BACKGROUND: The ability of cardiovascular biomarkers to predict the incidence of stroke subtypes remains ill-defined in the general population.Methods and Results: The blood levels of B-type natriuretic peptide (BNP) and high-sensitivity C-reactive protein (hs-CRP) and urinary albumin corrected by urinary creatinine (UACR) were determined in a general population (n=13,575). The ability to predict the incidence of ischemic stroke subtypes (lacunar, atherothrombotic, cardioembolic) for each biomarker was assessed based on the area under the receiver-operating characteristic curve (AUC-ROC) and using Cox proportional hazard modeling. The predictive abilities of UACR and hs-CRP for any subtype of ischemic event were found to be suboptimal. However, the ability of BNP to predict the incidence of cardioembolic stroke was excellent (AUC-ROC=0.81). When BNP was added to established stroke risk factors, the ability to predict cardioembolic stroke in terms of the AUC-ROC significantly improved (4-year follow-up, P=0.018; 8-year follow-up, P=0.009). Furthermore, when BNP was added to the JPHC score, the ability to predict cardioembolic stroke was significantly improved (net reclassification improvement=0.968, P<0.0001: integrated discrimination improvement=0.039, P<0.05). CONCLUSIONS: In the general population, plasma BNP was an excellent biomarker for predicting the incidence of cardioembolic stroke when used alone or in combination with established stroke risk factors.
BACKGROUND: The ability of cardiovascular biomarkers to predict the incidence of stroke subtypes remains ill-defined in the general population.Methods and Results: The blood levels of B-type natriuretic peptide (BNP) and high-sensitivity C-reactive protein (hs-CRP) and urinary albumin corrected by urinary creatinine (UACR) were determined in a general population (n=13,575). The ability to predict the incidence of ischemic stroke subtypes (lacunar, atherothrombotic, cardioembolic) for each biomarker was assessed based on the area under the receiver-operating characteristic curve (AUC-ROC) and using Cox proportional hazard modeling. The predictive abilities of UACR and hs-CRP for any subtype of ischemic event were found to be suboptimal. However, the ability of BNP to predict the incidence of cardioembolic stroke was excellent (AUC-ROC=0.81). When BNP was added to established stroke risk factors, the ability to predict cardioembolic stroke in terms of the AUC-ROC significantly improved (4-year follow-up, P=0.018; 8-year follow-up, P=0.009). Furthermore, when BNP was added to the JPHC score, the ability to predict cardioembolic stroke was significantly improved (net reclassification improvement=0.968, P<0.0001: integrated discrimination improvement=0.039, P<0.05). CONCLUSIONS: In the general population, plasma BNP was an excellent biomarker for predicting the incidence of cardioembolic stroke when used alone or in combination with established stroke risk factors.