Left Ventricular Ejection Fraction and Volumes: It Depends on the Imaging Method

Background and Methods In order to provide guidance for using measurements of left ventricular (LV) volume and ejection fraction (LVEF) from different echocardiographic methods a PubMed review was performed on studies that reported reference values in normal populations for two-dimensional (2D ECHO) and three-dimensional (3D ECHO) echocardiography, nuclear imaging, cardiac computed tomography, and cardiac magnetic resonance imaging (CMR). In addition all studies (2 multicenter, 16 single center) were reviewed, which included at least 30 patients, and the results compared of noncontrast and contrast 2D ECHO, and 3D ECHO with those of CMR. Results The lower limits for normal LVEF and the normal ranges for end-diastolic (EDV) and end-systolic (ESV) volumes were different in each method. Only minor differences in LVEF were found in studies comparing CMR and 2D contrast echocardiography or noncontrast 3D echocardiography. However, EDV and ESV measured with all echocardiographic methods were smaller and showed greater variability than those derived from CMR. Regarding agreement with CMR and reproducibility, all studies showed superiority of contrast 2D ECHO over noncontrast 2D ECHO and 3D ECHO over 2D ECHO. No final judgment can be made about the comparison between contrast 2D ECHO and noncontrast or contrast 3D ECHO. Conclusion Contrast 2D ECHO and noncontrast 3D ECHO show good reproducibility and good agreement with CMR measurements of LVEF. The agreement of volumes is worse. Further studies are required to assess the clinical value of contrast 3D ECHO as noncontrast 3D ECHO is only reliable in patients with good acoustic windows.

Assessment of left ventricular (LV) function and volumes is the corner stone of cardiac diagnostics. Several imaging methods are in clinical practice and one would assume that these methods would provide the same results. However, if the same patient with stable conditions is investigated with different methods, different results are obtained which may have an impact on patient management. This review gives an overview on comparative studies between echocardiographic modalities and cardiac magnetic resonance imaging (CMR), which is regarded as the reference method for LV volumes and ejection fraction (LVEF). 1 It will cover the normal val-ues, comparative studies with CMR and a critical assessment of the reproducibility data. A recent article by Dorosz et al. 2 has provided an extensive overview and a meta-analysis on studies comparing 2D and 3D echocardiography with CMR. Our review is complementary as it includes contrast echocardiography and provides a more comprehensive section on reproducibility, which has a major impact for clinical use.

Methods:
Normal values were selected from the guideline papers of the echocardiographic and radiological scientific societies or from articles which established the normal values. A pubmed review was carried out-including 18 studies, encompassing 1299 patients-comparing studies which included patients with abnormal LV function. Unlike the Dorosz article, the following review had an inclusion criteria of 30 patients or more per study and included contrast echocardiography and several more recent investigations in which Bland-Altman (BA) analysis was used as the method of calculating agreement with CMR in different patient groups. In these studies both patients with normal and abnormal hearts were included. In abnormal hearts the differences between the different imaging methods may become even more relevant.
BA analysis provides 2 parameters which allow to assess the agreement between different methods which measure the same parameter-bias and limits of agreement (LOA). Bias means the measurements with a specific echocardiographic technique are systematically different from the CMR measurements, which are regarded as the reference standard. A bias can be positive (= overestimation compared with the CMR measurements) or negative (= underestimation compared with the CMR measurements). For example, a bias of 5% means that the echocardiographic method overestimates the CMR measurements on an average by 5%. The LOA represent the degree of accuracy between the echocardiographic measurements and the CMR measurements. The LOA are calculated by 2 (or 1.96) standard deviations (SD) of the differences and covers the range of values which includes 95% of all the differences between the echocardiographic and CMR measurements. For example, with a bias of -5% and 2 SD = 10% the range is À15 to +5%. The smaller the LOA range the better is the agreement of the echocardiographic method when compared with CMR. 3 In the review, we also included studies which used other method for assessment of the inter-and intra-observer variability such as intra-class correlation coefficients (ICC), and coefficient of variability (CV).

