Author + information
- Received July 11, 2013
- Revision received January 9, 2014
- Accepted January 19, 2014
- Published online July 8, 2014.
- Benjamin C.F. Smith, MSc∗,†,
- Gary Dobson, MDCM, MSc†,‡,
- David Dawson, MSc∗,
- Athanasios Charalampopoulos, MD§,
- Julia Grapsa, MD, PhD∗,§ and
- Petros Nihoyannopoulos, MD∗,†∗ ()
- ∗Department of Cardiology, Hammersmith Hospital, Imperial College Healthcare NHS Trust, London, United Kingdom
- †School of Medicine, Imperial College, London, United Kingdom
- ‡Department of Anesthesiology, University of Calgary, Calgary, Alberta, Canada
- §National Pulmonary Hypertension Service, Hammersmith Hospital, Imperial College Healthcare NHS Trust, London, United Kingdom
- ↵∗Reprint requests and correspondence:
Prof. Petros Nihoyannopoulos, Department of Cardiology, Hammersmith Hospital, Du Cane Road, London W12 0HS, United Kingdom.
Background Quantitative assessment of right ventricular (RV) systolic function largely depends on right ventricular ejection fraction (RVEF). Three-dimensional speckle tracking (3D-ST) has been used extensively to quantify left ventricular function, but its value for RV assessment has not been established.
Objectives This study sought to prospectively assess whether 3D-ST would be a reliable method for assessing RV systolic function and whether strain values were associated with survival.
Methods Comprehensive 2-dimensional echocardiographic assessment, 3D-ST of the RV free wall, and measurement of RVEF was performed in 97 consecutive patients with established pulmonary hypertension (PHT) (RVEF 31.4 ± 9.6%, right ventricular systolic pressure [RVSP] 76.5 ± 26.2 mm Hg) and 60 healthy volunteers (RVEF 43.8 ± 9.4%, RVSP 25.9 ± 4.3 mm Hg).
Results Area strain (AS) (−24.3 ± 7.3 vs. −30.8 ± 7.2; p < 0.001), radial strain (23.2 ± 14.4 vs. 34.9 ± 18.2; p < 0.001), longitudinal strain (LS) (−15.5 ± 3.8 vs. −17.9 ± 4.4; p = 0.001), and circumferential strain (CS) (−12.2 ± 4.5 vs. −15.7 ± 6.1; p < 0.001) were all reduced in patients with PHT, compared with normal individuals. AS and CS strongly correlated to RVEF (r = 0.851, r = −0.711; p < 0.001). Systolic dyssynchrony index was greater in PHT (0.14 ± 0.06 vs. 0.11 ± 0.07; p = 0.003) and correlated to RVEF (r = −0.563, p < 0.001). AS (hazard ratio [HR]: 3.49; 95% confidence interval [CI]: 1.21 to 7.07; p = 0.017), CS (HR: 4.17; 95% CI: 1.93 to 12.97; p < 0.001), LS (HR: 7.63; 95% CI: 1.76 to 10.27; p = 0.001), and RVEF (HR: 2.43; 95 CI: 1.00 to 5.92; p = 0.050) were significant determinants of all-cause mortality. Only AS (p = 0.029) and age (p = 0.087) were predictive of death after logistic regression analysis.
Conclusions PHT patients have reduced RV strain patterns and more dyssynchronous ventricles compared with controls, which was relatable to clinical outcomes. AS best correlated with RVEF and provides prognostic information independent of other variables.
Serial echocardiographic examinations allow for monitoring anatomic and hemodynamic changes in patients with pulmonary hypertension (PHT). Right ventricular (RV) dysfunction in PHT is an important predictor of longevity (1), and deterioration in RV systolic function is crucial in patient management and follow-up.
Accurate assessment of RV systolic function is challenging by any method and more so using conventional echocardiography. Currently, it is recommended to use basic quantitative methods to evaluate RV dysfunction (2), though each has inherent drawbacks, limiting their specificity and reliability (3). Longitudinal strain (LS) of the RV has been measured in 2 dimensions (4–6) in populations with PHT, but the data are intrinsically limited within the 2-dimensional (2D) slice plane and, therefore, suffers from out-of-plane data loss.
