Background Biomarkers of a number of pathophysiological processes might improve the clinical diagnosis of stroke and transient ischaemic attack (TIA).

Methods We recruited symptomatic patients with suspected stroke or TIA <24 h after onset. We measured markers of inflammation, thrombosis; thrombolysis, cardiac dysfunction and cerebral damage. A panel of stroke physicians, neurologists and neuro-radiologists reviewed the clinical features, brain imaging and subsequent course of each patient to make the gold standard diagnosis. We constructed multivariate logistic models to analyse the data.

Results We recruited 405 patients with suspected stroke (285 stroke or TIA, 180 mimics) at a median of 7 h (IQR 3–19 h) after symptom onset. Higher levels of Ln NT pro-BNP (OR 2.2 (95% CI 1.5 to 3.0) 75th to 25th centile) and t-PA (OR 1.6 (1.2 to 2.2)) were associated with a diagnosis of TIA or stroke. Adjustment for neurological impairment and age attenuated these associations. The associations of higher levels of adiponectin (OR 1.8 (1.2 to 2.4)) and IL-10 (OR 1.1 (1.0 to 1.1)) with a diagnosis of mimic were robust to adjustment for neurological impairment, age and infection. A model with eight markers from systematic review had no better diagnostic performance than an emergency department clinician.

Conclusion We found no single marker or combination of markers that positively predicted a diagnosis of stroke or TIA independently of neurological impairment and age, or added usefully to the clinical diagnosis.