Tricuspid flow and regurgitation in congenital heart disease and pulmonary hypertension: comparison of 4D flow cardiovascular magnetic resonance and echocardiography

The PH group comprised only patients with pre-capillary pulmonary hypertension – either idiopathic pulmonary arterial hypertension or chronic thromboembolic pulmonary hypertension. In all PH patients, the diagnosis of pre-capillary PH had been previously confirmed by right heart catheterization. Pre-capillary PH was defined as mean pulmonary artery pressure of ≥ 25 mmHg and a pulmonary capillary wedge pressure of ≤ 15 mmHg [16]. All PH patients were on PH-specific therapy when entering the study. The CHD cohort consisted of two major groups: 1) patients with a systemic RV either after atrial switch procedure for transposition of the great arteries or congenitally corrected transposition of the great arteries, and 2) patients with isolated valvular pulmonary stenosis (PS) or tetralogy of Fallot with pulmonary valvular or homograft stenosis. Patients with contra-indications for CMR were excluded. A group of healthy subjects between 18 and 60 years old served as control population. Subjects were screened using physical examination, medical history and electrocardiogram and excluded if these investigations or subsequent CMR and echocardiogram showed any abnormalities.

The study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki. The medical ethics committees of all participating centers approved the study and written informed consent was obtained from all participants prior to inclusion.

Demographic data, electrocardiogram (ECG), basic echocardiographic measurements and functional capacity were obtained for each patient. Echocardiography and CMR were performed consecutively with less than 1 h in between investigations. Patients and controls did not use medication between investigations, nor did they perform any strenuous physical activity.

For planning purposes 4-chamber, 2-chamber (of both ventricles) and perpendicular views of each of the four cardiac valves were used – to ensure that the 4D-flow volume would enclose all valves during diastole and systole. Velocity data were acquired in three orthogonal directions. The 4D-flow CMR acquisition was based on the protocol previously described by Westenberg et al. [14]. The following acquisition parameters were used: TR/TE 7.3/3.9 msec, respectively; field of view 370 × 219 × 63 mm; 3D volume imaging with 63-mm slab thickness reconstructed into 28 3.5 mm slices; 10° flip angle; acquisition voxel 3.43 × 3.65 × 3.5 mm; reconstructed voxel size 2.9 × 2.9 × 3.5 mm; 1 signal acquired; retrospective gating with 20-30% acceptance window, with 30 phases reconstructed during 1 average cardiac cycle; maximal velocity encoding of 150 cm/s in all three directions. To reduce acquisition time, echo planar imaging was used with a factor of 5.

Analysis of through-plane flow across the TV was performed with Mass (version 5.1, Medis) In short, the TV plane was reconstructed in each cardiac phase using two orthogonal planes, perpendicular to the flow across the TV (Fig. 1 a-d). After the valve plane was reconstructed for each phase of the cardiac cycle, the through-plane flow was reformatted in five parallel planes with a slice gap of 5 mm. Subsequently, contours were drawn outlining the flow of the valve of interest (Fig. 1-e&f) for each phase. Through-plane motion correction using the velocity of myocardium was taken into account by a indicating a region of interest in the myocardium. Finally, forward and backward flow, regurgitant fraction and effective flow were derived from 4D-flow CMR. To illustrate the impact of measurement plane reconstruction perpendicular to the flow direction in 4D-flow CMR, we performed flow analysis using a static tricuspid valvular plane in 15 random datasets (i.e. in essence comparable to 2D-flow across the TV).

To assess the accuracy of the 4D flow measurement across the TV, 4D-flow derived TV effective flow was compared to the current reference: 2D-flow derived pulmonary effective flow. Severity of TR was classified according to regurgitation fraction: 0-10% (absent/trace); 10-20% (mild); 20-40% (moderate); > 40% (severe) [18].

Echocardiography was performed using a Toshiba Artida system (Toshiba, Tokyo, Japan) with a 5-MHz transducer. Analysis was performed offline in XCelera (version 4.1.1.1133 - 2013; Philips Healthcare). RV systolic pressure (RVSP) was calculated using the Bernoulli equation with maximum velocity of TR or by maximal pulmonary valve gradient (only in PS patients with valvular stenosis), plus estimated right atrial pressure. In systemic RV patients the systemic systolic blood pressure at rest was used as RVSP. Tricuspid annular plane systolic excursion (TAPSE) and systolic velocity using tissue (TDI S′) Doppler imaging were measured in 4-chamber view.

