- Email: [email protected]
Comparison of Left and Right Atrial Volume by Echocardiography Versus Cardiac Magnetic Resonance Imaging Using the Area-Length Method
Comparison of Left and Right Atrial Volume by Echocardiography Versus Cardiac Magnetic Resonance Imaging Using the Area-Length Method Matthew Whitlock, MD, Anuj Garg, MD, Jill Gelow, MD*, Timothy Jacobson, MD, and Craig Broberg, MD Increased atrial volumes predict adverse cardiovascular events. Accordingly, accurate measurement of atrial size has become increasingly important in clinical practice. The area-length method is commonly used to estimate the volume. Disagreements between atrial volumes using echocardiography and other imaging modalities have been found. It is unclear whether this has resulted from differences in the measurement method or discrepancies among imaging modalities. We compared the right atrial (RA) and left atrial (LA) volume estimates using the area-length method for transthoracic echocardiography and cardiovascular magnetic resonance (CMR) imaging. Patients undergoing echocardiography and CMR imaging within 1 month were identified retrospectively. For both modalities, the RA and LA long-axis dimension and area were measured using standard 2and 4-chamber views, and the volume was calculated using the area-length method for both atria. The echocardiographic and CMR values were compared using the Bland-Altman method. A total of 85 patients and 18 controls were included in the present study. The atrial volumes estimated using the area-length method were significantly smaller when measured using echocardiography than when measured using CMR imaging (LA volume 35 ⴞ 20 vs 49 ⴞ 30 ml/m2, p <0.001, and RA volume 32 ⴞ 23 vs 43 ⴞ 29 ml/m2, p ⴝ 0.012). The mean difference (CMR imaging minus echocardiography) was 14 ⴞ 14 ml/m2 for the LA and 10 ⴞ 16 ml/m2 for the RA volume. Similar results were found in the healthy controls. No significant intra- or interobserver variability was found within each modality. In conclusion, echocardiography consistently underestimated the atrial volumes compared to CMR imaging using the area-length method. © 2010 Elsevier Inc. All rights reserved. (Am J Cardiol 2010;106:1345–1350) An increased atrial volume is an independent predictor of adverse cardiovascular events, including stroke and congestive heart failure.1,2 Accordingly, accurate measurement of the atrial size has become increasingly relevant to clinical practice. Most clinical studies have used echocardiographic measurements, because echocardiography is the most accepted and best validated modality for atrial volume quantification.3 The area-length method has often been chosen for atrial volume quantification, because it is simple, relatively quick, and can be done using standard acquisitions. More recently, cardiac magnetic resonance (CMR) imaging has been used for atrial volume quantification. Because the long-axis views obtained by CMR imaging are similar to those obtained echocardiographically, the area-length method should give similar results for both modalities. To our knowledge, no studies have investigated this assumption. We sought to compare the area-length method of atrial volume measurements for echocardiography and CMR imaging for both the left atrial (LA) and right atrial (RA) volumes. Division of Cardiovascular Medicine, Oregon Health and Science University, Portland, Oregon. Manuscript received April 23, 2010; manuscript received and accepted June 9, 2010. *Corresponding author: Tel: (503) 494-8750; fax: (503) 494-8550. E-mail address: [email protected] (J. Gelow). 0002-9149/10/$ – see front matter © 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.amjcard.2010.06.065
Methods Patients ⬎18 years old who had been referred for CMR imaging at our institution were identified. The patients were eligible for inclusion if they had also undergone imaging with transthoracic echocardiography within 1 month before or after CMR imaging. The patients were excluded if any major clinical event was identified during the period between echocardiography and CMR imaging. Major clinical events were determined by chart review and included any cardiac surgery, hospital admission for heart failure or diuresis, completion of pregnancy, or transplantation. The patients were also excluded if either the echocardiogram or CMR image provided incomplete or inadequate views of the atria. In addition, 18 healthy volunteers were underwent echocardiography and CMR imaging on the same day, and the measurements and volumes for both echocardiography and CMR imaging were obtained using methods identical for the patients undergoing imaging, for separate analysis. The institutional review board at Oregon Health and Science University approved the present study. Echocardiographic apical 2- and 4-chamber views were obtained with the subject in the left lateral decubitus position using routine laboratory standards to demonstrate the LV apex and mitral valve. CMR imaging was performed with the patients in the supine position. A vertical long-axis www.ajconline.org
1346
The American Journal of Cardiology (www.ajconline.org)
Figure 1. Long-axis dimension (arrow) and area (shaded regions) of left atrium and right atrium by CMR imaging (Upper) and transthoracic echocardiography (Lower) from 2-chamber (Left) and 4-chamber (Right) views.
