OBJECTIVE To investigate age-associated changes in the circadian rhythm of the heart rate. METHODS The circadian rhythm was extracted from diurnal heart rate (HR) variations, and patterns of HR rhythm were compared in centenarians and controls. METHODS Centenarians living in the metropolitan area of Tokyo and in Aichi prefecture in 1992. METHODS Fifty centenarians underwent 24-hour ambulatory electrocardiogram (Holter) monitoring. The control group, comprised of 100 clinically healthy subjects who underwent similar Holter monitoring, was subdivided, by age, into two groups: the younger controls (age range 23 to 54 years, mean age 41) and the older controls (age range 55 to 82 years, mean age 69). METHODS Harmonic analysis was used to approximate the 24-hour RR interval (the interval between two neighboring R waves on the electrocardiogram) data obtained by Holter monitoring to a summation of three cosine waves with 24-hour, 12-hour, and 8-hour periods. The power of the period was adjusted for the goodness of curve-fit. The power of each period and the circadian acrophase (the timing of the peak in a 24-hour rhythm) were compared among the centenarians, older controls, and younger controls. HR rhythms were classified by k-means cluster analysis based on the power of the period. The prevalence of each pattern was compared among the three age groups. In the centenarians, the relationship between clinical parameters (activities of daily living, cognitive function, nutritional status, and present illness) and patterns of HR rhythm was investigated. RESULTS The power of the 24-hour period in the centenarians was significantly smaller than that in the older (P < .05) and younger (P < .001) controls. The power of the 8-hour period in the centenarians was significantly larger than that in the younger controls (P < .05). Advances or delays in the circadian acrophase were frequently observed in the centenarians compared with the younger controls. The power of each period did not differ between centenarians with (n = 11) and without (n = 39) overt diseases capable of altering HR rhythms. Five patterns of HR rhythm were identified: 24-hour period dominant (n = 84), 24-hour+12-hour period (n = 18), 12-hour period dominant (n = 11), 8-hour period dominant (n = 7), and low goodness of curve-fit (n = 30). The 8-hour period dominant pattern and the low goodness of curve-fit pattern were observed commonly in the centenarians, whereas the 24-hour period dominant pattern and the 24-hour+12-hour period pattern were observed frequently in the younger controls. Patterns of HR rhythm were not related to clinical parameters in the centenarians. CONCLUSIONS The circadian rhythm of HR changed with aging: there was reduction in the power of the 24-hour period, augmentation in the power of the 8-hour period, and a shift in the circadian acrophase.