Impact of aortic valve stenosis on myocardial deformation in different left ventricular levels: A three‐dimensional speckle tracking echocardiography study

Global systolic left ventricular (LV) myocardial function progressively declines as degenerative aortic valve stenosis (AS) progresses. Whether this results in uniformly distributed deformation changes from base to apex has not been investigated.


INTRODUCTION
Aortic valve stenosis (AS) progression leads to chronic left ventricular (LV) pressure overload.Initially, the LV adapts with changes in LV geometry and hypertrophy to maintain wall stress and cardiac output. 1,2[5] Myocardial strain is related to the extent of myocardial fibrosis, 6 and global longitudinal strain (GLS) assessed by two-dimensional (2D) ][8][9][10][11] Nevertheless, in AS, changes in longitudinal function may occur in different LV levels that are not captured by looking at the LV as a whole. 7Reduced basal longitudinal strain (LS) by 2D STE have been shown to be a more sensitive marker than GLS in predicting outcome in AS. 12 However, LV deformation occurs not only in the longitudinal direction, but in three dimensions, and is dependent on LV twist and rotation for efficient cardiac contraction and relaxation. 13,146][17] 3D STE, on the other hand, has better accuracy, reproducibility, and may overcome the limitations of 2D STE. 18In patients with AS, a few smaller studies using 2D and 3D echocardiography have shown increased apical rotation in severe AS. [19][20][21] However, there is still limited knowledge from 3D STE on how AS severity impacts myocardial deformation in different LV levels.Accordingly, the present study aims to explore the impact of AS severity on strain and twist measurements in different levels of the LV using 3D STE.

Study population
This cross-sectional study recruited patients from the outpatient clinic at Haukeland University Hospital, Bergen, Norway, who underwent serial echocardiographic examinations to follow progression of degenerative AS.Exclusion criteria were known coronary artery disease, previous cardiothoracic surgery, TAVR, coexisting aortic regurgitation, or other known valve diseases that could affect deformation values in any direction, based on hospital medical records.Patients with known atrial fibrillation were excluded based on the requirement of a regular heart rhythm to use 3D stitched images without artifacts.In addition, patients with known image quality too poor to be analyzed were excluded retrospectively.
Of 120 patients included, 7 were excluded because of irregular heart rate, 1 because of previously unknown myocardial infarction, and 27 patients due to poor image quality or volume stitching artifacts which precluded 3D STE analysis.This resulted in 85 patients eligible for the present analysis.
The study was approved by the Norwegian regional committee for medical health and research ethics (2014/1895/REK Nord).
All patients gave written informed consent.Self-reported information on medical history, cardiovascular risk factors and medication treatment was collected on a standardized questionnaire, and data were quality assured by study personnel.A 12-lead electrocardiogram and brachial blood pressure measurement both before and after the echocardiogram were undertaken in all patients by a study nurse.

Echocardiography
A standardized imaging protocol for 2D and 3D echocardiography using a Vivid E9 scanner with M5S probe (GE Vingmed Ultrasound AS, Horten, Norway) for 2D imaging, and 4 V probe for 3D imaging (GE Vingmed Ultrasound AS, Horten, Norway) was applied in all patients.
Volume-stitched 3D images were obtained using six beats with a target volume rate of 35 volumes per second (VPS) using harmonic imaging, as these were the optimal settings for deformation measurements found in previous in-vitro studies. 22,23The sector width was adjusted as narrowly as possible, while still covering the entire LV, optimized for post processing STE analysis.Patients were instructed to hold their breath during recording of two, four, and six heart cycles.To ensure quality, multiple 3D recordings were acquired.