view was obtained through the mitral valve and LV apex, identified from the CMR axial scout images. This was followed by a single-phase short-axis stack through the ventricle, prescribed from the vertical long-axis image, from which the LV volume and function was obtained using published standards.4 The 4-chamber plane was prescribed from the vertical long-axis and short-axis scout images, through both the mitral and tricuspid valve openings and LV apex. All images were obtained using steady-state free precession with retrospective gating. Standard 2- and 4-chamber views were evaluated off-line using commercially available software for both echocardiography (Syngo Dynamics Workstation 5.1.0, Siemens Medical Systems) and CMR imaging (CMR Tools, London, United Kingdom). One observer made all echocardiographic measurements and another did all the CMR imaging measurements. The observers were unaware of the findings of each other. Atrial diastole was determined by selecting the last frame in ventricular systole before mitral valve opening. The measurements were made according to published methods.5 The long-axis length of the left and right atria were defined by measuring the distance from the center of the mitral annulus to the posterior atrial wall. The atrial endocardial area was traced to exclude the atrial appendages and pulmonary or caval veins (Figure 1). Using the quantified length and area measurements, the LA and RA volumes from both echocardiography and CMR imaging were calculated. For the LA volume, we used the biplane area-length formula: volume ⫽ (0.848 ⫻ area4ch ⫻ area2ch/[(length4ch ⫹ length2ch)/2], where 4ch and 2ch is the 4-chamber and 2-chamber view, respectively.7 For the RA volume, the monoplane area-length formula was used: volume ⫽ 0.848 ⫻ (area4ch)2/length4ch. All volumes were indexed to the body surface area obtained at CMR imaging. Interobserver and intraobserver variability for the LA and RA volumes was examined for 10 patients who had
Figure 2. Correlation between LA volumes measured by echocardiography and CMR imaging.
undergone echocardiography and CMR imaging within a 24-hour period. The echocardiographic and CMR measurements were repeated by each observer, who were unaware of their first measurements (intraobserver variability). Each observer also measured the same patients using the other modality (interobserver variability). The differences were examined using the Bland-Altman method. Statistical analysis was performed using Statistical Package for Social Sciences, version 13 (SPSS, Chicago, Illinois). For the entire cohort, the mean ⫾ SD was calculated for the left and right volumes for both echocardiography and CMR imaging. The echocardiographic- and CMR-determined volumes were compared using Student’s t test, Pearson’s correlation, and Bland-Altman plots. The results are
Methods/Atrial Volume Comparison by Echocardiography and CMR Imaging
1347
Figure 3. Bland-Altman plot for LA echocardiography and CMR volume differences.
expressed as the mean ⫾ SD, and p ⬍0.05 was considered statistically significant. Results Of the 351 patients undergoing CMR imaging at our institution who were screened, 117 also had undergone transthoracic echocardiography within 1 month. Of these, 32 were excluded because of inadequate echocardiographic images (n ⫽ 17), inadequate CMR images (n ⫽ 7), or a major interval clinical event (n ⫽ 8). Thus, 85 patients were included in the present study. In 5 of these patients, the right atrium could not be accurately measured using echocardiography; thus, 80 patients were included for RA volume analysis. The mean patient age was 52 ⫾ 7 years, and 34% of the patients were women. The average body surface area of the patients was 2.0 ⫾ 0.3 m2. The mean interval between echocardiography and CMR imaging was 3 ⫾ 10 days (median 2). The clinical indications for CMR imaging included an assessment of viability (n ⫽ 26), a diagnosis of cardiomyopathy (n ⫽ 14), characterization of a cardiac mass (n ⫽ 6), clarification of echocardiographic findings (n ⫽ 8), an assessment of congenital heart disease (n ⫽ 10), suspected arrythmogenic right ventricular dysplasia (n ⫽ 6), and an assessment of aortic disease (n ⫽ 6). The average left ventricular ejection fraction of the patients was 49 ⫾ 20%, and the average right ventricular ejection fraction was 49 ⫾ 15%. The mean age of the healthy controls was 40 ⫾ 11 years, and 36% of the controls were women. The average body surface area of the controls was 1.8 ⫾ 0.2 m2. For the patients, the LA and RA volumes estimated by echocardiography were significantly lower than the corresponding CMR volumes. The mean LA volume was 35 ⫾ 20 ml/m2 by echocardiography and 49 ⫾ 30 ml/m2 by CMR imaging (p ⬍0.001). Similarly, the mean RA volume was 32 ⫾ 23 ml/m2 by echocardiography and 43 ⫾ 29 ml/m2 by CMR imaging (p ⫽ 0.012). The mean ratio (echocardiography/CMR) of the volume was 0.74 ⫾ 0.22 for the left atrium and 0.78 ⫾ 0.32 for the right atrium. A strong linear correlation (r ⫽ 0.9) was seen between
Figure 4. Correlation between RA volumes measured by echocardiography and CMR imaging.
