A prospective study of atrial fibrillation and stroke.

In a prospective study of 72 patients with stroke and atrial fibrillation, we classified strokes as cardioembolic or noncardioembolic based on arterial assessment using Doppler sonography and angiography. We analyzed and cross-tabulated 18 clinical characteristics and found four to be significantly associated with a cardioembolic mechanism: stroke with onset during activity and peak deficit at onset (p less than 0.008), previous infarct in a different vascular territory (p less than 0.01), previous transient ischemic attack in a different vascular territory (p less than 0.01), and transient ischemic attack lasting greater than 1 hour (p less than 0.02). Starting with these four characteristics, we used a step-down procedure to select variables for a logistic regression model. Only previous infarct in a different vascular territory (odds ratio = 7.38) and transient ischemic attack lasting greater than 1 hour (odds ratio = 7.89) were selected by the model. Using M-mode and two-dimensional echocardiography, we compared left atrial size in 46 patients with that in 78 controls who had atrial fibrillation without stroke. Left atrial size in patients and controls with mitral valvulopathy was significantly larger than that in patients and controls without mitral valve disease. There was, however, no difference in left atrial size between patients with nonvalvular atrial fibrillation and cardioembolic stroke and controls or patients with nonvalvular atrial fibrillation and noncardioembolic stroke. We concluded that some clinical characteristics are closely related to cardioembolic stroke and that left atrial enlargement reflects underlying cardiopathy rather than atrial emboli-forming capability.