Abstract 13142: Prediction of Embolic Stroke in People Without Atrial Fibrillation and Prior Stroke Using Measures of Atrial Myopathy: The Atherosclerosis Risk in Communities (ARIC) Study

Introduction: There is no effective way to predict embolic stroke (ES) in people without atrial fibrillation (AF). Atrial myopathy—characterized by left atrial (LA) remodeling—is associated with ES independent of AF.

Hypothesis: An ES prediction model for people without AF or stroke can be developed with measures of atrial myopathy.

Methods: ARIC study participants without AF, stroke, or anticoagulation at visit 5 (2011-13) were followed for 5 years. Covariates included CHA ₂ DS ₂ VASc variables, P-wave indices, echocardiographic measures of LA function and size, and N-terminal B-type Natriuretic Peptide (NTproBNP). The primary outcome was ischemic stroke. A secondary outcome of possible embolic stroke (PES) was defined as non-lacunar strokes and strokes with new AF within 1 year of stroke. Cox proportional hazards models were created for 5-year ischemic stroke (model A) and PES (model B) risk. Stepwise backwards selection (inclusion threshold P<0.20) generated parsimonious models. Model discrimination and calibration were evaluated by the c-statistic and Hosmer-Lemeshow χ ² statistic, respectively.

Results: 91 ischemic strokes (53 PES) were identified in 4154 participants (59.9% female, mean age 75.1(5.0) years). Age, sex, peripheral arterial disease, hypertension, abnormal P-wave axis, and left atrial reservoir strain were included in model A. Model B added NTproBNP. The C-statistics for models A and B were 0.653 (0.596-0.709) and 0.744 (0.680-0.808), respectively. Both models were well calibrated.

Conclusions: Prediction of ES may be achieved using measures of atrial myopathy in people without AF and prior stroke. Validation of our model is needed.