It is unknown whether 12-lead ECG can predict new-onset AF. In the present study, we identified patients with new onset AF from our digitally stored ECG database, and the P wave morphologies were analyzed in their preceding sinus rhythm recordings as the precursor state for AF. The P wave was analyzed in the most recent ECG recording of sinus rhythm preceding new onset AF within 12 months. The duration and amplitude of P waves were analyzed in 12 leads and compared between the 2 groups with the other clinical parameters. The study population consisted of 68 patients with new-onset AF and 68 age and sex-matched controls. Multivariate analysis revealed that the P wave amplitude in leads II and V1 (0.157 ± 0.056 versus 0.115 ± 0.057 mV, P = 0.032, and 0.146 ± 0.089 versus 0.095 ± 0.036 mV, P = 0.002) and P wave dispersion (56.9 ± 14.8 versus 33.5 ± 12.9 ms, P = 0.001) were significant independent factors for the prediction of new-onset AF. By using these factors, new-onset AF could be predicted with a sensitivity of 69.1% and specificity of 88.2%. P wave analysis is useful for predicting new onset AF.