Abstract WP220: Determinants for Cryptogenic Stroke: Clinico-radiologic Findings

Abstract

Background: Cryptogenic stroke accounts for 30-40% of all strokes. Aortic arch atheroma(AAA), patent foramen ovale(PFO) and paroxysmal atrial fibrillation(PAF) are common causes which require different treatment strategies. However not all patients could undergo extensive work-ups. We investigated clinico-radiological characteristics of these 3 common causes of cryptogenic stroke to find a model helping decision-making. Method: We enrolled 194 patients with cryptogenic stroke at the time of admission. After performing extensive work-ups including electrocardiography, transthoracic or transesophageal echocardiography, transcranial right-to-left shunt test, coronary computed tomography angiography or cardiac telemetry, patients were allocated into 3 groups; (1) AAA(n=25), (2) PFO(n=94), and (3) PAF(n=75). Vascular risk factors and findings on diffusion-weighted image were analyzed. Results: Clinico-radiologic features were different. Patients with PFO were younger (mean, 56.1[PFO] vs. 73.9[AAA] vs. 70.6[PAF] years; p <0.001) and hypertension was less common ( p <0.001). Smoking was more common in AAA group ( p =0.018). PAF-related strokes had higher NIHSS ( p <0.001) and a larger lesion (≥20mm, 81.1%[A-fib] vs. 16.0%[AAA] vs. 46.2%[PFO]; p <0.001). By contrast, AAA-related strokes showed small but multiple lesions scattered in multiple vascular territories ( p <0.001). Patients with PFO had subcortical lesions ( p =0.029) in posterior circulation ( p =0.004). Multivariate regression analysis revealed that age, NIHSS, hypertension, multiple(≥3) lesions and involvement of multiple vascular territories could differentiate 3 groups (R 2 =0.6581). We calculated probability for 3 groups and performed external validation. Conclusion: Patients with cryptogenic stroke showed distinct features which may provide clues for stroke mechanisms. Our data suggested that etiologic work-ups for cryptogenic stroke could be guided by these features.