Evaluation of P wave signal-averaged electrocardiographic filtering and analysis methods

The purpose of this study was to evaluate the signal processing and analysis methods currently in use for the P wave signal-averaged electrocardiogram and to define optimal parameters for its use. P wave signal-averaged electrocardiograms using the QRS as a trigger for alignment of the analysis window were obtained in 15 subjects with prior atrial fibrillation and 15 controls. Five methods of signal filtering (unidirectional, bidirectional, finite impulse response, least squares fit, and spectral fast-Fourier transform) and three filter frequencies (14, 29, and 60 Hz) were compared with logistic regression analysis. Analysis techniques, including P wave vector duration, individual orthogonal lead duration, and terminal root mean square voltage were also evaluated for the strength of their association with the occurrence of atrial fibrillation. The least-squares fit filter with bandwidth filtering of 29 to 250 Hz produced the strongest association with atrial fibrillation (odds ratio 26). A high correlation (r > 0.92) was noted among the individual orthogonal leads; however, neither individual leads nor total atrial activation determined from individual leads demonstrated a superior association with atrial fibrillation when compared with total vector P wave duration. Terminal P wave RMS volt ages were not significantly different between patients with prior AF and controls.