Role of bandpass filters in optimizing the value of the signal-averaged electrocardiogram as a predictor of the results of programmed stimulation
Cardiac Pacing, Artificial
Normal values for the signal-averaged electrocardiogram (SAECG) at 11 different high-pass filter settings were obtained from 100 normal subjects (group I). The filtered QRS duration and the duration of low amplitude signals less than 40 microV, but not the root mean square voltage of the last 40 ms (RMS40), showed normal distribution. A normal distribution for RMS40 could be obtained by transforming each value to its natural logarithm. The normal values were used in a systematic approach to optimize the accuracy of the SAECG to predict the results of programmed stimulation in 80 patients with spontaneous nonsustained ventricular tachycardia (VT). Fifty-two patients with no inducible VT (group II) and 28 patients with inducible sustained monomorphic VT (group III) were investigated. The 3 SAECG parameters at each high-pass filter in groups II and III were categorized as normal or abnormal and were evaluated singly or in combinations of 2 or 3. There was no combination that provided a sensitivity greater than 82% that could also be obtained by single determinations of low amplitude signals less than 40 microV at 25 to 40 Hz or RMS40 at 40 Hz. On the other hand, there were 267 different combinations that provided a maximal specificity of 98%. The best total predictive accuracy of a single parameter was 85%, provided by RMS40 at 40 or 60 Hz. The total predictive accuracy could be improved to 89% by 1 of 32 different combinations. The top combinations were mostly in triplets and included SAECG parameters recorded at different high-pass filter settings. The only 2 paired combinations with the best total predictive accuracy were RMS40 at 20 or 25 Hz paired with RMS40 at 40 Hz. Frequencies at both ends of the analyzed high-pass filter settings (less than 20 Hz and greater than 60 Hz) were not represented in the top predictive combinations. The SAECG parameters analyzed at 40 Hz were most frequently represented in the top predictive combinations, suggesting that the SAECG may have the best predictive accuracy at this filter setting. In summary, the combination of SAECG parameters analyzed at different filter settings can enhance the accuracy of the technique as a screening test for the results of programmed stimulation in patients with spontaneous nonsustained VT.