Variation in heart rate range by 24‐h Holter monitoring predicts heart failure in patients with atrial fibrillation

Abstract Aims The analysis of heart rate (HR) changes, such as the HR variability or HR turbulence, has been reported as a marker of cardiovascular events during sinus rhythm; however, those relationships during atrial fibrillation (AF) remain controversial, and those parameters are not commonly used in AF patients. We sought to investigate the relationship between a simple index focused on the HR and heart failure (HF) events in patients with permanent AF. Methods and results We enrolled 198 patients with permanent AF and evaluated the HR range, defined as the maximum HR minus the minimum HR on 24‐h Holter electrocardiogram recordings. The patients were divided into two groups, i.e., the larger (n = 101) and smaller (n = 97) HR range (HRR) groups, determined by the median value. The HF events were defined as hospitalizations for HF or urgent hospital visits due to exacerbations of one's HF status. The observation period of this study was set at 5 years from registration. The median age was 73 (68–77) years, and 29% were female. The median HRR was 84 (63–118) beats per minutes (bpm). During the observational period of 1825 days (median), HF events occurred in 37 (0.047 per patient‐year) patients. In a log‐rank test, the larger HRR group had more frequent HF events than the smaller HRR group (P = 0.0078). In the adjusted Cox proportional hazards model using the significantly different factors from the univariate analysis (Model 1) and factors and medications associated with HF (Model 2), the larger HRR group had a higher prevalence of HF events than the smaller HRR group for both models [Model 1, adjusted hazard ratio = 3.21, 95% confidence interval (CI) 1.593–6.708, P = 0.0009; Model 2, adjusted hazard ratio = 3.12, 95% CI 1.522–6.685, P = 0.002]. When analysed using the time‐dependent Cox proportional hazards model, the HRR was associated with HF with a statistically significant difference in both the univariate and multivariate analyses [hazard ratio = 1.01, 95% CI 1.006–1.020, P = 0.0002; Model 1, adjusted hazard ratio = 1.02, 95% CI 1.011–1.027, P