Author + information
- Received August 8, 2006
- Revision received October 10, 2007
- Accepted October 22, 2007
- Published online February 19, 2008.
- J. Tim Marcus, PhD⁎,⁎ (, )
- C. Tji-Joong Gan, MSc†,1,
- Jaco J.M. Zwanenburg, PhD§,2,
- Anco Boonstra, MD, PhD†,
- Cor P. Allaart, MD, PhD‡,
- Marco J.W. Götte, MD, PhD‡ and
- Anton Vonk-Noordegraaf, MD, PhD†
- ↵⁎Reprint requests and correspondence:
Dr. J. Tim Marcus, Department of Physics and Medical Technology, VU University Medical Center, P.O. Box 7057, 1007 MB Amsterdam, the Netherlands.
Objectives The purpose of this study was to explore in pulmonary arterial hypertension (PAH) whether the cause of interventricular asynchrony lies in onset of shortening or duration of shortening.
Background In PAH, leftward ventricular septal bowing (LVSB) is probably caused by a left-to-right (L-R) delay in myocardial shortening.
Methods In 21 PAH patients (mean pulmonary arterial pressure 55 ± 13 mm Hg and electrocardiogram–QRS width 100 ± 16 ms), magnetic resonance imaging myocardial tagging (14 ms temporal resolution) was applied. For the left ventricular (LV) free wall, septum, and right ventricular (RV) free wall, the onset time (Tonset) and peak time (Tpeak) of circumferential shortening were calculated. The RV wall tension was estimated by the Laplace law.
Results The Tonset was 51 ± 23 ms, 65 ± 4 ms, and 52 ± 22 ms for LV, septum, and RV, respectively. The Tpeak was 293 ± 58 ms, 267 ± 22 ms, and 387 ± 50 ms for LV, septum, and RV, respectively. Maximum LVSB was at 395 ± 45 ms, coinciding with septal overstretch and RV Tpeak. The L-R delay in Tonset was −1 ± 16 ms (p = 0.84), and the L-R delay in Tpeak was 94 ± 41 ms (p < 0.001). The L-R delay in Tpeak was not related to the QRS width but was associated with RV wall tension (p < 0.05). The L-R delay in Tpeak correlated with leftward septal curvature (p < 0.05) and correlated negatively with LV end-diastolic volume (p < 0.05) and stroke volume (p < 0.05).
Conclusions In PAH, the L-R delay in myocardial peak shortening is caused by lengthening of the duration of RV shortening. This L-R delay is related to LVSB, decreased LV filling, and decreased stroke volume.
In pulmonary arterial hypertension (PAH), leftward ventricular septal bowing (LVSB) is most prominent during early left ventricular (LV) diastole (1,2,3) and impairs LV filling (4,5,6,7). Leftward ventricular septal bowing has been assessed by cine magnetic resonance imaging (MRI) and echocardiography (1,8) and can be quantified by the radius of leftward curvature (9). By MRI and tissue Doppler imaging (6,10,11), a right ventricular (RV) delay in the time to peak strain was observed in PAH. Such a left-to-right delay induces a left-to-right transseptal pressure gradient that might be the mechanism causing LVSB (12). However, in these studies it was not yet clear whether this delay was caused by delayed RV onset or prolonged RV shortening.
The cause of either delayed or prolonged RV shortening is unknown, and knowledge of this cause could have implications for treatment. A first potential mechanism is an electrical conduction delay: RV overload and concomitant remodeling might well lead to a (partial) right bundle branch block (RBBB), left-to-right electrical dyssynchrony, and subsequent mechanical dyssynchrony. This would become manifest as a delayed RV time to onset of shortening in comparison with the LV. An alternative mechanism could be initiated directly by the mechanical pressure and volume overload, inducing increased RV wall tension and prolonged RV myocardial shortening. In this case, the time to onset of shortening would be similar for both ventricles, whereas the time to peak shortening of the RV wall would be delayed compared with the LV.
For the measurement of the left-to-right differences in the timing of shortening, MRI tagging and strain analysis provide a tool for accurate regional mapping of the onset and peak times in the myocardial wall (13). By this technique, any potential difference between RV and LV in onset times and peak times, to be denoted by the left-to-right (L-R) onset delay and the L-R peak delay, can be measured.
