Impact of surgical pulmonary valve replacement on ventricular strain and synchrony in patients with repaired tetralogy of Fallot: a cardiovascular magnetic resonance feature tracking study

Subjects in this study were all participants in a previously reported prospective, randomized, single-center trial comparing 2 techniques for surgical PVR [22]. These patients were > 10 years of age with repaired TOF or similar physiology presenting for PVR between February 2004 and October 2008. Other inclusion criteria in the prior study included chronic PR (regurgitation fraction by CMR ≥25%) and at least 2 of the following conditions: RV EDV index ≥160 ml/m², RV ESV index ≥70 ml/m², RV EF ≤45%, LV EDV index ≤65 ml/m², RV outflow aneurysm, or a clinical criterion such as exercise intolerance, symptoms and signs of heart failure, or cardiac medications. Exclusion criteria included any of the following: severe RV outflow obstruction, severe RV hypertension with RV pressure ≥ systemic pressure, and contraindications to preoperative CMR. Patients were randomized to PVR alone or a combination of PVR with RV remodeling surgery. The study protocol included a pre-operative and a 6-month post-operative CMR study.

For the current study, we analyzed the database of the prospective trial, selected only those patients with TOF, and performed feature tracking analysis of the CMR images. The Boston Children’s Hospital Committee on Clinical Investigation granted permission for this study and waived the requirement for informed consent.

The CMR protocol used for patients with repaired TOF has been previously published [23]. Briefly, studies were performed with a 1.5-T whole body scanner (Achieva, Phillips Healthcare Systems, Best, the Netherlands or TwinSpeed, GE Healthcare, Milwaukee, Wisconsin, USA) using surface coils selected based on patient size. The scanner manufacturer was different for the pre- and post-PVR CMR study in 4 patients. Imaging included breath-hold, electrocardiographically-gated, balanced steady-state free precession cine CMR acquisitions in 4-chamber and short-axis planes. A total of 12−14 slices were obtained in the short-axis plane to completely cover both ventricles. In all views, 30 images per cardiac cycle were acquired which yielded a temporal resolution of 20–40 ms, depending on heart rate. Ventricular volumes and blood flow were measured using commercially available software (QMASS and QFLOW, Medis, Leiden, the Netherlands) [23].

Feature tracking analysis for determination of strain was performed using commercially available software (Diogenes v. 1.1.02, TomTec Imaging Systems, Unterschleissheim, Germany), as previously described [12]. Circumferential strains for the LV and RV were measured from short-axis views at 3 levels: basal, mid-ventricular, and apical. The mid-ventricular level was first identified, and the basal and apical slices were specified to be equidistant from the mid-ventricular slice. The LV outflow tract was not included in the basal slice. Pre- and post-PVR images on each subject were compared to ensure similar slice locations. For short-axis views, the myocardium of each ventricle was divided equally into 6 segments at the mid and basal locations, and 4 segments at the apical location (Fig. 1a). Longitudinal strains for the LV and RV were measured from the 4-chamber view with division into 6 segments (Fig. 1b). Global peak strain for each view was calculated as the average of the peak strains for each segmental curve [17, 24]. Strain measurements were performed by the same investigator in all patients in order to promote consistency. In addition, they were repeated by that investigator and performed by a second investigator in a random sample of patients to assess intraobserver and interobserver agreement. Note that strain is calculated as follows: (final length – initial length) / initial length. Thus, shortening in the longitudinal and circumferential directions yields a negative strain number, and that the more negative the number, the better the shortening.

Various ventricular synchrony parameters based on feature tracking measurements have been reported; however, there is no consensus on the optimal approach. Hence, we used 3 techniques that have been reported in the literature as measures of ventricular synchrony [17, 25, 26]. Three ventricular synchrony parameters were calculated for each ventricle on each of the 3 short-axis views and the 4-chamber view based on the strain versus time data: 1) the maximum difference in time-to-peak strain among any 2 of the segments (latest versus earliest segment), 2) the standard deviation of the time-to-peak strain values for all segments, and 3) the cross-correlation delay (CCD). The time to peak strain was measured from the QRS trigger to the peak strain time. Using a custom built virtual instrument (LabVIEW 8.2, National Instruments, Austin, Texas, USA), the CCD was measured by iteratively shifting 1 curve in time relative to a second curve in a stepwise fashion, and calculating the normalized correlation coefficient between the curves for each time shift [25, 27]. The time shift that resulted in the maximum correlation coefficient was defined as the CCD between the 2 curves. Rather than relying on just the peak points, CCD takes into account the entire strain versus time curve to determine the temporal offset. For each ventricle on each of the 4 views, the CCDs were measured between opposing segments, and the largest value of the opposing segment CCDs was reported. In addition, the interventricular CCD on each of the 4 views was calculated by comparing the global strain versus time curves for the LV and RV.

