Comparison of Myocardial Deformation by Speckle-Tracking Echocardiography and Cardiac Magnetic Resonance in Patients with Fontan Circulation: Diagnostic Algorithm.

BACKGROUND: While the short- and median-term survival has improved considerably in patients with Fontan circulation, cardiac function and exercise capacity are still reduced and may deteriorate over time. Cardiac magnetic resonance (CMR) is the gold standard for the assessment of ventricular volume and function. Speckle-tracking echocardiography (STE) is a myocardial deformation technique to assess ventricular function, with promising results. The aim of our study is to validate STE and conventional echocardiography parameters and to compare them with CMR. Furthermore, we aimed to design a diagnostic algorithm applying some parameters in series for early detection of myocardial dysfunction.
MATERIALS AND METHODS: We performed a cross-sectional single-center study in 64 patients with Fontan circulation. Longitudinal and circumferential strain, strain rates, and conventional echocardiographic measurements were registered. Ventricular volumes and ejection fraction (EF) were obtained by CMR.
RESULTS: Seven patients presented ventricular dysfunction (EF <45% by CMR), without showing a significant correlation between STE parameters or conventional measures by echocardiography and CMR. After the application of the diagnostic algorithm with the optimal cutoff points (global longitudinal strain - 24.5%, global circumferential strain - 20%, and annular plane systolic excursion - 16.5 mm), we got a sensitivity rate and a negative predictive value of 100%. In 19 patients (40.1%), the absence of ventricular dysfunction was demonstrated without no false-negative cases.
CONCLUSIONS: STE should be considered a complementary diagnostic tool in Fontan patients. These suggested parameters applied in series are a useful tool for identifying early ventricular dysfunction and for diagnostic tests improvement with a fewer CMRs in the follow-up of these patients. Copyright:

INTRODUCTION

Despite improved short- and medium-term survival, ventricular dysfunction and heart failure in the Fontan circulation are unavoidable, and quantitative evaluation of single ventricle (SV) function is essential during follow-up.

Cardiac magnetic resonance imaging (CMR) is the gold standard technique, but its routine use is hampered in our health-care setting.[1] Speckle-tracking echocardiography (STE) is a well-established technique that has been shown to be able to identify early ventricular dysfunction in other congenital heart diseases.[2345678910111213]

The aim of our study is to validate STE and conventional echocardiography parameters and to compare them with CMR in SV. Furthermore, we aimed to design a diagnostic algorithm for early detection of myocardial dysfunction.

MATERIALS AND METHODS

This observational single-center study included 64 consecutive patients of pediatric age with Fontan circulation monitored in our department between December 2010 and December 2018. Clinical, echocardiographic, and CMR examinations were performed for all patients.

Inclusion criteria

Partial cavopulmonary connection procedure (Glenn or hemi-Fontan) prior to TCPC

Completion of the Fontan circulation after the age of 2.5

A minimum age of 3.5 at the time of study

At least 2 years of follow-up elapsed from the intervention to inclusion in the study

A maximum of 6 months between echocardiogram and CMR, without any catheter or surgical intervention between both studies.

Patients with defibrillators or pacemaker devices were excluded from this study.

The classification criteria we used were the predominant left or right ventricular morphology of the SV.

The study was approved by the local ethics committee, and written informed consent was obtained from all patients and/or their parents. The study protocol conforms to the ethical guidelines of the 1975 Declaration of Helsinki as reflected in a priori approval by the institution's human research committee.

Conventional echocardiographic measurements

An experienced pediatric cardiologist performed a transthoracic echocardiogram on all the patients following the pediatric guidelines of the American Society of Echocardiography,[14] using an iE33 ultrasound system (Philips Medical Systems®, Best, The Netherlands), and following the protocols established by our department, with a frame rate of 60–90 frames/second. Image quality and temporal resolution were adjusted according to standard techniques, optimizing the definition of the endocardial border. Xcelera software (Philips Medical Systems®) was used for offline analysis.

Left ventricle (LV) fractional shortening was calculated based on the parasternal short-axis plane at papillary muscle level from M-mode images.