The aim of this study is to explore in PAH whether the cause of the L-R delay lies in the onset of shortening, in the duration of shortening, or in both. The relative roles of conduction delay and prolonged shortening due to overload might then be revealed. In addition, the functional impact of the L-R mechanical asynchrony is determined by assessing its association with LVSB, LV filling, and stroke volume.
Patients and control subjects
Twenty-one patients with PAH, referred to the VU Medical Center for treatment, were recruited. The study was approved by the medical ethical committee of the VU Medical Center. Eleven healthy control subjects were included (age 38 ± 9 years, 6 women), with normal electrocardiogram (ECG) and QRS width of 80 ± 12 ms. In 2 control subjects, the pulmonary arterial pressure (PAP) was measured invasively, resulting in 24 ± 1 mm Hg, 8 ± 4 mm Hg, and 16 ± 2 mm Hg for systolic, diastolic, and mean values, respectively.
A 1.5-T Siemens Sonata whole body MRI system, equipped with a 6-element phased-array coil, was used (Siemens Medical Solutions, Erlangen, Germany). The Siemens ECG gating system was used, with no known delay between the true QRS complex and the QRS complex used to trigger the sequence. The MRI myocardial tagging with high temporal resolution (14 ms) was applied with Complementary Spatial Modulation of Magnetization (7 mm tag distance) and steady state free precession imaging. Parameters: three phase-encoding lines/beat, repetition time 4.7 ms, echo time 2.3 ms, no view sharing, flip angle 20°, voxel size 1.2 × 3.8 × 6.0 mm3. In all patients and control subjects this tagging cine was acquired in the mid-ventricular short-axis plane (13,14). In a subset of 9 patients (Patients #1 through #8 and #10) this tagging cine was also acquired in the basal and apical short-axis planes, to explore any effect of the longitudinal level on the timing of strain.
After the tagging acquisitions, the LV and RV were covered by a stack of short-axis cine MRI images, with steady state free precession imaging with a temporal resolution between 25 and 35 ms. In addition, cine images were acquired in the LV 3-chamber view showing the aortic and mitral valves and through the RV outflow tract showing the pulmonary valves. Finally, the flow was measured in the main pulmonary artery by MRI velocity quantification (temporal resolution 22 ms, velocity sensitivity 120 cm/s).
Timing parameters derived from strain
The tagged images were analyzed with the Harmonic Phase procedure (15). Circumferential shortening was calculated over time during the cardiac cycle. For the LV free wall, septum, and RV free wall, the onset time (Tonset) and peak time (Tpeak) of circumferential shortening were calculated related to the ECG R-wave by automated routines (13). Overstretch is circumferential lengthening beyond end-diastolic value (or “positive strain”). Overstretch was observed in the septum and quantified by the peak lengthening and the time to peak lengthening.
LV free wall, RV free wall, and septum definitions
The LV free wall was subdivided in 5 equal segments. The 2 segments of the LV wall that were in direct continuity with the septum were not included as part of the LV free wall. The RV free wall was delineated in the same way. The complete septum was taken for the calculation of the septal strain, from the anterior until the posterior connections with the ventricular wall. For the LV free wall, RV free wall, and septum, the strains and strain timing parameters were derived. The difference between RV and LV in time to onset of shortening is denoted by the L-R delay in Tonset and the difference in Tpeak by the L-R delay in Tpeak.
Timing parameters of the valves
The time to aortic valve closure was derived from the 3-chamber cine. The time to pulmonary valve closure was derived from the RV outflow tract cine. In Patients #8 through #11 this pulmonary valve timing was derived from the most basal short-axis cine that showed the valves during the last part of systole.
Global LV and RV parameters
The stack of short-axis cine images was used for the calculation of the LV and RV end-diastolic volumes (EDVs) and the RV end-systolic volume (ESV). The RV stroke volume was measured from the flow map in the main pulmonary artery. The maximal leftward septal curvature was measured at the most basal short-axis cine slice that still showed the LV and RV myocardium through the cardiac cycle. The septal coordinates were marked at the anterior and posterior insertions into the LV wall and at the middle of the septum. From these coordinates, the curvature was calculated (9).
