Stability, Precision, and Near-24-Hour Period of the Human Circadian Pacemaker

See allHide authors and affiliations

Science  25 Jun 1999:
Vol. 284, Issue 5423, pp. 2177-2181
DOI: 10.1126/science.284.5423.2177


Regulation of circadian period in humans was thought to differ from that of other species, with the period of the activity rhythm reported to range from 13 to 65 hours (median 25.2 hours) and the period of the body temperature rhythm reported to average 25 hours in adulthood, and to shorten with age. However, those observations were based on studies of humans exposed to light levels sufficient to confound circadian period estimation. Precise estimation of the periods of the endogenous circadian rhythms of melatonin, core body temperature, and cortisol in healthy young and older individuals living in carefully controlled lighting conditions has now revealed that the intrinsic period of the human circadian pacemaker averages 24.18 hours in both age groups, with a tight distribution consistent with other species. These findings have important implications for understanding the pathophysiology of disrupted sleep in older people.

Natural selection has favored endogenous circadian rhythmicity that, in the absence of periodic synchronizing cues from the environment, persists with an intrinsic period close to that of Earth's rotation in nearly all living organisms, including prokaryotes. Clock genes participating in transcriptional-translational feedback loops generate circadian oscillations in plants, insects, and mammals (1, 2), with a period (3–5) that is usually near 24 hours, is highly stable, and exhibits remarkably little interindividual variability within a given species—percent coefficients of variation (PCVs) of only 0.08% in the kangaroo rat, 0.3% in hamsters, 0.54% in the gila monster, and 0.7% in mice (3, 4, 6). An age-related shortening of circadian period, which is a determinant of the phase angle of entrainment, has been hypothesized to account for the circadian phase advance and early-morning awakening observed frequently in the elderly (7–11).

Quantification of circadian period in humans has yielded inconsistent results. Although the free-running circadian period of the human activity rhythm was believed to average more than 25 hours, as was initially reported nearly 40 years ago (12), it has since been reported to vary from 13 to 65 hours in normal subjects, with a PCV of 30.3% (13). The average free-running circadian period of the human body temperature rhythm has been reported to vary with both the experimental environment and the subjects' behavior, ranging from 24.2 to 25.1 hours (13–15). However, the generality of these findings has been limited by reports that activity (16, 17), knowledge of time of day (18), and exposure to ordinary indoor room light (19,20) can shift circadian phase or alter the observed free-running circadian period in humans and thus may have influenced those observations (21). Here, we assessed the intrinsic period of the circadian pacemaker in 24 young and older human subjects, each living for approximately 1 month in an environment free of time cues under conditions of controlled exposure to the light-dark cycle on a forced desynchrony protocol pioneered by Kleitman more than 60 years ago (22), using methodology detailed elsewhere (21,23).

We studied 11 healthy young men (mean age 23.7 years) and 13 healthy older subjects (9 men and 4 women; mean age 67.4 years) for 29 to 38 days (24). During the forced desynchrony protocol, the bedtime of each subject was scheduled to occur 4 hours later each day for ∼3½ weeks. Each subject's sleep-wake cycle was thus scheduled to a 28-hour “day” (Fig. 1). Rhythms driven by the circadian pacemaker were thereby desynchronized from each subject's sleep-wake cycle. In this way, exposure to both photic and nonphotic (25,26) synchronizers linked to the scheduled sleep-wake cycle was distributed evenly across all circadian phases (21). The 28-hour day length on this forced desynchrony protocol was (i) far enough outside the range of entrainment of the human circadian pacemaker so as to minimize the influence of the imposed schedule on the observed circadian period (21, 27, 28), and (ii) imposed consistently throughout the protocol. This was done to avoid the artifactual extension of the range of entrainment associated with the fractional desynchronization protocol (29), in which a gradually lengthening light-dark schedule was imposed (13). Also, to minimize the circadian resetting effects of ambient light (19, 28), we maintained constant low light levels during the scheduled wake episodes (Fig. 1). Several subjects returned for additional month-long studies so that we could compare the results of the forced desynchrony protocol with those of the classical free-running protocol (30).

