Right ventricular ejection fraction is better reflected by transverse rather than longitudinal wall motion in pulmonary hypertension

The local Ethics Committee of the VU University Medical Center approved the study protocol and all participants gave written informed consent. Between September 2004 and September 2008, 658 patients were referred to our hospital for evaluation of PH. PH was diagnosed according to a standard protocol [17], including right heart catheterization (RHC), and was confirmed when the mean pulmonary artery pressure at rest was >25 mmHg and the pulmonary capillary wedge pressure was < 15 mmHg. Inclusion criteria were: patients with PH with etiologies from WHO group 1 (pulmonary arterial hypertension) or group 4 (chronic thrombotic or embolic disease), and who had undergone CMR within 14 days after RHC had been performed. In total 123 patients met these criteria. Of this group, 15 patients were excluded because some CMR cines were lacking and 7 patients due to technically inadequate images. Thus, in total 101 patients were included in the study. In addition, a total of 29 healthy non-smoking, age and gender matched control subjects were included as a reference group. These healthy subjects had had no RHC.

In a subset of patients (n = 66), cardiopulmonary exercise testing was performed on an electromagnetically braked cycle ergometer (Rehcor, Lode Groningen, The Netherlands). A progressive increase in pedaling workload (5-20 W/min) was applied until maximum tolerance was reached. Peak oxygen consumption (VO_2,peak) was considered to express the patient's exercise capacity, which was measured using a metabolic cart (Vmax 229; Viasys, Yorba Linda, CA). To ensure accuracy, the measurements were time-averaged over a minimum of 20 seconds. The equipment was calibrated according the manufacturer's specifications.

CMR was performed by means of a 1.5T Siemens Avanto MRI system (Siemens Medical Solutions, Germany), equipped with a 6-element phased-array coil. ECG-gated cine imaging was performed using a balanced steady-state free precession pulse sequence, during repeated breath-holds. Long-axis slices were acquired in the four, three and two-chamber views.

The four-chamber image plane was localized by the following steps: a basal short-axis image at end-diastole was used as the first localizer. Orthogonal to this short-axis image, the four-chamber view was obtained by rotating the planning line to such an orientation that it passes through the middle of the mitral and tricuspid valvular rings. The LV vertical long-axis view was used as the second localizer, to assure that the planned four-chamber cine passes through the most apical point of the LV cavity.

Additionally, short-axis slices were obtained with a typical slice thickness of 5 mm and an interslice gap of 5 mm, fully covering both ventricles from base to apex. The MR parameters used were: temporal resolution between 35 to 45 ms, voxel size 1.5 × 1.8 × 5.0 mm³, flip angle 60°, receiver bandwidth 930 Hz/pixel, TR/TE 3.2/1.6 ms, matrix 256 × 156.

An apical four-chamber view was used to analyze longitudinal and transverse movements of the RV myocardium (Figure 1). The software for the analysis was implemented in MATLAB release R2008a (The MathWorks, Inc., Natick, United States).

Normal distribution of the data was verified using a normal probability plot and log transformed if necessary. All data are presented as mean ± SD, unless stated otherwise. A p-value < 0.05 was considered statistically significant. Data between controls and patients were compared using the 2-tailed Students t-test for unpaired data or using one-way ANOVA when more than two groups were tested. The Fisher exact test was used for categorical data. Linear regression was performed to test the relationships between the transverse measures (SFD and fractional-SFD) and RVEF, and between the longitudinal measures (TAPSE and fractional-TAAD) and RVEF. In the regression analysis only patients were included in order to avoid regression bias by the control group.

Receiver operating characteristic (ROC) analysis was used to test sensitivity and specificity of all measures to detect a RVEF less than the median RVEF value in patients. To correct for multiple testing, the threshold for significance was adjusted using Bonferroni correction for families of tests with 0.05 divided by the amount of tests giving the adjusted threshold for significance.

