Three and Two-Dimensional Cardiac Mechanic Parameters by Speckle Tracking Echocardiography are Predictors of Outcomes In Chagas Heart Disease


 BACKGROUND: Chagas disease (CD) is a neglected infectious disease associated with early mortality and substantial disability. Three-dimensional speckle tracking (3D STE) may play a role in the evaluation of CD. We aim to characterize new echocardiographic variables in patients with CD and to assess the hypothesis that 3D STE may predict outcomes. METHODS: Seventy-two patients with CD were included. Clinical and conventional 2D and 3D STE analysis were performed. Patients were followed up for sixty months. Clinical events were defined as hospitalization for heart failure, complex ventricular arrhythmias, heart transplant and all-cause death. RESULTS: Seventy-two patients were recruited and enrolled in three groups: left ventricular ejection fraction (LVEF) < 0.40 (N=22); 0.40 < LVEF < 0.50 (N=10) and LVEF > 0.50 (N=30). After a Cox model analysis, the top predictors of composite endpoints were 2D LV global longitudinal strain (GLS) ≤ -11.3% (AUC=0.87), 2D LV global circumferential strain (GCS) ≤ -10.1% (AUC=0.79), 3D LV GLS≤ -13% (AUC=0.82), 3D LV area strain ≤ -16% (AUC=0.81) and right ventricle (RV) GLS ≤ -17.2% (AUC=0.78). CONCLUSIONS: Patients with CD and mrLVEF were morphologically similar to the rLVEF patients despite the benign evolution as the pLVEF group. RV GLS, 2D LV GLS, 2D LV GCS, 3D LV GLS, and 3D LV area strain are strong predictors of sixty months outcomes in patients with CD.

Introduction strain 15,16 . It is well established in literature that endocardial bers, arranged longitudinally, are the rst to undergo functional damage. Hence, longitudinal myocardial strain has been considered the best predictor of cardiac events in patients with HF, in comparison to left ventricular ejection fraction (LVEF) and Tissue Doppler data 15,16 .
Nevertheless, literature data are scarce concerning cardiac mechanics evaluation by 3D STE, particularly in patients with CD. The present study aims to clinically characterize new echocardiographic variables concerning cardiac mechanics according to LVEF and evaluate its role in the prediction of clinical outcomes in patients with CD.

Methods
Eight hundred and eight patients with CD from the Cardiomyopathies and Heart Failure Clinics of a tertiary cardiological center were evaluated. Seventy-two consecutive patients with CD were recruited based on the following criteria: seropositivity for CD in two distinct laboratory tests (indirect hemagglutination assay, indirect immuno uorescence technique or ELISA -Enzyme-Linked Immunosorbent Assay) in sinus rhythm without other documented etiology. The exclusion criteria were age under 18 or beyond 70-years old, uncontrolled systemic arterial hypertension, diabetes mellitus, hypothyroidism, renal failure (serum creatinine > 1.5 mg/dL or glomerular ltration rate < 60mL/min/1.73m 2 ), hepatic failure, atrial brillation or frequent arrhythmias, coronary artery disease, patients with pacemakers, pregnancy and chronic obstructive pulmonary disease. This was a convenience sample because 3D STE is very sensitive technique and may be in uenced by several clinical conditions. All patients gave written informed consent, and the study was approved by the Hospital das Clínicas, Faculdade de Medicina, Universidade de Sao Paulo Ethical Committee. All patients underwent clinical, electrocardiographic and echocardiographic evaluation. All methods were performed in accordance with the relevant guidelines and regulations.

Echocardiographic Analysis
All studies were performed using the Vivid E9 (GE Healthcare Medical Systems, Milwaukee, WI, USA) with dedicated transthoracic transducers. Comprehensive conventional (2D Echo) and 3D Echo were performed according to joint recommendations from the American Society of Echocardiography 17 . Two and threedimensional speckle tracking analysis were performed o ine using a dedicated software (EchoPAC, BT12, GE Healthcare).

Clinical Follow-up
All patients were followed up sixty months. All patients were clinically evaluated by an assistant physician every three months or whenever necessary. Holter monitoring was assessed annually for each patient or whenever necessary. Patients were treated according the Brazilian Guidelines for CD 5 . Patients did not receive speci c pharmacological treatment for T.cruzi infection. Composite endpoints were de ned as hospitalization for heart failure (HF), complex ventricular arrhythmias (ventricular brillation and sustained ventricular tachycardia), heart transplant and death.