the LA volumes estimated by echocardiography and CMR, but the slope of the regression line was 0.6 (Figure 2). The mean LA volume difference (CMR minus echocardiography) was 14.7 ⫾ 14.3 ml/m2 (Figure 3). After excluding 3 patients with giant left atria (⬎100 ml/m2), the mean LA difference (CMR minus echocardiography) was still 13 ml/m2 (95% confidence interval 9 to 36 ml/m2). The LA volume quartiles for CMR imaging were ⬍60, 60 to 80, 81 to 100, and ⬎100 ml/m2. To determine whether agreement improved using a single standardized view, the monoplane LA volumes were calculated using the 4-chamber measurements alone. No change was seen in the analysis trends. Similar findings were obtained for the right atrium. Despite the strong linear relation between the RA volumes estimated by echocardiography and CMR imaging (r ⫽ 0.8), the slope of the regression line was 0.6 (Figure 4). After exclusion of 3 patients with a giant right atrium (⬎100 ml/m2), the slope of the regression line remained 0.6. The
1348
The American Journal of Cardiology (www.ajconline.org)
Figure 5. Bland-Altman plot for LA echocardiography and CMR volume differences.
Figure 6. Intraobserver variability for LA and RA volume measurements by CMR imaging and echocardiography.
Figure 7. Interobserver variability for LA and RA volume measurements by CMR imaging and echocardiography.
mean RA volume difference (CMR minus echocardiography) was 10 ⫾ 17 ml/m2 (Figure 5). We found similar results for same-day studies performed in 18 healthy volunteers. The mean LA volume by echocardiography was 23 ⫾ 6 ml/m2 compared to 48 ⫾ 5 ml/m2 by CMR imaging (p ⬍0.0001). The mean RA volume by echocardiography was 21 ⫾ 8 ml/m2 compared to 46 ⫾ 9 ml/m2 by CMR imaging (p ⬍0.0001). The mean ratio (echocardiography/ CMR) of the volume estimates was 0.50 ⫾ 0.13 for the left atrium and 0.46 ⫾ 0.15 for the right atrium. Regression analyses for the intraobserver and interob-
server variability are shown in Figures 6 and 7, respectively. No bias between reader 1 and reader 2 was identified for either modality. The agreement for the atrial volumes determined by CMR imaging was slightly stronger than for echocardiography (Table 1). To determine the major sources of discrepancy, we analyzed the length and area measurements separately, and their correlation with the volume estimates. For length, the ratio of the echocardiographically determined length/CMR determined length did not correlate with the ratio of the echocardiographically determined volume/
Methods/Atrial Volume Comparison by Echocardiography and CMR Imaging
1349
Table 1 Interobserver and Intraobserver variability for volume measurements by echocardiography (echo) and cardiac magnetic resonance imaging (CMR) using the area-length method Test Intraobserver CMR 1 vs CMR 2 Intraobserver Echo 1 vs Echo 2 Interobserver CMR Interobserver Echo
Absolute Difference CMR-Echo (ml/m2)
Percent Difference
R
Coefficient of Variability
p Value
Ratio ⫾ SD
6.2 ⫾ 9.5
10.7
0.97
10.5
0.58
1.09 ⫾ 0.13
5.2 ⫾ 10.4
12.8
0.94
15.7
0.58
0.93 ⫾ 0.15
5.0 ⫾ 11.2
9.7
0.96
11.5
0.67
1.05 ⫾ 0.11
6.1 ⫾ 17.0
21.4
0.85
26.1
0.53
1.19 ⫾ 0.42
CMR determined volume (r ⫽ 0.3). For the area, however, the ratio of the echocardiographically determined length/CMR determined length correlated with the ratio of the echocardiographically determined volume/CMR determined volume (r ⫽ 0.7). Therefore, discrepancies in the volume measurements between echocardiography and CMR imaging are better explained by differences in area than by differences in length. These findings were similar in the healthy controls. Discussion Our study has demonstrated a significant bias between echocardiography and CMR imaging in the quantification of atrial volumes using identical methods. Transthoracic echocardiography consistently underestimated the LA and RA volumes compared with CMR. The discrepancy was present across a wide range of atrial volumes, in both our patient population and normal controls. Despite the wide range of atrial volumes, a statistically significant difference was found between the mean values from the 2 modalities. These differences were not explained by interobserver differences, the development of a major clinical event between studies, or image quality. Our findings are consistent with those from previous studies assessing the measurement of atrial volumes using echocardiography and CMR imaging. The CMR estimation of atrial volumes using the biplane area-length method had been shown to overestimate the CMR measurements by short-axis quantification.6 In a comparison of CMR imaging with 2-dimensional and 3-dimensional echocardiography, echocardiography underestimated the