Of the 64 patients in the initial prospective trial, 36 had a diagnosis of TOF as well as pre- and post-operative CMR studies with suitable image quality for strain analysis. These patients constitute the study group, and their demographic and clinical data are summarized in Table 1. Thirteen patients were randomized to the PVR alone group and 23 to the PVR with RV remodeling surgery group. CMR examinations occurred at a mean of 4.5 ± 3.8 (range 0.03–12.7) months before PVR surgery and 7.3 ± 2.1 (range 5.2–10.8) months after surgery.

Key diagnostic testing results pre- and post-PVR are shown in Table 2. As expected, PVR led to virtual elimination of PR, and a significant decline in RV EDV and ESV. These changes yielded a slight decline in RV EF. The LV EDV and ESV both increased slightly with no significant change in the LV EF. There was no significant change in QRS duration or the heart rate at the time of CMR.

Global strain values pre- and post-PVR are shown in Fig. 2. In the LV, global circumferential strain at the base and apex improved following PVR; strain at the mid ventricle also tended to improve but this did not reach statistical significance. Global longitudinal strain was unchanged. In the RV, global circumferential strain at all 3 levels and global longitudinal strain were unchanged.

The impact of PVR on segmental strain in the LV and RV is shown in Fig. 3. Overall, changes in regional stain were minimal. Significant differences in circumferential LV strain were seen in 2 of 16 segments, and these were both improvements. Longitudinal strain changed in 1 of 6 segments and this was a decline. Significant differences in circumferential RV strain were seen in 2 of 16 segments, 1 increase and 1 decrease. Longitudinal strain changed in 1 of 6 segments and this was a decline.

Synchrony parameters derived from the segmental strain versus time curves pre- and post-PVR are shown in Table 3. For LV circumferential strain, all 3 parameters showed significantly improved synchrony post-PVR at the base. At the mid and apex levels, all parameters tended to improve and some met the significance threshold. There were no significant synchrony changes based on longitudinal strain.

For the RV, both of the time-to-peak parameters showed significantly improved synchrony at the base and worsened at the apex and longitudinally. At the mid location, there were no significant synchrony changes.

In addition to the synchrony parameters reported above, the raw time-to-peak strain values (rather than differences) for each segment were compared pre- and post-PVR. This analysis provides information on how PVR affects the timing of contraction relative to the cardiac cycle rather than assessing synchrony per se. All LV segments with a statistically significant change in time-to-peak strain following PVR were increases and these were 2 of the 6 longitudinal and 7 of 16 circumferential segments (Fig. 4a). Similarly, all RV segments with a statistically significant change in time-to-peak strain after PVR were increases, and these were 1 of the 6 longitudinal and 4 of the 16 circumferential segments (Fig. 4b).

Interventricular synchrony was assessed by comparing the global strain versus time curves for the LV and RV on each of the 4 views, and calculating the CCDs. The LV peak strain preceded RV peak strain in 128 of 144 segments (89%) pre-PVR and in 127 of the 144 segments (88%) post PVR. The interventricular CCD significantly increased after PVR at the mid-ventricular level (50.4 ± 26.4 ms vs. 66.7 ± 29.1 ms, p = 0.002), as a result of slightly earlier LV peak strain. There was no change in the interventricular CCD at the base (76.5 ± 53.8 ms vs. 90.5 ± 56.2 ms, p = 0.62), the apex (46.8 ± 34.4 ms vs. 49.0 ± 33.2 ms, p = 0.73), or the 4-chamber view (61.8 ± 109.6 ms vs. 52.3 ± 86.7 ms, p = 0.69).

Table 4 shows the intraobserver and interobserver agreement results for global strain and synchrony parameters. Intraobserver agreement was good for LV circumferential and longitudinal strain, and for RV circumferential strain (all ICCs ≥0.93), and acceptable for RV longitudinal strain (ICC 0.85). Interobserver agreement for these parameters was lower. Compared to global strains, time-to-peak strain and CCD generally had lower ICCs; nevertheless, intraobserver ICCs for these LV synchrony parameters were all ≥0.73. Agreement for RV synchrony values derived from longitudinal strain curves was only fair.

