Blood Markers for the Diagnosis and Prognosis of Stroke

Many blood markers have been associated with stroke. I set out to determine whether blood markers can be applied to: (i) improve the accuracy of the clinical diagnosis of stroke or TIA, and/or (ii) improve the prediction of poor outcome in patients who are still symptomatic at the time of admission with stroke or TIA. I systematically reviewed the existing literature on the diagnostic performance of a range of blood markers measured soon after stroke onset, to inform the choice of markers for my subsequent prospective studies in this thesis. Many studies had deficiencies in their design, which may have explained the apparently - and perhaps spuriously - impressive diagnostic performance of several markers. In the light of these data I was able to improve the design of my own studies and suggest how future studies of diagnostic markers could be improved. In order to define an appropriate comparator test for assessing the diagnostic accuracy of blood markers, I first examined the performance of emergency room nurses and doctors. I assessed the accuracy of their diagnosis of TIA or stroke ('acute cerebrovascular disease') in patients presenting with symptoms of suspected stroke, and compared them with a number of stroke diagnostic scales. In the 405 patients recruited to the study, the sensitivity of emergency department staff was 77% and specificity 58%. Each stroke diagnostic scale had a slightly better sensitivity, though worse specificity, than an emergency department clinician. I decided to use the diagnosis by an emergency department clinician of 'probable or definite acute cerebrovascular disease' as the best clinical performance reference standard. In blood taken from the same cohort of 405 patients, accredited research laboratories measured markers of inflammation, thrombosis, thrombolysis, cardiac strain and cerebral damage. Tissue plasminogen activator and loge N-terminal pro brain natriuretic peptide were associated positively with a diagnosis of acute cerebrovascular disease, though each marker did not add diagnostic value to the diagnosis of an emergency department doctor or nurse. I systematically reviewed the literature examining the association between the levels of blood markers with poor outcome (i.e. death or dependency) after stroke. I found that although almost all markers studied had a positive association with poor outcome, there were methodological problems with many studies, chiefly small sample size, publication bias or within study reporting biases, and lack of adjustment for important confounders such as age or stroke severity. With data from the Edinburgh Stroke Study, I examined the association between circulating markers of the inflammatory response (white cell count, interleukin-6, Creactive protein and fibrinogen) and poor outcome after stroke. After adjustment for age, whether the patient lived alone, was independent of activities of daily living, was orientated, able to lift both arms and able to walk, I found that higher levels of interleukin-6, white cell count and glucose were associated with poor outcome. The relevant test of a biological marker is not its predictive ability alone, but whether, when added to a validated predictive model based on clinical variables, it improves the prediction of outcome. No individual marker improved the prediction of poor outcome when added to a validated prognostic model based on clinical variables alone. From my cohort of 405 patients with suspected stroke 285 patients had a confirmed diagnosis. Follow up of these 285 patients with confirmed acute cerebrovascular disease showed that, after adjustment for neurological impairment and age, only interleukin-6 and N-terminal pro brain natriuretic peptide were significantly associated with death or disability at 3 months. Neither marker improved the predictions of a model to predict poor outcome based on clinical variables alone. To examine the relationship between circulating markers of the inflammatory response and recurrent stroke, myocardial infarction, and vascular death ('recurrent vascular events'), again I used data from the Edinburgh Stroke Study. After adjustment for clinical predictors (age, prior MI, stroke, or TIA and AF) I found that higher levels of interleukin-6, C-reactive protein and fibrinogen remained significantly associated with an increased risk of recurrent vascular events. However, the relationship with deaths from all causes was somewhat stronger for each marker, perhaps suggesting that higher marker levels were associated with debility rather than vascular events per se. In conclusion, I found no marker measured could improve on the diagnostic accuracy of an emergency department clinician for acute cerebrovascular disease, nor improve the prediction of poor outcome by a prognostic model based upon clinical variables. The work of this thesis does not support the routine use of blood markers as an aid to the diagnosis of, or the prediction of outcome of, acute stroke.