CMR atrial volumes by 14% to 37%.7 Although 2-dimensional and 3-dimensional atrial volumes calculated using the area-length method have correlated well,8 some studies have suggested that 3-dimensional echocardiography has better agreement with CMR imaging.9 Differences in measurement between imaging modalities could result from poor endocardial definition (“dropout”) and foreshortening with echocardiography. These limitations will become particularly evident when imaging the atria from standard apical views. The atria are located in the far field where “dropout” and distortion could be more significant, particularly of the lateral atrial wall.8 In addition, delineation of the pulmonary veins and appendices is likely to be less accurate with echocardiography. This has been supported by our finding that area
discrepancies accounted for more of the variability than length discrepancies. The echocardiographic volumes measured in our volunteers were consistent with the reported normal range.10 The normal values for the LA and RA area and length using CMR have been reported, although the volumes quantified from these measures were not included.11,12 Our results are consistent with these measurements; therefore, it is unlikely that our study included an inherent error to account for the discrepancy between modalities. The purpose of our study was not to demonstrate any limitation of atrial volume quantification using echocardiography, which has already been shown to have significant clinical importance. Rather, the data have shown that a comparison of one imaging modality to another is inherently flawed. A new reference range for interpreting CMR volumes is necessary. In normal volunteers, the CMR volume estimates were as great as 59 ml/m2, consistent with the lowest quartile of CMR determined LA volumes in our patient cohort. The mean ⫹ (1.96 ⫻ SD) was 59 ml/m2 for the left atrium and 64 ml/m2 for the right atrium. Therefore, we propose that the atrial volumes determined using CMR of ⬍60 ml/m2 can be considered normal, 60 to 80 ml/m2 mildly enlarged, 81 to 100 ml/m2 moderately enlarged, and ⬎100 ml/m2 severely enlarged. Our study was retrospective, and not all images were obtained concurrently. We allowed ⱕ30 days between echocardiography and CMR imaging. Although every attempt was made to exclude patients who incurred a major clinical change between the imaging studies, we could not exclude the possibility that the atrial dimensions between imaging studies resulted from an interval change in the cardiac hemodynamics. However, our findings were similar in patients who underwent echocardiography and CMR imaging on the same day. We have tried to consider any inherent methodologic error that might have accounted for the differences, including observer differences, and found none. Patient body position (supine vs lateral decubitus) differences might have been present but seem unlikely. Our study included patients with a wide variety of indications for cardiac imaging. Thus, the patients in our study were representative of the spectrum of those undergoing cardiac imaging in clinical practice. Even after excluding the patients at the far end of the spectrum with giant atria, our findings were unchanged.
1350
The American Journal of Cardiology (www.ajconline.org)
Accurate determination of the atrial volume has become increasingly relevant in clinical practice. Echocardiography significantly underestimates atrial volumes measured by CMR imaging, which must be recognized as more patients undergo both examinations. The discrepancies in the LA and RA volume observed between echocardiography and CMR imaging are significant and likely due to inherent differences in image acquisition. This must be considered when reporting and comparing atrial volumes using either method. 1. Tsang TS, Barnes ME, Gersh BJ, Bailey KR, Seward JB. Left atrial volume as a morphophysiologic expression of left ventricular diastolic dysfunction and relation to cardiovascular risk burden. Am J Cardiol 2002;90:1284 –1289. 2. Kizer JR, Bella JN, Palmieri V, Liu JE, Best LG, Lee ET, Roman MJ, Devereux RB. Left atrial diameter as an independent predictor of first clinical cardiovascular events in middle-aged and elderly adults: the Strong Heart Study (SHS). Am Heart J 2006;151:412– 418. 3. Lester SJ, Ryan EW, Schiller NB, Foster E. Best method in clinical practice and in research studies to determine left atrial size. Am J Cardiol 1999;84:829 – 832. 4. Grothes F, Mood JC, Bellenger NG, Smith GS, Klein HU, Pennell DJ. Interstudy reproducibility of right ventricular volumes, function, and masses with cardiovascular magnetic resonance. Am Heart J 2004;147:218 –223.