Figure 1

Experimental results from a 22-year-old man (subject 1111) living in an environment free of time cues on a 20-hour forced desynchrony protocol (left panel), a classical free-running protocol (center panel), and a 28-hour forced desynchrony protocol (right panel). The rest-activity cycle is plotted in a double raster format, with successive days plotted both next to and beneath each other and clock hour indicated on the abscissa. Baseline sleep episodes were scheduled at their habitual times (based on an average of their schedule during the week before laboratory admission). Thereafter, sleep/dark episodes (solid bars, light intensity <0.03 lux) were scheduled for 6.67 hours (33% of imposed day) in the 20-hour protocol, self-selected by subject (averaging 28% of cycle) in the free-running protocol, and scheduled for 9.33 hours (33% of imposed day) in the 28-hour protocol. During wake episodes, the light intensity was ∼15 lux (20- and 28-hour protocols) or ∼150 lux (free-running protocol). Constant routines (open bars) for phase assessments of the endogenous circadian temperature nadir (⊗) and the fitted melatonin maximum (▴) were conducted before and after forced desynchrony in all subjects except 1209, who began forced desynchrony immediately after the three baseline days. Period estimations were performed with the use of temperature data (continuously collected via rectal thermistor throughout all studies) and plasma melatonin and cortisol data (assayed from samples collected every 20 to 60 min during segments of the study in the 20- and 28-hour protocols). The estimated phase of the circadian temperature rhythm (dashed line) was determined by nonorthogonal spectral analysis (31, 32). The temperature period estimates are nearly equivalent under both forced desynchrony protocols (20-hour protocol, 24.29 hours; 28-hour protocol, 24.28 hours), independent of the imposed rest-activity cycle. However, the estimated temperature period (25.07 hours) observed during free-running conditions (with self-selected rest-activity cycle averaging 27.07 hours) was much longer.

Core body temperature, plasma melatonin, and plasma cortisol were sampled during the forced desynchrony protocols. Endogenous circadian period was estimated using a nonorthogonal spectral analysis (NOSA) technique, in which these data were fitted simultaneously with periodic components corresponding to both the forced period of the imposed sleep-wake cycle and the sought-for period of the endogenous circadian rhythm, together with their harmonics, using an exact maximum likelihood fitting procedure (31, 32).

The estimated intrinsic periods of the core body temperature, melatonin, and cortisol rhythms were highly correlated when analyzed within an individual subject (Table 1) (33), which supports the hypothesis that the circadian period measured in these studies reflects the intrinsic period of a central circadian pacemaker. Therefore, our estimate of the intrinsic period of the circadian pacemaker for each subject was computed by averaging the period estimates derived from each available variable. These intrinsic circadian period estimates from the 24 subjects were narrowly distributed, with nearly 90% of the estimates between 24.00 and 24.35 hours (Fig. 2). The average estimated (±SEM) intrinsic period was 24.18 ± 0.04 hours (PCV 0.54%) in the young men and 24.18 ± 0.04 hours (PCV 0.58%) in the older subjects (see Table 1).

Figure 2

Histogram of intrinsic circadian period (τ) estimates derived from young and older subjects. Intrinsic circadian period estimates of older subjects are indicated by solid bars, those of young subjects by open bars. Each subject's estimated intrinsic circadian period is reported as the average of the estimated periods from his or her core body temperature, melatonin, and cortisol rhythms (see Table 1).

Table 1

Intrinsic periods of the temperature (τt), melatonin (τm), and cortisol (τc) rhythms (expressed as hours:minutes) in young and older subjects in the 28-hour forced desynchrony protocol. For each subject, the estimated period of each of the three rhythms lies within the 95% confidence interval of the other two rhythms. τt, τm, and τc were highly correlated [Pearson correlation: τt versus τm, r = 0.951; τt versus τc, r = 0.982; τm versus τc, r = 0.984 (P < 0.0001 in all cases)]. Our composite estimate of the intrinsic period for each subject (τ) was computed by averaging τt, τm, and τc, if available. Constraints on the total blood collection volume and vascular access limited the number of older subjects for whom cortisol and melatonin data were available; also, in two young subjects (1145 and 1257), an inadequate number of blood samples were collected and analyzed for cortisol concentrations to obtain a reliable estimate of circadian period.

