Introduction: Accurate determination of stroke etiology is important in order to implement appropriate secondary stroke prevention. Approximately 25% are embolic strokes of undetermined source, a subset of which are eventually attributed to cardioembolism after prolonged outpatient electrocardiogram monitoring. We applied metabolite profiling to determine whether a metabolomic profile could identify a cardioembolic signature that could inform stroke of undetermined source.
Hypothesis: We hypothesized that a specific metabolomic profile reflected by circulating metabolites could be derived from patients with stroke due to cardioembolism.
Methods: Using liquid chromatography-tandem mass spectrometry, we analyzed 153 metabolites measured in plasma samples collected within 9 hours of stroke onset. Three hundred twenty six patients from an acute stroke cohort collected at two institutions were analyzed. Stroke subtypes were assigned using the Causative Classification System. Association with cardioembolic subtype was assessed and metabolites exceeding the Bonferroni corrected p value threshold were identified. Logistic regression was used to identify independent metabolite predictors of cardioembolic stroke.
Results and Conclusion: The following metabolites were significant in univariate analysis: 2-hydroxyglutaric acid, glutathione, glyceric acid, hippuric acid, threonine, ornithine, histidine, cotinine, pseudouridine, C18:2-carnitine, tricarboxylic acid metabolites, and tryptophan pathway metabolites (all p<6.5x10-3). Regression analysis identified the following multivariate predictors of cardioembolic stroke: hippuric acid (p=0.045), ornithine (p=6x10-5), histidine (p=4x10-6), and a composite sum of tricarboxylic acid metabolites (p=0.004). Metabolite profiling may augment traditional approaches to diagnosing cardioembolic stroke and may have implications for the diagnosis of embolic stroke of undetermined source.