Neurol Sci. 2026 Apr 18;47(5):435. doi: 10.1007/s10072-026-09021-4.
ABSTRACT
BACKGROUND AND AIMS: Embolic stroke of undetermined source (ESUS) is biologically heterogeneous and often presumed cardioembolic (CE). We investigated whether thrombus composition can capture etiologic heterogeneity within ESUS, clarifying its overall similarity to CE stroke and exploring whether selected radiological features are associated with atherothrombotic-like thrombus patterns.
METHODS: Retrospective single-center study of anterior-circulation ischemic stroke patients treated with mechanical thrombectomy (MT) (2022-2025) with available thrombus, classified as large-artery atherosclerosis (LAA), CE, or ESUS. Thrombus composition (RBCs, fibrin, leukocytes; % thrombus area) was analyzed using prespecified non-parametric hierarchical comparisons with Bonferroni correction (k = 3). Reference analyses compared LAA vs. CE; primary analyses ESUS vs. LAA and CE; secondary and tertiary analyses stratified ESUS by ipsilateral high-risk non-stenotic carotid plaque (hrNSCP - vs. hrNSCP+, defined by CTA-based Plaque-RADS score 3-4). Robustness was assessed using isometric log-ratio transformation and exploratory multivariable models.
RESULTS: Among 344 MT-treated patients, 159 (46.2%) were included (91 CE, 18 LAA, 50 ESUS). LAA and CE showed distinct thrombus profiles, with higher RBC and lower fibrin content in LAA compared with CE (p < 0.01). ESUS thrombi were descriptively CE-like, without significant differences versus CE or LAA after correction. After stratification, two distinct and biologically coherent thrombus phenotypes emerged: ESUS hrNSCP- (n = 42) thrombi overlapped with CE, whereas ESUS hrNSCP+ (n = 8) thrombi were indistinguishable from LAA; these patterns were confirmed by compositional and multivariable analyses.
CONCLUSIONS: When etiologic heterogeneity is taken into account, thrombus composition reveals distinct cardioembolic-like and atherothrombotic-like patterns within ESUS, supporting its role as a complementary tool for etiologic attribution alongside clinical and imaging data.
PMID:41998311 | DOI:10.1007/s10072-026-09021-4