Three-dimensional speckle tracking (3D-ST) is in principle better, compared with 2D-ST, because it is not slice-plane limited and delivers vectored data in 3 orthogonal planes from 1 analysis. There are as yet no published 3D-ST studies of the RV utilizing the association of 3D strain with clinical outcomes, and there are no suitably scaled studies of area, radial, or circumferential RV strain in patients with PHT. In this study, we hypothesized that: 1) 3D-ST would be a reproducible quantitative method for assessing RV systolic function; 2) RV strain may be reduced in patients with PHT; 3) RV strain may be related to measures of RV systolic function and pulmonary pressures; and 4) RV strain may be related to survival.
Patients were prospectively recruited via referral from the National PHT Centre at Hammersmith Hospital (London, United Kingdom) as part of their regular diagnostic assessment and follow-up. Disease etiology was categorized according to the 2008 Dana Point Classification (7). All patients with pulmonary arterial hypertension were treated with an endothelin receptor antagonist, a phosphodiesterase-5 inhibitor, or intravenous/inhaled prostacyclin analogs, or their combination, according to current guidelines (8).
Criteria for inclusion into the study were measurable tricuspid regurgitant velocity ≥2.7 m/s; established diagnosis of PHT (mean pulmonary arterial pressure from a catheter study of ≥25 mm Hg); and sinus rhythm. Patients with cardiomyopathy, congenital heart disease, arrhythmias, or suboptimal imaging were excluded. Asymptomatic healthy volunteers also were recruited from the community as controls and included if they were in sinus rhythm, had normal biventricular systolic and diastolic function, had structurally normal valves, and had measurable tricuspid regurgitation of <2.7 m/s. Patients were followed up after a minimum of 12 months with mortality data and assessed at 24 months. All subjects provided informed consent. The study was approved by the responsible ethics committee.
Comprehensive echocardiographic examination of the RV was performed by a senior echocardiographer accredited by the British Society of Echocardiography according to the most recent guidelines for the examination of the right heart (2).
Acquisition and analysis of 3D full-volume captures
Two 3D full-volume datasets for the RV were acquired using the matrix array 3D transducer with a center frequency of 3 MHz (PST-25SX, Toshiba Medical Systems Europe BV, Zoetermeer, the Netherlands), using the tissue harmonic mode with a 4-beat–triggered capture. The image was guided by a 5-plane view taken in the apical window and aligned so the entire RV was incorporated into the full-volume dataset. The depth was minimized, so only the chamber, walls, and tricuspid annulus filled the volume.
3D-ST was performed using the Toshiba 3DT speckle tracking software on the Toshiba Aplio Artida ultrasound system (version 3.00). Because the software is designed for the left ventricle, a modified methodology was devised, as detailed in the following text.
The long-axis A-plane was adjusted, cutting directly down the center of the RV, dividing the septum and RV free wall in half, and through the distal apical portion. The long-axis B-plane was aligned as close as possible to the interventricular septum and through the distal apical portion (Fig. 1), so that the anterior segments (labeled b, c, d in Fig. 2) could reliably be differentiated as RV free-wall. The remaining segments labeled (a, e, f in Fig 2) were discarded, because they could not reliably be differentiated into RV free wall or septum by the software.
Using the 3DT software and with the cineloop frozen to the end-diastolic frame, the endocardial border was manually traced in the biplane views and adjusted in the 3 coronal planes (Fig. 1). The epicardial border was then adjusted to correspond closely to the wall thickness. The software then tracked the speckles from the 3D dataset and calculated strain, volumes, and 3D right ventricular ejection fraction (RVEF). Peak systolic strain was used for analysis.
The systolic dyssynchrony index of the RV free wall was calculated from the segmental area strain (AS) in a similar manner to that described for the left ventricle (9). The time to the minimal AS for each segment was recorded and normalized to the cardiac period. Within each subject, the standard deviation of the 7 normalized segmental times was the systolic dyssynchrony index, a measure of dispersion of contraction.
Sample size determination
The determination of sample size was on the basis of the initial data of 28 patients and 10 controls that were previously reported (10). The sample size was calculated as n = 97, in order to achieve 90% power with significance of type I error: α = 0.01 and delta value of 0.4 (variance 5%) to detect a significant difference in RV strain between patients with PHT and healthy volunteers.
