Abstract A139: Plasma Proteomics Nominate Biomarker Candidates Of Acute Large Vessel Occlusion Strokes

Introduction: Early prehospital identification of patients with acute large vessel occlusion (LVO) is critical for timely triage to thrombectomy-capable centers. Objectives: We undertook a cross-hospital proteomics study to identify plasma biomarkers of LVO. Methods: We analyzed clinical and proteomics data from stroke patients aged ≥18 years using two prospective repositories from 2010 to 2020 at Grady Memorial Hospital (GMH) and Yale New Haven Hospital (YNHH). Plasma was collected at emergency room presentation before any intervention at GMH and within 72 hours of stroke onset at YNHH. Our primary objective was to identify differentially abundant protein (DAP) levels in LVO versus non-LVO strokes. Our secondary objective was DAPs in LVO versus acute ischemic stroke (AIS) without LVO, intracerebral hemorrhage (ICH), transient ischemic attack (TIA), and stroke mimics (MIM). We performed aptamer-based proteomics (7K SomaScan) on both cohorts. DAPs were identified using an unadjusted p-value <0.05 (Welch’s t-test), externally validated across independent cohorts, and adjusted for NIH stroke scale (NIHSS) in bivariate logistic regression. Biomarkers were added to NIHSS using forward selection. Prediction models assessed added value of biomarkers over NIHSS and prehospital scales for LVO classification. Results: We included 18 patients with LVO and 22 without LVO from GMH (mean age 63 years, 48% males) and 34 with LVO and 30 without LVO from YNHH (mean age 68 years, 58% males) (Figure 1). We also included 20 patients, each with ICH, TIA, and MIM from GMH. SomaScan quantified 7301 proteins. We identified 784 DAPs in GMH and 349 in YNHH, with 14 concordant DAPs (p <0.05, Figure 2A). Five concordant proteins (GH2, C1QL2, MFAP5, CD2, ACP2) remained independent predictors of LVO after NIHSS adjustment (Figure 2B). Prediction models incorporating GH2 and ACP2, along with NIHSS, significantly improved LVO classification (AUC: 0.91 GMH, 0.89 YNHH) compared to NIHSS alone (AUC: 0.71 GMH, 0.75 YNHH) (Figure 3A). Adding GH2 and ACP2 to prehospital scales (RACE, G-FAST, mG-FAST) increased AUCs to >0.88 (Figure 3B). GH2, C1QL2, and CD2 also classified LVO from non-LVO AIS, ICH, TIA, and MIM with AUCs >0.81 after NIHSS adjustment (Figure 3C). Conclusions: Our externally validated proteomics study identified concordant protein panels that enhance prehospital scales for LVO detection, aiding thrombectomy triage. These findings warrant validation in larger, multicenter cohorts.