5. Ujino K, Barnes ME, Cha SS, Langins AP, Bailey KR, Seward JB, Tsang TS. Two-dimensional echocardiographic methods for assessment of LA volume. Am J Cardiol 2006;98:1185–1188. 6. Sievers B, Kirchberg S, Addo M, Bakan A, Brandts B, Trappe HJ. Assessment of left atrial volumes in sinus rhythm and atrial fibrillation using the biplane area-length method and cardiovascular magnetic resonance imaging with TrueFISP. J Cardiovasc Magn Reson 2004;6:855–863. 7. Rodevan O, Bjornerheim R, Ljosland M, Maehle J, Smith HJ, Ihlen H. Left atrial volumes assessed by three- and two-dimensional echocardiography compared to MRI estimates. Int J Cardiovasc Imaging 1999;15:397– 410. 8. Jenkins C, Bricknell K, Marwick TH. Use of real-time three-dimensional echocardiography to measure left atrial volume: comparison with other echocardiographic techniques. J Am Soc Echocardiogr 2005;18:991–997. 9. Keller AM, Gopal AS, King DL. Left and right atrial volume by freehand three-dimensional echocardiography: in vivo validation using magnetic resonance imaging. Eur J Echocardiogr 2000;1:55– 65. 10. Wang Y, Gutman JM, Heilbron D, Wahr D, Schiller NB. Atrial volume in a normal adult population by two-dimensional echocardiography. Chest 1984;86:595– 601. 11. Anderson JL, Horne BD, Pennell DJ. Atrial dimensions in health and left ventricular disease using cardiovascular magnetic resonance. J Cardiovasc Magn Reson 2005;7:671– 675. 12. Sievers B, Kirchberg S, Franken U, Bakan A, Addo M, John-Puthenveettil B, Trappe HJ. Determination of normal gender-specific left atrial dimensions by cardiovascular magnetic resonance imaging. J Cardiovasc Magn Reson 2005;7:677– 683.
Methods Patients ⬎18 years old who had been referred for CMR imaging at our institution were identified. The patients were eligible for inclusion if they had also undergone imaging with transthoracic echocardiography within 1 month before or after CMR imaging. The patients were excluded if any major clinical event was identified during the period between echocardiography and CMR imaging. Major clinical events were determined by chart review and included any cardiac surgery, hospital admission for heart failure or diuresis, completion of pregnancy, or transplantation. The patients were also excluded if either the echocardiogram or CMR image provided incomplete or inadequate views of the atria. In addition, 18 healthy volunteers were underwent echocardiography and CMR imaging on the same day, and the measurements and volumes for both echocardiography and CMR imaging were obtained using methods identical for the patients undergoing imaging, for separate analysis. The institutional review board at Oregon Health and Science University approved the present study. Echocardiographic apical 2- and 4-chamber views were obtained with the subject in the left lateral decubitus position using routine laboratory standards to demonstrate the LV apex and mitral valve. CMR imaging was performed with the patients in the supine position. A vertical long-axis www.ajconline.org
1346
The American Journal of Cardiology (www.ajconline.org)
Figure 1. Long-axis dimension (arrow) and area (shaded regions) of left atrium and right atrium by CMR imaging (Upper) and transthoracic echocardiography (Lower) from 2-chamber (Left) and 4-chamber (Right) views.
view was obtained through the mitral valve and LV apex, identified from the CMR axial scout images. This was followed by a single-phase short-axis stack through the ventricle, prescribed from the vertical long-axis image, from which the LV volume and function was obtained using published standards.4 The 4-chamber plane was prescribed from the vertical long-axis and short-axis scout images, through both the mitral and tricuspid valve openings and LV apex. All images were obtained using steady-state free precession with retrospective gating. Standard 2- and 4-chamber views were evaluated off-line using commercially available software for both echocardiography (Syngo Dynamics Workstation 5.1.0, Siemens Medical Systems) and CMR imaging (CMR Tools, London, United Kingdom). One observer made all echocardiographic measurements and another did all the CMR imaging measurements. The observers were unaware of the findings of each other. Atrial diastole was determined by selecting the last frame in ventricular systole before mitral valve opening. The measurements were made according to published methods.5 The long-axis length of the left and right atria were defined by measuring the distance from the center of the mitral annulus to the posterior atrial wall. The atrial endocardial area was traced to exclude the atrial appendages and pulmonary or caval veins (Figure 1). Using the quantified length and area measurements, the LA and RA volumes from both echocardiography and CMR imaging were calculated. For the LA volume, we used the biplane area-length formula: volume ⫽ (0.848 ⫻ area4ch ⫻ area2ch/[(length4ch ⫹ length2ch)/2], where 4ch and 2ch is the 4-chamber and 2-chamber view, respectively.7 For the RA volume, the monoplane area-length formula was used: volume ⫽ 0.848 ⫻ (area4ch)2/length4ch. All volumes were indexed to the body surface area obtained at CMR imaging. Interobserver and intraobserver variability for the LA and RA volumes was examined for 10 patients who had
Figure 2. Correlation between LA volumes measured by echocardiography and CMR imaging.