The biplane Simpson method was used for the assessment of ejection fraction (EF) from apical four-chamber (4C) and two-chamber (2C) views in single LVs.

Right ventricle (RV) endocardial borders at end-diastole and at end-systole were traced from the apical 4C view to measure the RV fractional area change (FAC). Trabeculae, papillary muscles, and moderator band were included in the cavity area in systole and diastole. Tricuspid valve S' peak velocity was obtained from Doppler imaging on apical 4C view.

The analysis of the annular plane systolic excursion (APSE) was performed by placing the cursor in the lateral insertion of the atrioventricular valve of the dominant SV. Three measurements were taken in three different cardiac cycles, calculating mean value.

Qualitative valve regurgitation was assessed on color Doppler images following the echocardiographic criteria of the guidelines when the regurgitation was greater than mild to moderate.

Two-dimensional speckle-tracking echocardiography

All the images were recorded and saved digitally for subsequent offline analysis using QLab 10.7 software (Philips Medical Systems®. Best, The Netherlands) for the STE study.

Peak longitudinal strain and strain rate values were obtained based on 2C and/or 4C apical views. The peak circumferential strain and strain rate were calculated based on the parasternal short-axis plane at papillary muscle level, including these muscles and trabeculae in the ventricular cavity. The remnant ventricle or outlet chamber was excluded from the volume and mass analysis.

After initial manual demarcation of the endocardium, the system tracked the speckles throughout the cardiac cycle, performing automated myocardial segmentation and analysis of ventricular motion. Tracking during the cardiac cycle was visually verified and manually adjusted until all the myocardium was included and the speckles were accurately tracked. Three measurements were taken in three different cardiac cycles, calculating mean value, and thus determining global longitudinal strain (GLS) and global circumferential strain (GCS), as well as the corresponding global strain rates.

SV myocardium was automatically divided into six segments.[15] GLS and GCS were calculated as the arithmetic mean of the six segments obtained in 2C and/or 4C apical views and parasternal short-axis plane at papillary muscle level, respectively, needing at least 4 out of 6 valid segments for the analysis.

The lateral segments were always designated by the free wall of de SV, and the septal segments were designated by the remnant ventricle or outlet chamber.

Segments that did not have adequate tracking due to a poor acoustic window or a ventricular septal defect, were removed from de analysis after a visual verification.

Cardiac magnetic resonance

CMR studies were performed on all patients with equipment Panorama 1.0T Philips Medical Systems® or MAGNETOM Skyra Syngo MR E11 (3T) Siemens®. Steady-state free precession sequence and multislice and multiphase method were used to assess the ventricular volumes.

End-diastolic and end-systolic volumes and EF were estimated from the manual tracing of the endocardial and epicardial contours. Stroke volume, cardiac output, and cardiac index were also calculated. Ventricular dysfunction was defined as EF <45%.

Diagnostic algorithm

Algorithm was built based on the ability of echocardiographic parameters to predict ventricular dysfunction (EF < 45%) by CMR. GLS, GCS, and APSE were selected and ordered according to their correlation magnitude in our study population, as it is done in similar published studies.[161718]

Receiver operating characteristic (ROC) curves, areas under the curve (AUC), and cutoff points were analyzed for GLS, GCS, and APSE. Sensitivity, specificity, and positive and negative predictive values (NPVs) of the cutoff values were calculated for selected variables. The most suitable points were selected aiming the highest sensitivity and NPV, attempting the diagnosis of all patients with ventricular dysfunction.

Reproducibility

The intra- and inter-observer variability of the myocardial deformation parameters was calculated from 10 randomly selected patients. Duplicated measurements were made for GLS in the six segments from apical 4C view. To calculate intra-observer variability, the same observer repeated the measurements at least 2 weeks later to avoid recall bias. For inter-observer variability, another experienced pediatric cardiologist performed the measurements on the same 10 patients anonymously and independently.

Reproducibility was expressed as the intraclass correlation coefficient (ICC) with the relevant confidence intervals (CI), a coefficient of >0.75 being considered excellent, 0.6–0.74 good, and <0.4 poor.

