Right ventricular (RV) enlargement is used as a criterion for the treatment of RV outflow tract dysfunction in patients with congenital heart disease. Although RV volumes are most accurately measured by cardiac magnetic resonance (CMR), CMR is a limited resource. The aim of this study was to investigate whether simple echocardiographic measurements can adequately predict RV volumes below clinical thresholds, thereby reducing the need for CMR in some patients.
Children with repaired tetralogy of Fallot, double-outlet right ventricle, or truncus arteriosus who underwent CMR and echocardiography within a 4-week interval were retrospectively studied. From the four-chamber view, indexed RV lateral wall length, indexed RV end-diastolic perimeter length, and indexed RV end-diastolic area (RVEDAi), were measured. Results were compared with CMR indexed RV volume. The sensitivity and specifity of echocardiographic threshold values predicting RV volumes < 170 mL/m 2 were determined.
Fifty-one children (mean age, 12.7 ± 3.5 years; 25 male, 26 female) were reviewed. RVEDAi was correlated with CMR indexed RV volume ( r = 0.60, P < .0001). Indexed RV end-diastolic perimeter length and indexed RV lateral wall length were not correlated with CMR. RVEDAi < 20 cm 2 /m 2 had 100% specificity to predict indexed RV volume ≤ 170 mL/m 2 (area under the curve, 0.79), reducing the need for CMR in 15 of 51 patients (29%). A threshold RVEDAi of 22 cm 2 /m 2 would reduce the need for CMR in 21 of 51 patients (41%) at the expense of one false-negative result. The coefficients of variation were 14.7% for intraobserver variability and 9.6% for interobserver variability.
The specificity of echocardiography-measured RVEDAi can be set to predict RV volumes below a 170 mL/m 2 threshold in 100% of cases. This may reduce the need for CMR to determine RV volumes in ≥25% of patients with congenital heart disease, potentially reducing patient burden and costs.
Right ventricular (RV) enlargement is an indication for surgical or catheter pulmonary valve replacement or repair for RV outflow tract dysfunction in children with congenital heart disease (CHD).
Cardiac magnetic resonance (CMR) is currently the most accepted method to quantify RV volumes. However, CMR is relatively expensive, with limited availability in some centers. Although two-dimensional (2D) echocardiography is not sufficiently accurate to determine RV volumes for clinical decision making, it remains the most widely available modality for initial screening and serial follow-up. In patients who are otherwise asymptomatic but appear to have significant RV enlargement on the basis of echocardiography, clinicians are faced with the decision whether to perform CMR for the accurate determination of RV volumes. However, echocardiographic criteria to guide the need for more accurate RV volume quantification by CMR are lacking. RV enlargement by echocardiography is often suspected by subjective assessment, albeit with suboptimal interobserver variability. Current guidelines rely mainly on linear measurements of the RV body from the apical four-chamber view and the RV outflow tract from the parasternal short-axis view. However, these methods correlate weakly with RV size measured by CMR in children with CHD. Measurement of RV area from the apical four-chamber view correlates better with RV volumes measured by CMR and is readily measured using commercially available digital platforms.
We hypothesized that although 2D echocardiographic measurements are inadequate to determine absolute RV volumes, simple 2D echocardiographic indices can adequately predict RV volumes below a clinically determined threshold (e.g., 170 ml/m 2 ), thereby reducing the need for CMR in some patients. Accordingly, the objective of this study was to study the specificity of 2D echocardiographic indices to predict an RV end-diastolic volume indexed to body surface area (BSA) (RVEDVi) < 170 mL/m 2 .
We retrospectively reviewed echocardiographic, CMR, and demographic data of all children with repaired tetralogy of Fallot, double-outlet right ventricle, or truncus arteriosus who underwent CMR and echocardiography within 4 weeks of each other between 2004 and 2009 and who had no surgical or catheter interventions between echocardiography and CMR. Demographic and clinical data were extracted from the patients’ medical records. Patients were excluded if stored images were inadequate to visualize and trace the RV endocardium in the four-chamber view.