Only a few other reports have measured changes in strain that accompany PVR. A similar study from our group examined the impact of transcatheter pulmonary valve implantation on strain assessed by CMR feature tracking [17]. All patients (n = 31) had RV-to-pulmonary artery conduit dysfunction, and were divided into those with predominant pulmonary stenosis (n = 18) and those with predominant PR (n = 13). For the PR subgroup, PVR led to not only improved LV circumferential strain, as it did in our patients with surgical PVR, but also improved longitudinal strain. This difference in longitudinal strain response may be related to the fact that even in the PR subgroup, there was some degree of pulmonary stenosis, more than in the current study population. There was no change in RV circumferential or longitudinal strain with transcatheter PVR in the PR subgroup, in agreement with our results for surgical PVR.

Prior studies assessing the impact of PVR on ventricular synchrony are also scarce. Lurz et al. examined 20 patients with RV-to-pulmonary artery conduit obstruction, most with TOF, before and after transcatheter PVR [14]. Based on tissue Doppler echocardiography from an apical 4-chamber view, the time-to-peak strain for the RV free wall decreased and for the LV free wall was unchanged, leading to significantly less interventricular mechanical delay following PVR. Again, the presence of RV outflow tract obstruction confounds the comparison to our results. However, similar to our study, the QRS duration was unchanged highlighting the importance of assessing both mechanical and electrical measures of ventricular synchrony.

The clinical implications of improved LV circumferential strain and synchrony following PVR are uncertain but there are some data to suggest that this may be beneficial. Impaired LV systolic function in repaired TOF patients is fairly common [33, 34]. Both reduced LV global function as well as deformation indices have been shown to be important independent predictors of clinical deterioration, arrhythmia, and death in adults after TOF repair [3, 4, 24, 29, 34‐36]. Moreover, our group has shown that LV dyssynchrony is associated with ventricular tachycardia and death in repaired TOF patients [12]. It is thus possible that the improvements in these parameters seen with PVR may lead to a better prognosis, and further study is warranted. For the RV in TOF, global systolic dysfunction is common and associated with adverse outcomes [4, 24]. Abnormalities in RV deformational indices are also prevalent [10, 37, 38]. Our data showed no improvement in any of these parameters with PVR. This finding suggests that strategies other than PVR may need to be pursued to achieve meaningful changes in outcomes related to impaired RV function. To this end, our data supports efforts to maintain pulmonary valve function and optimize myocardial preservation at the time of initial repair as later PVR cannot be relied on to restore function. Similarly, antifibrotic pharmacologic therapy should be explored as studies have implicated fibrosis in the pathophysiology of myocardial dysfunction [39‐41]. The lack of improvement in RV functional indices may be because restoration of normal loading conditions occurred after the myocardium had sustained irreversible damage. Accordingly, future studies with larger numbers should assess whether earlier PVR leads to a better RV response. Lastly, our data showed that PVR did not lead to a convincing improvement in interventricular synchrony. Thus, for repaired TOF patients in whom interventricular dyssynchrony is thought to be playing a clinically important role, cardiac resynchronization therapy [42] and not simply PVR, may be needed. Future studies which focus on patients with marked interventricular dyssynchrony who undergo PVR are needed to definitely address this proposition.

Several limitations of this study are worth noting. Assessments were performed at a mean of 7 months following PVR. Available data suggests that RV volumes and, thus, EF remain stable between 1 and 5 years after PVR [43]. Nevertheless, we cannot exclude the possibility that changes in strain and synchrony might develop with longer follow-up even in the absence of prosthetic valve dysfunction. The distal RV outflow tract wall was not analyzed because in many patients it consisted of thin patch material from the initial TOF repair and was not amenable to feature tracking analysis. The temporal resolution of the cine CMR data was relatively low; consequently, small changes in synchrony parameters may have gone undetected. However, as the temporal resolution for each patient was similar on the pre- and post-PVR CMR studies, the synchrony changes which were identified are likely valid. The small sample size of the study may have limited the ability to detect small changes in strain and synchrony parameters. Similarly, the power of our study was not sufficient to assess the influence of age at initial repair and age at PVR on the impact of PVR. Finally, a comparison between the PVR alone versus the PVR with RV remodeling cohorts was not reported owing to the small numbers in each subgroup. We acknowledge that RV remodeling may have some impact on regional mechanics, but our study was not adequately powered to address this question.