Statistical analysis

The SAS 9.4 system (SAS Institute Inc. 2013. Cary, NC, USA) was used to conduct the statistical analysis. Quantitative variables were expressed as their mean, with standard deviation (SD). Qualitative variables were described in absolute and relative frequencies shown as percentages. For most cases, the Mann–Whitney U-test was used for the comparisons between continuous quantitative versus qualitative variables. The Pearson correlation coefficient was used for the quantitative correlations between systolic function parameters. Intra-observer and inter-observer variability was analyzed using Bland–Altman test and was expressed as ICC. For all calculations, P < 0.05 was considered statistically significant.

RESULTS

Patient characteristics

We included 67 consecutive patients who had undergone a TCPC in the pediatric age range. Clinical and anatomic variables are listed in Table 1.

Table 1

Subject characteristics and anatomy

Variable	n=64
Gender (male)	39 (60.9%)
Age at study (years)	15.56 (5.57)
Age at TCPC (years)	5.87 (2.42)
BSA (m2)	1.43 (0.39)
Resting HR (beats/min)	81 (45-130)
Resting SaO2 (%)	93 (74-98)
SBP (mmHg)	111 (83-139)
DBP (mmHg)	64 (42-97)
Dominant LV/RV morphology	38 (59.4%)/26 (40.6%)
Single ventricle diagnosis, n (%)
HLHS	12 (18.5)
Tricuspid atresia	22 (33.8)
Pulmonary atresia	15 (23.1)
DORV	10 (15.4)
DILV	7 (10.8)
Unbalanced AVSD	6 (9.2)
Other	16 (24.6)
Fontan fenestration	17 (26.6)

Data are expressed as median (range) or as n (%). TCPC=Total cavopulmonary connection, BSA=Body surface area, LV=Left ventricle, RV=Right ventricle, SBP=Systolic blood pressure, DBP=Diastolic blood pressure, HR=Heart rate, SaO2=Oxygen saturation, HLHS=Hypoplastic left heart syndrome, DORV=Double-outlet right ventricle, DILV=Double-inlet left ventricle, AVSD=Atrioventricular septal defect

The average age was 15.5 years (3.9–26.2 years) and the mean follow-up time since surgery was 9.7 (SD 4.7) years. Three patients with devices incompatible with CMR were excluded. All the echocardiography and CMR data were available and collected for 47 of the 64 patients. All the patients received corrective surgery entailing an extracardiac conduit, except for one, who had an intra-atrial lateral tunnel. None of the patients showed dehiscence or connection stenosis in the Fontan conduit or in the Glenn connection. The ascending and descending aorta gradients were normal, and none of the patients showed significant pulmonary, hepatic, or renal dysfunction. All patients were in sinus rhythm.

Association between echocardiographic parameters and cardiac magnetic resonance measures

Echocardiographic and CMR measurements are summarized in Table 2. CMR showed ventricular dysfunction (EF <45%) in 7 patients (11.1%). GLS and GCS were lower in patients with worse EF values, but we found no significant differences in strain and strain rate values between patients with normal and decreased SVEF. No significant association between EF assessed by CMR and conventional echocardiographic measures was found [Table 3].

Table 2

Echocardiographic and cardiac magnetic resonance measurements

Parameters	Mean (SD)	Minimum	Maximum
EF CMR (%)	53.84 (8.84)	26	76
EDV (ml/m2)	89.11 (28.13)	37.7	153.1
ESV (ml/m2)	41.29 (17.87)	11.2	99.57
GLS (%)	−19.22 (3.56)	−26.6	−12
GCS (%)	−16.04 (4.86)	−27	−7
GLR (s−1)	1.16 (0.37)	0.65	2.23
GCR (s−1)	1.21 (0.41)	0.36	2.04
APSE (mm)	12.83 (3.11)	7.2	18
M-mode SF (LV) (%)	33.87 (7.13)	19.8	52
EF Simpson (LV) (%)	53.61 (6.21)	41.5	65.6
FAC (RV) (%)	31.4 (3.97)	24.1	40.8
S’ peak velocity (RV) (cm/s)	6.99 (2.14)	5	13.4

