Preoperative Left Ventricular Global Longitudinal Strain Identifies Aortic Stenosis Patients with Improved Postoperative Recovery of Left Ventricular Geometry: A Prospective Cohort Study

Introduction The left ventricular ejection fraction (LVEF) is commonly used as a marker of aortic stenosis (AS) disease severity and to indicate surgical intervention. However, an LVEF <50% identifies mainly advanced disease. Hence, earlier detection of subclinical LV systolic dysfunction may improve clinical decision-making. The global longitudinal strain (GLS) can identify subclinical systolic dysfunction at earlier stages of AS progression even in the presence of preserved LVEF. To this end, we evaluated the preoperative prognostic significance of the LVGLS to identify patients who will undergo a more extensive postoperative LV reverse remodeling as a surrogate marker for clinical recovery. Methods We performed a prospective observational study based on detailed pre- and postoperative 2D transthoracic echocardiographic examinations, including strain analysis with speckle tracking. We screened 60 consecutive patients with severe AS and a preoperative LVEF ≥50% indicated for surgery; 39 patients met the study entry criteria and consented to their participation. Results The median age was 67 (range 30-79) years; 56.4% were female. At baseline, the GLS was 61.64±7.22%. Surgery led to an improvement in the GLS; the mean difference was 3.23% [95% CI=1.96 to 4.49%] during a median follow up time of 5 (interquartile range 4-6) months. The preoperative GLS correlated with the postoperative LV mass index (LVMI) r=0.526, P=0.001 and the intraventricular septal thickness in diastole (IVSd) r=0.462, P=0.003. Furthermore, patients with a normal GLS (≤-18.9%) at baseline experienced a better recovery of their LV morphology and systolic function during the postoperative course compared to those with an abnormal GLS (>-18.9%). The effect size, hedges g, was at least >0.75 for the LVMI, IVSd, intraventricular septal thickness in systole (IVSs), left ventricular posterior wall thickness in diastole (LVPWd) and LVEF, suggesting a clinically significant difference between subgroups at follow-up. Conclusion A normal preoperative left ventricular global longitudinal strain is associated with an improved left ventricular reverse remodeling and systolic function following surgery to resolve aortic stenosis.


INTRODUCTION
The assessment of left ventricular (LV) systolic function is crucial in the risk stratification of aortic stenosis (AS) patients [1] . Current guidelines recognise a ejection fraction (EF) <50% as a class I indication for aortic valve replacement (AVR) [2] . Nevertheless, clinical decision-making, according to the left ventricular ejection fraction (LVEF), has been questioned in recent years [3] . LVEF may remain normal for years, due to compensatory mechanisms, despite the occurrence of deep sub-clinical structural and functional myocardial changes that dictate disease progression, which can affect the clinical outcome [4] .
It is well established that patients with severe AS and significant LV systolic dysfunction can benefit from surgical aortic valve replacement (SAVR). Nevertheless, severe AS patients with LV systolic dysfunction are at higher risk to experience postoperative complications with an increased risk of intrahospital and longterm mortality in comparison to patients with preserved LV function.
The global longitudinal strain (GLS) is a more accurate marker of myocardial fibrosis compared to LVEF. In comparison to other indexes of LV systolic function, the parameters of LV longitudinal deformation are superior in detecting myocardial dysfunction and damage. Current evidence suggests that patients with severe AS have sub-clinical LV systolic dysfunction, resulting in distorted GLS values despite preserved LVEF [5] . We speculate that the normalisation of GLS values might signify a reversal of pathological myocardial changes and correlate with left ventricular reverse remodeling (LVRR). GLS can be routinely evaluated using 2D-speckle tracking echocardiography.
Consequently, GLS has the potential to be incorporated into clinical decision-making in patients with severe AS and to predict SAVR procedural success according to the postoperative LVRR.
This study evaluated the correlation between preoperative GLS and postoperative echocardiography parameters associated with LVRR. Furthermore, we analysed whether GLS can stratify patients who will experience more pronounced resolution of LV hypertrophy and improvement in systolic function following SAVR.

