Determining right ventricular (RV) function is challenging because of the complex anatomy of the right ventricle. Three-dimensional echocardiography (3DE) has achieved better estimation, but underestimations of volumes and ejection fraction (EF) has often been reported, and no previous study has synthesized these data. The investigators performed a meta-analysis on the bias and examined the related factors.
Studies comparing RV volumes and/or EF between 3DE and magnetic resonance imaging were eligible. A meta-analysis was performed to evaluate the systematic bias. The related bias was investigated using univariate and multivariate regression analysis.
Twenty-three studies including 807 subjects revealed underestimation of RV volumes ( P < .00001) and EF ( P = .03). Larger volumes and EF were associated with more underestimation. Older patient age was associated with overestimation of volumes and underestimation of EF.
This meta-analysis found underestimation of RV volumes and EF by 3DE and factors affecting the bias. These data provide a more detailed basis for improving the accuracy of 3DE for further clinical application.
Determining accurate right ventricular (RV) function is gaining more clinical importance, as RV dysfunction is associated with increased morbidity and mortality in patients with a variety of cardiopulmonary diseases. However, evaluation of the right ventricle using conventional two-dimensional echocardiography is still technically challenging because of its complex anatomic structure. Magnetic resonance imaging (MRI) has been used as the gold standard of quantitative RV assessment, but it is not used routinely in clinical settings because it is expensive and sometimes contraindicated. Three-dimensional echocardiography (3DE) is considered theoretically ideal for estimation of RV volumes and function, and in fact, its accuracy and reproducibility are better than those of two-dimensional echocardiography. This new echocardiographic method has been reported to be in good agreement with validation studies in vitro and in vivo. In clinical studies, however, it has often been reported that 3DE underestimates RV volumes and ejection fraction (EF) compared with MRI. These were relatively small studies, and there have been some contradictory reports that showed nonsignificant bias or overestimation in some of the values, creating a major discrepancy in this field. To date, there has been no systematic attempt to synthesize the current data on the existence, extent, and factors of the bias.
Therefore, the purpose of this meta-analysis was to assess the bias of 3DE compared with MRI in evaluating RV volumes and EF. In addition, we aimed to investigate patient, software, and hardware characteristics affecting the systematic bias.
We searched Medline using Medical Subject Headings through May 2010. The key words as text words and Medical Subject Heading terms used for the literature search were “magnetic resonance imaging,” “right ventricle,” and “3-dimensional echocardiography” and relevant phrases. Articles that included all these three key phrases were screened by title and abstract for relevance. The search was restricted to human studies in English. References of relevant articles were also reviewed. Abstracts without following full-text publications, reviews, comments, letters, and literature that was not original articles were excluded ( Figure 1 ).
Studies were considered eligible if they assessed the difference in RV end-systolic volume (ESV), end-diastolic volume (EDV), and/or EF between 3DE and MRI showing the mean difference and its standard deviation or if these data could be calculated from the original article. Two reviewers (Y.J.S. and T.S.) independently retrieved the full text of the relevant articles to assess eligibility.
Demographics included year of publication, number of subjects, mean age, background disease and conditions, hardware and software used for 3DE, and methods of endocardial contour tracking. Mean differences in RV ESV, EDV, and EF with standard deviations between 3DE and MRI were obtained from the text, tables, or graphs. Raw numeric data for each patient were obtained directly if the articles included the numeric data. If data were presented graphically, crude numbers were obtained from the plotted points on the graph. If one article included different methods or populations of 3DE validated by MRI, the results for each method and population were considered separate individual studies.
The principal summary measures were the differences of means in RV ESV, EDV, and EF between 3DE and MRI. The relationship between these two measurement methods and elements affecting the bias were also investigated.
Meta-Analysis of the Systematic Bias
Meta-analysis was performed using a random-effects model weighted by inverse variance to evaluate the systematic bias of RV ESV, EDV, and EF between 3DE and MRI. Subgroup analysis was performed for prespecified study groups : the use of matrix-array transducers, the use of a semiautomatic endocardial tracking system, mean patient age > 18 or < 18 years, and mean EDV > 200 or < 200 mL. Heterogeneity was assessed using Cochrane’s Q test via a χ 2 test and was quantified with the I 2 test. Testing for publication bias was performed using funnel plots.
Relationship Between Measurements by 3DE and MRI
Bland-Altman analysis was performed to assess relationship between the values of RV ESV, EDV, and EF measured by 3DE and MRI. Cutoff points of underestimation and overestimation were determined by the points at which the 95% confidence interval curves crossed the line of no difference. Studies that did not include graphical or numeric data of 3DE and MRI for each individual were excluded from this and the following analyses.
Factors Affecting Systematic Bias
Univariate regression analysis followed by multiple regression analysis was performed to investigate the factors related to the bias in RV ESV, EDV, and EF. The dependent variable was the difference in measurements between 3DE and MRI, and the independent variables included the values measured by MRI, patient age, use of a matrix-array transducer, and use of a semiautomatic contour tracking system. Only variables with P values ≤ .10 on univariate analysis were included in the multivariate analysis. The independence of the variables was evaluated by the variance inflation factor. A maximum variance inflation factor < 5 was used as a cutoff for independence.