Echocardiographic measurements
Conventional 2D echocardiographic images were analyzed using ImageArena (TomTec Imaging Systems, GmbH, Unterschleissheim, Germany).LV quantification was performed following the joint European Association of Cardiovascular Imaging (EACVI) and American Society of Echocardiography (ASE) guidelines. 24AS severity was assessed by peak jet velocity, mean aortic gradient, and aortic valve area using the continuity equation. 25,26EF was estimated using bi-plane method of disks.Filling pressure was estimated from ASE/EACVI diastolic function guidelines using the ratio of peak early transmitral blood velocity to average early mitral annular velocity (E/e'). 27All 2D echocardiograms were initially read by a single investigator (E.E) and later proofread by an experienced reader (E.G.).
3D STE data were analyzed using EchoPAC v202 (GE Vingmed Ultrasound AS, Horten, Norway).The best 3D recordings were selected using multi-slice short axis view, to ensure optimal image quality and to avoid images with any stitching artifacts.STE was analyzed by the software's 4D Auto LVQ-tool (Figure 1A & 1B).For this process, the LV in the three apical views from 3D were manually aligned with the axial axis.The apex and mid base were manually defined at both end-diastole (ED) and end-systole (ES) before the software automatically tracked the LV lumen.Any tracking inaccuracies were manually corrected.Any segment with tracking errors or artifacts was rejected.
If the software was unable to track more than three segments, the

Data analysis
3D STE analysis was performed in a standard LV model with 17 segments (Figure 2A).This model consists of six basal, six mid, and four apical segments as well as one segment for the apex. 24The EchoPAC software also provided 3D STE-derived rotation curves for the basal, mid, apical, and apex levels of the LV relative to the ultrasound probe (Figure 2B).Twist, which is defined as the difference in rotation between two different planes, was provided by the software, and found to be identical with basal rotation subtracted from apex rotation (Figure 3A).Torsion is defined as twist divided by the distance between the two planes, and was provided for the basal, mid, and apical LV levels.
Average 3D strain in longitudinal, circumferential, and radial direction as well as area strain was calculated for each LV level using a custom written Python program running Python 3 (Python Software Foundation, Wilmington, Delaware, USA).The global twist provided from the EchoPAC software was identical to The Python program also calculated twist for the six possible combinations [basal-apex (Figure 3A), basal-apical (Figure 3B), basal-mid (Figure 3C), mid-apex (Figure 3D), mid-apical (Figure 3E), and apical-apex (Figure 3F)].Apical-basal longitudinal strain ratio (ABr) was calculated by dividing the mean LS of the four apical segments by the mean six basal segments.
Negative peak values for longitudinal, circumferential, area strain, and positive peak values for radial strain were recorded for both regional (Figure 2B) and global (Figure 2C) curves.Positive peak rotation values were registered for the apex, apical and mid-levels, while negative peak rotation values were registered in the basal level.Peak negative twist rate was registered as peak untwist.

Statistics
Statistical analysis was conducted using R version 3.5.1 (The R Foundation for Statistical Computing, Vienna, Austria) and SPSS version 25 (SPSS Inc., Chicago, Illinois, USA).
The population was divided into AS severity classes based on peak aortic jet velocity: mild AS (< 3.0 m/s), moderate AS (3.0 -3.9 m/s), and severe AS (≥ 4.0 m/s). 25,26Results are presented as mean ± standard deviation.One-way analysis of variance (ANOVA) with Bonferroni post hoc test was used to show any difference between the groups for continuous variables.
Comparison of strain, rotation and torsion between different myocardial levels, was performed using mixed two-way ANOVA, with AS severity as group factor (P g ) and LV level as within-factor (P w ).Post hoc Bonferroni contrasts were used when appropriate.
The Greenhouse-Geisser adjustment of freedom was used for evaluation of the main effect when Mauchly's test of sphericity was violated.
Uni-and multivariable linear regression was used to identify any covariables.3D STE and LS at different levels were used as dependent variables while relative wall thickness (RWT), Peak jet velocity, E/e' , LV mass, sex, symptoms, and body mass index (BMI) were used as independent variables.For LS, multivariable linear regression using a stepwise procedure was first used to identify covariates, and then systolic BP and sex were forced into all models.Results were reported as standardized beta coefficients.
Intra-and interobserver variability was assessed by re-analyzing 20 randomly selected patients in a blinded manner and reported as intraclass correlation coefficient (ICC) and mean difference between the results.

Patient characteristics
The total study cohort included 32 patients with mild, 31 with moderate, and 22 with severe AS (Table 1).While most patients were  1).