METHODS
We prospectively screened 82 consecutive patients indicated for SAVR at our clinic between November 2016 and June 2017. The STROBE study flow diagram describing the inclusion and exclusion criteria is shown in Figure 1. The final study cohort consisted of n=39 patients. All patients provided a signed informed consent form to participate in this observational study. The Zan Mitrev Clinic's Ethics Committee approved the study.
We performed a medical history evaluation, physical examination, biochemical analysis and electrocardiography (ECG) and a coronary angiogram for the possibility of existing CAD.
Echocardiographic examinations were performed on a Philips Epiq 7 Cardiology Ultrasound Machine; saved recordings were examined using the Philips IntelliSpace Cardiovascular Portal platform.
We performed the measurements according to the guidelines and recommendations [2] . EF was calculated by Simpson's method, mean and peak trans-aortic gradients were determined using the continuous wave Doppler method, and aortic valve area was calculated with the continuity equation [6] .
Tissue Doppler imaging was used to obtain peak systolic (S') and peak early diastolic (E') mitral annular velocities measured on the septal and lateral sides of the mitral annulus and after the calculation of E/E' [7] . Relative wall thickness (RWT) was calculated according to the following formula 2 × PWd/LVEDd: "double the thickness of the posterior wall divided by left ventricular enddiastolic dimension" [8] .
The LV mass was calculated according to the linear method using the American Society of Echocardiography formula: LV mass=0.8 × (1.04[(LVIDd + PWTd + SWTd) 3 − (LVIDd) 3 ) + 0.6 g [9] , Where: LVIDd = Left ventricular internal diameter at end-diastole PWTd = Posterior wall thickness in diastole SWTd = Septal wall thickness in diastole and indexed to the body surface area (m 2 /kg). Mitral annular plane of systolic excursion (MAPSE) was measured in millimetres with M-mode echocardiography at four different points (septal, lateral, anterior and inferior) in apical 4and 2-chamber views; a value ≥10 mm was considered within the reference range [10] .
High-quality ECG-gated images were obtained and recorded with frame rate >50 frames/s in long apical axis 4-and 2-chamber views; recordings were subsequently analysed using 2-D speckle tracking, where segmental strains were presented as a bull's-eye map. A GLS value of −18.9% was set as the cut-off for normal longitudinal strain as recommended by the vendor [9] .

Ethics Approval and Consent To Participate
The Zan Mitrev Clinic's ethics committee approved the clinical practice and treatment procedures described in this case series.

Consent For Publication
The Zan Mitrev Clinic's ethics committee approved the publication of clinical data under the condition of full anonymity.

Availability of Data and Material
All data generated or analysed during this study are included in this published article and its supplementary information files.

Surgical Technique
Surgical aortic valve repair proceeds according to the standard protocol for cardiopulmonary bypass (CPB) placing patients under mild hypothermic conditions (32°C). Myocardial protection is achieved through continuous retrograde and antegrade cardiac perfusion with warm blood cardioplegia.

Statistical Analysis
Categorical parameters were summarised as absolute numbers and percentages. Continuous data are shown as mean±SD or median+interquartile range (IQR).
Continuous variables were evaluated using the D' Agostino-Pearson normality test -independent parametric data were analysed using the Student's t-test, and non-parametric continuous variables were assessed using the Mann-Whitney test for independent comparisons. Comparisons of preoperative versus postoperative data were performed using a paired T-test or Wilcoxon signed-rank test for non-parametric data.
Fisher's exact test was applied to evaluate the association between categorical variables with the outcome. Regression analysis between preoperative GLS and echocardiographic markers were performed using Pearson's or Spearman's correlation testing. We adopted the hedges' g as a measure for the effect sizes due to unequal and small samples sizes.
Data were analysed with the statistical softwares Graphpad Prism, version 7.03 and Statsdirect, version 3.1.20.

Patient Characteristics and Procedure Overview
Baseline patient characteristics are presented in Table 1. All patients had symptomatic aortic valve stenosis with clinical indication for SAVR according to European Society of Cardiology guidelines: AS severity based on aortic valve area index (AVAI), mean pressure gradient and peak jet velocity. Hypertension was the most frequent comorbidity (84.6% of patients). In 9 cases (23.1%), SAVR was combined with coronary artery bypass grafting (CABG) surgery. No cases of operative mortality were observed during a follow-up period of 219 patient-months. No patients were lost to follow-up.