Continuous values are represented as mean ± SD, whereas discrete variables are shown as percentages. Microsoft Excel 2007 (Microsoft Corporation, Redmond, WA) with the Excel Statistics add-in and Review Manager version 5.0 (The Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen, Denmark) were used for the analysis. A p value < .05 was considered statistically significant.
|Study ID||Investigator||Number of pts||Age (y) †||Pt characteristics||3DE view and method||Hardware||Software||Slice thickness||Tracking of endocardium|
|1||Vogel et al. (1997)||16||11 (5, 26.2)||3 pts with normal heart, 3 after surgical repair of CoA, 9 s/p ToF repair, and 1 s/p Mustard repair of complete ToGA||Apical or subxiphoid view, rotational or “fan-like” scanning to gain cross-sectional images, then reconstructed||Vingmed 800 annular array sector scanner §||TomTec #||2 mm||Manual|
|2||Papavassiliou et al. (1998)||13||5.9 (1.4, 12.9)||10 pts s/p ToF repair, 2 pts s/p hypoplastic left-heart syndrome repair, and 1 pt s/p ASD repair||Apical or subcostal view, internal rotation at increments of 5°-10° to gain cross-sectional images, then reconstructed||Internally rotating 5-MHz omniplane transthoracic transducer with Sonos 2500 ‖||TomTec 3.0 #||3–3.5 mm||Manual|
|3||Fujimoto et al. (1998)||15||28||Healthy volunteers||Apical view, rotating the transducer in 2° increments around 180° to gain cross-sectional images, then reconstructed||Toshiba SSH 160A ultrasound scanner ¶||TomTec #||2 mm||Manual|
|4||Prakasa et al. (2006)||36||Unable to calculate ∗||8 pts with ARVD/C, 20 family members of pts with ARVD/C, 8 pts with idiopathic ventricular tachycardia||Apical and modified apical views, matrix-array transducer with Sonos 7500 or an iE33 ultrasound machine ‡||TomTec 4D Echo-View, # not specified manual, 2 views|
|5||Nesser et al. (2006) (TEE)||20||57.6 (74, 23)||20 pts referred for TEE: 5 pts were normal, 6 pts had CAD (2 were post-MI), 3 had DCM, 1 had MVP with MR, 1 had muscular dystrophy, 1 had secundum ASD, 1 had aortic regurgitation, 1 had mitral valve prosthesis with TR, 1 had MS, 2 of 20 had AF||Midesophageal view, rotating transducer by 2° increments through 180° to gain cross-sectional images, then reconstructed||Standard 5.0-MHz multiplane transducer and ultrasound scanner (Vingmed CFM800 § )||TomTec 3.0 #||10–12 slices per study||Manual|
|6||Nesser et al. (2006) (TTE)||20||57.6 (74, 23)||Same as above||Parasternal, apical, or subcostal including low parasternal and a modified apical view, rotating transducer by 5° increments through 180° to gain cross-sectional images, then reconstructed||Standard 2.5-MHz transducer housed in a carriage device that allowed rotational movement with ultrasound scanner (Vingmed CFM800 § )||TomTec EchoScan 3.0 #||10–12 slices per study||Manual|
|7||Kjaergaard et al. (2006)||30||68.7 ± 9||Pts with history of inferior ST elevation MI, or history of pulmonary embolism and persistent symptoms at 6-mo follow-up, plus 10 healthy subjects||Apical views||s3 and x3 transducers with Sonos 7500 ‡||TomTec 4D Echo-View #||10 mm||Not specified|
|8||Gopal et al. (2007) (DS)||71||56 ± 14.3||Healthy volunteers||Modified apical view with DS method||2-MHz to 4-MHz matrix-array transducer ‡||TomTec #||10 mm||Manual|
|9||Gopal et al. (2007) (AR)||71||56 ± 14.3||Healthy volunteers||Modified apical view with AR method||2-MHz to 4-MHz matrix-array transducer ‡||TomTec #||7–10 slices per study||Manual|
|10||Jenkins et al. (2007)||50||62 ± 11||Pts s/p MI; all pts had LV regional wall motion abnormalities, with 70% of pts undergoing angiography; the majority had multivessel CAD; 41 pts had inferior involvement||Modified apical view||X4 matrix-array transducer with Sonos 7500 ‡||TomTec 4D Echo-View #||15° per slice||Semiautomatic|
|11||Niemann et al. (2007)||30||39 ± 22 for normal, 9 ± 6 in congenital heart disease||14 pts with grossly normal cardiac anatomy and 16 pts with congenital heart disease (3 ToF, 2 ToGA s/p Mustard, 2 ASD, 5 VSD, 2 CoA, 1 TA, 1 atrioventricular canal defect, 1 PDA)||Apical or subcostal views||Full matrix-array 3D transducer with Sonos 7500 ‡||TomTec 4D Echo-View #||Not specified||Semiautomatic|