Data are expressed as mean (SD). EF=Ejection fraction, CMR=Cardiac magnetic resonance, EDV=End-diastolic volume, ESV=End-systolic volume, APSE=Annular plane systolic excursion, GLS=Global longitudinal strain, GLR=Global longitudinal strain rate, GCS=Global circumferential strain, GCR=Global circumferential strain rate, SF=Shortening fraction, FAC=Fractional area change, RV=Right ventricle, LV=Left ventricle, SD=Standard deviation

Table 3

Myocardial deformation and conventional echocardiographic measurements in patients with preserved single ventricle ejection fraction and dysfunction by Cardiac magnetic resonance

Variables	Dysfunction	SVEF (P)	P
GLS (%)	−17.17 (4.82)	−19.86 (3.51)	0.47
GCS (%)	−13.42 (3.40)	−16.54 (4.79)	0.68
APSE (mm)	11.92 (3.45)	12.87 (3.19)	0.45
M-mode SF (LV) (%)	33.66 (11.37)	35.23 (6.77)	0.87
EF Simpson (LV) (%)	52.22 (6.91)	53.72 (6.26)	0.49
FAC (RV) (%)	29.45 (3.18)	31.56 (4.04)	0.44
S’ peak velocity (RV) (cm/s)	5.75 (0.49)	7.11 (2.2)	0.28

Data are expressed as mean (SD). EF=Ejection fraction, FAC=Fractional area change, GLS=Global longitudinal strain, GCS=Global circumferential strain, APSE=Annular plane systolic excursion, SVEF (P)=Single ventricle with preserved ejection fraction, SF=Shortening fraction, RV=Right ventricle, LV=Left ventricle, SD=Standard deviation

Patients were divided into two groups, with normal EF and decreased EF by CMR. GLS, GCS, and APSE were analyzed and are shown in Figure 1.

Figure 1

Longitudinal (a) and circumferential (b) strain and annular plane systolic excursion (c) in patients with normal ejection fraction (no dysfunction) and abnormal ejection fraction (dysfunction)

Conventional echocardiographic function parameters such as APSE, S' peak velocity, EF by biplane Simpson, and FAC were unrelated to myocardial deformation parameters [Table 4].

Table 4

Correlation between conventional echocardiographic variables and myocardial deformation parameters (P values)

Ventricle	Variables	GLS	GCS	GLR	GCR
LV	M-mode SF	0.639	0.802	0.181	0.255
	EF Simpson	0.251	0.103	0.141	0.461
RV	FAC	0.291	0.341	0.98	0.791
	S’ peak velocity	0.098	0.803	0.124	0.669
LV/RV	APSE	0.165	0.188	0.967	0.397

EF=Ejection fraction, FAC=Fractional area change, GLS=Global longitudinal strain, GCS=Global circumferential, GLR=Global longitudinal strain rate, GCR=Global circumferential strain rate, APSE=Annular plane systolic excursion, SF=Shortening fraction, RV=Right ventricle, LV=Left ventricle

In the same way, there was no statistically significant correlation between end-diastolic and end-systolic volumes and myocardial deformation parameters.

When analyzing GLS, GCS, and APSE parameters independently, we observed that they did not reach statistical significance. ROC curves and their AUC for GLS, GCS, and APSE were calculated and are shown in Figure 2. Different cutoff points with their respective sensitivity, specificity, NPV, and positive predictive value were analyzed.

Figure 2

Receiver operating characteristic curves of longitudinal and circumferential strain and annular plane systolic excursion. AUC = Area under the curve, CI = Confidence interval

ROC curve analysis for predicting a conventional CMR EF <45% showed that cutoff values of − 24.5% for GLS, −20% for GCS, and 16.5 mm for APSE yielded 100% sensitivity and 100% NPV. In addition, algorithm as a whole maintained a sensitivity and NPV of 100% [Table 5].

