Heart Rhythm. 2025 Oct 31:S1547-5271(25)03035-8. doi: 10.1016/j.hrthm.2025.10.054. Online ahead of print.
ABSTRACT
BACKGROUND: The effect of blood inflammatory biomarkers on atrial fibrillation (AF) incidence remains unclear.
OBJECTIVE: This study was to examine the effect of inflammatory biomarkers on AF incidence among senior adults.
METHODS: 3,265 healthy senior adults from the Guangzhou Heart Study were followed until end of 2024. Twenty blood inflammatory biomarkers were detected at baseline survey. AF was ascertained by electrocardiogram and clinical examination. Hazard ratio was calculated using Cox model. The prediction models were established using five machine learning methods.
RESULTS: 5.33% participants developed AF during the follow-up. After adjusting for confounders and comparing the highest with lowest tertiles, total bilirubin, direct bilirubin, and uric acid were associated with 1.80-fold, 1.68-fold and 1.63-fold risk of AF, whereas eosinophils, platelets, triglyceride-glucose index (TyG), and cholesterol were associated with 34%, 33%, 33% and 41% reduced risk. Every one-unit increment in red cell distribution width (RDW) was associated with an 11% increased risk of AF. The random survival forest (RSF) model with 14 predictors demonstrated the best predictive performance, with AUC, sensitivity, specificity, C-index, and Brier score being 0.832, 69.05%, 70.05%, 0.767, and 0.032 in training set, and 0.731, 68.75%, 61.97%, 0.697, and 0.031 in validation set.
CONCLUSION: Elevated baseline levels of RDW, total bilirubin, direct bilirubin, and uric acid increased the risk of AF, while elevated baseline levels of eosinophils, platelets, TyG, and cholesterol decreased the risk. The RSF model demonstrated excellent predictive performance for AF, but it was only internally validated and external validation is warranted to confirm its generalizability.
PMID:41177321 | DOI:10.1016/j.hrthm.2025.10.054