A blood-based biomarker panel to detect acute stroke.

BACKGROUND: The aim of this study was to develop an adjunctive, peripheral biomarker test to differentiate ischemic strokes, intracranial hemorrhages (ICHs), and stroke mimics in the acute setting. METHODS: Serum samples were collected from 167 patients who presented with an acute neurologic deficit within 24 hours of symptom onset. Patients were adjudicated to ischemic stroke, ICH, and mimic pathology groups based on clinical and radiographic findings. Samples were tested for levels of 262 potential markers. A multivariate Cox proportional hazards regression model of 5 biomarkers was built by stepwise selection and validated by bootstrapping. Its discriminative capacity was quantified by C index and net reclassification improvement (NRI). RESULTS: The final model consisted of eotaxin, epidermal growth factor receptor, S100A12, metalloproteinase inhibitor-4, and prolactin. It demonstrated a discriminative capacity for ischemic stroke versus mimic (C = .92), ischemic stroke and ICH versus mimic (C = .93), and ischemic stroke versus ICH (C = .82). The inclusion of biomarkers to a model consisting of age, race, and gender resulted in an NRI of 161% when detecting ischemic stroke versus mimic (P < .0001), an improvement of 171% when detecting ischemic strokes plus ICH versus mimic (P < .0001), and an improvement of 56% when detecting ischemic strokes versus ICH (P = .1419). CONCLUSIONS: These results suggest that information obtained from a 5-biomarker panel may add valuable information in the early evaluation and management of patients with stroke-like symptoms.

Duke Scholars

Author Yuliya Vladimirovna Lokhnygina Biostatistics & Bioinformatics, Division of Biostatistics

Author Daniel Todd Laskowitz Neurology, Neurocritical Care