Backgound The long-term age-related changes in circadian rhythm of heart rate variability (HRV), that is, autonomic nervous activity, remain unknown in elderly people. Methods and Results Holter monitoring was conducted twice at an interval of 15 years in 15 healthy elderly patients (age: 70.0±4.1 years, at first monitoring, female: 10) and assessed the age-related changes in 24-h mean and hourly mean normal sinus R-R interval (mean NN), HRV (high frequency (HF) component, low frequency (LF) component and LF/HF) and the circadian rhythms. As a result, 24-h mean mean NN (0.976±0.115 vs 0.903±0.117 (s), p=0.0019), LF/HF (1.681±0.731 vs 0.962±0.442, p=0.0022), and LF (278.88±176.43 vs 179.19±132.33 (ms²), p=0.0039) significantly decreased 15 years later, although 24-h mean HF (221.20±138.89 vs 310.78±296.73 (ms²), p=0.1102) increased slightly. The hourly mean NN closely correlated with hourly HF and LF/HF throughout circadian rhythms both at first and second monitoring. In the morning hours, amplitude rates of all HRV indices increased significantly 15 years later. Conclusion In elderly people, age-related changes in the 24-h mean heart rate (HR) were conversely dissociated from those of the 24-h mean HRV. However, the close correlation between hourly HR and HRV was preserved, even in very elderly patients. Additionally, the amplitude rates in HRV in the morning increased with age. These age-related changes of HR and HRV might be characteristic of elderly people. (Circ J 2006; 70: 889 - 895)