Statistical Analysis
Continuous numeric variables were expressed as mean ± standard deviation (SD). Normally distributed data were compared using a 2-sample Student's t-test and 1-way analysis of variance. Non-normally distributed data were compared with Mann-Whitney U test and Kruskal-Wallis test. Categorical variables were expressed as frequency (percentage) and were compared using the chi-square test (x 2 ) or the Fisher exact test, when appropriate. A two-tailed p value of 0.05 was considered signi cant.
Intraobserver and interobserver reproducibility of 2D STE parameters were evaluated in 10 randomly selected subjects and evaluated using concordance correlation coe cients (CCC) and Bland-Altman analysis.
Intraobserver variability was assessed by having one observer re-measure after 30 days. Another observer blinded to the rst observer's measurements evaluated the randomly selected exams for interobserver variability analysis. To determine the optimal cutoff value of prognostic STE parameters for predicting composite endpoints, mortality and hospitalization for heart failure, receiver-operating characteristic (ROC) curves were used.
Survival curves according to the LVEF were obtained in a Kaplan-Meier analysis and compared using the logrank test. Uni and multivariate Cox regression models were use for estimations of the predictors of outcomes.

Results
Eight-hundred and eight medical electronic reports of patients with CD followed in the single center where the study was performed were analyzed. Seventy-two patients were recruited according to the inclusion and exclusion criteria. The great majority of patients were excluded due to uncontrolled hypertension, diabetes mellitus, irregular cardiac rhythm, hypothyroidism and renal failure. Patients were enrolled in three groups: patients with LVEF < 0.40 (reduced LVEF, rLVEF) (N=32); patients with LVEF ≥ 0.40 and ≤ 0.50 (mid-range or mildly reduced LVEF, mrLVEF) (N=10); patients with LVEF > 0.50 (preserved LVEF, pLVEF) (N=30).

Clinical Characteristics and Conventional Echocardiographic Evaluation
Clinical, anthropometric and electrocardiographic data are described in Table 1. Conventional 2D Echo variables are comprised in Table 2. There were no difference between the groups regarding gender distribution, mean age and anthropometric variables.      Two-dimensional STE (2D STE) 2D STE data are shown in Table 3. 2D STE feasibility was very high in all the groups. LV GLS values were able to distinguish the groups according to the LVEF. RV GLS and RV free wall longitudinal strain values were lower in the group rLVEF in comparison to the other groups. LV Peak systolic CS and LV displacement were lower in the group rLVEF in comparison to the group pLVEF.
With exception of LV basal displacement, there were no differences regarding LV peak systolic CS and displacement between the groups rLVEF and mrLVEF. RV GLS and free wall strain were similar between the groups with mrLVEF and pLVEF.
Three-dimensional Echocardiographic results 3D Conventional Echo and STE parameters are shown in Table 4. There were no statistical differences between rLVEF and mrLVEF regarding left atrium (LA) indexed volume, LV end systolic indexed volume, sphericity index. 3D STE feasibility was very good in all the groups. GLS was different between the groups.
There were no statistical differences between mrLVEF and pLVEF regarding 3D LV GLS, LV GCS, LV GCS or LV area strain. Figure 1 depicts 3D LV GLS, LV GRS, LV GCS and LV area strain of a patient with CD and severe LV dysfunction. Meyer curves are shown in Figure 2. One patient in pLVEF died of non-cardiovascular cause. There were no clinical events in mrLVEF. There were nine hospitalizations for decompensated HF, two episodes of complex ventricular arrhythmias, two heart transplants and seventeen deaths in rLVEF.

Composite Endpoints
In the overall patients, bilateral cardiac remodeling was associated with composite endpoints. Larger atrial and ventricular diameters and volumes (absolute or indexed) as well as reduced biventricular function evaluated by 2D and 3D Echo, reduced velocities at tissue Doppler and increased E/e´ ratio were associated with composite endpoints. 2D LV GLS, 2D LV GCS, RV GLS, 3D LV GLS and 3D LV area strain were the strongest predictors of composite endpoints (Table 5). Death from all causes and Hospitalization for Heart Failure As for composite endpoints, bilateral cardiac remodeling was associated with mortality. Larger atrial and ventricular diameters and volumes (absolute or indexed) as well as reduced LV and RV function (TAPSE) evaluated by 2D and 3D Echo, reduced velocities at tissue Doppler and increased E/e´ ratio were associated with death from all causes and hospitalization for heart failure. Interestingly, RV S´ wave, FAC were not associated with overall mortality.
2D LV GLS, 2D LV GCS, RV GLS, 3D LV GLS and 3D LV area strain were the strongest predictors of death from all causes and hospitalization for HF (Table 5).

