Although bepridil is effective for conversion of long-lasting persistent atrial fibrillation (AF) to sinus rhythm, it sometimes takes a long time to interrupt AF and there is no feasible index to predict its efficacy.In 60 patients with long-lasting persistent AF, bepridil (100-200 mg/day) was administered and continued for 8 weeks while body surface ECG was recorded every 2 weeks. The fibrillation cycle length (FCL) was evaluated using the spectral analysis of the fibrillation waves in each ECG. AF was interrupted in 32 patients receiving bepridil. The conversion was observed at 2 weeks in 4, at 4 weeks in 7, at 6 weeks in 7, and at 8 weeks in 14 patients. When comparing these responders and nonresponders, clinical background characteristics other than the dosage of bepridil did not show any difference and neither did temporal changes in QT parameters and heart rate. In contrast, the FCL and ΔFCL (prolongation in FCL from baseline) became significantly larger in responders than in nonresponders at later observation points (FCL: 177 ± 17 versus 164 ± 19 ms, P = 0.018, and ΔFCL: 38 ± 16 versus 22 ± 12 ms, P < 0.001, at 4-week point; FCL: 188 ± 17 versus 169 ± 19 ms, P = 0.004, and ΔFCL: 49 ± 18 versus 27 ± 14 ms, P < 0.001, at 6-week point).Repetitive evaluation of FCL using spectral analysis of fibrillation waves can be a feasible index to predict the efficacy of bepridil therapy.