Results: Normal Values:
The normal values for LV volumes and LVEF are shown in Tables I-III which have been extrapolated from different references (mean AE 2SD), with particular importance placed on the lower limits of LVEF, as these are clinical indicators for LV impairment. There are major discrepancies between modalities and processing techniques. There are also differences between gender and various ethnicities. [4][5][6][7][8][9][10][11][12][13][14][15] Studies Comparing Echocardiographic Methods with CMR: Only 2 multicenter studies have been performed to compare CMR with echocardiographic   which highlights the importance of adequate training in the utilization of 3D ECHO for LV function assessment. 17 There are 16 single center studies which satisfy the inclusion criteria of this review; 3 of which included ≥100 patients (highlighted by a box in Figs. 1-3). Whereas native and contrast echocardiography can be performed with all state of the art scanners, there are currently only 4 commercially available systems for 3D ECHO with their specific analysis software (Philips Ultrasound Ltd., Bothell, WA, USA, GE Healthcare, Salt Lake City, UT, USA, Siemens Healthcare, Erlangen, Germany, and Toshiba America Medical Systems Inc., California, USA). In addition, there is one commercially available software for off line analysis (TomTec Imaging Systems, Untershleissheim, Germany). No studies have been performed to compare measurements obtained on scanners from different manufacturers. The single center studies included a total of 1087 patients and healthy volunteers. The bias and LOA for measurements of LVEF, EDV, and ESV are shown in Figures 1-3. [16][17][18][19][20][21][22][23][24][25][26][27][28][29][30][31][32][33] The findings in single and multicenter studies can be concluded as follows: 1 2D contrast echocardiography is superior to 2D noncontrast echocardiography regarding agreement of volume and LVEF measurements. The volumes measured with 2D contrast echocardiography are closer to the corresponding CMR measurements than those obtained with noncontrast echocardiography. Contrast 2D echocardiography is particularly useful in patients with poor acoustic windows. 16,19,20 2 Most studies showed the superiority of noncontrast 3D over noncontrast 2D ECHO. [16][17][18][19][20][21][22][23][24][25][26][27][28][29][30][31][32][33] In particular the measured volumes deviated less from the CMR    method compared with biplane Simpson obtained from a 3D dataset. Voxel count was also superior to multiplane measurements of LVEF. 23,25 There was no significant difference in LVEF estimation between the QLAB and TomTec voxel methods. 29,31 However, the TomTec volume measurements were closer to CMR than the QLAB measurements. 24   infusion during 3D ECHO acquisition. The LVEF was not different with both methods, but the agreement of EDV and ESV became worse when a contrast agent was used, the bias (LOA) for contrast EDV was À34 mL compared to À5.7 mL for native 3D ECHO. It was suggested by the authors that this negative impact of values relative to the reference method may have been due to bubble destruction, resulting from the high density of scanlines required for full volumetric acquisition. 26 In a recent study of Thavendiranathan and colleagues the reproducibility of noncontrast 3D ECHO exceeded that of 2D and 3D contrast echocardiography. But, this study included only patients with good image quality and no CMR measurements were performed. 34 Observer Variability in the Comparative Studies: Different methods of statistical analysis were used to assess the reproducibility of tests between 2 different observers and of repeat tests for the same observer; ICC, BA method, CV, and percentage difference of the mean (MD). 3,35 Pearson's correlation coefficient was also used in one article. Generally, using ICC as the statistical test for assessing reproducibility, 3D ECHO was more reproducible than 2D ECHO. With BA analysis there was an obvious difference between the 2 methods, although, BA was not used very often for comparison in 3D ECHO. The most frequently used test for 3D ECHO was MD, defined as the absolute difference between corresponding repeated measurements expressed in percentage of their mean, which showed an improvement of 3D ECHO as compared with 2D ECHO (Tables IV-VI). The reproducibility of CMR measurements is better than that measured with noncontrast 2D ECHO in most studies. But with contrast echocardiography there are only minor differences in particular for LVEF-in the multicenter study of Hoffman et al. contrast 2D ECHO had a better variability using ICC than did CMR (0.91 vs. 0.86, respectively) when the onsite reader and 2 offsite readers were compared. 16 The inter-and intraobserver variability of CMR measurements are dependent on the expertise of the readers. 36 The inter-observer variability of LVEF measurements can be improved from 7.2% to 3.7% after training. 36 In Table VI the studies are listed in which the reproducibility was reported for CMR. The variability of computed tomography (CT) and radionuclide ventriculography (RNV) reported by separate studies are listed in Table VII; all of which reported excellent variability except one study which reported a correlation coefficient of 0.6 for LVEF by RNV. This is not surprising as CT utilizes the same methods for border delineation and volume calculation as CMR using images with higher spatial resolution. 37 Discussion: This is the most comprehensive review on echocardiographic methods for measurement of ejection fraction. We have indeed included the entire spectrum of available echocardiographic methods for assessment of LV function. A recent review and meta-analysis of Dorosz et al. 2 provided an excellent summary of native 2D and 3D ECHO, but did not include other frequently used technologies such as contrast echocardiography. There are only 2 more recent articles comparing 2D and 3D ECHO with CMR (Chang et al. and Thavendiranathan et al. 27,33 ). That warrants no new meta-analysis on this topic. For the comparison of contrast echocardiography or M mode echocardiography with CMR there were not enough studies to justify a meta-analysis.