Longitudinal strain
GLS was significantly lower in severe AS patients compared to mild and moderate AS (p = .002)(Table 2 and Supplementary figure 1).Basal and mid LS followed a parallel pattern with lower values in the severe than mild AS group (p < .001and p = .004for basal and mid, respectively) (Table 3).Apical LS was lower in the severe than moderate AS group (p = .027),while mild AS did not differ significantly from either (Figure 4A).

Circumferential strain
Global circumferential strain (GCS) showed no difference between the groups of AS severity.Circumferential level strain was highest in the apical level in all AS severity groups, but no difference was found between the AS severity groups.

Area strain
Area level strain was higher in the apical than in the basal level.Both mid and basal levels showed lower area strain for severe than mild AS.

Radial strain
Radial strain was higher in the apical than basal level for the moderate and severe AS groups and was lower in severe than mild AS for mid and basal level.

Rotation
Peak rotation, presented in Table 3, showed a gradient in rotation from base towards apex with positive rotation (counterclockwise) at apex level and negative rotation (clockwise) at the basal level during systole, when observed from the apex.There was a significant difference between each level except between the apex and apical level, which only differed in the moderate AS group.Between the groups, basal rotation was significantly higher in severe AS (Figure 4B and Supplementary figure 2).

Twist
Peak twist increased with AS severity for apical-basal and midbasal twist (Table 2 and Supplementary figure 3).Peak untwist was higher in patients with severe AS for Apical-Basal and Apical-Mid untwist.

Torsion
Although global torsion showed borderline significance between the severity groups, there was no difference in torsion by severity for different levels (Table 3).

Covariate analysis
Using multivariable linear regression analysis in the total study population, partially different independent covariates of 3D STE GLS and LS at basal, mid, and apical level were identified ( particularly associated with male sex, higher BMI, and higher peak aortic jet velocity.Furthermore, lower basal LS was associated with higher filling pressure (E/e') and higher LV mass.Lower mid LS was independently associated with higher RWT and presence of symptoms.Lower apical LS was associated with male sex and higher systolic blood pressure (all p < .05).Increased apical-basal twist was associated with higher RWT, higher E/e' , higher peak jet velocity, higher BMI, lower GCS, but not with symptoms (Table 5).

Reproducibility
Inter-and intraobserver variability was excellent for 3D STE data.
Reproducibility for 3D strain, rotation, and torsion is presented in Supplementary table 1.

