Comparison with the algorithmic methods for AP localization
Majority of AP localization algorithms were based on a limited number of cases, often partially pre-selected. For example, the sample size was 120, 157, 140, 135, 93 and 148 in Ferrari et al., Boersma et al, D’Avila et al., Arruda et al., Fitzpatrick et al., and Millstein et al. algorithms, respectively. (10-13;15;16) Small sample size can create an illusion that a particular QRS pattern is 100% specific for some location – as it is not seen at other AP locations. However, a bigger sample might show that other cases with the same ECG pattern had successful ablation at other locations as well. The current study, based on the largest cohort of overt AP ever studied, fully supports the concept that preexcited ECG pattern does not point to a particular location but rather reflects a wider area, albeit with the site of maximum AP frequency surrounded by a probability gradient. Representative illustration of this phenomenon can be found onSupplementary Figure 4 , which presents outcomes of analysis of the same ECG with the D’Avila algorithm and with WPW24.com app.
Algorithms are trying to create an ideal match between AP location and QRS pattern. Imperfections of that match are covered by algorithm’s sensitivity and specificity. Several independent validation studies showed that sensitivity/specificity/accuracy of algorithmic methods are lower than initially reported. These studies are in agreement with the current results, because low accuracy of an algorithm indicates that the APs were not limited to the single predicted locations but probably there was a wider spread of APs. Consequently, algorithmic methods are a somewhat misleading by creating an illusion of single location even when their sensitivity, specificity and accuracy are low - indicating contrariwise. In contrast, WPW24.com, instead of pretending precision and pointing to one location is showing where the accessory pathways were actually ablated in all studied cases with the QRS pattern that is similar or identical to the one with which the current patient had presented.
Almost all algorithms are based on the analysis of baseline preexcited ECG, occasionally, on ‘most preexcited’ baseline ECG – that is on a QRS morphology that results from fusion between native conduction and AP conduction. That might be another reason for the divergence between the original studies and the independent validation studies – as the degree of preexcitation for the same localization varies and it is impossible to determine how much preexcited ECGs were used in the original study. Full preexcitation pattern maximizes the localizing information contained in the preexcited QRS polarity because the confounding impact of conduction via His-Purkinje system on QRS is nearly eliminated. Moreover, full preexcitation standardizes the ECG patterns for comparison between studies. It was shown in our previous work that using fully preexcited ECG increases accuracy of AP localization, even for algorithms that were designed for analysis of baseline preexcitation. (6)