Clinical impact of various geometric models for calculation of echocardiographic left ventricular mass

Abstract

Background: M-mode echocardiographic left ventricular mass calculated using a thick-wall prolate ellipsoidal model is widely used in clinical and epidemiologic studies. Doubts regarding the ability of this approach to obtain a precise estimate of left ventricular weight across a wide range of values have recently been raised and an alternate thin-wall ellipsoidal model has been proposed to gain greater precision.

Objective: To compare thin-wall and thick-wall (American Society of Echocardiography and Penn convention) models for calculation of left ventricular mass.

Design: Validation, cross-sectional, and longitudinal studies.

Participants: Necropsy data and living cohorts from Naples, New York City, and Umbria region of Italy (PIUMA registry).

Results: The average thin-wall left ventricular mass was slightly greater than the necropsy left ventricular weight (mean 225 versus 220 g), whereas no difference was detected using regression-adjusted thick-wall methods. Use of the thin-walled model slightly overestimated left ventricular mass relative to both thick-wall models at the lowest left ventricular mass while slightly underestimating the highest values. Comparison of Cox proportional hazard models in two longitudinal studies demonstrated that there was a substantial equivalence among methods, with a marginally better performance of thick-wall models for cardiovascular risk stratification (P < 0.05 in one study).

Conclusions: Although it is imperfect, because it is based on simplifying geometric assumptions, computation of left ventricular mass on the basis of M-mode echocardiographic left ventricular dimensions using thick-wall prolate-ellipsoidal models is valuable for identification of left ventricular hypertrophy and for cardiovascular risk stratification of patients with essential hypertension. Calculation of left ventricular mass by use of a thin-wall prolate-ellipsoidal geometry does not yield appreciably different results than those which are obtained by use of thick-wall models.