DISCUSSION
The present cross-sectional study using 3D STE gives novel information on the association of AS severity and other clinically important confounders with myocardial deformation at different LV levels.As demonstrated, increasing AS severity is associated with a reduction in LS, particularly in the basal and mid LV level.Other regional myocardial deformation markers like circumferential, radial, and area strain as well as torsion, showed no significant difference.Using 3D STE, apical LS may seem to be preserved in patients with both mild and moderate AS, while patients with severe AS have reduced apical LS.This hints to a possible compensatory role for apical LS in preserving GLS in mild and moderate AS.Furthermore, we found that apical-basal twist was higher in patients with severe AS compared to mild and moderate AS.
Our findings expand results from previous studies by 2D STE, demonstrating GLS to be reduced with increasing severity, 7,8 and lower basal LS in symptomatic AS 28 by revealing that regional and global changes in LV mechanics in AS are not only related to the severity of AS, but also to LV remodeling and presence of cardiovascular risk factors in the individual patient.
Schueler et al. were the first to explore the feasibility and usefulness of 3D strain for the determination of changes in LV mechanics after transcatheter aortic valve implantation (TAVR). 29In their study, 44 elderly patients underwent 2D and 3D LV deformation imaging before, and 6 months after TAVR.The authors showed that 3D LS significantly increased 6 months after TAVR, while no significant changes were observed in twist and rotation, or 2D strain.Of note, 3D strain values at basal and apical LV levels were not reported.
Several studies have shown that rotation and twist is higher in patients with more severe AS. 19,20,30 Using 2D STE, Holmes et al. found increased apical rotation to be associated with poor survival in severe AS, suggesting apical rotation as a compensatory mechanism to preserve cardiac output in severe LV outflow obstruction. 19In the present study, only basal rotation was higher in the severe AS group.This could possibly be explained by the nature of a twisting object, where the middle part has close to stationary rotation, if the two ends rotate the same amount in opposite directions.However, the apex rotation showed much wider dispersion than the other levels of rotation in the present study, reflecting 3D STE measurement limitations involving the apex.Our findings of increased basal rotation contrasts with a smaller 2D STE study by van Dalen et al., 30 which reported similar basal rotation in 24 normal controls and 48 patients with AS.The same study also found significantly higher apical rotation and twist in patients with AS than control subjects.
The present study compared six different combinations of twist measurement.Ideally twist would be measured from apex to the basal level.Because of the apex rotation dispersion, twist involving apex also showed much wider dispersion than for other twist combinations.Twist measured one level down, from apical to basal level, gave much less dispersion.Comparison of apical-basal twist between the groups also showed a significant increase in twist for the severe AS group, while none of the twist combinations involving the apex showed any significant difference between the groups.Although not found in this study, a potential loss of apical function in severe AS could also affect apical rotation, which has been reported in symptomatic AS patients. 31LV rotation, however, may not be considered a strict deformation parameter, as it is relative to the observer, namely, the probe.LV twist and torsion, on the other hand, can be considered as true deformation parameters, as their relative measurements are both within the myocardium.
Interestingly, diverse clinical and echocardiographic covariates of 3D STE LS were identified at different levels, expanding previous knowledge from 2D STE studies assessing covariates of GLS in AS.Peak jet velocity and LV mass were identified as determinants of GLS but also of basal LS, and RWT as determinant of mid-level LS. 7,8,32 In a prospective study of 220 patients with severe asymptomatic AS, higher LV mass index was associated with impaired GLS. 33In that study, neither GLS nor LV mass index differed between groups of patients that developed symptoms during follow-up.In the present cross-sectional 3D study, presence of symptoms was associated with reduced LS at the mid-level.In addition, male sex was a major determinant of GLS and apical LS, as previously reported from a 2D study. 7In contrast to the findings of Ng et al. we found no association between age and GLS or LS at any level. 8Higher RWT was associated with lower mid LS, but not with lower GLS, as previous reported in a 2D STE study by Dahl et al. 32 In addition, higher body mass index was independently associated with lower 3D GLS, expanding findings in a 2D STE study in 44 patients with severe AS that found obesity to be associated with lower GLS and with more endomyocardial fibrosis in endocardial biopsies. 34

Clinical implications
As a clinical tool, 3D STE strain, twist and other parameters related to LV mechanics have the potential to identify subclinical dysfunction as well as monitor disease progression and response to treatment.In a recent survey conducted by the EACVI, 35 it was shown that > 90% of European echo labs had access to 3D echocardiography, but the majority of centers reserved the technique for selected cases and specific indications.The most frequent use was assessment of LV dimensions and function before cardiac device implantation or surgery for valvular heart disease, or as screening for cardiotoxicity from chemotherapy.
AS is the most common valvular heart diseases requiring valve replacement or intervention and the incidence is increasing.The optimal timing of intervention for asymptomatic patients is challenging.Hence, the assessment of LV function by 3D LV deformation has potential as a valuable supplement to the routine follow-up of AS patients who have normal conventional EF and are apparently asymptomatic.The present study suggests that basal LS may be important as an early marker of LV deterioration in patients with AS and EF > 50%.

