3D Echo systematically underestimates right ventricular volumes compared to cardiovascular magnetic resonance in adult congenital heart disease patients with moderate or severe RV dilatation

Patients were recruited in a prospective, consecutive manner from the Adult Congenital Heart Disease out-patient program at our institution. Patients were eligible if they had undergone a CMR examination in the preceding 12 months, or if the responsible physician indicated that CMR exam would be performed in the subsequent 6 months. Recruitment was restricted to 3 groups; those with previously repaired Tetralogy of Fallot (ToF) in infancy or childhood (without subsequent pulmonary valve replacement); those with a right ventricle in the systemic position (e.g., Senning, Mustard or congenitally corrected transposition patients) and those with repaired coarctation (CoA) of the aorta. Patients within the first 2 groups were expected to have a range of RV dilatation and dysfunction, and for the purpose of analysis were treated as a single group. The coarctation group were included as an internal control arm with the expectation that right ventricular size and function would be normal or near-normal. The study was approved by the Institutional Review Board and all subjects gave written informed consent.

All patients underwent a full 2D trans-thoracic echocardiogram with particular focus on the right ventricle according to the recommendations of the American Society of Echocardiography [4]. All images were acquired by 2 British Society of Echocardiography-accredited sonographers with over 40 years of echocardiography experience between them (JF, GW). The standard 2D clinical examination was followed by additional full volume 3D image data acquisition (Philips IE33, Phillips Medical Systems, The Netherlands) by one of two experienced cardiac sonographers. Three dimensional image acquisition was performed from modified standard views (most often apical 4 chamber) in order to maximize visualization of the RV. The 3D echo datasets were analysed offline on a dedicated workstation (TomTec v1.2, TomTec Imaging Systems, Germany). Two dimensional multiplanar reconstructions of the RV were semi-automatically generated with manual correction to produce data sets in the short axis, four chamber and RV inflow-outflow orientation. End diastole and end-systole were selected manually by review of individual image phases and contouring was performed by semi-automatic border detection after manual placement of key seed points to define the RV apex and tricuspid annular plane (Figure 1). Resulting contours were checked for accuracy and corrected as necessary. Precise details of this process with the TomTec software package have been published recently [5].

CMR examinations were performed on a Philips 1.5 T Intera magnet (Philips Medical Systems, The Netherlands). Steady state free precession (SSFP) images were acquired in the short axis plane from the atrioventricular groove to the cardiac apex. Cine acquisitions were performed with vectorcardiographic ECG gating over 6-10 heartbeats at held end-inspiration. Technical parameters were: slice thickness 8 mm no gap, 25 cardiac phases, TR 3.2 ms, TE 1.6 ms, FoV 320 × 320 cm, reconstruction matrix 256 × 256 cm. CMR studies were analysed offline using QMass V6.1 software (Medis, The Netherlands). All RV contours were performed on the short axis cine stack from the pulmonary valve to the RV apex, with trabeculation assigned to the blood pool[6]. Selection of end systole/diastole was performed manually by visual assessment of smallest/largest RV cavity size in each cardiac phase. RV volumes were calculated by the method of summated discs according to Simpson's rule [7].

Data are presented as mean (± SD). Descriptive statistics were used for normally distributed data. Non parametric statistics were employed to compare differences in volumes and function where appropriate. Fishers exact test was used for the comparison of proportions. Intra- and inter- observer analysis were tested for both techniques after an interval of 8 weeks by the original readers and a further reader analyzing a random selection of 50% of the cases from each imaging modality. Intra-class correlation coefficients were calculated to rate both inter and intra-observer variability in measurements of RV size and function. The method of Bland and Altman was used for assessment of systematic bias between methods of measurement [8]. P < 0.05 was taken to indicate statistical significance.

We acknowledge several limitations to our study. Firstly our sample size was relatively small (though not dissimilar to the published literature in this area) and in particular lacked power to clearly define the progressive limitations of 3D echo in reference to graded increases in RV end-diastolic volume. Secondly, although our patient recruitment was prospective and consecutive the 3D echo was not performed on the same day as the CMR study; however the mean time difference was only 12 weeks between studies in which time one would not expect any significant systematic change in RV cavity size in a stable outpatient population. Thirdly, since 3D echo currently acquires a 'pyramid' of data over multiple sequential heartbeats we did not attempt to recruit any patient who was not in sinus rhythm. This is also a relative limitation of CMR which generally requires a stable R-R interval for routine segmented cine imaging. Fourthly we acquired CMR images in end-inspiration rather than end expiration as is done more routinely. This is because 3D echo windows were generally superior at end-inspiration and we wished to match the two techniques as far as possible. Limited volunteer data suggests this entails a risk of 'over-sampling the RV' since the end inspiratory position of the diaphragm may vary by as much as 25% [28]. However the risk of scanning the same slice twice on sequential breatholds as a result is mainly a problem for scanning in the axial plane. We contoured the RV from the short axis plane in order to obviate this concern. Finally, we acknowledge a greater experience with both acquisition and post-processing of CMR images than 3D echo, which may have introduced a learning bias into our results. This is, however, the case whenever a new modality is compared to an existing reference standard and, as shown above, our results are not discordant with those seen in several studies of both adult congenital and acquired heart disease populations.