Background & purpose: Stroke diagnosis could be challenging in acute phase. We aimed to determine the diagnostic accuracy of total leucocyte count (TLC) and C-reactive protein (CRP) in differentiating Ischemic Stroke (IS) and Hemorrhagic Stroke (HS) using a predictive model approach.

Methods: It is an ongoing, prospective, diagnostic test study conducted at Neurology Department of All India Institute of Medical Sciences, New Delhi, India. Predictive models were developed by combining biomarkers and clinical variables in logistic regression models. The biomarkers were assayed by using the Enzyme Linked Immuno Sorbent Assay (ELISA) technique. Accuracy was evaluated using receiver operating characteristic (ROC) curves. Entire statistical analysis was performed in STATA version 13.0 software.

Results: 181 stroke patients have been recruited (105-IS & 76-HS) in the study till date. Average time from symptom onset to blood collection was 24 hours. TLC >8450 (OR 1.98; 95% CI 1.01 to 3.86) & CRP>27.05 (OR 3.20; 95% CI 1.06 to 9.64) individually differentiated IS and HS; however statistical significance was lost when TLC was included in the logistic regression model, but it was only integrated by clinical variables with a predictive accuracy of 82.71%. CRP>27.05 (OR 4.59; 95% CI 1.18 to 17.81) together with systolic blood pressure and age in the prediction model differentiated IS and HS with a predictive accuracy of 78.43% having a sensitivity of 94% & specificity of 60%.

Conclusion: Using a predictive model approach, CRP along with clinical variables had the potential to be used as a biomarker for differential diagnosis of stroke. Addition of new biomarkers in prediction model is required to increase its predictive accuracy.

DIAGNOSTIC ACCURACY OF TOTAL LEUCOCYTE COUNT AND C-REACTIVE PROTEIN AS BLOOD BASED PROTEIN BIOMARKERS IN DIFFERENTIATING HEMORRHAGIC STROKE FROM ISCHEMIC STROKE