Author + information
- Received June 1, 2013
- Revision received August 2, 2013
- Accepted August 9, 2013
- Published online October 1, 2013.
- Victor Chien-Chia Wu, MD∗,
- Masaaki Takeuchi, MD∗∗ (, )
- Hiroshi Kuwaki, MD∗,
- Mai Iwataki, MD∗,
- Yasufumi Nagata, MD∗,
- Kyoko Otani, MD∗,
- Nobuhiko Haruki, MD∗,
- Hidetoshi Yoshitani, MD∗,
- Masahito Tamura, MD∗,
- Haruhiko Abe, MD∗,
- Kazuaki Negishi, MD†,
- Fen-Chiung Lin, MD‡ and
- Yutaka Otsuji, MD∗
- ∗Second Department of Internal Medicine, University of Occupational and Environmental Health, School of Medicine, Kitakyushu, Japan
- †Menzies Research Institute Tasmania, Hobart, Tasmania, Australia
- ‡Department of Second Section of Cardiology, Chang Gung Memorial Hospital, Chang Gung University College of Medicine, Taipei, Taiwan
- ↵∗Reprint requests and correspondence:
Dr. Masaaki Takeuchi, Second Department of Internal Medicine, University of Occupational and Environmental Health, School of Medicine, 1-1 Iseigaoka, Yahatanishi-ku, Kitakyushu 807-8555, Japan.
Objectives The hypothesis of this study was that minimal left atrial volume index (LAVImin) by 3-dimensional echocardiography (3DE) is the best predictor of future cardiovascular events.
Background Although maximal left atrial volume index (LAVImax) by 2-dimensional echocardiography (2DE) is a robust index for predicting prognosis, the prognostic value of LAVImin and the superiority of measurements by 3DE over 2DE have not been determined in a large group of patients.
Methods In protocol 1, we assessed age and sex dependency of LAVIs using 2DE and 3DE in 124 normal subjects and determined their cutoff values (mean + 2 SD). In protocol 2, 2-dimensional (2D) and 3-dimensional (3D) LAVImax/LAVImin were measured in 556 patients with high prevalence of cardiovascular disease. After excluding patients with atrial fibrillation, mitral valve disease, and age <18 years, 439 subjects were followed to record major adverse cardiovascular events (MACE). Patients were divided into 2 groups by the cutoff criteria of LAVI in each method.
Results In protocol 1, there was no significant age and sex dependency for each 2D and 3D LAVI. In protocol 2, during a mean of 2.5 years of follow-up, MACE developed in 88 patients, including 32 cardiac deaths. Kaplan-Meier survival analyses showed that all 4 LAVI cutoff criteria had significant predictive power of MACE. After variables were adjusted for clinical variables and left ventricular ejection fraction, all 4 methods were still independently and significantly associated with MACE, but 3D-derived LAVImin had the highest risk ratio. 3D LAVImin also had an incremental prognostic value over 3D LAVImax.
Conclusions LAVIs by both 2DE and 3DE are powerful predictors of future cardiac events. 3D LAVImin tended to have a stronger and additive prognostic value than 3D LAVImax.
Remodeling of the left atrium caused by pressure and/or volume overload reflects severity and chronicity of underlying pathologic conditions rather than instantaneous left ventricular (LV) diastolic dysfunction and filling pressure (1–3). Thus, the degree and extent of left atrial (LA) dilation are closely coupled with disease severity. Several previous studies demonstrated that left atrial volume (LAV) measured by 2-dimensional echocardiography (2DE) is a robust index for predicting future cardiovascular events in various clinical scenarios (4–7). Although normal range and cutoff values of LAV using 2DE have been described in guidelines, the values established only maximal LAV and maximal LAV corrected to body surface area (maximal left atrial volume index [LAVImax]) (2). Recently, many researchers have focused on the importance of minimal LAV because its size is determined by the direct exposure of LV diastolic pressure and reflects more reliably underlying pathology (8). Biplane 2DE determination of LAV is convenient and currently the standard method of choice (9). However, it is sometimes inaccurate due to the complexity of LA shape and the inability to obtain optimal cutting planes (10).
Transthoracic 3-dimensional echocardiography (3DE) has the potential for more accurate determination of LAV because it does not rely on geometric assumptions (11,12). However, few studies investigated the prognostic utility of LAV measurement by 3DE (13). We hypothesized that minimal left atrial volume index (LAVImin) measured by 3DE is the best index for predicting future cardiovascular events.
