be an effective noninvasive metric to predict the oxygen demand of myocardium after TAVR without a need to perform invasive coronary hemodynamic measurements.

Among different geometrical parameters, strong correlations were observed between mean coronary flow and valve to coronary distance, left coronary ostium height, and aortic root diameter with mean coronary flow. As shown in Figure 7E and 7F, mean coronary flow was strongly correlated with coronary ostium height (r=0.79; P<0.001) and aortic root diameter (r=0.58; P<0.001). Moreover, a strong correlation was observed between mean coronary flow and valve to left coronary distance (Figure 7C) (r=0.78; P<0.001). This correlation was less significant for valve to right coronary distance (Figure 7D) (r=0.5; P<0.05). In summary, our findings showed that the restricted gap between the expanded valve and coronary inlet might impede the coronary blood flow. We also found that patients with lower coronary ostium height are at higher risk of reduced coronary flow after TAVR.

Based on the preliminary correlation analysis with all patient‐specific variables against total mean coronary flow, we identified key parameters for multivariate regression as left coronary height, LV workload, aortic root diameter, valve‐to‐left‐coronary distance, and valve‐to‐right‐coronary distance. The overall regression for total mean coronary flow pre‐TAVR was significant (R²=0.67; F(3, 22)=14.6; P<0.001) (Table S2) with the only significant predictor being the LV workload. On the other hand, after TAVR, the fitted model that also included valve‐to‐coronary distances was found to be a better predictor for total mean coronary flow (R²=0.88; F(5, 20)=30; P<0.001) (Table S3) with the strongest predictors being the valve‐to‐left‐coronary distance (F=11.4; P<0.01), LV workload (F=11.4; P<0.01), and the aortic root diameter (F=4.5; P<0.05). The valve‐to‐right‐coronary distance along with the left coronary height did not significantly predict the total mean coronary flow.

Induced extravascular compressive forces caused by AS reduces the coronary perfusion pressure that is associated with reduced systemic arterial compliance.⁴⁵ After TAVR, immediate increase of pressure and flow downstream of aortic valve (ascending aorta) is expected as results of AS removal.⁴⁵ Our results confirm that the expected enhancement in pressure and coronary flow occurs during systole. However, the diastolic flow in coronaries might not improve or may even decrease for many patients. Such decrease in coronary blood flow is associated with reduced capacity to augment myocardial oxygenation, leading to LV dysfunction, increased apoptosis (which is linked to myocardial fibrosis and is an independent indicator of mortality), and sudden death.^{45, 46}

We observed that in addition to the global hemodynamics and aorta pressure, the geometrical details including the valve to aorta distance, valve to coronary distance, implantation depth, and aortic root shape drastically affect the local hemodynamics before reaching coronary ostium (Figure 7). Although the focus of recent studies has been on the risk assessment for coronary obstruction,⁴⁷ our findings suggest that if the distance between stent and coronary ostium is restricted (even if not fully obstructed) after TAVR, the coronary flow will be impeded. On the other hand, although the number of patients who need a second TAVR is increasing as the procedure expands toward younger patients, the true risk of coronary obstruction is not known yet.⁴⁸ This means the proper coronary filling mechanism will be more complicated in these populations with higher risk of mortality.

Another interesting finding of this study was that TAVR might reduce the wall shear stress for the coronary arteries during diastole due to the reduced blood flow. This makes the coronary arteries more susceptible to atherosclerosis, due to the low wall shear stress‐induced inflammatory activation of endothelium mainly at the inner bend of curved arteries, ostia of branches, and lateral walls of bifurcations.⁴⁹

The progression mechanism of AS is very similar to that of atherosclerosis in CAD, and there is a high cooccurrence of both diseases in the same patient.⁵⁰ AS and CAD share several important cellular mechanisms including lipid deposition, inflammatory cell infiltration, cytokine release, and calcification,^{51, 52} which explains why CAD is present in ≈50% of the patients with AS.⁴² Our study revealed that the reduction of TAWSS for majority of patients during diastole might increase the current existing risk of CAD through enlargement of plaque area, increased plaque eccentricity, and reduced vessel area.⁵³

Coronary artery bifurcation has a unique local flow and wall shear stress condition because of patient‐specific anatomy of main and side branches.⁵⁴ The unfavorable hemodynamic alterations at the branch points of bifurcation followed by reduced shear stress could promote endothelial cell dysfunction, which is attributed to the progression of atherosclerosis.⁵⁴ It has been established that there is a strong association between biomarkers for vascular disease and flow patterns.⁵⁵ Interestingly, our results showed that for all patients, regardless of flow alterations after TAVR, coronary bifurcation is exposed to a considerably lower wall shear stress region in comparison to right coronaries or other branches (which might contribute to plaque progression). Previous studies noticed that within the coronary vascular system, plaque was mainly found in areas of flow divergence (such as bifurcation) and speculated that distinctive wall shear stress profiles may be the cause of this plaque concentration.⁵⁶ This is a critical concern for TAVR patients as there is a high risk of restenosis for bifurcation lesions after percutaneous coronary intervention.

LV workload is an effective metric for LV functional assessment that is closely dependent to the coronary perfusion and the coronary flow.¹⁴ Increased LV workload promoted cardiac remodeling with an increase in myocardial oxygen demand, which makes myocardium vulnerable to ischemia.⁵⁷ AS increases the LV workload in order to compensate for the reduced ejected flow, and after TAVR, a reduced LV workload is expected.¹⁴ However, our results showed that the reduction in aortic valve pressure gradient post‐TAVR was not associated with reduced LV workload for a large number of patients. Indeed, either a single pathology or an amalgamation of pathologies following TAVR procedure could contribute to an increased LV workload, including worsening or emerging of mitral valve regurgitation, paravalvular leakage, hypertension, and heart failure as well as increased forward LV stroke volume. Among 31 patients in this study, the LV workload increased post‐TAVR for 17 patients. In 82% of these patients, worsening mitral valve regurgitation (from mild to moderate) and increased forward LV stroke volume or increased blood pressure (systole and diastole) or presence of paravalvular leakage were responsible for a rise in LV workload. As 1 example, in patient #4, presence of moderate mitral regurgitation and paravalvular leakage, increased systolic blood pressure, and increased forward stroke volume caused a rise in LV workload from 1.23 J (pre‐TAVR) to 2.23 J (post‐TAVR).

Another interesting finding of our study was that the coronary flow was correlated with LV workload for both pre‐ and post‐TAVR states. Although we have previously shown the usefulness of LV workload metric for patients undergoing TAVR,¹⁴ this study demonstrated the significance of the interplay between LV workload and coronary flow. Our findings are indeed in line with the critical role of effective coronary circulation for cardio protection, as the reduction in coronary blood flow will lead to myocardial infarction over time.⁵⁸

This study was performed on 31 patients who underwent TAVR in both pre‐ and post‐intervention states. Future studies must be conducted on a larger population of such patients to further confirm the clinical findings of this study. Another limitation of this study was that it focused only on the Edwards Sapien 3 balloon‐expandable valve and did not consider other valve types. This means that the investigation of design factors was limited to only 1 type of valve. In the future, research will be conducted to broaden the scope of the study by including a wider range of self‐expandable and balloon‐expandable valves. This will enable a more thorough examination of the effect of design factors on hemodynamic outcomes. Although we have reported all the relevant periprocedural complications, our study did not focus specifically on follow‐up clinical outcomes such as mortality or myocardial infarction. Future study should investigate whether the simulated coronary blood flow (considering different geometries⁹⁶) translates into impact on clinical outcomes such as mortality or other adverse events.