TR was visualized in a parasternal 2-chamber RV, parasternal short axis aorta orientation, subcostal and in the modified apical 4-chamber view – if possible an apical 2-chamber-RV view was also used. To quantify TR the following parameters were taken into account: vena contracta, early tricuspid inflow velocity, hepatic vein systolic flow reversal, color Doppler flow jet area and density of TR Doppler signal [19]. The severity of TR was graded none/trace, mild, moderate or severe based on these (semi-) quantitative and qualitative assessment - by an experienced imaging-cardiologist (F.M.) blinded to 4D-flow CMR results.

All demographic and baseline data are listed in Table 1. In short, PH patients were older than the control group and CHD patients. Patients with systemic RVs had the highest RVSP. Basic volumetric measurements are also listed in Table 1; RV volumes were increased in systemic RV and PH patients compared to controls.

4D-flow and 2D-flow CMR measurements are depicted in Table 2. Acquisition time for the 4D-flow CMR dataset was 3.5-7 min; acquisition time of cine CMR images necessary for post-processing was approximately 7 min. Lastly, acquisition time for echocardiographic images related to assessment of TR was approximately 15 min. Analysis time for 4D-flow CMR across the TV was approximately 35 min per patient; analysis with a static plane (2D TV flow) was approximately 10 min. Analysis of TR grade on echocardiography took on average 6 min per patient.

TV 4D-flow CMR analysis was possible in 67/67 patients (100%). Effective flow measured across the TV by 4D-flow CMR could be compared to 2D-flow derived effective flow across the PV in 85/88 subjects, due to poor quality of 2D-flow images in three subjects (Fig. 2a-c). The intra-class correlation coefficient between both measurements was 0.90 (95% confidence interval 0.85-0.94; p < 0.001) and the R² was 0.83. Mean difference in 4D-flow derived TV effective flow vs 2D-flow derived PV effective flow was −1.6 ml (p = 0.083) with limits of agreement: −20.0 to 16.8 ml.

Analysis of TV effective flow, using a static annular plane – similar to TV 2D flow method – was performed in 15 datasets. Intra-class correlation coefficients for TV 2D-effective flow vs. PV 2D-effective flow, TV 2D-effective flow vs. TV 4D-effective flow and TV regurgitant fraction using 2D flow vs. 4D-flow were respectively 0.666 (p < 0.001), 0.767 (p < 0.001) and 0.389 (p = 0.002). Effective flow was significantly overestimated by TV 2D-flow compared to both other methods (p < 0.001; Fig. 3), regurgitant fraction was significantly underestimated (p < 0.001; Fig. 3).

Reproducibility measurements are listed in Table 3 and are shown in Fig. 4. Both intra- and interobserver measurements yielded good intra-class correlation coefficients (all > 0.91 and p < 0.001) for TV 4D-flow CMR. Mean intra- and interobserver differences in forward flow and effective flow were small with good limits of agreement (Table 3). The mean difference and limits of agreement for measurement of TR (%) were acceptable for both intra- and inter-observer measurements (1.08% [−7.90; 10.06] and −1.10% [−7.96; 5.76], respectively).

Reproducibility of echocardiography showed a kappa value of 0.57 (95% CI 0.30-0.84) for intra-observer agreement and 0.66 (95% CI 0.39-0.92) for inter-observer agreement. During second observation TR grade was classified differently in 6/15 (= 40%) of patients by the same observer, and in 5/15 (= 33%) by a second observer.

Severity of TR could be graded by 4D-flow CMR in all 67 patients, and in 65/67 (97%) patients by echocardiography. Of these patients, 40/65 patients (61.5%) showed consistent results for echo and 4D-flow CMR, but 25/65 (38.5%) were classified differently by at least 1 grade using 4D-flow CMR compared to echocardiography (Table 4). Both methods showed only moderate agreement; with a linear weighted kappa of 0.52 (95%-confidence interval 0.37-0.67).

A trace to mild TR grade by echocardiography excludes moderate or severe TV regurgitation by 4D-flow CMR in 95% of patients (=negative predictive value). However, moderate or severe TR grade by echocardiography corresponded to a moderate or severe TR grade by 4D-flow CMR in only 60% of patients (=positive predictive value). Of the 14 patients with moderate or severe TR on echocardiography or 4D-flow CMR, 2 could not be classified by echocardiography and 9/12 (75%) were classified differently by both methods.