Accordingly, the aims of this study were to establish a normal range of LAVImax and LAVImin and their cutoff values (mean + 2 SD) in normal controls using 2DE and 3DE and to evaluate the utility of these cutoff criteria to predict future major cardiovascular events (MACE) and cardiac death in a larger number of patients.
In protocol 1, 124 healthy subjects >18 years of age (mean age 44 ± 16 years; range 18 to 85 years; 56 men) were enrolled as a normal control group. They were recruited from hospital employees and their relatives and volunteers through advertising. In protocol 2, a total of 556 patients who underwent both clinically indicated 2DE and 3DE for underlying cardiac diseases were randomly selected from our 3DE database. Exclusion criteria included atrial fibrillation, mitral valve disease, and age <18 years. Clinical characteristics including hypertension, diabetes mellitus, hyperlipidemia, smoking, chronic kidney disease (CKD) and coronary artery disease were evaluated for all patients based on established criteria. CKD was defined as an estimated glomerular filtration rate <60 ml/min/1.73 m2. The study was approved by the ethics committee of the University of Occupational and Environmental Health Hospital, and written informed consent was obtained from all subjects at the time of echocardiographic examinations.
2DE was performed using a commercially available ultrasound machine and transducer (iE33 and S4/S5-1 transducer, Philips Medical Systems, Andover, Massachusetts). Three consecutive beats in the apical 4- and 2-chamber views including the entire left atrium were acquired, with specific attention directed at ensuring that the long axis of the left atrium was maximally delineated and the difference between long-axis diameter from the 2 imaging planes was <5 mm. LAVs were determined using the biplane Simpson method at end-systolic frames just before mitral valve opening (maximal left atrial volume [LAVmax]) and at end-diastolic frames coincided with the R-wave on the electrocardiogram (minimal left atrial volume [LAVmin]). In each view, the LA wall was traced, excluding the LA appendage and pulmonary veins. LAV was calculated using the following formula on the Xcelera workstation (Philips Medical Systems):where A and B are the diameters of transverse axis perpendicular to the long axis of the left atrium from the apical 4- and 2-chamber views, and L is the long-axis diameter. Values were then indexed to the body surface area (LAVImax and LAVImin).
A fully sampled matrix-array transducer (X3-1 or X5-1) (Philips Medical Systems) was used to acquire the 3-dimensional (3D) full-volume datasets from apical approach during held respiration. To ensure inclusion of the entire LAV within the pyramidal scan volume with a relatively higher volume rate, the datasets were acquired using multibeat acquisition, wherein 4 wedge-shaped subvolumes (93° × 21°) were acquired during a single 5-s to 7-s breath hold. A 3D volumetric assessment of LAV was performed using commercially available quantitative software (3DQ Adv, QLAB, version 9.0, Philips Medical Systems). From the full-volume datasets, 2 orthogonal long-axis and 1 short-axis views of the left atrium at end-systole were selected. Manual adjustment was required to ensure that the long axis of the left atrium was maximally delineated in both long-axis views. Then, 5-point marking, including the septal, lateral, anterior, and inferior corners of the mitral annulus and the roof of the LA wall, were performed. The software automatically determined the LA wall in 3D space using the deformable shell model and made time domain LAV curve, from which LAVmax and LAVmin were determined. Manual adjustment of the LA wall border was performed by cropping the 3D datasets when inadequate tracking of the LA wall was observed.
Follow up information was obtained regularly in the outpatient clinic. Telephone contact of patients, physicians, and next of kin was performed if the patient had been treated in the other hospital. The primary endpoint was cardiac death. The secondary endpoint was MACE including cardiac death, nonfatal myocardial infarction, stroke, and admission due to heart failure (HF).
Intraobserver variability was determined by having the observer repeat the LAV measurement from the same dataset 1 month apart using 2-dimensional (2D) and 3D echocardiographic images in 25 randomly selected patients. Interobserver variability was determined by having a second observer perform these measurements in the same 25 patients.