Normal Values:
It is important to have a reference point from which to compare values. In a perfect world one normal range should apply for all cardiac imaging tools in calculating ejection fraction and volumetric measurements. However, it is becoming apparent that due to the differences in methodology and algorithms between diagnostic modalities a fixed value is not possible, and so it is necessary to develop a range of normal values corresponding to specific modalities. This may even be the case for various software packages used in the same diagnostic tool (Tables I-III). The reference values are based on studies involving cohorts as low as 60 patients ranging up to 1200.
In circumstances such as monitoring of treatment with potentially cardiotoxic drugs (trastuzumab), accurate assessment of LVEF is crucial. However, if measurements are used interchangeably between different tests, which may be    occurring in current practice, then interpretation may become difficult and the information could be misleading. A normal ejection fraction for CMR may correspond with a mildly compromised ventricle in 2D ECHO, and so on. Thus, the difference between these measurements may be the difference between whether a patient does, or does not, qualify for therapeutic intervention. The differences in normal values are particularly large when EDV and ESV are compared (Tables II  and III). It should be acknowledged that if techniques are used interchangeably an improvement or deterioration may be observed on an individual basis which does not reflect the patient's underlying pathology.
Comparison of Echocardiographic Studies with CMR: CMR has been regarded the reference standard for measurement of LV volumes and ejection fraction because of its high image quality and volumetric data. 1 There are well-performed ex vivo studies which have demonstrated the validity of CMR measurements. 38 The bias and 95% LOA between the dog heart model data and different methods for LV volume determination were between 4.94 AE 12.11 mL and 1.71 AE 18.11mL. High image quality with good segmentation of blood and tissue as well as a volumetric dataset are the prerequisites for accurate measurements of LV volumes. We have not included several older techniques which were used prior to the introduction of Simpson's biplane method in 2D ECHO, such as M-mode and linear methods of LV functional assessment, including Quinones et al., Dumesnil et al., Baran et al., The early methods for assessment of LVEF such as the Quinones method involved diameter measurements on 2D echocardiograms to calculate LVEF based on a mathematical calculation assuming an ellipsoid shape. While innovative at the time, the Quinones method could not adapt to other LV morphologies other than ellipsoid shapes. These methods are not recommended any more by the American and European Societies of Echo- CMR = cardiac magnetic resonance imaging; ICC = intra-class correlation; MD = mean difference expressed as a percentage of the mean (% AE 2SD); EDV = end-diastolic volume; ESV = end-systolic volume; LVEF = ejection fraction; CI = confidence interval.  cardiography. 7,43 This is due to proven inaccuracies in ventricles with abnormal shapes and regional wall-motion abnormalities. To our knowledge there are no studies comparing these methods with CMR. One of the difficulties facing LV functional assessment is that LVEF may be a moving target as a beat to beat variability has been reported up to 5.8 AE 1.7%. LVEF varies with BP, inotropic state and heart rate. To obtain reliable comparison of LVEF measurements from 2 different methods it is mandatory to examine the patient under the same hemodynamic conditions. The effect of the beat to beat variation can only be minimized by taking multiple measurements and averaging the results. 44,45 However, in reality this is often not carried out due to time constraints and high clinical loads.
There is no systematic difference in the measured ejection fraction between the echocardiographic methods and CMR (Fig. 1). As already reported for the normal values there are major differences in volumes between echocardiographic methods and CMR. With the use of contrast agents these differences in volume measurement have reduced, however, this is still not to the level where results can be considered interchangeable.
Regarding volumes 3D ECHO has been demonstrated to show a large improvement toward the values of CMR, and as such, is the most accurate ultrasound technique for determining LV function, although the total number of patients included in trials is still small and very good acoustic windows are needed. 46 The most promising technique in echocardiography is certainly contrast 3D, however, the excellent results demonstrated by the Jenkins group in 2009 could not be reproduced by Caiani et al., and so further investigation is required. 47 Table VIII summarizes the advantages and limitations of the different echocardiographic methods. The reproducibility of the echocardiography techniques showed a marked improvement with the introduction of contrast 2D ECHO and 3D ECHO in both intra-observer and inter-observer methods and comes close to CMR. However, we think there is not yet enough data to provide benchmarks for quality assessment. The differing tests used for variability assessments make it difficult for a reliable assessment to be made-in particular as some studies do not include either intra-or inter-observer calculations. Considering the importance of accurate assessment of LV function it is remarkable that there has been only a limited body of comprehensive studies which allow to define the differences between the different imaging methods. In particular, the data on the reproducibility are not satisfactory. The scientific societies should encourage studies or registries to broaden the database and to provide guidelines on how to perform validation studies.
The BA analysis appears to be an ideal test to analyze differences between methods or between observers. Based on the available studies the different imaging techniques for assessment of LVEF and volumes are not interchangeable. If followup measurements are necessary they should be performed with the same method.
In this review, only 2 of the 18 studies reviewed as validation studies for echocardiography were multicenter studies. Most of the data are from single center studies, which are subject to referral bias. Thus, the reproducibility may be overestimated. Further investigation from larger cohorts is needed.

Is Visual Assessment an Alternative?
Visual assessment of LV function on 2D echocardiograms has been used in many hospitals; for example, by estimating the LVEF in 5% steps such as 30-35% or just classifying the LV function as normal, mildly, moderately, or severely impaired. The reason for using a visual rather than a quantitative assessment is the extra time needed to calculate LV volumes and difficulties to trace the endocardial borders on still frames. To our knowledge no studies using visual assessment have been published in contrast echocardiography, where endocardial borders usually are well seen. Although visual assessment of global LV function has been reported to be "reasonable" among experienced readers, the actual interobserver variability was 5.8%. 48 This does not allow the use of visual assessment for follow-up studies of LVEF and volumes. In CMR a comparison of visual and quantitative assessment of LVEF showed a major underestimation (8.4%) with visual assessment. Therefore, it was recommended to use quantitative analysis for accurate assessment of LV function. 49