Data are expressed as the mean ± SD for normally distributed data, and as median with interquartile range when variables are not normally distributed. Normal distribution of each variable was assessed by a histogram and the Kolmogorov-Smirnov test. Comparison between 2 groups for categorical variables was by Fisher exact test, and by independent samples t test for continuous variables. Correlation between variables was assessed by Spearman correlation coefficient. The sensitivity and specificity of strain vector cutoff points were decided by the criterion value indicated by the Youden index from the receiver-operating characteristic (ROC) curve to predict significant RV impairment (RVEF <40%) and mortality outcome. Kaplan-Meier plots to assess mortality risk were then formed using the cutoffs found. The contribution of factors to outcome was assessed with logistic regression using a stepwise backward likelihood ratio method of regression. The p value above which factors were removed was 0.10. Test–retest reproducibility for radial strain (RS), LS, and circumferential strain (CS) was by a Bland-Altman analysis. All tests are double sided, and the cutoff value of statistical significance was 0.05. Statistical analyses were performed using SPSS version 17.0 (SPSS, Chicago, Illinois), except for the ROC and Kaplan-Meier tests, which were performed using MedCalc version 12.4.0 (MedCalc Software, Ostend, Belgium).
Between November 2009 and July 2012, 138 PHT patients and 107 healthy volunteers were prospectively enrolled into the study. Forty-one patients and 47 controls were excluded (Fig. 3), leaving 97 PHT patients and 60 controls for analysis. The baseline characteristics for both groups are presented in Table 1. Patients were on average older than controls. There were no subjects with severe tricuspid regurgitation or any degree of pulmonary stenosis. Of the patients, 60 (62%) were diagnosed with pulmonary arterial hypertension, 31 (32%) with chronic thromboembolic PHT, and 6 (6%) secondary to left heart disease.
On 3D full-volume imaging, the volume rate was less in the controls than it was for patients (13.7 ± 0.8 volumes/s vs. 15.1 ± 2.6 volumes/s; p < 0.001), but mean resolution did not differ between groups (0.48 ± 0.06 mm/pixel vs. 0.49 ± 0.06 mm/pixel; p = 0.513).
RV strain values
In patients, RV strain was universally reduced for all vectors (Table 2). When strain vectors were plotted against RVEF (Fig. 4), there were moderate-strong correlations for all strain vectors, with the strongest correlations for AS (r = −0.850, p < 0.001) and CS (r = −0.707, p < 0.001), then LS (r = −0.540, p < 0.001) and RS (r = 0.444, p < 0.001). Looking at the patient population in isolation, AS (r = −0.865, p < 0.001) and CS (r = −0.759, p < 0.001) remained strongly correlative, with lesser relationships for LS (r = −0.547, p < 0.001) and CS (r = 0.284, p = 0.005).
Tricuspid annular plane systolic excursion (TAPSE) weakly correlated with LS (r = −0.355, p < 0.001), AS (r = −0.293, p < 0.001), and RS (r = 0.249, p = 0.002) across both arms. For the PHT group alone, there were slightly better correlations for LS (r = −0.462, p < 0.001), AS (r = −0.366, p < 0.001), and RS (r = 0.269, p = 0.008), with CS reaching statistical significance (r = −0.208, p = 0.043), albeit weakly correlative.
Age was only weakly correlative with LS (r = 0.177, p = 0.022) and AS (r = 0.164, p = 0.035) across the entire study population. There were no significant correlations for age in patients.
Right ventricular systolic pressure (RVSP) was estimated from tricuspid regurgitant velocity in all patients and controls, and RV end-diastolic pressure could be estimated in a subset of 77 PHT patients and 38 controls who had measureable pulmonary regurgitation. RVSP weakly correlated to RS (r = −0.329, p < 0.001) and AS (r = 0.286, p < 0.001), then LS (r = 0.196, p = 0.014) and CS (r = 0.189, p = 0.017). Within the PHT patients alone, RVSP was only weakly correlated to RS (r = −0.213, p = 0.037).
