Twenty one studies that evaluated 58 single biomarkers and seven combinations were included in the review. Accuracy data were available for 21 biomarkers. Approximately 3,000 stroke patients and 1,500 controls were included in the review. Four of the 21 studies assessed biomarkers in patients with suspected stroke; the rest compared patients with stroke to a control group. Six of the 21 studies used a pre-specified diagnostic threshold; the rest derived the cut-off threshold from the cohort examined. Only two of the 21 studies reported assessment of the biomarker blind to stroke status, although all stroke diagnoses appeared to have been blind to biomarker results. Nine studies reported the delay between symptom onset and the drawing of blood for biomarkers (range 30 minutes to five days).
Single biomarkers: Five had reported sensitivities over 90% (NDKA, PARK7, UFD-1, NMDA receptor (NR) 2 fragment and NR2A/B antibodies) and 14 had reported specificities over 90% (PARK7/ RNA-BP, UFDP, NDKA, GSTP, ischemia modified albumin (IMA), visin like protein (VLP-1), beta globin DNA, NR2 fragments, S100 beta, FABP, neurone specific enolase (NSE), NR2A/2B antibodies, myelin basic protein (MBP) and thrombomodulin). Most markers were assessed in only one cohort of patients.
Combinations of markers: Data were extracted for five different combinations of markers, but none of the studies reported the regression equation for the combined panel of markers that would have allowed the calculation of the additional informative value provided by each biomarker.
Discrimination between ischaemic and haemorrhagic stroke: Five studies assessed six individual biomarkers and two panels of markers for discrimination between ischaemic and haemorrhagic stroke. Most studies defined a haemorrhagic stroke as the case-positive diagnosis. The two studies that reported a positive test as a diagnosis of ischaemic stroke both had high sensitivity and specificity (absolute values not reported); the markers assessed were apoprotein C1, apoprotein C3 and NR2A/2B antibodies.
Sensitivity and specificity values, with 95% CIs, for each data set were illustrated on forest plots, but numerical values were not reported.