By advertising and word of mouth, the investigators recruited 337 volunteers who signed initial consent for the study. The study was approved and undergoes continuing annual review by the UCSD Human Research Protections Program (IRB) and the affiliated IRB of the VA San Diego Healthcare System. It was conducted in accord with the principles expressed in the Declaration of Helsinki. The target ages for young adults were 18–30 years and for older adults were 60–75 years. We sought both older and young-adult participants who were aerobically fit and in good general health, so that they would be capable of undergoing the exercise condition if randomized to that treatment. An inclusion criterion was regular participation in aerobic exercise for ≥20 min/day, ≥3 times/week at an intensity of ≥60% of maximal effort. Many of the volunteers were quite successful competitive endurance athletes (particularly those in the older group). All volunteers underwent medical histories, physical examinations, blood sugar, cholesterol and lipoprotein screening, and physician-supervised monitored exercise to verify the absence of EKG abnormalities. About 1/3 of the initial volunteers were dropped during the screening process for exercise safety considerations (e.g., high cholesterol or EKG abnormalities during monitored exercise) or because they decided they did not wish to complete the protocol. Also, potential participants were excluded if they took medications thought to influence melatonin or cortisol (e.g., melatonin, beta blockers, high doses of aspirin, corticosteroids). More older than young participants were recruited, because it was predicted a larger N might be needed for adequate power to detect a PRC in the older age group.
Preliminary screening studies included sleep and medical history forms and the Pittsburgh Sleep Quality Index (PSQI) [17]. Some of the PSQI results have been reported elsewhere [18]. For 7 days before entering the laboratory, participants wore the Actillume-I wrist actigraph for continuous 24-h recording of activity and illumination exposure. Sleep-wake was inferred by validated algorithms [19, 20]. During the same week, participants completed home sleep logs estimating sleep time and quality, and completed a baseline Center for Epidemiologic Studies Depression Scale (CESD) [21]. The CESD was repeated both on the first and last days in the laboratory and one week later, to measure any mood effects of the interventions. Participants were asked to abstain from alcohol and caffeine for 2 days before entering the laboratory.
Participants first entered the Circadian Pacemaker Laboratory at about 09:30 and were assigned to individual studio apartments with sound and light isolation. They were asked to remain in their rooms or in a hallway with illumination limited to 50 lux for the duration of their time in the laboratory, from 4.7 to 5.6 days. They were not permitted in distant parts of the laboratory near windows or daylight. During their entire time in the laboratory, they were instructed to follow a special ultra-short sleep-wake cycle, consisting of 30 min in bed in complete darkness with sleep encouraged, followed by 60 min out of bed in background illumination, which was maintained at <50 lux in the usual direction of gaze. The ultra-short sleep wake cycle is a protocol used successfully by several laboratories to reduce sleep and light masking of circadian rhythms [22–26]. Although maintained in standardized lighting and ultra-short sleep-wake cycles, participants' social interactions were not restricted. Visitors and contacts with staff were permitted, along with reading, watching television (less than 10 lux), craft projects, working at computer games (less than 8 lux), telephone calls, and preparing meals. Strenuous exercise was not permitted.
Baseline observations were continued for the first 30–53 h, of which the final 24 h were analyzed for baseline circadian assessments. Almost all participants were randomized to receive bright 3000 lux light stimuli or exercise when they first entered the laboratory, without being advised in advance of what treatment they would experience at what times of day. Because of difficulties recruiting healthy participants, 7 older volunteers were invited to enter the light protocol several months after having completed the exercise protocol to which they had initially been randomized. After a baseline of varying length, participants commenced bright light exposures centered at one of 8 times: 0100, 0400, 0700, 1000, 1300, 1600, 1900, or 2200 h. The 8 protocols are illustrated in Fig. 1. A 3-h block of bright light treatment was administered at the same time of day for 3 days. The bedrooms of about 18 m2 were painted with white reflective paint. The ceilings had 8 recessed fixtures, each with a diffuser covering six 4-foot T12 cool white 4100, 40-watt fluorescent bulbs (Philips F4C Advantage X). The lights were controlled externally. For bright light treatments, all bulbs were lit, whereas for 50 lux, only one dimmer bulb was used. The ceiling fluorescent lighting provided approximately 3000 photopic lux to the cornea in a horizontal direction of gaze (see Fig. 2). Structured block randomization was employed so that approximately equal numbers were assigned to each of the 8 bright light stimulus times.