ROC curves for prediction of cardiovascular outcomes
Receiver-operator characteristic (ROC) curves of the strong predictors of cardiovascular events were built for the ve years of follow up (Figure 3). The cut offs of the STE parameters that yield the prediction of outcomes in ve years of follow up are depicted in Table 6.

Reproducibility
The intraobserver and interobserver reproducibility of most of the STE parameters was excellent, as re ected by high CCC (Supplemental File; Table 7), with exception of 2D LV GCS and 2D LV GCS. Bland-Altman analysis demonstrated good intraobserver and interobserver agreement.

Discussion
To the best of our knowledge, this is the rst study to evaluate 3D conventional and strain analysis in patients with CD from normal LVEF up to reduced LVEF. So far, only 2D STE parameters have been described in different stages of CD and most studies evaluated patients with normal LVEF 18-21 . Very few studies have reported 3D Echo conventional parameters in patients with CD but none have included 3D STE analysis 22,23 .
The main ndings of our study were: 1) 3D conventional and 2D STE analysis in mrLVEF were anatomically and functionally similar to rLVEF; 2) RV GLS, 2D LV GLS, 2D LV GCS, 3D LV GLS, and 3D LV area strain were strong predictors of outcomes in patients with CD, with superior value to other conventional parameters provided by cardiac dimensions, Tissue Doppler indexes or TAPSE and fractional area change for prediction of overall mortality or composite endpoints.
Our results are in disagreement with previous studies suggesting that in patients with the indeterminate form of CD with pLVEF, global and segmental longitudinal systolic strain is reduced compared with healthy subjects, thus indicating that it could be a sensitive technique to detect early myocardial damage 19,21 . In other published investigation, only regional LV longitudinal strain was reduced in indeterminate form of CD with pLVEF that is in concordance with our results. Also, in a recent study, with patients in different CCM stages, 2D LV GLS was the more accurate measurement regarding stage A differentiation from the stages B, C, and D 24 . However, reports of 2D LV GLS in CD are con icting and still remain a controversial issue. To explain the abnormal strain, myocarditis and in ammatory in ltrate was found in 15% and 37% respectively of indeterminate phase 25 . In another publication, in 60.6% of patients with indeterminate phase, it was found abnormalities like ber degeneration, volume changes, interstitial edema, in ammatory in ltrates and brosis 26 . Besides, it was reported even in patients with Chagas disease with preserved or minimally impaired ventricular function cardiac brosis in 3% and 11% respectively using late gadolinium enhancement on cardiac magnetic resonance (CMR) 27 . Moreover, some studies support the use of 2D LV GLS in surveillance providing a window of opportunity for early intervention and preventing heart failure patients.
In our study, 3D conventional Echo and 2D STE parameters evidenced similar results comparing rLVEF and mrLVEF. Hitherto, the pathophysiology of rLVEF and mrLVEF heart failure patients is incompletely understood and, consequently, the reasons for this nding are unclear. In other etiologies, mrLVEF patients may include a heterogeneous population with patients that partially recovered the cardiac function under guidelinesdirected medical treatment, or patients that have not yet evolved to rLVEF, or patients that never will follow to rLVEF. Previously, we reported improvement in CCM under guidelines-directed medical treatment 27  from CD patients showed that the clinical evolution of the disease was correlated with a continuous progression of ber destruction, brosis, myocardial in ammation, and a reduction in performance 2 . These ndings may indicate that, possibly, patients with CD mrLVEF could be sharing with rLVEF some pathophysiological mechanisms related to 3D conventional Echo and 2D STE parameters but for all characteristics. In fact a recent meta-analysis concluded that signi cant differences in hospitalization and mortality were detected between mildly or mid-range heart failure and the other subtypes of heart failure including diverse etiologies but not included CD 28 . Nevertheless, CD is a very peculiar etiology that is not included in the studies. Herein, there is an extreme paucity of data concerning patients with mildly or midrange Chagas heart disease and no study so far evaluated 3D cardiac mechanics in patients with mrLVEF and CD.
Our ndings that RV GLS, 2D LV GLS, 2D LV GCS, 3D LV GLS, and 3D LV area strain were strong predictors of outcomes in patients with CD are in concordance with an increasing number of studies that have suggested that 2D LV GLS is superior to LVEF as a measure of LV function and as a predictor of mortality and cardiac events in other etiologies, mainly ischemic cardiac disease 29  In respect to the RV GLS, several studies showed that this parameter provides strong additional prognostic value to predict overall and cardiovascular mortality in rLVEF patients with other etiologies, essentially ischemic cardiac disease,. The predictive value was even higher than parameters evaluated by CMR as RV ejection fraction and RV strain 31 . In concordance in our study, RV GLS was also a strong predictor of outcomes. Values of RV GLS under -17.2% and -14.9% were associated with composite endpoints and mortality, respectively. Similarly, an RV GLS under -19% was related to all-cause or cardiovascular mortality in rLVEF patients with ischemic, hypertensive or idiopathic cardiomyopathies 31 . As shown in previous studies in other etiologies and potentially also for CCM, RV GLS and RV free wall LS have performed better than conventional parameters for the prediction of outcomes. This may be probably because the use of the longitudinal strain for RV as for LV evaluation allows the analysis of the deformation of the endocardial bers, which might be more sensitive to reduced coronary perfusion and increased wall stress and are usually affected earlier in myocardial diseases.
Concerning 3D analysis, our ndings that 3D LV GLS and 3D LV area strain are strong predictors of outcomes in CCM suggest a very innovative clinical application for this technique. Until now, some studies evaluated the prognostic value of 3D Strain in a variety of clinical scenarios including valvular heart disease, ischemic heart disease and chronic renal failure, showing that the reduction of 3D values was associated with poor outcomes 32 . Particularly, 3D LV area strain is a very promising index that quanti es endocardial area change, integrating longitudinal and circumferential deformation, allowing a more detailed evaluation of the different types of myocardial bers and enabling a better understanding of the pathophysiology of cardiomyopathies 31,32 . In a recent meta-analysis that aimed to determine normal ranges of 3D Strain, the authors reported that the mean value of 3D LV GLS was 19.1%, ranging from 15.8 to 23.4% among the studies 33 . In our study, 3D LV GLS values and 3D LV area strain under -13% and -10%, were associated with composite endpoints and mortality, respectively.

