Long-term prevention of atrial fibrillation is not constantly realized by single-site right atrial pacing, and the beneficial role of multisite atrial pacing is still being studied. Accordingly, we compared the effectiveness of dual site and single site atrial pacing in 83 patients (50 men, 33 women, aged 69 +/- 10 years), who received a DDD device for primary sinus node dysfunction or bradycardia with documented atrial fibrillation. Inclusion criteria for dual site pacing were a sinus P wave > or = 120 ms and at least two episodes of documented paroxysmal AF in the 6 months preceding implantation. Dual site atrial pacing (high right atrium-coronary sinus ostium) was performed in 30 cases, and was compared to 53 single site paced patients, 21 with a P wave > or = 120 ms and 32 with a P wave < 120 ms. The basic pacing rate was programmed at 68 +/- 4 beats/min (range 60-75 beats/min). Sinus P wave (133 +/- 20 vs 95 +/- 9 ms; P < 0.001), paced P wave (107 +/- 14 vs 99 +/- 15; P < 0.05), number of antiarrhythmic drugs used (2.4 +/- 1.2 vs 1.6 +/- 1.5, P < 0.05), and the duration of symptoms (8.1 +/- 4.5 vs 3.8 +/- 2.4 years; P < 0.001) were significantly higher in dual site patients. The other characteristics were similar. During the follow-up of 18 +/- 15 months (range 3-30 months), paroxysmal AF was documented in 33 patients. Among these patients, 13 developed permanent AF following at least one episode of paroxysmal AF. When comparing dual site patients and single site patients with a P wave duration > or = 120 ms, paroxysmal AF incidence was lower in the dual site group (9/30 patients vs 15/21 patients, P < 0.01), as well as permanent AF (1/30 patients vs 8/21 patients, P < 0.01). By contrast, comparison between dual site patients and the group of single site patients with a P wave duration < 120 ms did not evidence any significant differences in paroxysmal (9/30 patients vs 9/32 patients) and permanent (1/30 patients vs 4/32 patients) AF incidences. Dual site seems better able than single site atrial pacing to improve the natural history of patients with a prolonged P wave, reducing the incidence of paroxysmal and permanent AF. No benefit could be expected in patients with a normal P wave duration.