undergone echocardiography and CMR imaging within a 24-hour period. The echocardiographic and CMR measurements were repeated by each observer, who were unaware of their first measurements (intraobserver variability). Each observer also measured the same patients using the other modality (interobserver variability). The differences were examined using the Bland-Altman method. Statistical analysis was performed using Statistical Package for Social Sciences, version 13 (SPSS, Chicago, Illinois). For the entire cohort, the mean ⫾ SD was calculated for the left and right volumes for both echocardiography and CMR imaging. The echocardiographic- and CMR-determined volumes were compared using Student’s t test, Pearson’s correlation, and Bland-Altman plots. The results are
Methods/Atrial Volume Comparison by Echocardiography and CMR Imaging
1347
Figure 3. Bland-Altman plot for LA echocardiography and CMR volume differences.
expressed as the mean ⫾ SD, and p ⬍0.05 was considered statistically significant. Results Of the 351 patients undergoing CMR imaging at our institution who were screened, 117 also had undergone transthoracic echocardiography within 1 month. Of these, 32 were excluded because of inadequate echocardiographic images (n ⫽ 17), inadequate CMR images (n ⫽ 7), or a major interval clinical event (n ⫽ 8). Thus, 85 patients were included in the present study. In 5 of these patients, the right atrium could not be accurately measured using echocardiography; thus, 80 patients were included for RA volume analysis. The mean patient age was 52 ⫾ 7 years, and 34% of the patients were women. The average body surface area of the patients was 2.0 ⫾ 0.3 m2. The mean interval between echocardiography and CMR imaging was 3 ⫾ 10 days (median 2). The clinical indications for CMR imaging included an assessment of viability (n ⫽ 26), a diagnosis of cardiomyopathy (n ⫽ 14), characterization of a cardiac mass (n ⫽ 6), clarification of echocardiographic findings (n ⫽ 8), an assessment of congenital heart disease (n ⫽ 10), suspected arrythmogenic right ventricular dysplasia (n ⫽ 6), and an assessment of aortic disease (n ⫽ 6). The average left ventricular ejection fraction of the patients was 49 ⫾ 20%, and the average right ventricular ejection fraction was 49 ⫾ 15%. The mean age of the healthy controls was 40 ⫾ 11 years, and 36% of the controls were women. The average body surface area of the controls was 1.8 ⫾ 0.2 m2. For the patients, the LA and RA volumes estimated by echocardiography were significantly lower than the corresponding CMR volumes. The mean LA volume was 35 ⫾ 20 ml/m2 by echocardiography and 49 ⫾ 30 ml/m2 by CMR imaging (p ⬍0.001). Similarly, the mean RA volume was 32 ⫾ 23 ml/m2 by echocardiography and 43 ⫾ 29 ml/m2 by CMR imaging (p ⫽ 0.012). The mean ratio (echocardiography/CMR) of the volume was 0.74 ⫾ 0.22 for the left atrium and 0.78 ⫾ 0.32 for the right atrium. A strong linear correlation (r ⫽ 0.9) was seen between
Figure 4. Correlation between RA volumes measured by echocardiography and CMR imaging.
the LA volumes estimated by echocardiography and CMR, but the slope of the regression line was 0.6 (Figure 2). The mean LA volume difference (CMR minus echocardiography) was 14.7 ⫾ 14.3 ml/m2 (Figure 3). After excluding 3 patients with giant left atria (⬎100 ml/m2), the mean LA difference (CMR minus echocardiography) was still 13 ml/m2 (95% confidence interval 9 to 36 ml/m2). The LA volume quartiles for CMR imaging were ⬍60, 60 to 80, 81 to 100, and ⬎100 ml/m2. To determine whether agreement improved using a single standardized view, the monoplane LA volumes were calculated using the 4-chamber measurements alone. No change was seen in the analysis trends. Similar findings were obtained for the right atrium. Despite the strong linear relation between the RA volumes estimated by echocardiography and CMR imaging (r ⫽ 0.8), the slope of the regression line was 0.6 (Figure 4). After exclusion of 3 patients with a giant right atrium (⬎100 ml/m2), the slope of the regression line remained 0.6. The
1348
The American Journal of Cardiology (www.ajconline.org)
Figure 5. Bland-Altman plot for LA echocardiography and CMR volume differences.
Figure 6. Intraobserver variability for LA and RA volume measurements by CMR imaging and echocardiography.
