Reproducibility of the signal-averaged electrocardiogram.

Variation in measurements made from signal-averaged electrocardiograms was examined. Averaging 200 beats in 18 normal subjects, pairs of high-frequency QRS and low-amplitude signal durations correlated at immediate, short-term, and long-term intervals. The percent high-frequency (60 Hz to 120 Hz) voltages in the late potential region had modest correlations. However, 95% confidence intervals of differences in paired measurements were as much as 7% for high-frequency QRS, 20% for a low-amplitude signal, and 53% for percent high-frequency, similar at all time intervals. With electrocardiograms averaged to 0.3 microV noise, high-frequency QRS and low-amplitude signals prolonged, but variation was similar to that of 200 beat pairs. In contrast, low noise reduced percent high-frequency and lessened variation to 29%. Therefore variation in signal-averaged electrocardiographic measurements was considerable (high-frequency voltage greater than durations). Noise did not appreciably influence variation in durations but was critical to consistent voltage measurements in the frequency domain.