Limitations
One of the main limitations of 3D STE compared to 2D is its lower temporal resolution.It can be argued that the extra added dimension reduces the need for a high temporal resolution due to the ability to track speckles in all directions and follow the speckles in 3D space. 15,22,36However, if the volume rate becomes too low, measurements and peaks may be underestimated.For the same reason, the usefulness of time derivative on 3D STE deformation to produce strain rate and twist rate is limited.Assessment of time to untwist was not included in the present analysis for this reason.This study used a volume rate of 36 ± 6, which was considered an optimal volume rate for 3D STE in our previous in vitro studies, 22,23 and gave good reproducibility of 3D strain and twist against sonomicrometry as a gold standard.
The low temporal resolution makes measurements dependent on curve derivation, such as strain rate and untwist rate, unreliable, and peak values will be inaccurate.
Of the 120 patients initially enrolled, only 70% were included in the final study, mostly because of poor 3D STE tracking.This was mainly due to poor acoustic visibility and stitching artifacts, which may limit the use of 3D STE in clinical practice.
The present study was not designed to compare 3D STE versus 2D STE, but rather to explore 3D STE findings.Furthermore, the study did not include a healthy control group without AS.
The patient population in this study was not screened for transthyretin cardiac amyloidosis (ATTR-CM).Studies in the later years have shown ATTR-CM to be much more abundant than previously anticipated, especially in the AS population, where some studies have shown ATTR-CM to be prevalent in up to 16 %, 37 especially in those with the subgroup of low-flow, low-gradient AS. 38 As ATTR-CM is known to affect basal LS with apical sparing, 39 it cannot be excluded that our findings of reduced basal strain with preserved apical strain may have been confounded by ATTR-CM.
Although few studies have explored level-specific deformation in AS by 3D STE, high apical to basal LS ratio has been associated with poorer outcome in two previous studies using 2D STE. 40,41ter-vendor difference for 3D strain measuring has been reported, 42 and our results should therefore not be extrapolated to other vendors.Since twist planes may be placed at different levels of the LV by different 3D STE algorithms, twist measurements are particularly vulnerable to inter-vendor differences.
Of clinical importance, the presence of symptoms in our cohort was based on patient self-reports, and systematic exercise testing was not included in this protocol.Lack of physical activity, especially in the aging group of patients, could mask apparent symptoms.
Our participants did not undergo cardiac magnetic resonance (CMR).Earlier CMR studies using myocardial tagging and feature tracking imaging have demonstrated normal values from deformation measurements in different myocardial levels and myocardial layers. 43,44However, the use of CMR is more resource demanding and less available compared to echocardiography.

CONCLUSION
This study shows that 3D STE can reveal regional and global changes in LV mechanics related to the severity of AS, LV remodeling, and cardiovascular risk factors.Patients with severe AS had lower GLS and higher apical-basal twist than those with mild and moderate AS.
When looking at regional deformation of LV levels, deformation changes seem to be more prevalent in the basal LV levels.Apical level LS tends to have a biphasic pattern with higher ABr for moderate than mild AS, while apical LS is lower in severe than moderate AS.Generally, apex measurements seem to be less accurate, with wider dispersion than the other LV levels.These echocardiographic findings may add to the conventional methods applied today in the work-up of patients with AS.
views of the LV.Example of 3D echocardiographic image in patients with mild (A) and severe (B) AS with corresponding LS curves and bull's-eye plot.
Patient and 2D echocardiographic characteristics.
TA B L E 1 Abbreviations: AS, aortic valve stenosis; ANOVA, analysis of variance; BMI, body mass index; BSA, body surface area; BP, blood pressure; E/e' , filling pressure; IVSd, septum thickness at end diastole; LVEF, left ventricle ejection fraction; LVIDd, left ventricle internal diameter at end diastole; PWDd, posterior wall thickness at end diastole; RWT, relative wall thickness.a Dfference with mild AS. b Difference with moderate AS.asymptomatic, 13 reported dyspnea (New York Heart Association functional class II-III) in the questionnaire, 6 in the moderate AS, and 7 in the severe AS group.None reported angina pectoris or syncope.The E/e' ratio and RWT were higher in patients with severe AS (p = .017),while sex, age, BMI, and blood pressure did not differ (Table

Table 4
3D STE strain, rotation, and torsion at different LV levels.
). Lower GLS was TA B L E 3 w < .001,p g = .266,p i = .078 Longitudinal strain and LV rotation by severity and LV level.Longitudinal strain at apical, mid, and basal level (A); Left ventricular rotation at apex, apical, mid, and basal level (B).Vertical lines represent standard error of mean; * p < .05 vs. Mild AS group; **p < .05 vs. Moderate AS group.Univariate and multivariate association of longitudinal strain in different levels and GLS.Covariates of apical-basal twist in uni-and multivariable linear regression analyses.