Figure 7. Interobserver variability for LA and RA volume measurements by CMR imaging and echocardiography.
mean RA volume difference (CMR minus echocardiography) was 10 ⫾ 17 ml/m2 (Figure 5). We found similar results for same-day studies performed in 18 healthy volunteers. The mean LA volume by echocardiography was 23 ⫾ 6 ml/m2 compared to 48 ⫾ 5 ml/m2 by CMR imaging (p ⬍0.0001). The mean RA volume by echocardiography was 21 ⫾ 8 ml/m2 compared to 46 ⫾ 9 ml/m2 by CMR imaging (p ⬍0.0001). The mean ratio (echocardiography/ CMR) of the volume estimates was 0.50 ⫾ 0.13 for the left atrium and 0.46 ⫾ 0.15 for the right atrium. Regression analyses for the intraobserver and interob-
server variability are shown in Figures 6 and 7, respectively. No bias between reader 1 and reader 2 was identified for either modality. The agreement for the atrial volumes determined by CMR imaging was slightly stronger than for echocardiography (Table 1). To determine the major sources of discrepancy, we analyzed the length and area measurements separately, and their correlation with the volume estimates. For length, the ratio of the echocardiographically determined length/CMR determined length did not correlate with the ratio of the echocardiographically determined volume/
Methods/Atrial Volume Comparison by Echocardiography and CMR Imaging
1349
Table 1 Interobserver and Intraobserver variability for volume measurements by echocardiography (echo) and cardiac magnetic resonance imaging (CMR) using the area-length method Test Intraobserver CMR 1 vs CMR 2 Intraobserver Echo 1 vs Echo 2 Interobserver CMR Interobserver Echo
Absolute Difference CMR-Echo (ml/m2)
Percent Difference
R
Coefficient of Variability
p Value
Ratio ⫾ SD
6.2 ⫾ 9.5
10.7
0.97
10.5
0.58
1.09 ⫾ 0.13
5.2 ⫾ 10.4
12.8
0.94
15.7
0.58
0.93 ⫾ 0.15
5.0 ⫾ 11.2
9.7
0.96
11.5
0.67
1.05 ⫾ 0.11
6.1 ⫾ 17.0
21.4
0.85
26.1
0.53
1.19 ⫾ 0.42
CMR determined volume (r ⫽ 0.3). For the area, however, the ratio of the echocardiographically determined length/CMR determined length correlated with the ratio of the echocardiographically determined volume/CMR determined volume (r ⫽ 0.7). Therefore, discrepancies in the volume measurements between echocardiography and CMR imaging are better explained by differences in area than by differences in length. These findings were similar in the healthy controls. Discussion Our study has demonstrated a significant bias between echocardiography and CMR imaging in the quantification of atrial volumes using identical methods. Transthoracic echocardiography consistently underestimated the LA and RA volumes compared with CMR. The discrepancy was present across a wide range of atrial volumes, in both our patient population and normal controls. Despite the wide range of atrial volumes, a statistically significant difference was found between the mean values from the 2 modalities. These differences were not explained by interobserver differences, the development of a major clinical event between studies, or image quality. Our findings are consistent with those from previous studies assessing the measurement of atrial volumes using echocardiography and CMR imaging. The CMR estimation of atrial volumes using the biplane area-length method had been shown to overestimate the CMR measurements by short-axis quantification.6 In a comparison of CMR imaging with 2-dimensional and 3-dimensional echocardiography, echocardiography underestimated the CMR atrial volumes by 14% to 37%.7 Although 2-dimensional and 3-dimensional atrial volumes calculated using the area-length method have correlated well,8 some studies have suggested that 3-dimensional echocardiography has better agreement with CMR imaging.9 Differences in measurement between imaging modalities could result from poor endocardial definition (“dropout”) and foreshortening with echocardiography. These limitations will become particularly evident when imaging the atria from standard apical views. The atria are located in the far field where “dropout” and distortion could be more significant, particularly of the lateral atrial wall.8 In addition, delineation of the pulmonary veins and appendices is likely to be less accurate with echocardiography. This has been supported by our finding that area
discrepancies accounted for more of the variability than length discrepancies. The echocardiographic volumes measured in our volunteers were consistent with the reported normal range.10 The normal values for the LA and RA area and length using CMR have been reported, although the volumes quantified from these measures were not included.11,12 Our results are consistent with these measurements; therefore, it is unlikely that our study included an inherent error to account for the discrepancy between modalities. The purpose of our study was not to demonstrate any limitation of atrial volume quantification using echocardiography, which has already been shown to have significant clinical importance. Rather, the data have shown that a comparison of one imaging modality to another is inherently flawed. A new reference range for interpreting CMR volumes is necessary. In normal volunteers, the CMR volume estimates were as great as 59 ml/m2, consistent with the lowest quartile of CMR determined LA volumes in our patient cohort. The mean ⫹ (1.96 ⫻ SD) was 59 ml/m2 for the left atrium and 64 ml/m2 for the right atrium. Therefore, we propose that the atrial volumes determined using CMR of ⬍60 ml/m2 can be considered normal, 60 to 80 ml/m2 mildly enlarged, 81 to 100 ml/m2 moderately enlarged, and ⬎100 ml/m2 severely enlarged. Our study was retrospective, and not all images were obtained concurrently. We allowed ⱕ30 days between echocardiography and CMR imaging. Although every attempt was made to exclude patients who incurred a major clinical change between the imaging studies, we could not exclude the possibility that the atrial dimensions between imaging studies resulted from an interval change in the cardiac hemodynamics. However, our findings were similar in patients who underwent echocardiography and CMR imaging on the same day. We have tried to consider any inherent methodologic error that might have accounted for the differences, including observer differences, and found none. Patient body position (supine vs lateral decubitus) differences might have been present but seem unlikely. Our study included patients with a wide variety of indications for cardiac imaging. Thus, the patients in our study were representative of the spectrum of those undergoing cardiac imaging in clinical practice. Even after excluding the patients at the far end of the spectrum with giant atria, our findings were unchanged.