Continuous data are expressed as mean ± SD. Categorical data are presented as a number or percentage. All statistical analyses were carried out using commercial software (JMP, version 9.0, SAS Institute Inc., Cary, North Carolina; SPSS, version 17, SPSS Inc., Chicago, Illinois). Categorical variables were compared using the Fisher exact test or chi-square test whenever appropriate. A Student t test was used to test the differences in continuous variables between 2 groups. Linear regression analysis was used to study the relationship between 2 parameters. Bland-Altman analysis was performed to determine bias and limits of agreement between 2 measurements. Kaplan-Meier survival analysis was used to plot cardiac death and MACE. Differences between survival curves were obtained by the log-rank test. For multivariate analysis on cardiac death and MACE, a separate Cox proportional hazard model was used. To evaluate the influence of 2D LAVImax, 2D LAVImin, 3D LAVImax, and 3D LAVImin in addition to clinical variables and left ventricular ejection fraction (LVEF), 4 models were constructed. Incremental value of LAVI was also assessed in 4 modeling steps using nested regression model. The first step consisted of fitting a multivariate Model 0 of age, sex, hypertension, diabetes mellitus, CKD, and coronary artery disease. Then, LVEF was included in the second step. Next, LAVImax was included in the third step. Finally, LAVImin was included in the fourth step. The change in overall log-likelihood ratio chi-square was used to assess the increase in predictive power after the addition of LAVI. Harrell's C of the each model was used as an analogous overall measure of discrimination for predicting survival time (14). A p value <0.05 was considered significant. Intraobserver and interobserver variability was calculated as the absolute differences between the corresponding 2 measurements in the percentage of their mean. Interobserver and intraobserver reproducibility was determined by intraclass correlation coefficient (ICC).
No significant age and sex dependency of LAVImax and LAVImin by both 2DE and 3DE were observed. Although there was no significant difference between 2D LAVImax and 3D LAVImax (p = 0.3609), 3D LAVImin was significantly larger compared with 2D LAVImin (p = 0.0005). Cutoff values in individual methods were determined as mean + 2 SD (Table 1). A significant but moderate correlation of LAVIs was noted between 2DE and 3DE (Fig. 1) (LAVImax: r = 0.67, p <0.0001, LAVImin: r = 0.51, p <0.0001).
A total of 117 patients were excluded due to the presence of atrial fibrillation (n = 85), mitral valve disease (n = 38) and age <18 years (n = 4). Thus, the final group consisted of 439 patients. Baseline clinical characteristics and echocardiographic parameters of study patients are shown in Tables 2 and 3. 2D measurements of both LAVImax and LAVImin were possible in all patients, and 3D measurements of LAVIs were possible in all but 1 patient. The mean volume rate of 3DE was 18 ± 2/s (range 14 to 23). All patients had been followed for a median of 916 days (maximum 1,815 days). Eighty-eight patients experienced MACE, including 32 cardiac deaths, 3 nonfatal myocardial infarctions, 35 HF admissions, and 18 strokes. Using their own cutoff values, patients were divided into normal or abnormal groups in each method of 2D LAVImax, 2D LAVImin, 3D LAVImax, and 3D LAVImin measurement. Regarding cardiac death, 2D LAVImax, 2D LAVImin, and 3D LAVImin showed a statistically significant difference of survival rate between the normal group and abnormal group. However, 3D LAVImax failed to show a significant difference in cardiac death between the 2 groups (Fig. 2). For MACE, all 4 methods showed a significant difference in survival curve between the normal and abnormal groups (Fig. 3).
After adjusting for age, sex, hypertension, diabetes mellitus, CKD, coronary artery disease, and LVEF, the independent associations of outcome were analyzed using a multivariate Cox proportional hazards model (Table 4). To avoid problems due to collinearity, LAVmax and LAVmin parameters were evaluated in separate models. All 4 LAVI parameters did not provide any significant statistical difference for future cardiac death. However, for MACE, all 4 LAVIs remained at statistically significant power, with 3D LAVImin having the lowest p value (p = 0.0002) and highest hazard ratio (3.00) compared with the other 3 LAVIs.
Figure 4 shows nested regression model and Harrell's C-statistic. For cardiac death, the addition of LAVImax and LAVImin by both 2DE and 3DE had no incremental value over Model 0 + LVEF (Figs. 4A and 4B). For MACE, although the addition of both 2D and 3D LAVImax showed significance for incremental value in addition to Model 0 + LVEF, only 3D LAVImin offered additional significant incremental value (Figs. 4C and 4D).
To determine the effect of HF and CKD on prognostic value of LAVIs, Kaplan-Meier analysis was performed in subgroups with and without HF or CKD (Table 5). None of the LAVI cutoff criteria showed a significant difference in survival for cardiac death and MACE in patients with HF. Although 2D LAVImax showed significance for predicting cardiac death in patients without HF, all 4 LAVI cutoff values discriminated significantly MACE rates in patients without HF. All cutoff values showed significant prognostic power for MACE in patients with CKD. In addition, 2D LAVImax, 2D LAVImin, and 3D LAVImin have predictive value for cardiac death in patients without CKD.