Table 5

Sensitivity, specificity, positive and negative predictive values, and cutoff points for selected variables

	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)	Cutoff point (%)
GLS	100	14.3	12.7	100	−24.5
GCS	100	23.8	15.8	100	−20
APSE	100	23.3	17.9	100	16.5 mm
Total	100	45.2	17.8	100

GLS=Global longitudinal strain, GCS=Global circumferential, APSE=Annular plane systolic excursion, PPV=Positive predictive value, NPV=Negative predictive value

Flow diagram was built applying the selected variables according to the established order and best cutoff values [Figure 3]. Only those patients with values below the cutoff point selected for each variable, continued in the algorithm. At the end of the process, the absence of ventricular dysfunction was confirmed for 19 patients (40.1%), without no false-negative cases. All patients with dysfunction by CMR were correctly diagnosed by this algorithm.

Figure 3

Flow diagram for the 47 patients with TCPC who had all measures for GLS, GCS, and APSE. GLS = Global longitudinal strain, GCS = Global circumferential strain, APSE = Annular plane systolic excursion, FP = False positive, FN = False negative, TP = True positive, TN = True negative

The intra- and inter-observer variability analyzed for GLS using STE ranged from good to excellent, with an ICC of 0.62 (CI: 0.13–0.87) and 0.77 (CI: 0.38–0.93), respectively.

In 64.3% of the echocardiograms, it was impossible to adequately track one or two segments because of the suboptimal window, mainly located in the basal septal region. This area is frequently absent due to the presence of large ventricular septal defects and was consequently excluded from the analysis. In CMR, no study was excluded due to limitations in terms of image quality.

DISCUSSION

Several previous studies in patients with Fontan physiology have demonstrated no correlation between conventional echocardiographic measurements such as Simpson method, S' peak velocity, and APSE and SVEF by CMR. This fact suggests that these parameters are not sensitive enough for detection of early ventricular dysfunction in SV with an abnormal mass/volume ratio.[56] These findings were found in our cohort as well.

No statistically significant differences were found in myocardial deformation and conventional echocardiographic measurements between patients with and without dysfunction by CMR. This might be explained because EF data were similar to those described as normal values (ventricular dysfunction was present just in seven patients), and the range of values was not wide.

As shown in previous publications, myocardial deformation parameters are reduced in patients with acquired and congenital heart disease over the years.[891011] Patients with Fontan palliation could have reduced STE parameters, but global and segmental strain and strain rate measurements showed significant heterogeneity in all functionally SVs.[192021] The underlying hypothesis may be related to the complex anatomy, the remnant ventricle size, and the arrangement of the myocardial fibers and their contractility.

In our case, values of GLS and GCS were slightly decreased compared to healthy pediatric population[222324] but very similar to those reported by other authors as Lopez et al.[20]

Recently, Koopman et al.[18] compared strain measures by STE with ventricular function by CMR, and no correlation between GLS by STE and volumes, cardiac index, and EF by CMR was found. The mean age of included patients, as well as mean follow-up time since surgery, was very similar in our study. In addition, we obtained analogous GLS and longitudinal strain rate values to those obtained by Koopman et al., and no significant differences between EF by CMR and STE parameters were found. A possible explanation might be that STE, more specifically GLS, in early stages, is not a useful tool for detecting decreased EF by CMR.

Regarding the GCS and strain rate that we have calculated, we think that for a complete evaluation of the SV, it is necessary to take into account circumferential myocardial deformation parameters in the global ventricular analysis, as the compensatory changes that the ventricle undergoes as part of its adaptive process should be considered into the study of ventricular contractility. Interestingly, GCS values were lower than GLS: −16.04% (4.86) vs. −19.22% (3.56), respectively, and a similar trend toward lower circumferential strain values in patients with lower EF was observed. This might be explained by an inadequate compensatory ventricular hypertrophy of longitudinal fibers that would result in abnormal contractility despite having nearly preserved GLS values. Another explanation could be the underestimated GCS values by a worse acoustic window on the short-axis view.