RV wall tension
Our estimation of RV wall tension starts from the law of Laplace for a thin-walled sphere (16):
Wall tension = 0.5 × pressure × radius
The RV pressure during shortening is estimated by the systolic PAP. The RV-radius of curvature is difficult to measure directly because of the RV’s irregular shape. Therefore, we estimate this radius from the RV ESV by assuming that this volume can be described by a sphere in PAH patients. Then the mean RV-radius is 0.620 × (RV-ESV)1/3. Finally, to be able to compare different patients with different body sizes, the RV radius is normalized to BSA (body surface area). The estimation of RV wall tension then becomes:
RV wall tension = 0.5 × PAPsystolic × RVradius/BSA
Values are expressed as mean ± SD. First, the timing parameters were tested for a normal distribution by the Shapiro-Wilks test. Then, comparisons between LV and RV timing parameters were performed with the paired-samples t test (2-tailed). By the same test, the Tpeak in the RV wall was compared with the times of LVSB, septal overstretch, and pulmonary valve closure. Comparisons between patients and control subjects were performed by independent samples t testing (2-tailed). The relations between the L-R delay in Tpeak versus ECG-QRS width, PAP, and RV wall tension were tested by linear regression. The relations between septal curvature, stroke volume (SV), and LV EDV versus the L-R delay in Tpeak were also tested by linear regression. In these regression tests, the L-R delay in Tpeak was normalized for the R to R interval (RR) of the individual patient.
The interobserver variation was determined for the Tpeak of the RV by Bland-Altman analysis, for a subset of 10 patients.
Sixteen patients were diagnosed as having idiopathic PAH, whereas 5 had chronic thromboembolic PAH. The systolic, diastolic, and mean PAPs were 87 ± 20 mm Hg, 33 ± 9 mm Hg, and 55 ± 13 mm Hg, respectively, as measured via right heart catheterization with a Swan-Ganz catheter. The medication at the time of the MRI is listed in Table 1. The ECG-QRS width was 100 ± 16 ms. On the basis of the ECG morphology, 3 patients had an incomplete RBBB, and 1 patient had a complete RBBB. The RV SV was 53 ± 19 ml, RV EDV 202 ± 69 ml, RV free wall mass 91 ± 56 g, LV mass 114 ± 27 g, LV EDV 105 ± 28 ml, LVEF 51 ± 12%, BSA 1.85 ± 0.26 m2, and leftward septal curvature was 0.14 ± 0.05 cm−1. The LV EDV in the patients was smaller than in the control subjects (105 ± 28 ml vs. 158 ± 36 ml, p = 0.001).
Images and strains
Figure 1 shows 3-chamber cine images, short-axis cine images, and short-axis tagged images at the time of aortic valve closure and at the time of maximal LVSB. In the patients, peak circumferential shortening of the LV and RV free walls was −14 ± 4% and −14 ± 3%, respectively (p = 0.88). Peak LV circumferential shortening in the patients was smaller than in the control subjects (−14 ± 4% vs. −20 ± 2%, p < 0.001). For the septum, peak shortening was −11 ± 3%, and the maximal overstretch was 6 ± 2%. Figure 2 shows the circumferential shortening curves during the cardiac cycle for the LV and RV free walls and the septum. The LV and RV start simultaneously, but the RV reaches its peak later than the LV. The septum shows overstretch (positive shortening) at the same time that the RV reaches its peak shortening.
In Figure 3, the same plot is given for a healthy control subject. In this control, the RV peak is not later than the LV peak, and the septum does not overstretch.
The results of the timing parameters are given in Figure 4 and Table 2. In Table 2 the data from the control subjects are included. The differences between the timing parameters are presented in Table 3. As shown in Table 3, there is no L-R delay in Tonset, in contrast to the large L-R delay in Tpeak of 94 ± 41 ms (p < 0.001). In addition, TpeakRV is >Tpulmcl by 59 ± 40 ms, meaning that the RV free wall shows considerable post-systolic shortening, which does not contribute to ejection. The time of septal overstretch (Tstretch) is not different from the TpeakRV (p = 0.93). Also the time of LVSB and the TpeakRV are not different (p = 0.60). In the patients with an RBBB, the L-R delay in Tpeak was not different from the L-R delay in patients without an RBBB. In the control subjects, no L-R delay in timing was observed.
