Abstract
In acute heart failure (AHF) with atrial fibrillation (AF), major bleeding, ischemic stroke, and death compete as the first clinical event. Whether anemia and renal dysfunction together define a single bleeding-risk construct is unclear. We retrospectively analyzed 712 consecutive patients hospitalized with AHF and AF in a multicenter registry (2014-2020). Hemoglobin < 12 g/dL and creatinine clearance (CrCl) < 50 mL/min defined four mutually exclusive groups. Fine-Gray models evaluated bleeding-first, stroke-first, and death-first events, with hemoglobin and CrCl also entered simultaneously as continuous covariates. Incremental value for 365-day bleeding-first was assessed in one common cohort with bootstrap optimism correction. Of 712 patients (mean age 79.1 years; 48.6% female; 81.7% on oral anticoagulants), 138 had neither abnormality, 40 anemia only, 236 renal dysfunction only, and 298 both. Bleeding occurred first in 81 patients (11.4%), stroke in 43 (6.0%), and death in 140 (19.7%). Adjusted bleeding-first risk was highest with both abnormalities (subdistribution hazard ratio [sHR] 3.67; 95% confidence interval [CI] 1.39-9.68; p = 0.009 versus neither), but neither the contrast against anemia only (sHR 2.42; 95% CI 0.74-7.91; p = 0.14) nor the anemia × renal dysfunction interaction (p = 0.98) was significant. Modeled continuously, lower hemoglobin was associated with bleeding-first (sHR 1.22 per 1 g/dL decrease; 95% CI 1.06-1.40; p = 0.006) and lower CrCl with death-first (sHR 1.20 per 10 mL/min decrease; 95% CI 1.07-1.35; p = 0.002). Relative to the model with anemia alone, neither adding renal dysfunction (optimism-corrected change in area under the curve [ΔAUC] 0.0138; 95% CI - 0.0473 to 0.0576) nor using the combined binary phenotype (0.0145; 95% CI - 0.0168 to 0.0432) showed clear improvement in prediction. The bleeding association persisted in anticoagulated patients and after excluding bleeding supported only by a hemoglobin fall. Although patients with both anemia and renal dysfunction had the highest observed bleeding-first risk, their coexistence showed neither a supported interaction nor improved 365-day prediction beyond anemia alone. Lower hemoglobin and lower CrCl showed different observed association patterns across the competing first events, favoring component-level interpretation over a unitary anemia-renal risk construct.