Intra-observer and inter-observer variability of the endocardial wall measurements were assessed using the analysis of agreement method described by Bland and Altman [19]. To this end, the same observer repeated CMR measurements of 10 patients and 10 control subjects within a period of one month to determine the intra-observer variability. A second observer repeated the same measurements to obtain the inter-observer variability.

All statistical analyses were performed with SPSS Statistics 15.0 (SPSS Inc., Chicago, United States).

There was no difference between the 101 PH patients and 29 control subjects with respect to age (PH = 50 ± 15 years vs. control = 46 ± 19 years, p = 0.082) and gender (PH = 68% female vs. control = 67% female, p = 0.35). Table 1 summarizes the clinical and hemodynamic characteristics of the PH patients. Most patients were from WHO group 1 (80%), with the remainder being from group 4 (chronic thromboembolic pulmonary hypertension). Approximately 60% of all patients were medically treated for PH at enrollment, and many of them went through one or more regimens.

Baseline CMR measurements of PH patients and control subjects are presented in Table 2. Values of RVEF were significantly different between the groups, but no significant differences were found for LVEF. However, there was a significant difference for LVEDV between the two groups.

Figure 1 illustrates four-chamber and short-axis views of a healthy subject and two patients with moderate or severe PH.

Figure 2 shows results of fractional-SFD, evaluated for seven levels ranging from apex to base, and fractional-TAAD in patients and controls. In PH patients, virtually no regional variation in fractional-SFD was found. In contrast, control subjects showed the highest fractional-SFD at the apical segment, and the lowest at the basal segment. At all levels, there were significant differences in transverse movements between patients and control subjects (Table 3). From Table 3 it can be seen that for all subjects, SFD values are consistently smaller than TAPSE, but that this does not hold for fractional-SFD compared to fractional-TAAD due to the shorter end-diastolic transverse dimension (SF) compared to the longitudinal dimension (TAA).

No significant differences were found in longitudinal and transverse motion between the patients from WHO group 1 and 4 (data not shown).

Fractional-SFD was correlated positively with RVEF at each level of the RV free wall (Table 4), but the strongest correlations were found at level 4 (mid-level; R² = 0.70, p < 0.001; Figure 3) and level 5 (R² = 0.66, p < 0.001). Weaker relationships were found for SFD, fractional-TAAD and TAPSE (Figure 3 and Table 4).

Figure 4 and Table 5 show the results of ROC-analysis, indicating that fractional-SFD at mid RV is a sensitive and specific indicator of depressed RVEF below the median value in patients (< 35%, range 10-61%, range control subjects: 41-77%). Compared to fractional-SFD, there were statistically significant differences among the areas under the curves for SFD (p = 0.005), fractional-TAAD (p = 0.003) and TAPSE (p < 0.001) after Bonferroni correction.

In a subset of patients (n = 66) exercise capacity was measured using VO_2,peak and VO_2,peak-predicted (Table 1). Figure 5 illustrates VO_2,peak-predicted below and above the median value (44%, range: 10-110%) in comparison with fractional-SFD and fractional-TAAD. A significant difference was found for fractional-SFD (p = 0.002) but not for fractional-TAAD.

Figure 6 shows the intra- and inter-observer variability using Bland-Altman plots. The intra- and inter-observer variability for all longitudinal and transverse measures was not statistically significant.

A significant relationship between fractional-SFD and RVEF was found for all levels, with the strongest relation at mid RV level (R² = 0.70, p < 0.001). The relationship between TAPSE and RVEF, however, was much lower (R² = 0.21, p < 0.001). Although this value corresponds to what was found by Kjaergaard et al. [20] (R² = 0.23), other studies have reported higher correlations (R² > 0.38) [10, 11]. This difference might be explained by the inclusion of control subjects in their regression analysis.

From a practical perspective, the measurement of fractional-SFD appears clinical useful. A cutoff value of 17.7% was used to differentiate between patients with a low (< 35%) or preserved (>35%) RVEF (sensitivity of 84.2%; specificity of 88.1%), which had a significantly higher discriminatory power compared to SFD, fractional-TAAD or TAPSE (Figure 4; Table 5). Because RVEF is of prognostic importance [3, 4] and our results showed that fractional-SFD has a high predictive value for RVEF, fractional-SFD might be useful in clinical practice.

