Sex-related electrocardiographic differences in patients with different types of atrial fibrillation: Results from the SWISS-AF study

Abstract

Background: Sex-related electrocardiographic differences are a well-known phenomenon, but not their expression in patients with atrial fibrillation (AF). In this study we aim to assess the presence of significant sexrelated differences in ECG features, with particular attention to P-wave parameters, of a large cohort of patients affected by different types of AF. Methods: A 5-min resting 16-lead ECG was evaluated for 1119 AF patients in sinus rhythm. The durations of the main ECG waves and intervals were measured for both atrial and ventricular activity. Moreover, the beat-to-beat P-wave variability was computed for lead II and for the first principal component (PC1) computed across the 16 leads. The percentage of variance explained by PC1 was computed. Results: Males compared to females showed significantly longer RR interval (1.02 ± 0.16 s vs 0.97 ± 0.15 s, p b .001), PQ interval (191 ± 34 ms vs 183 ± 35 ms, p = .008), QRS duration (105 ± 17 ms vs 98 ± 13 ms, p = .021), significantly lower percentage of variance explained by PC1 and P-wave variability. Males with paroxysmal AF compared to females with paroxysmal AF had significantly longer RR interval (1.01 ± 0.17 s vs 0.96 ± 0.14 s, p b .001), shorter QTc (388 ± 27 ms vs 402 ± 27 ms, p b .001), lower P-wave variability in PC1. Males with persistent AF compared to females with persistent AF had significantly shorter QTc interval (396 ± 30 ms vs 407 ± 26 ms, p = .019), longer PQ interval (194 ± 35 ms vs 182 ± 30 ms, p = .037), higher V1 terminal force (2.1 ± 1.2 mV*ms vs 1.8 ± 1 mV*ms, p = .007), lower percentage of variance explained by PC1. Conclusions: AF patients present with several sex-related ECG differences. Consequently, sex should be taken into account when developing ECG algorithms identifying patients at risk for AF progression.

Beat-to-beat P-wave Variability Increases from Paroxysmal to Persistent Atrial Fibrillation

“…The recordings were filtered (high-pass filter, Butterworth, 4th-order, fc = 0.5 Hz; low-pass filter, FIR, Hamming window, fc = 80 Hz and non-linear notch filter fc= 50 Hz) in order to reduce the noise influence on the variability estimate [10]. Then, the first three principal components (PCs) of the ECG leads were added to the analysis.…”

“…A similar procedure was applied on the P-waves on lead II (where the atrial activity is more visible and positive): all P waves were cross-correlated and the 20 most correlated ones were averaged to compute a template. Then, all P-waves were aligned according to the lag at which the cross-correlation function with the template was maximal [10], [12]. Abnormal and false Pwaves were discarded using a clustering procedure on each lead: the PCA was applied on all P-windows and the coefficients of the first 2 PCs were used to characterize each P-window on a 2-D plane.…”

“…where Pn is the n-th P-wave and N the total number of Pwaves in the recordings. Furthermore, a similarity index (SI) was computed as the cosine of the angle between two consecutive P-waves [10]. This procedure was repeated for each lead and principal component.…”

“…In order to evaluate the spatial Pwave similarity, for each beat the PCA was applied on the 15 P-waves (one for each lead) and the percentage of variance explained by the first 2 components computed as a measure of spatial similarity. The median value for all the P-waves was computed [10].…”

A Real-time ECG Classification Scheme Using Anti-aliased Blocks with Low Sampling Rate

“…Batch normalization are replaced by instance normalization (IN) [9]. Age and sex are introduced into the feature set due that many ECG parameters are age-and sex-dependent [10,11].…”