Participants continued to wear the Actillume wrist actigraphs throughout their time in the laboratory. Oral temperatures were taken with fast-reacting high-resolution electronic thermometers every 30 minutes. Because of superior circadian goodness of fit, only those oral temperature measurements obtained every 90 min immediately after awakening were used to obtain circadian analyses. Since these latter temperature measurements were each made in bed after 30 min. lying in bed, temperature was measured in a sort of constant routine which would minimize any effects of posture, activity, or meals. We did not think that the advantages of rectal temperature recording would outweigh the inconvenience and risks to participants. During the baseline and again after the final bright 3000 lux light stimulus, every urine voiding was collected. With few exceptions, participants provided a urine specimen each 90 min during lights-on, drinking at least 200 cc every 90 min to maintain steady production. The volume of each urine sample was measured and aliquots (2 ml) were immediately frozen and then soon transferred to -70°C, where the samples were stored for later assays of 6-sulphatoxymelatonin (aMT6s) and urinary free cortisol. Visual-analog 100 mm line ratings were given on 8 scales every 3 h: these scales were ALERT, SAD, TENSE, EFFORT, HAPPY, WEARY, CALM, SLEEPY, AND OVERALL. Monk and colleagues have validated similar scales in time-isolation laboratory settings [27]. The CESD inventory was repeated near the beginning and towards the end of the laboratory stay.
aMT6s
The aMT6s assays were performed using Bühlmann 96 well ELISA kits (EK-M6S) purchased from ALPCO, Ltd. (Windham, NH). At the usual dilution of 1:200, the analytical sensitivity of the EIA was 0.35 ng/ml and the functional least detectable dose was 1.3 ng/ml for coefficients of variation (CVs) <20%. In our laboratory, control urine samples averaging 4–6 ng/ml gave intra- and inter-assay CVs of 4% and 7%, respectively. All samples from an individual participant were run at the same time and wherever possible on the same 96-well plate. Selected samples (especially peak or "circadian night" samples measuring > 38 ng/ml or samples < 1 ng/ml) were assayed repeatedly at either increased (1:800 to 1:3200) or decreased (1:25 to 1:100) dilution when necessary to obtain more accurate estimates or to clarify irregular circadian patterns in excretion rate (ng/hr).
From the aMT6s concentration, the urine volume, and the collection times, the aMT6s excretion rate (ng/h) was computed for each collection interval (the interval between one voiding and the next one) and subsequently associated with each 5-min interval within the collection interval. From this time series of 5-min intervals, the circadian analyses were computed (see below).
Urinary free cortisol
Urine samples were assayed for free cortisol using DSL-2100 Active Cortisol RIA kits (Diagnostic Systems Laboratories, Inc. Webster, Texas). Because our 90 min sampling protocol typically yielded somewhat dilute urine, the urine sample volume in the RIA was increased to 75 μl combined with 25 μl of zero calibrator, adjusting the volume of kit standards and controls accordingly (e.g. 25 μl standards plus 75 μl deionized water). A low dose control (mean 1.3 μg/dL) run in triplicate in 12 assays gave intra- and inter-assay coefficients of variation (CVs) of 6.8% and 8.7%, respectively. Samples measuring <0.16 or >20.0 μg/dL when run at 75 μl were reassayed using either 250 μl or 25 μl of sample to obtain more accurate estimates. As with aMT6s, the cortisol concentrations were used to infer cortisol excretion for each 5 min interval. Because the urine integrates the pulsatile secretion of cortisol into blood, fewer urine samples than blood samples are needed to obtain a precise assessment of the phase of the circadian system. However, interim analyses suggested that urinary cortisol was not yielding more reliable circadian information than aMT6s, so cortisol was not assayed for the final third of laboratory studies.
