The Variability in Circadian Phase and Amplitude Estimates Derived from Sequential Constant Routines

Both the constant routine (CR) and the dim light melatonin onset have been suggested as reliable methods to determine circadian phase from a single circadian cycle. However, both techniques lack published studies quantifying the intercycle variability in their phase resolution. To address this question eight healthy male subjects participated in two CRs, 7 days apart. Circadian phase was determined using 3-min samples of core body temperature and two hourly urinary sulphatoxy melatonin excretion rates. Phase and amplitude were estimated using simple (24 h) and complex (24 + 12 h) cosinor models of temperature data and the onset, offset, and a distance-weighted-least-squares (DWLS) fitted acrophase for the melatonin metabolite. The variability in phase estimates was measured using the mean absolute difference between successive CRs. Using the simple 24 h model of temperature data, the mean absolute phase difference was 51 min (SD = 35 min). Using the complex model, the mean absolute phase difference was 62 min (SD = 35 min). Using the DWLS fitted acrophase for the melatonin metabolite, the mean absolute phase difference between CR1 and CR2 was 40 min (SD = 26 min). The results indicate that for CRs a week apart, the mean absolute difference in an individual's phase estimate can vary by 40-60 min depending on the choice of dependent measure and analytic technique. In contrast to the intraindi-vidual variability, the group results showed considerably less variability. The mean algebraic difference between CRs, using temperature- or melatonin-derived estimates, was less than 5 min, and well within the range of normal measurement error.