The short-term and long-term reproducibility of spectral turbulence and late potential variables of the signal-averaged ECG in a population sample of healthy subjects and the impact of gender, age, and noise

Abstract

Previous methods for frequency domain analysis of the signal-averaged ECG (SAECG) have had low reproducibility. The reproducibility of time domain late potential analysis and spectral turbulence analysis was evaluated with 2 immediately consecutive SAECG recordings in 121 randomly selected subjects without heart disease (short-term) and also in 47 subjects after 1 month (longterm). A test was late potential positive if 2 or more of 3 variables were outside the reference limits and spectral turbulence positive if the score was 3 or 4. The short-term reproducibility was high for the filtered QRS duration (FQRSD), root mean square amplitude of the last 40 ms (RMS40) and high frequency low amplitude signals less than 40 μV (HFLAS40) of the time domain and total QRS duration (TQRSD), power spectral density of the last 40 ms and the late potential duration of time domain analogous analyses. The Spearman rank order correlation coefficients were 0.89, 0.88, and 0.84 and 0.97, 0.91 and 0.97, respectively. The reproducibility of the spectral score variables varied, and the correlation for the low slice correlation ratio was 0.71, spectral entropy 0.61, interslice correlation mean 0.58, and interslice correlation SD 0.28. A diagnostic inconsistency between 2 tests occurred in 0 (0%) subjects in late potential analysis if FQRSD was required for positivity, and in 7 (6%) otherwise, and in 9 (7%) of spectral turbulence analysis. If the spectral variable mean peaks per slice, with a correlation of 0.89, replaced interslice correlation SD in the spectral score, diagnostic inconsistency occurred in 0 (0%) subjects. The reproducibility seemed higher in women and in younger people but significantly only for interslice correlation subjects mean and HFLAS40. The long-term reproducibility did not differ significantly from shortterm for any variable. In conclusion, the reproducibility was high in all time domain and time domain analogous variables. It varied among the spectral turbulence score variables and was very low for interslice correlation SD. The reproducibility of the spectral score improved substantially if this variable was replaced by mean peaks per slice.