AS (r = 0.328, p < 0.001) and RS (r = −0.310, p < 0.001) correlated best with RV end-diastolic pressure, with weaker correlations for LS (r = 0.203, p = 0.025) and CS (r = 0.189, p = 0.017). There were no significant correlations for RV end-diastolic pressure when patients were evaluated in isolation.
There were no differences between the 3 Dana Point subcategories of PHT for RVSP, RV end-diastolic pressure, and RVEF, or for any of the strain vectors.
Indices of segmental dyssynchrony for area strain
Normalized time to peak segmental AS was longer for PHT patients than it was for controls (0.48 ± 0.08 vs. 0.44 ± 0.08; p = 0.004). The systolic dyssynchrony index was greater in PHT patients than in controls (0.14 ± 0.06 vs. 0.11 ± 0.07; p = 0.003) and was inversely correlated to ejection fraction (r = −0.566, p < 0.001; and r = −0.561, p < 0.001 for patients alone). These data suggest that patients with PHT had more dyssynchronous ventricles than controls, and that dyssynchrony was related to RV dysfunction.
Mortality outcome data
Mean follow-up time was 33 ± 8 months (14 to 44 months). During this time frame, 21% of the patients from the PHT arm died (20 of 97). Nonsurvivors had reduced AS (−21.2 ± 5.8% vs. −25.1 ± 7.4%; p = 0.033), CS (−10.1 ± 3.1% vs. −12.8 ± 4.7%; p = 0.020), LS (−13.7 ± 3.6% vs. −16.0 ± 3.7%; p = 0.016), and RS (17.4 ± 11.6% vs. 24.7 ± 14.6%; p = 0.040). There were no significant differences between survivors and nonsurvivors for RVEF, RVSP, TAPSE, and systolic dyssynchrony index.
Strain vectors and RVEF were entered into a ROC model to evaluate probability of 24-month mortality (Table 4, Fig. 5). AS, CS, and LS cutoffs most accurately predicted mortality. Kaplan-Meier survival probability was then assessed using the ROC cutoff values (Table 5, Fig. 6). AS, CS, LS, and RVEF were significant determinants of all-cause mortality.
TAPSE, RV systolic annular velocity, RVEF, age, sex, and minimum AS were entered as independent variables into a logistic regression model and processed using a stepwise backward-likelihood ratio method. This resultant model was statistically significant and contained only AS (p = 0.029) and age (p = 0.087) as predictors of death, suggesting the superiority of the 3D-derived AS over other variables.
Intraobserver variability in patients and controls
Repeated measures of RS, LS, and CS values were performed in 117 of 159 PHT patients and controls who had 2 suitable 3D full-volume captures, with intraobserver variability as assessed by Bland-Altman graphs presented (Table 6, Fig. 7). When the 2 repeated measures were compared, the most reproducible strain vector was LS, then CS, followed by RS.
This is the first comprehensive study to our knowledge using 3D-ST as a novel quantitative measure of RV function in PHT patients relating to clinical outcome. All strain vectors were significantly reduced in PHT patients, with strong associations for AS and CS with RVEF, and with lesser, but still significant, correlations for RS and LS (Central Illustration). There were also moderate correlations for AS and LS with TAPSE.
The strong correlative findings for AS and RVEF are interesting, given that this is a composite vector representing deformation of the endocardial surface (11) and may prove to be an important addition to the echocardiographic assessment of RV systolic function. Cutoff values for impaired RVEF for AS and CS might be valuable for risk stratifying patients and help with therapeutic decision making.
Reduced strain is also associated with adverse outcome (12,13). Fine et al. (13) reported increased mortality in those with decreased RV LS in a cohort of 575 patients referred for echocardiographic assessment for PHT. Complementing this, we also found that nonsurvivors had significantly reduced strain values, not only longitudinally, but also in the transverse plane. Radial and circumferential strain for the RV has not previously been reported elsewhere.
Our study also defines cutoff values, which may be clinically relevant for mortality risk, where lower-magnitude strain values below the cutoffs for AS, CS, and LS in particular were associated with considerable increases in risk. Consequently, it may be suggested that improved RV strain may reduce mortality. This was illustrated in a cohort of patients with pulmonary arterial hypertension, where medical therapy improved LS in some patients, and a 5% improvement in LS had a >7-fold reduction in mortality risk at 4 years (14). Anecdotal evidence from studies of bilateral lung transplantation (15) and post-endarterectomy patients (4) suggests physiological changes in RV loading conditions in impaired RVs also have immediate positive consequences for RV systolic strain.