CMR was performed on a 1.5-T unit (Avanto; Siemens Medical Solutions, Erlangen, Germany), using a standardized protocol. RVEDVi was measured from a stack of short-axis cine images, using commercially available software (QMass version 7.2; Medis Medical Imaging Systems, Leiden, The Netherlands). RV volume was taken as analyzed for the clinical study and was not reanalyzed. BSA was calculated using the Haycock formula.
Echocardiography was performed using an iE33 (Philips Medical Systems, Andover, MA) or a Vivid 7 (GE Healthcare, Milwaukee, WI) system, with probe frequencies appropriate for patient size and with simultaneous electrocardiographic recording. Ultrasound machine settings are standard across the laboratory, with adjustment by the sonographer to optimize images during the exam for image depth, sector width, compression, and gain. All data were digitally stored in Digital Imaging and Communications in Medicine format and measurements performed offline using comercially available software (Syngo Dynamics; Siemens Medical Solutions). Echocar-diographic measurements analyzed included RV end-diastolic diameter from the parasternal long-axis view, RV lateral wall length indexed to BSA from the apical four-chamber view, RV end-diastolic perimeter indexed to BSA (RVEDPi), and RV end-diastolic area indexed to BSA (RVEDAi) from the apical four-chamber view. RVEDPi and RVEDAi were measured by tracing the RV endocardial border, excluding trabeculations, at end-diastole (largest RV area) ( Figure 1 ). In some patients, part of the RV apex was missing from the image sector. If this area was small, we extrapolated the lines to complete the RVEDAi. If the area was large (more than just the RV apex tip) or if the RV endocardium was not adequately visible to trace, the patient was excluded as detailed previously. We further investigated whether expressing the RV measurements as ratios to the analogous left ventricular measures (e.g., RVEDAi/left ventricular end-diastolic area indexed to BSA) would improve predictive capability. Indexed left ventricular end-diastolic area was derived from the best available view of the left ventricle. This was not always the same four-chamber view in which the RV measurements were derived.
Statistical analysis was performed using Excel version 2007 (Microsoft Corporation, Seattle, WA) and SPSS (SPSS, Inc., Chicago, IL). Data are presented as mean ± SD. Pearson’s correlation coefficients were calculated to determine correlations between echocardiographic variables and RV volume by CMR. Sensitivity, specificity, and positive and negative predictive values of the echocardiographic variables tested were determined from 2 × 2 contingency tables using a CMR RV volume of 170 mL/m 2 as the cut point. Receiver operating characteristic curve analysis was used to determine sensitivity and specificity characteristics of RVEDAi versus CMR RVEDVi. To test measurement variability, RVEDAi was measured by two independent observers for interobserver variability and twice by the same observer after an interval of 8 weeks in 10 randomly selected patients with tetralogy of Fallot. The coefficient of variation (the absolute difference in percentage of the mean of repeated measurements) was used to evaluate intraobserver and interobserver variability for RVEDAi measurement. P values < .05 were considered statistically significant. The study was approved by the institutional research ethics board.
Demographic and Clinical Data
Of 63 eligible patients, 12 (19%) were excluded because of inadequate echocardiographic RV imaging, leaving 51 patients in the study group. Inadequate RV imaging resulted from either incomplete imaging of the RV apex or lateral wall or poor image quality leading to inadequate delineation of the RV endocardium. The study group characteristics are shown in Table 1 . The mean age was 12.7 ± 3.5 years (range, 0.5–17 years). The ratio of male to female patients was 25:26. Forty-five of the 51 patients (88%) had repaired tetralogy of Fallot or double-outlet right ventricle. Of the patients with double-outlet right ventricle, all but one had normally related great vessels and subaortic ventricular septal defect. One patient had malposed great vessels and a subpulmonary ventricular septal defect before surgery. Six patients (12%) had repaired truncus arteriosus. None of the patients had more than mild aortic valve insufficiency.