The importance of transverse wall movements in ejection was first acknowledged by Rushmer et al. [15]. They described that although longitudinal movements are easily identified, they are probably much less important for ejection than compression of the RV chamber by movement of the free wall toward the septum (bellows action). However, only a few studies have been performed on (regional) RV transverse wall motion and its contribution to RV function [12‐14, 21, 22] and none of these were performed in PH. Moreover, while some reports examined transverse movements just below the tricuspid valve [21, 22], our results, in PH, showed that loss of transverse motion at this level was lower compared to the other regions (Figure 2; Table 3). Evidence for a hypokinetic apex in PH was shown earlier by strain analysis employed by tissue Doppler echocardiography [23‐25]. However, this has not been observed using MR myocardial tagging [26].

Considering the transverse measures at mid RV a remarkably large difference exists between PH patients and control subjects (figure 2 and table 3). Moreover, this difference is much larger than is found for the longitudinal measures. To be able to explain these results accurately, the definitions of the measures need to be considered in more detail.

SFD is defined as the difference between end-diastolic and end-systolic SF dimension, and thereby quantifies disturbances of RV free-wall movements. Additionally, paradoxical leftward septal bowing increases the end-systolic SF dimensions and, as a consequence, also affects SFD. Thus, especially in severe patients the existence of a septal bowing may contribute to the large differences seen between patients and controls.

Fractional-SFD is defined as SFD divided by the end-diastolic SF dimension. The unusually high end-systolic pressure in the lumen of the RV in patients with PH affects the cross-sectional shape of the cavity. Whereas it normally tends to be crescent shaped in cross section, it adopts a more circular cross section when contracting against pressure that approaches or exceeds that of the left ventricle (Figure 1). This alteration of geometry mainly affects transverse rather than longitudinal dimensions, and therefore contributes more to reduction of fractional-SFD rather than fractional-TAAD. In addition to any impairment of RV myocardial contractility, this "transverse" dilation may contribute to the large difference between PH patients and controls.

Despite several anatomical studies, the relationship between fiber orientation and right-heart mechanics is still not completely clear [27, 28]. Presently, there is increasing consensus that both the septum and RV free wall play an important role in RV contraction. The septum consists of oblique longitudinal fibers with spiral architecture [29], resulting in the twisting motion required for efficient ejection against increased vascular resistance. In contrast, the predominant transverse fiber orientation of the RV free wall leads to circumferential compression or bellows action, which maintains RVEF with normal pulmonary artery pressure [28, 30, 31].

In the setting of pulmonary vascular disease, it has been shown that the fiber orientation of the septum becomes more transverse [30]. An altered fiber orientation in the RV free wall is also likely to occur as the RV dilates. This dilation is profound toward the apical segments and results in an increased apical angle (Figure 1) [32]. Changes in fiber direction in the free wall are supported by a study of Pettersen et al. [33] in which MR strain analysis was applied to patients with a systemic RV. There results indicated a predominant circumferential over longitudinal free wall shortening at mid RV, while the reverse has been observed in healthy control subjects [33, 34].

Therefore, we hypothesize that disturbed RV wall motion in PH can be explained by impaired RV myocardial contractility due to altered fiber orientation. However, further studies are needed to explain how changes of fiber geometry contribute to both longitudinal and transverse wall motion.

Some limitations of the current study should be noted. Firstly, geometry and heavily trabeculated myocardium of the RV make it sensitive to errors in the determination of endocardial definition. Secondly, four-chamber views were acquired with equal geometric orientation at end-diastole and end-systole, without accounting for through plane motion. It is unknown whether RV torsion may affect the end-systolic dimension. A previous study reported regional differences in rotation in the systemic RV. However, when values were averaged over different segments, this resulted in almost absent rotation at the basal and the apical level with no absolute global ventricular torsion [35].