Determination of sources of error and improvement in accuracy of left ventricular mass measurement by echocardiography

Abstract

Abnormal Left ventricular mass (LVM) prognosticates adverse cardiovascular events. Conventionally, LVM measured by echo assumes a prolate ellipsoid (PE) shape; however, it poorly correlates with reference standard of cardiac magnetic resonance imaging (CMR) derived LVM. PE model assumes LVL = 2 × LVID. We developed a new echo LVM formula based on LV length and tested for accuracy against CMR. A retrospective study of consecutive patients with an echocardiogram and CMR within 3 months. Derivation (n = 170) and validation cohorts (n = 54) were used to test the new formula. Following analysis of correlation of interventricular septum (IVS), LV internal dimension (LVID), posterior wall (PW) and LVL between echo and CMR, a novel paraboloid-shape linear regression (PLR) model was derived. LVM by both models were compared to CMR. Poor correlation observed between actual and assumed LVL (0.52 with CMR; 0.44 with echo). Strong correlation was noted between echo and CMR measured LVL, LVID, IVS (r > 0.80) and a moderate correlation with PW (r = 0.62). Strong correlation of LVL was harnessed to develop PLR model, which significantly decreased paired error in derivation cohort (from 64 ± 42 to 22 ± 21 gm) and validation cohort (from 63 ± 46 to 25 ± 18 gm). Furthermore, it demonstrates significant reduction in absolute, relative errors and variability along with superior correlation in both cohorts. Between echo and CMR, LVL demonstrates one of the best correlation among LV dimensions. The assumption, LVL = 2 × LVID appears inaccurate. PLR model incorporates LVL and significantly improves accuracy, reduces variability of LVM.

Keywords: Cardiac magnetic resonance imaging; Echocardiography; Error; Left ventricular mass.