View this table:

The intrinsic period we observed does not appear to have been dependent on the length of the imposed sleep-wake cycle. The intrinsic period of the core body temperature rhythm derived from subjects 1111 and 1507 on both a 20- and a 28-hour forced desynchrony study were nearly equivalent: 24.29 and 24.28 hours, respectively, for subject 1111 (see Fig. 1) and 24.26 and 24.16 hours, respectively, for subject 1507. Estimates of the intrinsic period of the core body temperature data alone from the combined group of 24 subjects on the 28-hour forced desynchrony protocol (mean ± SEM = 24.17 ± 0.03 hours) and from a series of 14 subjects studied on a 20-hour forced desynchrony protocol from two other experiments (mean ± SEM = 24.15 ± 0.04 hours) (34) were not significantly different (P = 0.6211). One older subject (1507) also participated in a 42.85-hour forced desynchrony protocol and exhibited a temperature period of 24.15 hours, as compared to a period of 24.16 hours on the 28-hour forced desynchrony protocol. These results are also consistent with the 24.20-hour temperature period estimate of an additional young man (1134) who participated in an 11-hour forced desynchrony experiment. Thus, the observed circadian period was equivalent on various imposed sleep-wake and associated light-dark cycles (11, 20, 28, or 42.85 hours). In contrast, when two of the same subjects participated in classical free-running studies in which they self-selected their exposure to a light-dark cycle (light, ∼150 lux; dark, <0.03 lux), the observed period of the temperature cycle was substantially longer [subject 1111, 25.1 hours (Fig. 1); subject 1105, 25.0 hours].

We hypothesize that the longer, more variable circadian period of the temperature rhythm observed in such classical free-running protocols (35) [averaging 25.1 hours (PCV 2.5%) among free-running subjects whose activity-rest cycle was synchronized with their body temperature rhythm and 24.9 hours (PCV 0.8%) among internally desynchronized subjects (13)] occurs because both synchronized and spontaneously desynchronized free-running subjects preferentially select room light exposure before the circadian temperature minimum, and darkness after that minimum (36,37), thereby systematically eliciting light-induced phase delays and minimizing light-induced phase advances (28,38). We thus hypothesize that this unequal distribution of the sleep-wake and associated light-dark cycle across circadian phases in the free-running protocol (as compared with their more equal distribution in the forced desynchrony protocol) was responsible for the overestimation of circadian period derived from the free-running protocol. This hypothesis is supported by (i) the results of simulations using Kronauer's mathematical model of the resetting effect of light on the human circadian pacemaker, which indicate that such feedback effects of ordinary room light alone can lengthen the apparent circadian period observed under classical free-running conditions by more than 0.7 hours (28); (ii) the observation of a shorter average endogenous circadian temperature period derived from free-running subjects when their rest-activity cycle spontaneously desynchronizes from their body temperature cycle, and thereby distributes light exposure more evenly across all circadian phases (13, 28, 36, 39); and (iii) the results of subjects 1105 and 1111, who each exhibited a much longer apparent circadian period when studied on the classical free-running protocol than when studied on the forced desynchrony protocol in dim light.

Unlike the highly variable, much longer circadian period estimates derived from body temperature data in classical free-running human studies, the much smaller coefficient of variation (PCV 0.55%) and the nearer-to-24-hour mean value (24.18 hours) of the intrinsic circadian period estimates derived from all three variables in these forced desynchrony studies is consistent with coefficients of variation and mean values for circadian period estimates observed in other mammals and derived from mathematical modeling of data from human circadian studies (28, 40). These results suggest that the intrinsic period of the human circadian pacemaker is likely to be under the same tight genetic control as has been demonstrated for a wide variety of other species (1, 3–5). Precise estimation of the circadian period is critical for pursuing the possible genetic basis of circadian rhythm sleep disorders.

Our results on the forced desynchrony protocol, together with those of others (14, 15, 41), are in contrast to those of Wever, who observed an average circadian temperature period of 24.8 hours in subjects living in constant conditions in whom internal desynchronization was forced by an imposed 20-, 28-, 30-, or 32-hour cycle of ordinary room light alternating with absolute darkness (13). However, naps around the time of the temperature nadir (42), which are associated with reduced retinal light exposure, may have exerted feedback effects that influenced the estimate of the average circadian period (28), because in that study the timing of sleep was not restricted to the scheduled dark episodes.

Although the group average period estimate from our series is similar to that recently reported using a forced desynchrony protocol of only 5 days in duration (41), the interindividual variability of circadian period estimates derived from that much shorter protocol was significantly greater than it was for estimates derived from our 3- to 4-week protocol (F test,P < 0.0001), with more than half of the period estimates from that study outside the 95% confidence interval of the present study (41, 43).

