Variability of digitized echocardiography: size, source, and means of reduction.

Digitization of M-mode echocardiograms provides useful information on left ventricular function, but its variability has been assessed rarely. Inter- and intraobserver (technical) variability of readers digitizing the same cardiac cycles, and variability between beats, days, and subjects (biologic) was determined. Technical variability was small for both standard dimensions (correlation coefficient r values 0.82 to 1.00) and rates of change (r values 0.70 to 0.98). Biologic variability was large with expected normal differences between 2 measurements (95% confidence limits) of 11 to 55% and 37 to 106% for standard dimensions and rates of change, respectively. By averaging measurements of 5 beats from each of 2 days, the expected normal differences are reduced to 6 to 32% and 23 to 63% for standard dimensions and rates of change. This study emphasizes the large biologic variability in rates of change of digitized left ventricular measurements. Normal variation between studies can be reduced and real physiologic or pathologic changes perceived best if many beats from more than 1 day are measured.