Background
Left ventricular hypertrophy (LVH) is highly prevalent in patients on hemodialysis [
1‐
4]. LVH, as measured by the left ventricular mass index (LVMI), is a strong predictor of cardiovascular disease (CVD) [
3,
4]. Consequently, a reliable and valid method to detect LVH is needed for both clinical and scientific implications [
5,
6]. For the assessment of left ventricular mass (LVM), cardiac magnetic resonance imaging (CMR) has been established as the most accurate and reproducible method [
7‐
9]. However, given its limited availability and high cost, CMR is not practical for clinical use in large-scale clinical studies. In contrast, the two-dimensional (2D) targeted M-mode echocardiography is preferred in the clinical context because of its widespread availability, low cost, simple handling, and extensive evidence base [
10]. Nevertheless, echocardiographic linear measurement and LVM calculation by cube function formulas have their own limitations [
10‐
13]. The current recommended formula from the American Society of Echocardiography (ASE) is based on special geometric assumptions, which may become inaccurate in the presence of asymmetric hypertrophy, eccentric remodeling, or distortion of left ventricular (LV) geometry and may lead to an incremental overestimation of LVMI [
3,
14‐
17]. Teichholz et al. designed a formula that includes a volume-correcting function in order to minimize the error inter alia in patients with LVH [
18]. A recent CMR study investigating patients with aortic stenosis demonstrated that the Teichholz (Th) formula had a lower tendency to overestimate the value within a population with increased LVMI [
17].
Here, we investigated the performance of two echocardiographic formulas, ASE and Th, in calculating LVMI in patients on hemodialysis.
Discussion
This study assessed the diagnostic accuracy of echocardiography in determining the LVMI in hemodialysis patients. The two echocardiographic formulas correlated moderately with CMR and overestimated LVMI compared to CMR measurements. The overestimation of LVM using the Th formula was less severe in patients with higher LVMI values. Therefore, these analyses showed an advantage of the Th formula in calculating LVMI in chronic hemodialysis patients with LVH but not in the absence of LVH.
This phenomenon might be explained by the intrinsic difference between the formulas for estimating the left ventricular volume (LVV). This becomes evident once the true LV geometry differs from the assumed ellipsoidal geometry. Though CMR imaging or 3D echocardiography measures myocardial volume to quantify LVM, 2D targeted linear m-mode quantification by echocardiography relies on geometric assumptions considering a left ventricle shape approximately equivalent to a prolate ellipsoid (LVV = ¾πr
1r
2r
3) [
27]. The basic assumption of the cube function is that the minor radii (r
1 and r
2) of the prolate ellipsoid are half the major radius (r
3) [
28]. In patients with increasing LVM, there is probably an incremental change in the configuration as a result of an increasing short axis accompanied by a gradual decrease in the long axis [
18]. Prior studies using the ASE formula have also shown that measurement of LVMI by 2D targeted M-mode echocardiography results in overestimation of LVMI [
15,
29‐
33]. Thus, the ASE formula is recommended for use in patients without major distortions in LV geometry [
5]. The approach developed by Teichholz et al. for correcting systematic errors in the LVV formula attempts to take changes of configuration into account and correct the cube function formula to LVV = [7.0/2.4 + r](D)
3 as an adaption to the observed curvilinear changes in the enlarging left ventricle [
17,
18]. This component of the Th formula may also be the reason for the declining trend in ΔLVMI described above.
Notably, the original study by Teichholz et al. analyzed the relationship between the long and short axes using an out-of-use angiographic technique. However, a recent CMR study confirmed these findings, demonstrating that the left ventricle tends to become more spherical with increasing LVM [
17]. This study enrolled 99 Asian patients with aortic stenosis and compared the LVMI calculated by the ASE and Th formulas. The authors demonstrated a significant overestimation of both echocardiographic formulas compared to CMR. However, a comparison of the two echocardiographic formulas revealed a significantly lower overestimation of LVMI by the Th formula in patients with LVH compared to those without, which is in good agreement with our findings in hemodialysis patients.
In contrast to these findings, a necropsy validation study by Devereux et al. demonstrated a systematic underestimation of LVM by the Th formula, which was particularly obvious in patients with LVH [
21]. Interim methodological and technological changes may explain this inconsistency. First of all, animal necropsy validation studies of CMR seem to indicate that assessing LVM by manual planimetry underestimates LVM [
33]. Secondly, the echocardiography ultrasound frequency has changed, and fundamental imaging (FI) was replaced by harmonic imaging (HI) to improve the definition of pericardial and epicardial borders [
6,
21]. This FI technique has been shown to underestimate LVM values compared to CMR, whereas the HI technique, which we used, seemed to provide higher LVM values [
34]. Another point to take into consideration is the exclusion of the PMT of the LV by CMR. Studies have demonstrated that PMT contributes 6.2 to 15.5% [
35] of the total LVM and significantly affects LVM quantification [
35,
36]. For example, Janik et al. compared one-dimensional LVM based on the ASE formula to the three-dimensional method derived by CMR using manual planimetry. The inclusion of PMT yielded significantly lower mean LVM differences between these methods, and the difference was 3-fold higher among patients with concentric and eccentric hypertrophy [
35]. All of these aspects could make an additional contribution to the discrepancy of CMR and TTE measurement of LVM, which cannot necessarily be explained by methodological differences alone [
33].
LVM is traditionally indexed by body surface area (BSA) [
22] but other indexations have also been used [
6]. In dialysis patients, indexation by height
2.7 has been shown to be somewhat superior to BSA with regard to predicting cardiovascular mortality [
3]. However, several studies showed that indexation by BSA has a similarly high prognostic value and is still recommended by ASE [
13]. Given that the focus of this study was on the comparison of two methods we chose indexation by BSA which also allowed us to compare our findings with previous studies.
Our study is the first to compare the determination of LVMI by echocardiography using the Th and ASE formulas to CMR findings in hemodialysis patients. The strengths of this analysis are that the echocardiographic and CMR studies were performed according to standard operating procedures defined before the start of the MiREnDa study [
20].
The limitations of our study are the limited sample size and low percentage of woman. We acknowledge the lack of patients with higher LVMI for consistent evaluation of the trend of overestimation towards higher LVMI, and the limitations of our cut-off value for the Th formula. Geometric changes in patients with LVH measured by CMR could not be analyzed, as it was not part of the MiREnDa CMR protocol. In addition, the LVM measured by CMR was calculated by exclusion of PMT, although there are recent data for normal values with inclusion of PMT [
37]. A further limitation is the lack of any reproducibility analysis. Furthermore, our findings likely cannot be extrapolated to patients receiving peritoneal dialysis or to non-white hemodialysis patients.
Acknowledgments
We thank the participants in the MiREnDa trial, the local dialysis center and regional university center staff, the safety and data monitoring committee (Jan C. Galle, David Petroff, and Bernhard O. Böhm), the Center for Clinical Trials at the University Hospital of Würzburg, the Department of Pharmacy at the University Hospital Würzburg, and Christian Ritter at the Institute for Diagnostic and Interventional Radiology.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.