In our patient cohort, effective flow across the TV measured with 4D-flow CMR showed excellent correlation to effective flow across the PV measured with 2D-flow CMR. This is true for both patient groups (ICC 0.870) as well as healthy controls (ICC 0.895). Conversely, TV flow measured without reconstruction of a measurement plane perpendicular to the flow (similar to 2D flow CMR), significantly overestimated effective flow volume compared to the PV effective flow. Our results are in line with previous reports on 4D-flow CMR derived TV flow in patients without RV heart disease, by Westenberg et al., and those with tetralogy of Fallot without TR, by van der Hulst et al. [14, 15]. The limits of agreement for effective flow (i.e. stroke volume) between TV 4D-flow CMR and PV 2D-flow CMR varied between −20.0 to +16.8 ml. These values are acceptable, as even for repeated 2D-flow measurements of semi-lunar valves there is considerable inter-observer and interscan variation [20, 21]. For example, Kondo et al. reported a relative difference of 7.0 ± 5.6% for repeated PV 2D-flow CMR measurements [21]. The difference between 4D-flow CMR derived TV flow and 2D-flow CMR derived PV flow can further be explained by several other factors: 1) 2D-flow suffers from some through plane movement; 2) 2D-flow across the PV was obtained during end-expiratory breath-hold while 4D-flow CMR was obtained during free-breathing and 3) interscan variability (although time between acquisitions was < 10 min).

Thus far, reproducibility measurements have been mostly obtained in patients with structurally normal hearts or focused on other applications of 4D-flow imaging (i.e. peak flow velocity and wall stress) [14, 22]. In patients with RV dilation and hypertrophy, TV geometry is distorted [4]. Hence, reproducibility of 4D-flow CMR derived TV flow in this population cannot be extrapolated from healthy subjects or patients with structurally normal hearts. The present study fills this gap and demonstrates good intra-observer and inter-observer agreement with acceptable limits of agreement and excellent ICC coefficients (all >0.90). The reproducibility in our cohort is comparable to limits of agreement reported in TV flow of healthy controls by Westenberg et al. and 2D-flow imaging of pulmonary or aortic valves [20, 21].

We demonstrate a discrepancy between 4D-flow CMR and echocardiographic TR grading in our patient cohort. Twenty-five out of 65 patients (38.5%) were classified differently by at least one grade using quantitative 4D-flow CMR TR measurements compared to echocardiographic assessment. Importantly, this number was even higher in patients with severe or moderate TR (namely 9/12 = 75% of patients). With the lack of a gold standard it is difficult to be sure which of these techniques provides the most accurate results. However, TV flow assessment by 4D-flow CMR has been validated both in vitro and in vivo [10, 14, 23] and TR grading by echocardiography has shown less consistent results [5, 7, 8], which may indicate 4D-flow CMR is more reliable. Adding to this, agreement between repeated measures was very good (ICC > 0.9) for repeated 4D-flow measurements, but only moderate to considerable (Kappa 0.57 and 0.66) for repeated TR grade by echocardiography. Importantly, a trace to mild TR grade by echocardiography still excludes moderate or severe TV regurgitation by 4D-flow CMR in 95% of patients.

As is the case for many other measurements of RV size and function, echocardiography is an excellent screening tool. However, for assessment of degree of TV regurgitation, 4D-flow CMR seems of added value. If moderate or severe TR is seen on echocardiography, additional assessment of TR by means of 4D-flow CMR may help or guide clinical decision-making. In patients in whom surgery is considered, TR grading using 4D-flow CMR may improve selection of patients that potentially benefit from TV annuloplasty [24‐27]. Moreover, the detailed anatomical information obtained with 4D-flow CMR (Fig. 1) may be helpful in identifying the mechanism of regurgitation and planning of intervention.

In both pulmonary hypertension patients and in CHD the degree of TR is related to symptoms and mortality [1‐3, 28]. Accurate determination of the degree of TR can help physicians better understand a patients’ physiology and ensure timely referral of patients for surgery or transplant.

This study has limitations. First, 4D-flow CMR post processing and analysis remains time-consuming and labor-intensive (35 min), limiting its use in clinical practice for many institutions. Currently, validation of semi-automated software is underway – shortening post-processing times and making analysis less user-dependent [29]. Furthermore, in addition to observer variability, interscan variability will need to be evaluated in future studies.

The number of patients with moderate to severe TR was limited, which may limit its application in these patients.

We chose to use 2D-flow CMR derived PV flow as reference measurement for stroke volume, introducing interscan variability and through plane motion. Invasively measured stroke volumes across the pulmonary valve were not available. We did not use 4D-flow CMR for PV flow as many of our patients had severe PS with a large degree of aliasing or a homograft – for which 4D-flow CMR is not yet validated.