Regression analysis of timing parameters
By regression analysis, there was no relationship between the L-R delay in Tpeak and the L-R delay in Tonset (p = 0.91). Also, the L-R delay in Tpeak was not associated with the ECG-QRS width (p = 0.65) or PAP-systolic (Table 4). As shown in Table 4 and Figure 5, there was an association between the L-R peak delay and the RV wall tension (p = 0.01, r = 0.55). The L-R delay in Tpeak was related to leftward septal curvature (p < 0.05) and was negatively related to LV EDV (p = 0.02, r = 0.50) and RV stroke volume (p = 0.023, r = 0.49), as shown in Table 4.
For the subset of 9 patients with base, mid, and apical coverage, the L-R difference in Tpeak was 85 ± 35 ms, 110 ± 51 ms, and 48 ± 57 ms at basal, mid, and apical levels, respectively. The effect of the level was not significant. Also, the Tpeak was measured in the RV anterior, RV lateral, and RV inferior subregions; no effect of these subregions was found.
The interobserver variation in the Tpeak of the RV was given by a correlation coefficient of 0.88 with p < 0.001 and a bias of −5 ms with 95% confidence limits of agreement of −47 and +37 ms, respectively.
The results showed that in PAH there is a 94-ms L-R delay in Tpeak of shortening, which is caused by lengthening of the duration of RV shortening rather than a delay in the onset of RV shortening.
Cause of L-R asynchrony in Tpeak
Because the L-R peak delay was not related to an L-R onset delay or the QRS width, it is unlikely that electrical conduction delay would be the only dominant factor responsible for the L-R peak delay. Instead, the mechanism of the prolonged RV systole in PAH is probably the increased RV wall tension as shown by the correlation between L-R peak delay and RV wall tension (Fig. 5). This mechanism is supported by measurements in rat cardiac trabeculae, which provided evidence that an increased load of myocytes leads to a slower and prolonged shortening velocity (17).
The mechanism of LVSB is now better documented: maximal LVSB coincides with peak RV shortening and overstretch of the septal wall. Thus it is unlikely that there is compression of the septum, as suggested earlier (5). The overstretch indicates that LVSB is a result of higher pressure in the RV than in the LV, owing to the ongoing shortening in the RV free wall, whereas the LV free wall is already in its relaxation phase.
Impaired RV systole
The observed RV SV (53 ± 19 ml) is lower than normal reference values of 88 ± 19 ml (18) and showed a negative correlation with the L-R peak delay. This effect of the L-R peak delay can be explained: owing to the mechanical asynchrony between the RV free wall and the septum, the RV contraction is very inefficient in its late phase. As shown in the results, the Tpeak of the RV free wall is 120 ms after the Tpeak of the septum, and when finally the RV free wall reaches its peak shortening, the septum is shifted to the left. The observation that the TpeakRV is 59 ms later than the closure of the pulmonary valves (post-systolic shortening) further illustrates the inefficiency of this last part of RV myocardial shortening. The effective systole of the RV is estimated to begin at the onset of RV shortening (mean 52 ms) and to run until pulmonary valve closure (mean 328 ms), thereby taking 276 ms. With 59 ms post-systolic shortening, total RV shortening time is 335 ms, and thus (59/335) × 100% = 18% of total RV shortening time is wasted, and energy is dissipated in the nonfunctional LVSB. Thus the observed L-R asynchrony in PAH can be considered as an independent factor that has a negative effect on RV stroke volume. This is in line with earlier Doppler-echo observations showing that RV dyssynchrony is related to RV dysfunction in PAH (10,11,19) and also in line with the architectural disadvantage of a leftward bowed septum (20).
In principle the loss of RV forward stroke volume might also be caused by tricuspid regurgitation. However we found no significant difference between the RV SV derived from (RV EDV − RV ESV) and the RV SV derived from forward flow. This indicates that the volumetric contribution of tricuspid regurgitation to the loss of forward flow was minor.
Impaired LV diastole
As mentioned in the results, the L-R delay in Tpeak predicted leftward septal curvature and thereby had a negative effect on LV EDV. The negative relation between LVSB and LV filling was shown earlier in larger patient groups (7,12). This is confirmed in the present study, where the septum bowed maximally to the left at 395 ms, whereas the aortic valves had already been closed at 311 ms. Our observed negative association between L-R delay in Tpeak and LV EDV supports the concept that the L-R asynchrony plays a key role in the LV filling impairment. The impaired LV filling and the ineffective RV systolic function both contribute to the loss of stroke volume, as displayed in the flowchart in Figure 6.