Brazilian Journal of Cardiovascular Surgery
The mean (± SD) GLS was −16.58±4.25%, and the LV fractional shortening was 36.95±9.92% before surgical intervention.
We observed LV hypertrophy in all patients. The mean (± SD) left ventricular mass index (LVMI) in males was 216.3±48.45 g/ m², of which 15 out of 17 had concentric hypertrophy. In female patients, the mean (± SD) LVMI was 150.3±47.39 g/m² with 19 out of 22 having concentric hypertrophy at baseline.
After a median follow-up time of 5 (IQR 4-6) months, several parameters associated with AS severity were improved (Supplemental Tables 1 to 4). SAVR resulted in notable improvements in overall LV dimensions and LV systolic function. Furthermore, a significant reduction of E/e' values and increase in average MAPSE (Supplemental Table 2) point to improved diastolic and systolic function after SAVR.
Finally, we observed a gradual normalisation of LV strains compared to baseline values, in particular, the GLS, which improved by 3.23% [1.96 to 4.49%], P=0.0001 (Supplemental Table 4; Figures 2 and 3).

Left Ventricular Global Longitudinal Strain as a pre-SAVR predictive marker for Left Ventricular Mass Regression
We determined the association between preoperative LVGLS and various other echocardiography parameters linked to LVRR at follow-up.
Next, we examined the prognostic significance of GLS. For this purpose, the cohort was stratified into two sub-groups according to the recommendations of the vendor Philips Epiq 7 for a GLS cut-off value of −18.9%; "normal" = GLS ≤−18.9% group (n=13) and the "abnormal"= GLS >−18.9% group, n=26. The sub-groups were equally matched for age, BMI and BSA (Supplemental Figure 3D).
At baseline, both sub-groups had similar LV morphology and systolic functions, except significant differences in GLS (Figures 3C to E). We observed a trend suggesting higher NYHA status in the abnormal GLS sub-group. However, this difference did not reach statistical significance. On the other hand, NYHA class was associated (r p =0.477; P=0.0023) with the extent of preoperative LV hypertrophy; patients with NYHA class 3 had a more pronounced LV hypertrophy compared to those with NYHA class 2, mean difference in LVMI was 59.71 g/m 2 (95% CI 17.21 to 102.4 g/m 2 ), P=0.0072 ( Figure 3F).
All patients successfully underwent SAVR and experienced an uncomplicated postoperative course. Patients who required concomitant CABG experienced a similar postoperative LVRR as those who underwent isolated SAVR (Supplemental Figures 1B to D) The clinical condition of the whole cohort improved at followup; notably, the "normal GLS" subgroup (≤−18.9%) had a more favourable LVMI compared to the "abnormal" subgroup (>−18.9%) ( Figure 3B). Furthermore, the normal GLS sub-group were more likely to recover a healthy LVMI compared to those with an abnormal GLS, OR 12.27 (95% CI 2.257 to 50.45), P=0.0021 ( Figure 3G). In conclusion, patients with a normal GLS experience a faster normalisation of LV dimensions after SAVR.

DISCUSSION
Aortic stenosis (AS) is the most common degenerative valvular heart disease; a result of immune-mediated calcification of the valve leaflets. Failure of compensatory mechanisms over time leads to clinical symptoms as a consequence of progressive fibrosis, impairment of myocardial contractility and reduced LVEF [1] . However, reduced LVEF (<50%) manifests in the late stages of the disease and has been associated with irreversible myocardial dysfunction and is an independent risk factor for sudden death in patients with severe AS.
To this end, more sensitive markers for the early stages of AS that overcome the limitations of LVEF-guided clinical decision-making are warranted [11] . Early detection of subclinical systolic dysfunction may optimise the timing of aortic valve intervention [12] . GLS is a promising and sensitive marker of the subclinical systolic dysfunction [13,14] .
In the present study, we assessed the prognostic significance of LVGLS in a cohort of severe AS patients with preserved LVEF. We demonstrate that 1) the baseline GLS value strongly correlates with several echocardiography parameters linked to LV hypertrophy and LV systolic functions. Moreover, 2) patients with a normal GLS (<−18.9%) experienced a more pronounced LVRR during the early postoperative period. In contrast, LVRR extent was significantly impaired in patients with an abnormal GLS (≥−18.9%) at baseline.
Notably, sub-groups were matched for age, gender and BMI; both groups were similar concerning the degree of AS severity, LV hypertrophy, LVEF and NYHA class at baseline. The main difference between sub-groups before SAVR was their sub-clinical LV systolic function as determined by speckle-tracking echocardiography strain analysis ( Figure 3C and Supplemental Figure 1A).
SAVR forms the cornerstone of AS management. It has been shown that the degree of LV hypertrophy and extent of LVRR after SAVR determines long-term clinical prognosis [15] .
Even though the evaluation of GLS can detect subclinical systolic dysfunction with high sensitivity [13,14] , strain analysis by speckle tracking echocardiography is currently not implemented in major clinical practice guidelines [2] . Nonetheless, our data suggest that the preoperative GLS has the potential to classify patients who will experience accelerated LVRR after SAVR. Thus, a normal GLS is associated with better postoperative recovery of normal LV geometry.
Our data align with recent reports that advocate the inclusion of LVGLS analysis to improve risk stratification of patients with severe AS, facilitate clinical decision-making and timing of aortic valve replacement [16] . Of importance, the LVGLS is independently associated with all-cause mortality in AS patients [17] .
Heterogeneity in GLS measurements among published clinical studies hampers the inclusion of the LVGLS analysis in clinical practice guidelines; the recorded measurements vary among different vendors because of proprietary differences in the software used to calculate deformation [18] . Reported reference values of GLS vary from −15.9% to −22.1% [19] . Nevertheless, our Finally, we evaluated the postoperative values of candidate echocardiography outcome measures associated with LVRR ( Table 3). The most prominent effects between the "normal" and "abnormal" subgroups at follow-up were observed for markers of LV geometry.    The results advocate the use of the LVGLS as an early and more sensitive marker to establish AS stenosis severity, facilitate indication for AVR and classify patients with a better postoperative prognosis following surgery.

