Introduction: A rapid diagnostic tool that appropriately triages stroke and non-stroke patients would decrease patients’ time to treatment and therefore lower the risk of morbidity. Single blood protein biomarkers for stroke have not shown clinical utility. Our objective in this biomarker discovery study is to examine a wide array of proteins to identify a panel that stratifies stroke patients from stroke-mimicking patients.
Hypothesis: Serum from ischemic stroke patients contains proteins identified by the SOMAscan biomarker discovery assay that can stratify stroke from non-stroke patients.
Methods: Patients presenting to the emergency department with stroke-like symptoms had peripheral blood drawn upon arrival according to Stroke Alert protocols. Once informed consent was obtained, serum was analyzed with the SOMAscan assay to quantify the expression of 1305 proteins. These data were analyzed using the semi-supervised machine learning algorithm, random forest analysis (RFA), to determine a combination of proteins which best predicted the final patient discharge diagnosis.
Results: Twenty-two stroke patients, 9 transient ischemic attack (TIA) patients, and 23 patients with other stroke-mimicking symptoms participated in the study. Normal age- and sex-matched serum controls were obtained commercially. RFA identified four panels encompassing a total of nine proteins to stratify patients between the stroke, TIA, other stroke mimic, and control groups. Applying these panels to the same training set of patient data, one panel of two proteins distinguished stroke from stroke-mimics with a sensitivity of 50% and specificity of 100%. A second panel of three proteins identified stroke from non-stroke patients with a sensitivity of 78% and specificity of 95%. A provisional patent application describing these panels is underway.
Conclusion: Ischemic stroke patients and stroke-mimic patients exhibit a differential expression of serum proteins at the time of clinical presentation. The simultaneous analysis of two or three proteins in a biomarker panel exhibits high specificity and may be clinically useful as a stroke “rule in” tool for emergency personnel. This would facilitate triage and avoid inappropriate treatment of non-stroke patients.