An angiopoietin 2, FGF23, and BMP10 biomarker signature differentiates atrial fibrillation from other concomitant cardiovascular conditions

作者：Winnie Chua, Victor Roth Cardoso, Eduard Guasch, Moritz F. Sinner, Christoph Al-Taie, Paul Brady, Barbara Casadei, Harry J.G.M. Crijns, Elton Dudink, Stéphane Hatem, Stefan Kääb, Peter Kastner, Lluı́s Mont, František Neháj, Yanish Purmah, Jasmeet S. Reyat, Ulrich Schotten, Laura C. Sommerfeld, Stef Zeemering, André Ziegler, Georgios V. Gkoutos, Paulus Kirchhof, Larissa Fabritz · 发表于：Scientific Reports · 年份：2023 · DOI：10.1038/s41598-023-42331-7 · 被引用次数：30 · 研究领域：Cardiovascular Function and Risk Factors、Atrial Fibrillation Management and Outcomes、Cardiovascular Health and Disease Prevention

Early detection of atrial fibrillation (AF) enables initiation of anticoagulation and early rhythm control therapy to reduce stroke, cardiovascular death, and heart failure. In a cross-sectional, observational study, we aimed to identify a combination of circulating biomolecules reflecting different biological processes to detect prevalent AF in patients with cardiovascular conditions presenting to hospital. Twelve biomarkers identified by reviewing literature and patents were quantified on a high-precision, high-throughput platform in 1485 consecutive patients with cardiovascular conditions (median age 69 years [Q1, Q3 60, 78]; 60% male). Patients had either known AF (45%) or AF ruled out by 7-day ECG-monitoring. Logistic regression with backward elimination and a neural network approach considering 7 key clinical characteristics and 12 biomarker concentrations were applied to a randomly sampled discovery cohort (n = 933) and validated in the remaining patients (n = 552). In addition to age, sex, and body mass index (BMI), BMP10, ANGPT2, and FGF23 identified patients with prevalent AF (AUC 0.743 [95% CI 0.712, 0.775]). These circulating biomolecules represent distinct pathways associated with atrial cardiomyopathy and AF. Neural networks identified the same variables as the regression-based approach. The validation using regression yielded an AUC of 0.719 (95% CI 0.677, 0.762), corroborated using deep neural networks (AUC 0.784 [95% CI 0.745, 0.822]). Age, sex, BMI and three...