The intraobserver variability for the measurements was 2D LAVmax (5.4 ± 4.2%), 3D LAVmax (5.1 ± 3.9%), 2D LAVmin (8.4 ± 7.3%), 3D LAVmin (9.4 ± 8.2%). The corresponding interobserver variability was 2D LAVmax (11.0 ± 11.5%); 3D LAVmax (6.2 ± 4.0%), 2D LAVmin (15.1 ± 17.1%), 3D LAVmin (9.8 ± 11.6%), respectively. The intraobserver ICC for 2D LAVmax, 3D LAVmax, 2D LAVmin, and 3D LAVmin were 0993, 0.998, 0.995, and 0.996, respectively. The interobserver ICC for 2D LAVmax, 3D LAVmax, 2D LAVmin, and 3D LAVmin were 0.987, 0.995, 0.990, and 0.996, respectively.
There were 3 major findings in this study. 1) 2D LAVImax, 2D LAVImin, and 3D LAVImin were powerful predictors of future cardiac death and all 4 methods were robust index for predicting future MACE. 2) After adjusting for age, sex, hypertension, diabetes mellitus, CKD, coronary artery disease, and LVEF, all 4 methods were independently associated with MACE but not cardiac death. 3) 2D LAVImin failed to show significant incremental value in addition to 2D LAVImax for predicting MACE, whereas 3D LAVImin showed significant incremental value in addition to 3D LAVImax for predicting MACE.
Assessment of LA size has evolved from 1-dimensional M-mode for LA diameter (15), 2D biplane area-length, or Simpson method for LA volume to 3D semiautomatic atrial border tracing for LA volume with increasing accuracy and reproducibility (12,16). Insights from cardiac computed tomography and anatomic studies showed that the left atrium has a complex shape with an irregular ellipsoid LA free wall and oblique planar interatrial septal wall (17,18). Pitfalls exist in 2DE determination of LAV because: 1) the LA and LV long axes do not appear in the same cutting plane; 2) it is not always guaranteed that apical 4- and apical 2-chamber views are exactly 90° perpendicular to each other; and 3) the 2D cutting plane obtained often does not bisect the center of the LA short-axis view (10). For these reasons, observer variabilities of LAV by 2DE measurements were reported to be larger than those by 3DE measurements (10,11,19). Compared with multidetector computed tomography as standard reference, a previous study revealed that there was 19% underestimation of LAV by 2DE but only 8% underestimation by 3DE (11). Similarly, compared with cardiac magnetic resonance as standard reference, there were 31-ml and 16-ml underestimations of LAVmax and LAVmin by 2DE but only 1-ml underestimation and 0-ml difference by 3D LAVmax and LAVmin (12).
Recommendation for determining LA size was focused on the maximal volume measurement (1,2,6). LAVImax has been shown to be a potent biomarker for first-ever ischemic stroke (20). LAVImax provided independent information over clinical and other echocardiographic variables for predicting mortality in patients with HF (21). In another large sample study of 36,561 patients with preserved LVEF for an average follow-up of 1.7 ± 1.0 years, LAVImax predicted mortality risk irrespective of LV geometry (22).
In recent investigations, researchers found the LAVImin correlated more with cardiovascular events and was predictive of outcomes. LAVmin but not LAVmax was found to be a significant predictor of the first atrial fibrillation or flutter after being entered in a multivariate model adjusted for covariates (23). Another study also showed that LAVmin measured by 3DE was the best independent predictor of adverse cardiovascular events among 2DE- and 3DE-derived LV and LA parameters (13). To investigate the prevalence, risk factors, and possible cardiac predictors of silent brain infarcts in the community, an epidemiological study showed that LAVmin was identified as a better correlate of LV diastolic function than LAVmax (8), and LAVImin and LA total emptying fraction were more strongly associated with subclinical brain infarction and white matter hyperintensity volume than LAVImax (24). Although both LAVmax and LAVmin increased gradually with progression of LV diastolic function, the increase in LAVmin with worsening diastolic function was more pronounced and showed an increase in even mild diastolic dysfunction, whereas LAVmax increased in later stages of diastolic dysfunction (8).
This study was unique because we first established cutoff values of 2D and 3D LAVImax/LAVImin for healthy subjects, and then we determined the utility of each criterion for the prediction of prognosis of both cardiac death and MACE in a large patient group. Kaplan-Meier analysis revealed that 2D LAVImax/min and 3D LAVImin had significant power to predict cardiac death. However, we could not demonstrate the significance in 3D LAVImax, possibly due to the small number of cardiac deaths (n = 32). On the other hand, all 4 methods showed significant power to predict future MACE, with 3D LAVImin being the best due to the highest chi-square value. This is in agreement with previous studies that showed that LAVImin is more reflective of chronic LV diastolic pressure/volume overload and is a better predictor of cardiac events than LAVImax (8,25).