|Age (y)||12.7 ± 3.5 (0.5–17)|
|BSA (m 2 )||1.4 ± 0.38 (0.3–2.1)|
|Weight (kg)||47.9 ± 18.8 (5.9–92.9)|
|Height (cm)||150.3 ± 22.8 (62–184)|
|Children with tetralogy of Fallot or double-outlet right ventricle||45/51 (88%)|
|Children with truncus arteriosus||6/51 (12%)|
CMR and Echocardiographic Measurements
Table 2 summarizes the various echocardiography and CMR measurements. RVEDAi by echocardiography correlated moderately well with CMR RVEDVi ( r = 0.60, P < .0001). There was no correlation between indexed RVEDPi ( r = 0.02, P = .90) or indexed RV lateral wall length ( r = 0.03, P = .80) and CMR RVEDVi.
|RVEDVi (mL/m 2 )||153 ± 52 (85 to 302)|
|RV end-systolic volume index (mL/m 2 )||82.2 ± 34.3 (28 to 170)|
|RV ejection fraction (%)||48.7 ± 9.6 (28 to 79)|
|LV end-diastolic volume index (mL/m 2 )||87.8 ± 22.5 (30 to 192)|
|LV end-systolic volume index (mL/m 2 )||35.7 ± 9.9 (20 to 63)|
|LV ejection fraction (%)||59.3 ± 8 (34 to 75)|
|Cardiac mass index (g/m 2 )||57.5 ± 13.7 (32 to 96)|
|Pulmonary regurgitation fraction (%)||31.9 ± 15.7 (0 to 64)|
|Pulmonary regurgitation volume (L/min/m 2 )||1.9 ± 1.1 (0 to 5.8)|
|RVEDAi (cm 2 /m 2 )||22.9 ± 5.2 (11.5 to 36.6)|
|RVEDPi (cm 2 /m 2 )||17.6 ± 4 (11.3 to 31.6)|
|RV lateral wall length index (cm/m 2 )||9.6 ± 2.4 (6.3 to 17.4)|
|M-mode RV dimension (cm)||3.0 ± 0.8 (1.5 to 4.6)|
|RV end-diastolic dimension Z score||3.4 ± 1.7 (−0.7 to 6.6)|
|Tricuspid regurgitation severity|
Sensitivity and specificity characteristics of RVEDAi by echocardiography to predict CMR RVEDVi ≥ 170 mL/m 2 are shown in the receiver operator characteristic curve ( Figure 2 ) and in Table 3 . RVEDAi < 20 cm 2 /m 2 had 100% specificity to exclude RVEDVi ≥ 170 mL/m 2 (area under the curve, 0.79; P = .001; Figure 2 ). Fifteen of the 51 patients (29%) had RVEDAi < 20 cm 2 /m 2 . In these patients, CMR could have been avoided at the time, on the basis of the echocardiographic measurements ( Figure 3 ). A threshold RVEDAi of 22 cm 2 /m 2 would reduce CMR in 21 of 51 patients (41%) at the expense of one false-negative result ( Figure 3 ). Similar results would be obtained at a clinical threshold of 150 mL/m 2 and an RVEDAi of 20 cm 2 /m 2 (14 of 51 patients = 27%; Figure 3 ).
|RVEDVi < 170 mL/m 2||RVEDVi > 170 mL/m 2|
|RVEDAi < 20 cm/m 2||15||0||15||PPV: 15/15 = 1|
|RVEDAi > 20 cm/m 2||18||18||36||NPV: 18/36 = 0.5|
|Sensitivity: 15/33 = 0.45||Specificity: 18/18 = 1|
|RVEDAi < 22 cm/m 2||20||1||21||PPV: 20/21 = 0.95|
|RVEDAi > 22 cm/m 2||13||17||30||NPV: 17/30 = 0.57|
|Sensitivity: 20/33 = 0.67||Specificity: 17/18 = 0.94|