On the other hand, global radial strain and radial strain rate were not included because they usually show high variability in children. Therefore, they are not very helpful parameters in the clinical routine without reference values in patients with univentricular circulation.[16]

No correlation between conventional echocardiographic parameters and strain and strain rate values could be observed in our study. This contrasts with the published data from Cho et al.,[25] where GCS by STE showed a strong correlation with EF by conventional echocardiographic parameters in adult patients with preserved EF and ventricular dysfunction by ischemic disease.

Manual tracing of the endocardial contour resulted in high variability of EF values due to the presence of prominent trabeculations and a difficult visualization of the apical region.

Our study suggests that SVEF assessed by CMR is not correlated with conventional echocardiography measurements and strain and strain rate values as analyzed by STE. When our diagnostic algorithm was applied, all patients with ventricular dysfunction by CMR were correctly diagnosed. According to our results and other published data, we assumed that variables could not be used individually for predicting ventricular dysfunction by CMR.[18] When analyzing parameters independently, we observed that GLS, GCS, and APSE presented the P values closest to statistical significance. The selected cutoff points for GLS and GCS (−24.5% and − 20%) were similar to those considered normal for general pediatric population.[222324] Sensitivity and NPV of 100% were a priority to ensure the correct diagnosis for all patients with ventricular dysfunction. This fact made more difficult to diagnose patients with normal EF that were classified as dysfunctional ventricles. However, using the highest sensitivity, we did not obtain any false-negative cases and diagnostic algorithm seemed to be safe for our patients.

This algorithm may be an easy and objective way for ventricular function assessment for population with complex and variable ventricular anatomies and would be a useful tool for identifying early ventricular dysfunction using an available, faster, and cheaper technique without longer analysis time. Furthermore, this would allow us to decrease sedation needs in children, number of CMR performed, and the use of radiological contrast, with the consequent optimization of diagnostic test and processing time.

Limitations

The modest sample size of our study is a limiting factor to take into account, although it is considerably larger when compared to other studies using myocardial deformation in SV patients with a TCPC.[12212627]

Furthermore, ventricular dysfunction was present in few patients in our cohort (7, 11.1%), so it may result in a lack of statistical power to detect certain differences.

This is a single-center study with a retrospective design. These facts mean that the extrapolation of its results may be limited. On the other hand, we used the same software and protocols with the same observers obtaining less heterogeneous results. TCPC was performed for all patients but one, using the same surgical technique (extracardiac conduit). Thus, the results obtained are not confused by changing surgical techniques over time.

Because echocardiograms and CMRs were not performed simultaneously, some factors such as preload conditions, drugs, and different hemodynamic situation, must be considered.

Intra-observer variability of longitudinal strain presented lower values than inter-observer variability, but they were in the range of those published by other authors who used similar methods.

All patients presented adequate image quality to be included in the analysis, but 17 out of 64 patients (26.5%) did not have all the measurements for being included in diagnostic algorithm due to retrospective nature of our study.

Furthermore, the cross-sectional design of our study resulted in a limitation to determine the possible clinical implications, so larger and prospective studies are needed to improve clinical utility in Fontan patients.

The results obtained in our study allow us to conclude that the SV EF assessed by CMR may not be correlated to conventional echocardiography measurements and strain and strain rate values analyzed by STE for contemporary patients after Fontan intervention.

STE should be considered as a complementary diagnostic technique for assessment and monitoring ventricular function and would detect early failure of univentricular circulation.

Conventional echocardiographic measurements and strain parameters used individually were not helpful when detecting a decreased EF by CMR.

However, when we apply myocardial deformation by STE and conventional echocardiographic parameters in series, they may be a useful tool for identifying early ventricular dysfunction using an available, faster, and cheaper technique with optimization of resources and diagnostic tests.

Further studies are needed to analyze the clinical impact of these parameters in Fontan patients.

Ethical clearance

The study was approved by the local ethics committee, and written informed consent was obtained from all patients and/or their parents. The study protocol conforms to the ethical guidelines of the 1975 Declaration of Helsinki as reflected in a priori approval by the institution's human research committee.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.