OBJECTIVE: To assess the relation between P wave duration and the occurrence of atrial fibrillation in hypertrophic cardiomyopathy (HCM). DESIGN: High resolution signal averaged electrocardiography was used in 110 patients with HCM to determine whether patients at risk for paroxysmal atrial fibrillation could be detected during sinus rhythm by measuring P wave duration. Filtered P wave duration was measured manually, over an average of 300 beats per patient. RESULTS: During follow up, 18 patients (16%) had at least one one clinical episode of paroxysmal atrial fibrillation, including four patients whose initial episode followed the signal averaged electrocardiogram. The 18 patients with overt atrial fibrillation had greater P wave duration than the 69 patients who did not develop atrial fibrillation: 150 (SD 20) v 126 (14) ms; P < 0.001. Another 23 patients without clinically evident atrial fibrillation had prolonged bursts of the arrhythmia on Holter recording, and showed a P wave duration (138 (15) ms) intermediate between patients with and without clinical atrial fibrillation. In assessing risk for atrial fibrillation, P wave duration > or = 140 ms was associated with sensitivity, specificity, and positive predictive accuracy values of 56%, 83%, and 66%. Multivariate analysis showed that duration of the P wave was the only independent variable associated with occurrence of atrial fibrillation; if P wave duration > or = 140 ms was combined with left atrial enlargement > 40 mm, the specificity and positive predictive accuracy for atrial fibrillation increased to 93% and 78%. CONCLUSIONS: Measurement of P wave duration in sinus rhythm by high resolution signal averaged electrocardiography may be a useful non-invasive clinical tool for identifying patients with HCM likely to develop electrical instability and atrial fibrillation.