Novel Models for the Prediction of Left Atrial Appendage Thrombus in Patients with Chronic Nonvalvular Atrial Fibrillation

Predicting left atrial appendage thrombus (LAAT) in chronic nonvalvular atrial fibrillation remains challenging despite the fact that several predictive models have been proposed to date. In this study, we sought to develop new and simpler models for LAAT prediction in chronic nonvalvular atrial fibrillation. The study enrolled 144 patients with chronic nonvalvular atrial fibrillation who underwent transesophageal echocardiography for LAAT detection. We examined the association of LAAT incidence with the CHA(2)DS(2)-VASc score and echocardiographic parameters pertaining to the left atrium (LA), including diameter, volume index, strain, and strain rate measured on speckle tracking echocardiography. LAAT was found in 24.3% of patients (39/144). The following parameters had good diagnostic performance for LAAT: LA volume index >57 mL (area under the curve (AUC), 0.72; sensitivity, 77.1%; specificity, 64.2%), LA positive strain ≤6.7% in the four-chamber view (AUC, 0.84; sensitivity, 77.1%; specificity, 77.1%), and LA negative strain rate >−0.73 s(−1) in the four-chamber view (AUC, 0.83; sensitivity, 85.7%; specificity, 70.6%). The CHA(2)DS(2)-VASc score alone had a low predictive value for LAAT in this population (χ(2) = 3.53), whereas the combination of CHA(2)DS(2)-VASc score with LA volume index had significant association and better predictive value (χ(2) = 12.03), and the combination of CHA(2)DS(2)-VASc score with LA volume index and LA positive strain or negative strain rate in the four-chamber view had the best predictive ability for LAAT (χ(2): 33.47 and 33.48, respectively). We propose two novel and simple models for noninvasive LAAT prediction in patients with chronic nonvalvular atrial fibrillation. These models combine the CHA(2)DS(2)-VASc score with LA volume index and LA longitudinal strain parameters measured on speckle tracking echocardiography in the four-chamber view. We hope these simple models can help with decision-making in managing the antithrombotic treatment of such patients, whose risk of stroke cannot be determined solely based on the CHA(2)DS(2)-VASc score.