Clinical Implications
In our study, it was observed high feasibility of 2D and 3D strain analysis in all the groups. Despite the signi cant LV dilatation, especially in the group with rLVEF, which could limit the analysis of all LV segments, the feasibility was close to 100%. These ndings demonstrate the great applicability of this technique, even Therefore, we hypothesize that, eventually, the analysis of biventricular longitudinal strain could add incremental data regarding mortality and composite endpoints, predicting outcomes in patients with CD, as previously described, shedding light on complexes mechanisms of cardiomyocytes contraction imbalance and pathophysiology of patients with Chagas cardiomyopathy as a useful tool for prognostication.
Thereby, cardiac mechanics provide more re ned and accurate information related to cardiac dysfunction and derangement than conventional parameters expressed by morphological variables as cardiac diameters and volumes and LVEF and might be related to myocardial brosis as evaluated by CMR but with expressive lower cost, mainly considering the economic issues in the developing countries, where CD is more prevalent.

Limitations
Due to the high prevalence of unknown or poorly controlled hypertension, in addition to diabetes mellitus and hypothyroidism, the vast majority of patients could not be included in this study, limiting the nal number of patients. However, this extreme caution is mandatory when evaluating new techniques, without de nite cutoff values and clinical applications. The small number of patients is a limitation as well as the fact that this was a single-center study.
Another limitations concern the limited availability of speci c equipment besides more substantial time to acquire the datasets and perform all the measurements. Currently, conventional 3D Echo and 3D STE are available in a very limited number of research centers of echocardiography. Besides, both techniques require Page 22/28 a learning curve and expertise until an adequate examination is possible. Nowadays, there is still inter-vendor variability in 3D STE measurements and spatial and temporal resolution are not yet adequate to perform reliable segmental analysis. Finally, there still considerable interobserver variability of 3D strain measurements but that was not a limitation in this study.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.