Fig. 3 -Overview of echocardiography correlated with the GLS. Linear regression line represents the correlation between the GLS and A) left ventricular mass index and B) intraventricular thickness at diastole. Correlation coefficients are embedded in the graphs. C) Scatter plots depict preoperative (open symbols) and postoperative (closed symbols) data of the global longitudinal strain based subgroups, stratified according to a "normal GLS" (≤−18.9%) cohort (n=13) (circles) and "abnormal GLS" (>−18.9%) cohort (n=26) (squares), D) left ventricular mass index and E) intraventricular septal thickness at diastole. F) Scatter plots show the preoperative left ventricular mass index according to the NYHA class. G) Contingency graph presents the percentage of patients who experienced a full recovery of LV geometry within a median follow-up period of 5 (IQR 4-6 months). The odds ratio is embedded in the graphs
Accordingly, patients with a normal GLS (<−18.9%) experienced a faster recovery of normal LV geometry and substantial improvement in their LV functionality (Supplemental Table 6). In contrast, abnormal GLS values at baseline translated to a modest LVRR, characterised by a slower normalisation of their LV geometry during the follow-up period.
Based on our data, one could hypothesise that normal GLS values pertain to LV cardiomyocytes with the intrinsic capacity to overcome interstitial fibrosis [1] and attenuate signalling effectors underlying pathologic remodelling.
Resolution of pathological LV hypertrophy is generally considered a marker of favourable prognosis and prolonged survival [18] ; on the other hand, recent reports suggest that a rapid improvement of the LVMI is paradoxically linked to a high(er) 30-day mortality [19,20] . There were no early mortality cases in our study; currently, our long-term follow-up is still ongoing, and long-term mortality can not be assessed yet. Given the complexities of LV mass regression post-AVR [21] , the limitations of 2D imaging-derived calculations [19] , SV adaptions [21] it will be of high interest to determine potential differences in mid-to-longterm mortality rates between the "normal" and "abnormal" GLS subgroups.
Our study has several limitations that require further discussion; first, our results pertain to a single-centre, observational retrospective study. A more controlled prospective setting, e.g. stratification based on NYHA class, classification of AS stages [22] and strict comparison based on comorbidities such as coronary artery disease, might strengthen the data. Also, a subgroup analysis comparing patients with or without concomitant CABG revealed no differences in the extent of LVRR based on postoperative values of GLS, LVMI and IVSd (Supplemental Figure 1).
Second, the study was powered to detect an effect size of >0.5. Consequently, the sample size for the sub-group comparisons was too small to confirm other trends with a smaller effect size; for instance, the association between the preoperative NYHA class and GLS value.
Third, the subgroups should be followed over a more extended period to confirm a possible association between the postoperative LVRR with adverse cardiovascular events and (allcause) mortality.
Despite these limitations, our conclusions are based on highly significant mean/median differences (Supplemental Tables 1 to 4, correlation coefficients (Supplemental Table 5) and large effect sizes, >0.75 (Supplemental Table 6), which combined may be interpreted as clinically relevant observations.

CONCLUSION
In conclusion, a normal preoperative LVGLS is associated with improved postoperative recovery of normal LV geometry