Echocardiography in serial evaluation of left ventricular systolic and diastolic function: importance of image acquisition, quantitation, and physiologic variability in clinical and investigational applications

Abstract

Quantitative echocardiography is frequently used for serial evaluation of left ventricular performance. This prospective study was designed to determine the extent to which the acts of image acquisition and quantitation, and the subjects themselves, affect total variability in two-dimensional and Doppler echocardiographic indexes of left ventricular morphology and performance. Therefore, two technicians and two readers acquired and analyzed 60 echocardiograms from 15 normal subjects, each of whom was studied four times (twice on each of two visits). Analysis of variance based on generalizability theory was used to estimate the magnitude of these variability sources by calculating standard deviations (SD) and used to estimate their contribution to total variability. Of the two-dimensional echocardiographic indexes tested, ejection fraction varied least (SD, 6.6%) and left ventricular mass varied most (SD, 35.3 gm). Of the Doppler indexes, normalized early diastolic filling velocity integral varied least (SD, 8.4%) and deceleration time varied most (SD, 48.6 msec). Technical (image acquisition and quantitation) variability contributed most (and subject variability least) to total variability of stroke volume (68%) and deceleration time (67%). Technical variability contributed least (and subject variability most) to variability of ejection fraction (43%) and diastolic filling time (25%). The acts of image acquisition and quantitation varied more between than within technicians and readers. Peak atrial filling velocity and the ratio of peak early to atrial filling velocity significantly differed between technicians. Left ventricular ejection fraction, left ventricular mass, peak atrial filling velocity, early filling integral, and deceleration of early filling differed significantly between readers. Therefore the acts of image acquisition and quantitation, and subject variability itself, all contribute to total variability in echocardiographic indexes. Changes seen on clinical studies should be interpreted as abnormal only when exceeding the total variability originating from these sources. Generalizability theory allows one to tailor strategies to reduce variability. These strategies include increasing the number of observations, readers, and technicians for any given "baseline" study and using the same readers and technicians for sequential follow-up studies.