A prospective study was conducted to investigate the validity of the hypothesis that P-wave dispersion (Pd) may be a clinically useful predictor of progression from paroxysmal to persistent atrial fibrillation (AF). Two hundred four consecutive patients with a diagnosis of paroxysmal AF were studied. Standard 12-lead electrocardiography and echocardiography were performed on all patients at the time of entry into the study. Pd was measured as the difference between maximum and minimum P-wave duration in any of the 12 leads. Mean follow-up was 66 +/- 8 months. Group I included patients (n = 132) in whom paroxysmal AF did not progress to persistent AF, and group II included those (n = 72) who developed persistent AF. In group II, age, percentage of men, percentage of patients with diabetes mellitus, maximum P-wave duration, Pd, and left atrial dimension were significantly higher than in group I (p <0.05). Multivariate logistic regression analysis using these 6 factors identified age (odds ratio 2.18, 95% confidence interval 1.41 to 3.41, p <0.01) and Pd (odds ratio 1.91, 95% confidence interval 1.51 to 2.44, p <0.01) as independent predictors of a transition to persistent AF. Pd >or=40 ms predicted progression to persistent AF with sensitivity of 71%, specificity of 77%, positive predictive value of 63%, negative predictive value of 83%, and accuracy of 75%. In conclusion, Pd was a clinically useful predictor of progression from paroxysmal to persistent AF.