Proteomics to Identify New Blood Biomarkers for Diagnosing Patients With Acute Stroke.

Background: Blood biomarkers are a potential tool for early stroke diagnosis. We aimed to perform a pilot and exploratory study on untargeted blood biomarkers in patients with suspected stroke by using mass spectrometry analysis.

Methods And Results: This was a prospective observational study of consecutive patients with suspected stroke admitted within 6 hours of last being seen well. Blood samples were collected at admission. Patients were divided into 3 groups: ischemic stroke (IS), intracerebral hemorrhage (ICH), and stroke mimics. Quantitative analysis from mass spectrometry data was performed using a supervised approach. Biomarker-based prediction models were developed to differentiate IS from ICH and ICH+stroke mimics. Models were built aiming to minimize misidentification of patients with ICH as having IS. We included 90 patients, one-third within each subgroup. The median age was 71 years (interquartile range, 57-81 years), and 49 participants (54.4%) were women. In quantitative analysis, C3 (complement component 3), ICAM-2 (intercellular adhesion molecule 2), PLGLA (plasminogen like A), STXBP5 (syntaxin-binding protein 5), and IGHV3-64 (immunoglobulin heavy variable 3-64) were the 5 most significantly dysregulated proteins for both comparisons. Biomarker-based models showed 88% sensitivity and 89% negative predictive value for differentiating IS from ICH, and 75% sensitivity and 95% negative predictive value for differentiating IS from ICH+stroke mimics. ICAM-2, STXBP5, PLGLA, C3, and IGHV3-64 displayed the highest importance score in our models, being the most informative for identifying patients with stroke.

Conclusions: In this proof-of-concept and exploratory study, our biomarker-based prediction models, including ICAM-2, STXBP5, PLGLA, C3, and IGHV3-64, showed 75% to 88% sensitivity for identifying patients with IS, while aiming to minimize misclassification of ICH. Although our methodology provided an internal validation, these results still need validation in other cohorts and with different measurement techniques.