W. Whiteley, W. L. Chong, A. Sengupta, P. Sandercock



Publication year




Periodical Number






Author Address

Full version

Background and Purpose-The performance of validated prognostic clinical models in acute ischemic stroke might be improved by addition of data on blood biomarkers. Methods-We searched Medline and EMBASE from 1966 to January 2007 for studies of blood markers in patients with ischemic stroke and an assessment of outcome (death, disability, or handicap). We adopted several strategies to reduce bias. Results-Studies were generally small (median number of subjects, 85; interquartile range, 49 to 184). Few had evidence of a sample size calculation (7 of 82 [9%]) or reported blinding to whether patients had stroke (21 of 82 [26%]). Of the 66 studies reporting a measure of association, 10 did not adjust for age or stroke severity, 14 adjusted for age, 7 adjusted for severity, and 35 adjusted for both; 30% (20 of 66) used a data-dependent threshold to predict good or bad outcome. There was evidence of within-study reporting bias and publication bias. Cardiac markers showed the most consistent association with poor outcome. Conclusions-Blood biomarkers might provide useful information to improve the prediction of outcome after acute ischemic stroke. However, this review showed that many studies were subject to bias. Although some markers had some predictive ability, none of the studies was able to demonstrate that the biomarker added predictive power to a validated clinical model. The clinical usefulness of blood biomarkers for predicting prognosis in the setting of ischemic stroke has yet to be established. (Stroke. 2009; 40: e380-e389.)