Using multivariate Cox proportional hazards model adjusted for age, sex, major cardiovascular risk factors, and LVEF, none of LAVI cutoff criteria showed prognostic utility for future cardiac death, which again may be due to the small number of cardiac deaths that occurred in our study group. On the other hand, all 4 methods of LAVI determinations showed a robust predictive power for future MACE, with 3D LAVImin having the most significant p value and highest hazard ratio among all.
We found that 3D LAVImin, but not 2D LAVImin, had incremental predictive value for future MACE over the model that already included age, sex, major clinical risk factors, LVEF, and LAVImax. Incremental value of 3DE reinforced the superiority of 3DE over 2DE, as reported in recently published studies (8,13,24), for the prognosis of MACE.
A significant predictive value of LAVI for cardiac death and MACE could not be demonstrated in patients with HF because these patients were already associated with increased LAV; hence, no independent prognostic power from the established LAVI cutoff criteria could be observed. However, in patients without HF, all 4 LAVI cutoff values showed significance for predicting future MACE. Patients with CKD are associated with a higher risk of death and cardiovascular events (26–28). All LAVI criteria were useful for predicting MACE in patients with CKD, suggesting the incremental predictive value of LAVIs in patients who are already at high cardiovascular risk.
This study supported LAVImax as a powerful predictor of MACE, but even more so to verify that LAVImin has greater significance in the prognosis of future MACE. LAV measurements by 3DE are superior to 2DE against the reference standard (9,10,17). We also demonstrated that 3D LAVImin has a stronger and additive prognostic value over 3D LAVImax. The routine echocardiography should therefore include assessment of LAVImin for predicting cardiovascular events, especially in patients who already have higher cardiovascular risk profiles, determined by either 2DE or 3DE. In addition, serial LAVI follow-up can provide a simple, quick, yet quantifiable log of clinical progress and treatment effect (29).
First, our study subjects are a high-risk population with higher prevalence of cardiovascular disease. Therefore, the results of our study may not be appropriately extrapolated to the general population. Second, we could not determine statistically significant superiority of 3D LAVI over 2D LAVI for future cardiovascular events. Due to collinearity, it is possible that a higher power is needed to detect differences in the prognostic value of 2D and 3D parameters. Third, patients were enrolled in retrospective fashion; thus, temporal relation to specific cardiac events may at times be difficult to assess. Fourth, the study population consisted entirely of Japanese subjects, and the relevance of this study to other ethnic backgrounds awaits further research. Last, only patients with echocardiographic images that could be adequately used for 2DE and 3DE analyses were enrolled in the study and may result in selection bias.
Both LAVImax and LAVImin determined by 2DE and 3DE were highly prognostic indexes for predicting future cardiovascular events in a high-risk population. Our study showed that 3D LAVImin seemed to be the best predictor of future MACE. Thus, it should be useful to establish 3D LAVImin as the best predictor in larger, multicenter, prospective study.
All authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- 2-dimensional echocardiography
- 3-dimensional echocardiography
- chronic kidney disease
- heart failure
- intraclass correlation coefficient
- left atrial
- left atrial volume
- left atrial volume index
- maximal left atrial volume index
- minimal left atrial volume index
- maximal left atrial volume
- minimal left atrial volume
- left ventricular
- left ventricular ejection fraction
- major adverse cardiac events
- Received June 1, 2013.
- Revision received August 2, 2013.
- Accepted August 9, 2013.
- 2013 American College of Cardiology Foundation
- Abhayaratna W.P.,
- Seward J.B.,
- Appleton C.P.,
- et al.
- Lang R.M.,
- Bierig M.,
- Devereux R.B.,
- et al.
- Simek C.L.,
- Feldman M.D.,
- Haber H.L.,
- Wu C.C.,
- Jayaweera A.R.,
- Kaul S.
- Moller J.E.,
- Hillis G.S.,
- Oh J.K.,
- et al.
- Russo C.,
- Jin Z.,
- Homma S.,
- et al.
- Mor-Avi V.,
- Yodwut C.,
- Jenkins C.,
- et al.
- Tsang T.S.,
- Abhayaratna W.P.,
- Barnes M.E.,
- et al.
- Lim T.K.,
- Dwivedi G.,
- Hayat S.,
- Majumdar S.,
- Senior R.
- Fatema K.,
- Barnes M.E.,
- Bailey K.R.,
- et al.
- Russo C.,
- Jin Z.,
- Liu R.,
- et al.
- Weiner D.E.,
- Tighiouart H.,
- Amin M.G.,
- et al.