RV dyssynchrony in PHT
Our study demonstrated that patients with PHT had more dyssynchronous RV contraction compared with controls, with increased dyssynchrony for worsening impairment. This is a new way of measuring dyssynchrony in the RV and confirms previous observations in PHT patients (16). The additional appeal of this measurement is that it is calculated from AS data already obtained, so in the future, it could be added into existing analysis packages.
Anatomy of the RV free wall and strain
Until recently, strain analysis of the RV has focused on the longitudinal direction because of the fact that the RV free wall is best visualized in the apical view, whereas avoiding transverse strain directions because of its thin cross section. This has been justified because of the predominant longitudinal arrangement of myocardial fibers (17) and the inference that longitudinal motion is the predominant contributor to RV systolic function.
In both patients and controls, CS had the smallest magnitude of the 3 tangential vectors. Anatomically, this makes sense, given that the arrangement of RV myocardial fibers does not lend itself to circumferential shortening because of the lack of the middle layer (3). Despite this, the strong association with systolic impairment may be an important observation in the failing ventricle. It is possible that this reduction in strain is due to circumferential stretch and subsequent reduction in contractility. Because the myocardial arrangement is oblique in the superficial subepicardium (17), this finding may signify failure of the outer layer to contract circumferentially.
Similarly, RV anatomy may explain the predominance of radial and longitudinal deformation in both our cohorts. The superficial, obliquely arranged subepicardial fibers create an inwards wave-like contraction (radial axis), whereas the deeper, longitudinally arranged subendocardial fibers produce a wave-like base-to-apex (longitudinal) contraction (17,18). RS also is said to be the predominant variable in systolic volume variation (19) and had the largest reduction in magnitude for all the strain vectors when patients were compared with controls.
Relation with RV systolic pressures
Given the multitude of influences on RV strain, it is not surprising that the correlations for strain and RV pressures were not as strong as for those of functional parameters. We found only moderate correlations for RV strain vectors and RVSP and RV end-diastolic pressure but a definite difference between patients and controls. This is in line with 2D-ST studies in PHT patients, who demonstrated reduced LS and slightly stronger, but still moderate, correlations of LS to RVEF (5,6). As yet, there are no published data correlating to RS, CS, or AS for the RV.
Comparison with previous 2D studies
There are no large-scale studies or indexing to sonomicrometry for the categorization of normal RV strain values. Instead, normal LS by 2D-ST is derived from a meta-analysis of healthy controls from a series of 5 studies (2). In this meta-analysis, the mean LS value of −29% with a lower limit of −18% in healthy patients was greater than mean LS for controls in this study (−15.5 ± 3.8%). This overestimation is consistent with the discordant values found in studies of the left ventricle when comparison was made between 2D and 3D methods (20). Given the complex anatomy of the RV, the measurement of strain by 3D-ST methodology may better reflect the true nature of contraction, because it includes a larger portion of the RV and includes more strain vectors.
Segmentation of the RV free wall
This study was performed using proprietary software installed onto the Toshiba Artida Aplio. The 3DT function allows for analysis of all myocardial segments and all strain vectors using the data from a single 3D full-volume capture. Originally developed for the left ventricle, segmentation is somewhat limiting for use in the RV. It is not possible to define the location of the septal–free wall border with regard to the fixed segmentation, making it difficult to reliably separate each segment from each other. It was therefore decided to exclude the septum and segments, which could not be defined as purely RV free wall from analysis, because their inclusion could have tainted the data.
Although several studies have included the septum in the calculation of RV “global” strain (5,6), it is our opinion that this was not appropriate, given the RV dysfunction characterized in this population. The ventricular septum is divided into left and right components, which cannot reliably be differentiated by echocardiography (3). Therefore, any inclusion of the septum will necessarily include left ventricular strain values that may not directly affect the right ventricle. At one extreme, reduced septal strain in patients with concomitant left ventricular failure could further feed the decrease in “global” RV strain (5).
