Abstract 24: Predictors of Occult Atrial Fibrillation in Cryptogenic Stroke

Abstract

Background and Purpose: Prolonged cardiac monitoring may identify paroxysmal atrial fibrillation (AF) in patients with cryptogenic stroke. We aimed to identify clinical, echocardiographic, and neuroimaging features which may increase the efficiency of detecting AF on cardiac monitoring. Methods: We studied a retrospective cohort of 227 subjects with cryptogenic ischemic stroke referred for 28 day mobile cardiac outpatient telemetry (MCOT). Patients with large artery disease or high risk sources of cardioembolism were excluded. We reviewed medical records, brain images, and echocardiograms, blinded to MCOT results. Acute and/or chronic infarctions were characterized by size, location, and as cortical, subcortical, or both; wedge-shaped; lacunar; borderzone; and/or multiple territories. Cardiac features included left atrial (LA) size, ejection fraction, aortic arch atheroma, and PFO. Variables were tested in univariate analyses and further incorporated in a multivariate logistic regression model to determine independent predictors of detecting AF. Results: The cohort age was 62.9±2.9 years, 42% were men, and 53% were white. Median CHADS was 3 and CHADS2Vasc was 5. Infarcts were >1.5 cm in 62% of subjects, predominantly cortical in 47%, subcortical in 39%. Only 9% were single, deep, and <1.5 cm. LA size was 3.6±0.7 cm and ejection fraction was 61±9%. MCOT detected AF in 30 (13%) patients. In multivariate analysis, AF was only associated with age>60 (OR 3.6 [1.2-10.4], p=0.02) and prior (chronic) cortical or cerebellar infarction (OR 3.3 [1.3-8.6], p=0.013) (C-statistic 0.72). There was no association with any other clinical, echocardiographic, or radiographic parameter. AF was detected in 32% of patients with age >60 and the presence of prior cortical or cerebellar infarction, compared to 4% with neither of these factors. Conclusion: Atrial fibrillation is detected on MCOT in a substantial minority of cryptogenic stroke patients. Age>60 and the presence of prior cortical or cerebellar strokes are predictive of detecting AF in these patients. Other brain and cardiac characteristics were not found to be helpful. These data may aid in the selection of patients for prolonged arrhythmia monitoring.