|12||Lu et al. (2008)||18||10.6 ± 2.8 (6, 18)||Healthy volunteers||Apical view||2-MHz to 4-MHz X4 matrix-array transducer with Sonos 7500 ‡||TomTec 4D Echo-View #||5 mm||Manual|
|13||Iriart et al. (2009)||34||31 ± 14 (17, 54)||22 surgically repaired ToF and 15 healthy volunteers||Modified apical view||1-MHz to 4-MHz 4Z1c matrix-array transducer with Sequoia ultrasound system ∗∗||TomTec 4D #||Not specified||Semiautomatic|
|14||Khoo et al. (2009) (DS)||28||16.5 (12, 25)||Pts with congenital cardiac disease||Modified apical window with DS method||X3-1 or X7-2 transducers, Phillips ie33 ‡||TomTec EchoView 5.4 and 4D RV analysis #||10 mm||Manual|
|15||Khoo et al. (2009) (AR)||28||16.5 (12, 25)||Pts with congenital cardiac disease||Modified apical window with AR method||X3-1 or X7-2 transducers, Phillips ie33 ‡||TomTec EchoView 5.4 and 4D RV analysis #||22.5°||Manual|
|16||Khoo et al. (2009) (ABD)||28||16.5 (12, 25)||Pts with congenital cardiac disease||Modified apical window with automated border detection method||X3-1 or X7-2 transducers, Phillips ie33 ‡||TomTec EchoView 5.4 and 4D RV analysis #||NA||Semiautomatic|
|17||Khoo et al. (2009) (MAB)||28||16.5 (12, 25)||Pts with congenital cardiac disease||Modified apical window with manual adjustment of the detected borders||X3-1 or X7-2 transducers, Phillips ie33 ‡||TomTec EchoView 5.4 and 4D RV analysis #||NA||Semiautomatic|
|18||Grewal et al. (2010)||25||35 ± 14||22 pts with previous intracardiac repair of ToF and 3 pts with severe PR related to previous pulmonary valvotomy||Apical window||Full matrix-array 3D transducer||TomTec 4D #||Not specified||Semiautomatic|
|19||Grapsa et al. (2010) (PAH)||60||42.8 ± 18.3||60 PAH||Apical views||X-4 transducer||TomTec 4D #||7 mm||Semiautomatic|
|20||Grapsa et al. (2010) (Normal)||20||39 ± 16.2||20 normal subjects||Apical views||X-4 transducer||TomTec 4D #||7 mm||Semiautomatic|
|21||Sugeng et al. (2010)||28||53 ± 18||9 CHF, 7 secondary PAH, 5 primary PAH, 4 congenital heart disease, 3 CAD||Apical window||X3-1 matrix-array transducer||TomTec 4D RV #||NA||Semiautomatic|
|22||van der Zwaan et al. (2010)||50||26.3 ± 10.5||ToF (21), PS (5), PA (3), ToGA (7), aortic valve etiology (10), and others||Modified apical view||X3-1 transducers, Phillips ie33 ‡||TomTec 4D RV 4.0 #||Not specified||Semiautomatic|
|23||Leibundgut et al. (2010)||88||50 ± 16||IHD (19%), idiopathic CM (11%), myocarditis (25%), ARVD/C (13%), storage disease (4%), aortic dilatation (8%)||Modified apical view||X3-1 matrix-array transducer||TomTec 4D RV #||NA||Semiautomatic|
∗ Original article included 8 of 23 pts with ARVD/C. The mean age of these 23 pts was 38 ± 10 years. The mean ages of family members and pts with idiopathic ventricular tachycardia were 30 ± 13 and 39 ± 10 years, respectively.
Meta-Analysis of the Systematic Bias
There was significant underestimation by 3DE for RV ESV (−5.5 mL, P < .00001), EDV (−13.9 mL, P < .00001), and EF (−0.9%, P = .03) ( Figures 2 A– 2 C). There was no significant trend in accuracy over time. Subgroup analysis revealed more underestimation of the volumes in studies with mean ages < 18 years and with mean EDVs > 200 mL, whereas the use of matrix-array transducers and semiautomated tracking systems were not significantly associated with the bias. There was significant heterogeneity ( I 2 = 91%, 96%, and 80% for RV ESV, EDV, and EF, respectively; heterogeneity by Q test, P < .05 for all measurements) among the included studies. Funnel plots showed no asymmetry.
Relationship Between Measurements by 3DE and MRI
Raw data could be extracted from 15 of the 17 articles (excluding studies by Gopal et al. and Jenkins et al. ); therefore, a total of 488 subjects were included in this analysis (the population in the individual study ranged from 13 to 88; average, 31). The Bland-Altman analysis between measurements by 3DE and MRI with 95% confidence interval of the estimated mean revealed significant underestimation by 3DE in RV ESV, EDV, and EF when these values were >62 mL, >89 mL, and >44%, respectively. Overestimation of RV ESV was observed when the volume was <25 mL ( Figures 3 A– 3 C).