Abstract 50: RNA Expression for Diagnosis of Stroke Etiology Differentiating Large Artery and Cardioembolic Stroke: Analytical Validation of Testing From the BASE Clinical Trial

Abstract

Background: An accurate test to differentiate large artery stroke (LAS) patients from those with cardioembolic stroke (CES) would be of significant clinical utility. Using the Biomarkers of Acute Stroke Etiology (BASE) trial (NCT02014896) dataset, our purpose was to utilize blood gene expression signatures for accurately differentiating LAS from CES acute stroke etiologies. Methods: The BASE trial enrolled suspected stroke patients presenting to 20 hospitals within 24 hrs of symptom onset. Final gold standard diagnosis and stroke etiology were determined by an adjudication committee using all hospital data but blinded to RNA test results. Whole blood, obtained in PAX tubes, was frozen at -20C within 72 hrs and analyzed at a core lab (Ischemia Care, Dayton, OH) using Affymetrix HTA microarrays. Genes on the HTA microarray were filtered to eliminate genes with low expression or high CV (> 10%) when run on replicate samples leaving 9,513 potential signature genes. A two-way random forest classifier was built through cross validation of the training data resulting in a 45 gene diagnostic signature. Results: This is a planned interim cohort study of the 1700 patients enrolled in the BASE trial that does not include lacunar strokes, TIA, or stroke mimics. Overall, 222 patients were enrolled with NIHSS>5, 70 (32%) with LAS and 152 (68%) with CES; 59% were male, and median (IQR) age was 70.7 yrs (62.0, 80.2). Median (IQR) time from symptom onset to blood collection was 1200 (448, 1568) minutes. Coexistent pathology at presentation included atrial fibrillation 90 (48%), hypertension 153 (82%), hyperlipidemia 87 (47%), diabetes 60 (32%), and coronary artery disease 70 (37%). Patients were randomly divided into training (148), early symptom onset (<18hrs) validation (39) and a late symptom onset (>18 hrs) validation (35). The diagnostic gene signature results in the early validation cohort distinguished LAS from CES; C-statistic 0.78 (0.50-1.0, 95% CI), sensitivity 0.90 (0.55-1.0, 95% CI) and specificity of 0.70 (0.43-1.0, 95% CI). Conclusion: Early RNA expression differentiates large artery stroke patients from those with cardioembolic stroke, and may have therapeutic and secondary prevention implications.