Abstract 4143300: Atrial Function Improves Stroke Prediction in Patients with Atrial Fibrillation

Background: Stroke contributes to significant morbidity and mortality in atrial fibrillation (AF). Clinical risk scores lack predictive accuracy and mechanistic links to thrombo-embolization. Hypothesis Adding MRI-derived parameters to stroke risk calculators improves their predictive accuracy. Aims Demonstrate an association between left atrial emptying fraction (LAEF), long axis strain (LAS) and stroke in AF patients. Incorporating the above-mentioned LARF and LAS into the CHA 2 DVASc score improves its predictive accuracy. Methods: AF patients undergoing cardiac MRI were identified from the University of Washington Cardiac Arrhythmia Data Repository (CADRe). Pts with prior AF ablation were excluded from the study. LAEF and LAS were quantified using a semi-automated contouring of the LA wall was done through image processing software (CVI42®) (Panel A). Results: 417 pts (average age 66±13years; 64% males) were included. The average CHA 2 DVASc (stroke component removed) score was 2.1±1.4. LAEF was 46.18%±11.64%, and LA LAS was 24.28%±13.45%. Univariate analysis revealed a significant association between stroke/TIA and LAEF (OR= 0.962, p= 0.024); but not with LAS (OR=0.962, p=0.091) or CHA 2 DVASc (OR=1.084; p=0.553). LAEF remained significantly associated with stroke in multivariate analysis, adjusting for age, gender, congestive heart failure, hypertension, coronary artery disease, and diabetes, (OR= 0.963, p= 0.042). Three other models including composite CHA 2 DVASc with and without LAEF with their respective ROC curves and AUC are shown in Panel B. Adding LAEF improved the AUC of CHA 2 DVASc (AUC 0.564 vs 0.653; p=0.048) or its components (AUC 0.539 vs 0.632; p=0.043). Using a cut-off LAEF of 46.53% (AUC: 0.71 CI: 0.61,0.81; Sensitivity: 68%; Specificity: 74%) resulted in the reclassification of 47(11.3%) pts to a higher risk category and none (0%) to lower risk (Panel C). Conclusion: Left atrial emptying fraction is independently associated with stroke or TIA, and improves the predictive accuracy of clinical risk scores in AF pts.