Biomarkers for Prediction of Stroke and Bleeds in Atrial Fibrillation: Ready for Prime Time?

Abstract

Strokes associated with atrial fibrillation (AF) have a 30-day mortality rate of 24%, and the majority of patients who survive have residual neurological deficits that challenge their ability to live independently.1–3 The ability to accurately stratify risk of ischemic stroke has been evolving for the past 25 years.4 Current international guidelines have embraced the CHA2DS2-VASc scheme (a clinical prediction rule based on age, sex, history of heart failure, history of hypertension, history of known atherosclerosis, diabetes mellitus, and history of stroke or transient ischemic attack) because of improved stratification of lower risk.5 Lifelong anticoagulant therapy is recommended for patients with scores of ≥2. Despite its widespread adoption, the ability of the CHA2DS2-VASc scheme to predict ischemic stroke risk in diverse populations is suboptimal (C statistic, 0.64; 95% confidence interval [CI], 0.58–0.70).6 From a statistical standpoint, risk models with a C statistic >0.80 are generally considered to be very good, between 0.70 and 0.80 good, and between 0.50 and 0.70 weak in their ability to discriminate risk. Biomarkers offer the potential to overcome the crude nature of dichotomous variables that do not capture severity or duration of disease (eg, hypertension, heart failure, diabetes mellitus) or individual biological response to these insults. An individual biomarker profile may more accurately reflect the pathophysiological underpinnings of thrombogenicity and susceptibility to AF itself. In this issue, Berg and colleagues7 report their external validation of the ABC scores (based on age, biomarkers, and clinical history) for stroke and bleeding within the ENGAGE AFTIMI 48 trial population (Effective Anticoagulation With Factor Xa Next Generation in Atrial Fibrillation–Thrombolysis in Myocardial Infarction 48).7–10 High-sensitivity troponin T, N-terminal B-type natriuretic peptide, age, and history of stroke/transient ischemic attack were the strongest predictors of stroke and systemic embolism, yielding a C index of 0.67 (95% CI, 0.65–0.70). In contrast, the CHA2DS2-VASc scheme resulted in a C index of 0.59 (95% CI, 0.57–0.62), although the authors correctly highlight the higher risk of the ENGAGE AF trial participants, the majority of whom were taking anticoagulant therapy at baseline. A recent comparative effectiveness review6 concluded that the CHADS2 (a clinical risk prediction rule based on congestive heart failure, hypertension, age ≥75 years, history of diabetes mellitus, and previous stroke or transient ischemic attack symptoms), CHA2DS2-VASc, and ABC-stroke score were comparable, although the authors acknowledged that data for the ABC score were limited. The present study by Berg and colleagues7 importantly extends the evidence supporting a multimarker strategy for prediction of stroke risk in AF because of the improved discrimination and precision of the ABC-stroke score across CHA2DS2-VASc categories. Levels of these cardiac structural proteins likely more accurately reflect the unique individual © 2019 American Heart Association, Inc.