BackgroundHeart failure (HF) is common among patients with atrial fibrillation (AF), and accurate risk assessment is clinically important. ObjectivesThe goal of this study was to investigate the incremental prognostic performance of N-terminal pro–B-type natriuretic peptide (NT-proBNP), high-sensitivity cardiac troponin T (hs-cTnT), and growth differentiation factor (GDF)-15 for HF risk stratification in patients with AF. MethodsIndividual patient data from 3 large randomized trials comparing direct oral anticoagulants (DOACs) with warfarin (ARISTOTLE [Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation], ENGAGE AF-TIMI 48 [Effective Anticoagulation With Factor Xa Next Generation in Atrial Fibrillation–Thrombolysis In Myocardial Infarction 48], and RE-LY [Randomized Evaluation of Long-Term Anticoagulation Therapy]) from the COMBINE-AF (A Collaboration Between Multiple Institutions to Better Investigate Non-Vitamin K Antagonist Oral Anticoagulant Use in Atrial Fibrillation) cohort were pooled; all patients with available biomarkers at baseline were included. The composite endpoint was hospitalization for HF (HHF) or cardiovascular death (CVD), and secondary endpoints were HHF and HF-related death. Cox regression was used, adjusting for clinical factors, and interbiomarker correlation was addressed using weighted quantile sum regression analysis. ResultsIn 32,041 patients, higher biomarker values were associated with a graded increase in absolute risk for CVD/HHF, HHF, and HF-related death. Adjusting for clinical variables and all biomarkers, NT-proBNP (HR per 1 SD: 1.68; 95% CI: 1.59-1.77), hs-cTnT (HR: 1.39; 95% CI: 1.33-1.44), and GDF-15 (HR: 1.20; 95% CI: 1.15-1.25) were significantly associated with CVD/HHF. The discrimination of the clinical model improved significantly upon addition of the biomarkers (c-index: 0.70 [95% CI: 0.69-0.71] to 0.77 [95% CI: 0.76-0.78]; likelihood ratio test, P < 0.001). Using weighted quantile sum regression analysis, the contribution to risk assessment was similar for NT-proBNP and hs-cTnT for CVD/HHF (38% and 41%, respectively); GDF-15 provided a statistically significant but lesser contribution to risk assessment. Results were similar for HHF and HF-related death, individually, and across key subgroups of patients based on a history of HF, AF pattern, and reduced or preserved left ventricular ejection fraction. ConclusionsNT-proBNP, hs-cTnT, and GDF-15 contributed significantly and independently to the risk stratification for HF endpoints in patients with AF, with hs-cTnT being as important as NT-proBNP for HF risk stratification. Our findings support a possible future use of these biomarkers to distinguish patients with AF at low or high risk for HF.