One of the perceived limitations of 3D-ST is the low temporal and spatial resolution, which could underestimate the peak strain value, but will not hide the pattern of contraction. Its strength lies in the fact that there are more speckles per segment and less speckles lost to out-of-plane motion, so temporal resolution may not be as important. Given that 3D-ST does not need to constantly source “new” speckles, the measurement of strain might be regarded as being more accurate.
A large proportion of strain analyses had to be discarded because of inadequate imaging with 21% of PHT patients and 36% of controls being excluded. This included patients with poor apical acoustic windows, image dropout, and in some cases, the RV being too large for the 3D sector.
Despite there being no direct validation for use in the RV, 3D-ST has been validated for RS, LS, CS (21), and AS (11) of the left ventricle using sonomicrometry with good correlation. At present, there is only pooled data for normal values for LS using 2D-ST (2). The future use of 3D-ST would benefit from validation using sonomicrometry as has been done for the left ventricle (21), but this is an invasive measurement that is not appropriate for use in humans. More appropriate might be 3D magnetic resonance imaging tagging for future validation studies.
This is the first study to our knowledge using 3D-ST to characterize RV systolic function in a cohort of patients specifically with PHT. The PHT patient population had reduced RV strain and more dyssynchronous ventricles than controls. Within the PHT patient group, those with more abnormal strain values had poorer outcomes. Reduced AS, LS, and CS were associated with increased mortality risk. The new measurement of AS had strong associations with RVEF, whereas only AS and, to a lesser degree, age were predictors of death, suggesting the superiority of 3D–derived AS over other variables. Thus, AS may be an important additional measurement in the assessment of RV systolic function in predicting outcomes.
COMPETENCY IN PATIENT CARE: Patients with pulmonary hypertension are at risk of developing right heart failure, and more accurate methods for assessment of RV function could improve clinical outcomes.
COMPETENCY IN MEDICAL KNOWLEDGE: Evaluation of RV function by echocardiography and cardiac magnetic resonance imaging is challenging because the RV walls are thin and the chamber is crescent shaped. Three-dimensional speckle tracking echocardiography addresses some of the limitations of these techniques.
TRANSLATIONAL OUTLOOK 1: Speckle tracking data may help risk stratify patients with pulmonary hypertension and guide clinical management. Future studies may permit earlier detection of RV dysfunction.
TRANSLATIONAL OUTLOOK 2: Proprietary speckle tracking software has been designed for evaluation of the left ventricle. Development of software specifically for assessment of the RV could improve the diagnostic value of this modality.
In addition to the PHT patients and controls who volunteered for this study, the authors acknowledge the support of the staff in the Hammersmith Hospital Echocardiography Department (London, United Kingdom), with particular thanks to Taryn Coulter, Alejandro Rendon-Sanchez, and Wing-See Cheung, who assisted with patient recruitment and imaging. The authors also would like to thank Willem Gorissen of Toshiba Medical Systems (Zoetermeer, the Netherlands) for providing technical information.
The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- area strain
- circumferential strain
- longitudinal strain
- pulmonary hypertension
- receiver-operating characteristic
- radial strain
- right ventricle/ventricular
- right ventricular ejection fraction
- right ventricular systolic pressure
- speckle tracking
- tricuspid annular plane systolic excursion
- Received July 11, 2013.
- Revision received January 9, 2014.
- Accepted January 19, 2014.
- American College of Cardiology Foundation
- Rudski L.G.,
- Lai W.W.,
- Afilalo J.,
- et al.
- Sheehan F.,
- Redington A.
- Simonneau G.,
- Robbins I.M.,
- Beghetti M.,
- et al.
- Galie N.,
- Hoeper M.,
- Humbert M.,
- et al.
- Smith B.,
- Grapsa J.,
- Dawson D.,
- et al.
- Haeck M.L.,
- Scherptong R.W.,
- Ajmone Marsan N.,
- et al.
- Fine N.M.,
- Chen L.,
- Bastiansen P.M.,
- et al.
- Ho S.Y.,
- Nihoyannopoulos P.
- Meier G.D.,
- Bove A.A.,
- Santamore W.P.,
- et al.
- Carlsson M.,
- Ugander M.,
- Heiberg E.,
- et al.
- Seo Y.,
- Ishizu T.,
- Enomoto Y.,
- et al.