The aim of this study was to use fast Fourier transform analysis to clarify the characteristics of human atrial fibrillation (AF) waves with respect to the duration of AF and the effect of class I antiarrhythmic drugs. Twenty-two patients (10 paroxysmal AF, 12 persistent AF) without organic heart disease were studied by conventional electrophysiological methods. Electrograms were recorded from the right atrial free wall during AF and spectral analysis was performed for 35s (16 consecutive 4096-ms epochs with 50% overlap) and the fibrillation cycle length (FCL) was calculated from the peak frequency. Mean FCL and SD were determined from 16-epoch data, and the temporal variability of FCL was defined as the SD of FCL. Paroxysmal AF had a longer mean FCL than persistent AF (178+/-26ms vs 139+/-16 ms, p<0.001) and AF duration had a significant inverse correlation with mean FCL (r=-0.79, p<0.001). The temporal variability of FCL was significantly greater in paroxysmal AF than in persistent AF (p<0.05) and there was a significant positive correlation between the mean FCL and the temporal variability of FCL (r=0.66, p<0.001). In 8 of 18 patients given a class I antiarrhythmic drug (cibenzoline or procainamide), AF was terminated and in those patients the mean FCLs before administration of class I drugs were significantly greater than in patients without AF termination. With respect to mean FCL before drug administration, conversion occurred in 100% of patients with FCL > or =168 ms and in 17% of those with FCL <168 ms. A longer duration of AF shortens the mean FCL, which is consistent with atrial electrical remodeling. Class I drugs prolong the mean FCL above a critical level and will terminate AF, which can be estimated from the mean FCL before drug administration.