Even though none of the subjects in our experiments were allowed to nap, we observed a consistent period averaging 24.18 hours, contrary to a prior report of a 24.7-hour circadian period in non-napping subjects (14). We hypothesize that this discrepancy was observed because the sleep episodes of the non-napping subjects in that earlier report were not evenly distributed across circadian phases (14) and were therefore apt to induce feedback effects on the pacemaker (25, 28).

Interestingly, scheduling subjects to a non–24-hour rest-activity cycle alone is not in itself sufficient to assess the intrinsic circadian period in human subjects: The body temperature cycle of subjects scheduled to a 27-hour rest-activity cycle, but not shielded from exposure to Earth's 24-hour light-dark cycle, remained entrained to the 24.0-hour day (44). However, lack of knowledge of the time of day may not be so critical when a non–24-hour schedule is behaviorally imposed. Estimates of the endogenous circadian period of the melatonin rhythm (24.27 hours, PCV 0.84%) derived from a field study of submariners living undersea (and hence shielded from bright outdoor but not artificial light) for 6 weeks while maintaining an 18-hour naval duty schedule were only about 0.1 hour longer than the results reported here from subjects studied in our controlled laboratory environment (45), even though the submariners knew the time of day and only the work hours (but not the sleep or meal times) were scheduled to an 18-hour routine in that field study.

The circadian period of blind subjects not entrained to the 24-hour day while they are living in society has been reported to average 24.3 to 24.5 hours (26, 46), somewhat longer than we now report for sighted subjects. This apparent discrepancy may be due to (i) the influence of the nonuniform distribution of nonphotic synchronizers associated with the self-selected rest-activity cycle of blind subjects [and of sighted subjects living in constant darkness (13)], which has been shown to affect circadian period estimates in other mammals living in constant darkness (25); (ii) the inclusion in the group average of only those blind subjects with longer than average circadian periods who were unable to maintain entrainment via weaker nonphotic synchronizers (25,26), coupled with the classification of all blind subjects whose period estimates were indistinguishable from 24 hours as entrained, resulting in their exclusion from the group average; or (iii) aftereffects of entrainment to the 24-hour day in the sighted subjects (47). The final possibility would suggest that prior entrainment to the 24-hour day in sighted people might shorten the circadian period observed upon release from entrainment.

In the present experiment, contrary to a prior assessment of the temperature rhythm under classical free-running conditions (8), we did not detect a significant difference in the intrinsic circadian period between the healthy young and older subjects studied; the average intrinsic period (±SEM) in the young men was 24.18  ±  0.04 hours, versus 24.18  ±  0.04 hours in the older men and women (P = 0.961) (Table 1), consistent with recent reports in both male and female Syrian hamsters studied throughout their life-span (48). However, with the number of subjects we studied and the observed variability in the intrinsic period, we only had the power (presumed α = 0.05; power = 0.90; standard deviation = 0.15 hours) to detect a difference in circadian period greater than 9 min between the young and older subjects in our study. Given our estimates of the distribution of circadian periods in young and older subjects seen in Fig. 2, it remains possible that a much larger series of such studies might detect a small age-related difference in average circadian period.

Despite comparable circadian periods, the older subjects in this study exhibited the characteristically earlier entrained circadian phase and earlier morning awakening typically found in this age group relative to young subjects (23). Therefore, it is unlikely that the systematic age-related advance in circadian phase and the time of spontaneous awakening can be attributed—at least in this healthy group—to an age-related shortening of circadian period (7–11). The recent report of a similar estimate of circadian period in adolescents (49) further supports the conclusion that this pacemaker property remains stable with age. Putative mechanisms for age-related changes in sleep-wake timing and consolidation include age-related changes in the sleep-homeostatic process and its interaction with the circadian and entrainment processes (23). However, these results do not preclude the possibility that abnormal circadian entrainment might be due to an abnormal circadian period in some older individuals, as has been reported (10).

These results contribute to understanding circadian entrainment in both young and older people and have practical implications for understanding the pathophysiology of, and developing treatments for, circadian rhythm sleep disorders, including the dyssomnia of night shift work, transmeridian travel, both delayed and advanced sleep phase syndrome, and disrupted sleep in older people. These data reveal that the human circadian pacemaker is as stable and precise in measuring time as that of other mammals, and they suggest that understanding of the molecular mechanisms regulating circadian period in other species may well apply to humans (50).


Stay Connected to Science

Navigate This Article