Comparison of signal-averaging electrocardiographic systems using device specific criteria in 104 normal subjects

The advent of several signal-averaged electrocardiogram (SAECG) systems for late potential (LP) assessment warrants comparisons to assess intersystem reproducibility and variability. Simultaneous SAECGs on two systems, Arrhythmia Research Technology (ART) and Marquette (MEI), were performed on 104 normal volunteers (53 males, age 44 +/- 14 years), and analyzed filtered QRS duration (TFQRS), root mean square voltage (RMS40), and low amplitude signal duration (LAS40), filtered at 40-250 Hz. The Gomes criteria (TfQRS > 114 msec, RMS40 < 20 microV and LAS40 > 38 msec) were used as criteria for LP. The data was also analyzed using the recently proposed system specific criteria for MEI (TFQRS > 120 msec, RMS40 < 20 microV and LAS40 > 38 msec). Where appropriate, statistical analysis was performed using simple linear and Spearman's rank correlation, analysis of variance, Finn's R and McNemar's test.

The means +/- SD for ART and MEI were: TFQRS: 97.2 +/- 8.9 vs 108.2 +/- 7.2 msec (R = 0.76), RMS40: 31.8 +/- 17.8 vs 45.3 +/- 19.9 microV (R = 0.53), and LAS40: 32.2 +/- 8.4 vs 30 +/- 7.4 (R = 0.54). When the Gomes criteria were applied, the number of subjects identified by each system as abnormal were: TFQRS = 3 vs 22 (P < 0.001), RMS40 = 20 vs 8 (P = 0.004), LAS40 = 21 vs 9 (P = 0.004), TFQRS/RMS40 = 3 vs 6 (P = 0.38), TFQRS/LAS40 = 3 vs 7 (P = 0.22), RMS40/LAS40 40 = 17 vs 8 (P = 0.02), and all three criteria = 3 vs 6 (P = 0.38) for ART vs MEI, respectively. Percent agreement was 81.7% for TFQRS and 84.6% for RMS40 and LAS40 when single criteria were applied. Agreement improved when combined criteria were utilized (87.5%-95.2% for any two criteria and 95.2% for all three criteria). The intersystem agreement that was not due to chance was 0.63-0.69 for single criteria and 0.75-0.90 for combined criteria. Disagreement was highly significant for the three criteria when used singly and for RMS40 and LAS40 combined. Disagreement was not significant when TFQRS was used in combination with > or = one other criteria. When the MEI criteria were applied, there was a decrease in the number of subjects identified by the MEI system as abnormal, using the TFQRS criteria singly or in combination. Percent agreement for system specific TFQRS measurements was 94.2% for single criteria and 97.1% for combined criteria. The intersystem agreement that was not due to chance improved (88-0.94). Disagreement between system specific criteria for TFQRS was not significant (P > 0.05).

Our data indicate that although there is a general correlation between ART and MEI measurements, variability is substantial, leading to significant differences when the criteria for LP are applied, especially for single parameter determinations. Thus, there is a need to establish system specific normal ranges and more accurate criteria for LP parameters.