The measured values of LV EDV were shown to be smaller than the healthy control values. Also the observed LV free wall peak shortening was smaller than in the healthy control subjects. This is well explained by the Frank-Starling effect: the LV myocardial muscle fibers are not stretched to their optimal length and thus are not able to perform their optimal shortening. This underfilled LV further contributes to the leftward septal bowing and inefficient RV systole. In addition, Gurudevan et al. (21) evaluated several different indicators of LV filling and showed that the LV underfilling is in large part responsible for the impaired LV relaxation pattern.
In clinical practice, understanding the meaning of LVSB is relevant, because it is much easier to measure than the strain-derived properties of LV, RV, and septal wall. The LVSB can directly be measured and timed from MRI cine imaging (9) or from echocardiography. This easily obtained timing of maximal LVSB coincides with the RV time to peak shortening. The time to LV peak circumferential shortening can be estimated by the time of aortic valve closure. Thus the L-R delay in peak shortening can be estimated from the time interval between aortic valve closure and LVSB. This provides an easy measure to follow individual PAH patients during treatment, also when MRI is not available.
Another potential implication can be derived from the key role that the L-R delay in peak shortening plays in the loss of LV and RV performance. Although conduction delay is not the cause of the L-R mechanical delay, the mechanical synchrony between the LV and RV might be improved by earlier electrical activation of the RV free wall with pacing. This early activation will shift the RV contraction period to an earlier time in the cardiac cycle and possibly also shorten it. Both effects might reduce RV post-systolic shortening, thereby improving both the RV efficiency and LV filling. However, this has not been proven. Whether this might be an effective approach to improve cardiac function in PAH must be tested first in animals, before it might be considered in patients.
The role of electrical conduction delay was indirectly estimated from the ECG-QRS width and the onset times of shortening. The true role of conduction delay still needs more exploration.
The RV load was estimated by the wall tension. The assumption that the RV in PAH can be described by a sphere needs confirmation. The estimation might be improved by taking the wall thickness into account to calculate wall stress. However, the RV wall thickness is difficult to define, owing to its trabeculated endocardial border.
In PAH, no L-R delay was observed in the onset times of shortening, whereas a large L-R delay was observed in the times to peak shortening. Thus the L-R delay in myocardial peak shortening is caused by lengthening of the duration of RV shortening. An increased RV wall tension rather than electrical conduction delay is related to this interventricular mechanical asynchrony. Because wall tension is the product of pressure and radius, it means in practice that those PAH patients with an increased RV pressure combined with an enlarged RV volume will have increased RV wall tension and thereby more L-R mechanical asynchrony in Tpeak. This asynchrony is associated with leftward septal bowing, LV underfilling, and decrease in RV stroke volume.
- Abbreviations and Acronyms
- body surface area
- end-diastolic volume
- L-R delay
- left-to-right ventricular delay in circumferential shortening
- left ventricle/ventricular
- leftward ventricular septal bowing
- magnetic resonance imaging
- pulmonary arterial hypertension
- pulmonary arterial pressure
- right bundle branch block
- right ventricle/ventricular
- time to onset of circumferential shortening
- time to peak of circumferential shortening
- Received August 8, 2006.
- Revision received October 10, 2007.
- Accepted October 22, 2007.
- American College of Cardiology Foundation
- Dong S.J.,
- Smith E.R.,
- Tyberg J.V.
- Beyar R.,
- Dong S.J.,
- Smith E.R.,
- Belenkie I.,
- Tyberg J.V.
- Nelson G.S.,
- Sayed-Ahmed E.Y.,
- Kroeker C.A.G.,
- et al.
- Vonk Noordegraaf A.,
- Marcus J.T.,
- Gan C.T.,
- Boonstra A.,
- Postmus P.E.
- Gan C.T.,
- Lankhaar J.W.,
- Marcus J.T.,
- et al.
- Stojnic B.B.,
- Brecker S.J.,
- Xiao H.B.,
- Helmy S.M.,
- Mbaisourroum M.,
- Gibson D.G.
- Zwanenburg J.J.,
- Gotte M.J.,
- Kuijer J.P.,
- et al.
- Norton J.M.
- Van Heuningen R.,
- Rijnsburger W.H.,
- Ter Keurs H.E.
- Saleh S.,
- Liakopoulos O.J.,
- Buckberg G.D.
- Gurudevan S.V.,
- Malouf P.J.,
- Auger W.R.,
- et al.