1350
The American Journal of Cardiology (www.ajconline.org)
Accurate determination of the atrial volume has become increasingly relevant in clinical practice. Echocardiography significantly underestimates atrial volumes measured by CMR imaging, which must be recognized as more patients undergo both examinations. The discrepancies in the LA and RA volume observed between echocardiography and CMR imaging are significant and likely due to inherent differences in image acquisition. This must be considered when reporting and comparing atrial volumes using either method. 1. Tsang TS, Barnes ME, Gersh BJ, Bailey KR, Seward JB. Left atrial volume as a morphophysiologic expression of left ventricular diastolic dysfunction and relation to cardiovascular risk burden. Am J Cardiol 2002;90:1284 –1289. 2. Kizer JR, Bella JN, Palmieri V, Liu JE, Best LG, Lee ET, Roman MJ, Devereux RB. Left atrial diameter as an independent predictor of first clinical cardiovascular events in middle-aged and elderly adults: the Strong Heart Study (SHS). Am Heart J 2006;151:412– 418. 3. Lester SJ, Ryan EW, Schiller NB, Foster E. Best method in clinical practice and in research studies to determine left atrial size. Am J Cardiol 1999;84:829 – 832. 4. Grothes F, Mood JC, Bellenger NG, Smith GS, Klein HU, Pennell DJ. Interstudy reproducibility of right ventricular volumes, function, and masses with cardiovascular magnetic resonance. Am Heart J 2004;147:218 –223.
5. Ujino K, Barnes ME, Cha SS, Langins AP, Bailey KR, Seward JB, Tsang TS. Two-dimensional echocardiographic methods for assessment of LA volume. Am J Cardiol 2006;98:1185–1188. 6. Sievers B, Kirchberg S, Addo M, Bakan A, Brandts B, Trappe HJ. Assessment of left atrial volumes in sinus rhythm and atrial fibrillation using the biplane area-length method and cardiovascular magnetic resonance imaging with TrueFISP. J Cardiovasc Magn Reson 2004;6:855–863. 7. Rodevan O, Bjornerheim R, Ljosland M, Maehle J, Smith HJ, Ihlen H. Left atrial volumes assessed by three- and two-dimensional echocardiography compared to MRI estimates. Int J Cardiovasc Imaging 1999;15:397– 410. 8. Jenkins C, Bricknell K, Marwick TH. Use of real-time three-dimensional echocardiography to measure left atrial volume: comparison with other echocardiographic techniques. J Am Soc Echocardiogr 2005;18:991–997. 9. Keller AM, Gopal AS, King DL. Left and right atrial volume by freehand three-dimensional echocardiography: in vivo validation using magnetic resonance imaging. Eur J Echocardiogr 2000;1:55– 65. 10. Wang Y, Gutman JM, Heilbron D, Wahr D, Schiller NB. Atrial volume in a normal adult population by two-dimensional echocardiography. Chest 1984;86:595– 601. 11. Anderson JL, Horne BD, Pennell DJ. Atrial dimensions in health and left ventricular disease using cardiovascular magnetic resonance. J Cardiovasc Magn Reson 2005;7:671– 675. 12. Sievers B, Kirchberg S, Franken U, Bakan A, Addo M, John-Puthenveettil B, Trappe HJ. Determination of normal gender-specific left atrial dimensions by cardiovascular magnetic resonance imaging. J Cardiovasc Magn Reson 2005;7:677– 683.