Circadian Analyses
Separate analyses were done for the last 24 h of baseline 90-min sleep-wake cycle, before light treatment, and for the comparable final 24 h of follow-up laboratory 90-min cycle (starting 6 h after the end of light treatment to minimize transients). For measures such as urinary aMT6s, urinary cortisol, oral temperature, and actigraphic minute-by-minute scored sleep, the best-fitting 24-h cosine was estimated with a least-squares technique. Then the acrophases (peak of the fitted curve) and mesors (mean of the fitted curve) were obtained as the estimates of the daily mean excretion and circadian timing. Baseline results from some of the first participants in this study, combined with some of the participants who would be assigned to exercise, have been reported previously [18, 28]. Each phase response resulting from bright light stimuli was then computed, e.g., as the acrophase of the baseline minus the acrophase of the follow-up interval. A negative phase shift indicated a delay, e.g., that the acrophase of the rhythm occurred at a later clock time after the stimulus than during baseline. A positive phase shift would indicate an advance, e.g., that the acrophase was at an earlier time at follow-up than at baseline. The phase shifts from baseline to follow-up were then related to the time lag between the center time of the 3-h light stimulus and the baseline acrophase of aMT6s (or temperature, cortisol, etc.) to form the phase-response curves for each variable studied with each phase reference.
At the same time that these experiments were performed, a separate group of men and women of similar ages were exposed to the 90-min ultra-short sleep-wake cycle in the same laboratory with no more than 50 lux light exposures [29]. These subjects appeared to free-run with a period averaging 24.38 h [29]. Thus, for the participants exposed to bright light, the phase shifts were interpreted as relative advances or delays in reference to their mean phase shift, which approximated the free-running delay among the untreated subjects. The mean of the participants undergoing bright light stimuli was regarded as the best estimate for their free-running trend, because the untreated subjects were selected by different criteria and were in the laboratory for a shorter duration, so their estimated free-running period might have been more affected by transients.
To further describe changes in circadian phase and waveform, we estimated the circadian timing of nocturnal aMT6s onsets and offsets algebraically from upward (onset) and downward (offset) crossings of the mesor (ng/h), calculated from 24 h cosine fits to the data (Fig. 3). Shifts in onset and offset times were also computed. To aide interpretation of the PRC data in relation to clock timing in the home environment, some of the figures plotted phase shifts to light on a 24 h abscissa titled Circadian Clock Time (Figures 4, 5, 6). The abscissa Circadian Clock Time references the timing of light stimulation to a phase marker (i.e., aMT6s acrophase or onset), and then displays the environmental time scale corresponding to when the mean phase marker occurred at baseline. The mean phase markers used are also located on the time scale as asterisks. This form of display illustrates our best estimate of the mean environmental clock time at which the stimuli were given, adjusted for variations in each participant's baseline phase.
To test the null hypothesis that there were no phase-response curves, that is, no phase-shifts dependent on the timing of the 3000 lux light stimuli, we used both the PRC bisection test [30] and factorial ANOVA. The PRC bisection test locates the best bisection of the circular distribution of initial phases to maximize the contrast between advances and delays. In general, the best bisection will be at the inflection from delays to advances. The test then determines if the bisection separates advances and delays significantly better than would occur in a random distribution. These tests were performed for all 106 participants and on subgroups of older and young adults, male and female. The inflection points of the PRCs from delay to advance were estimated with the PRC bisection analyses. The amplitudes of PRCs were contrasted between older and young adult groups, men and women, using methods derived from the PRC bisections [30]. Because the PRC bisection test was a new approach, these tests were confirmed by factorial ANOVA, allocating the phase shifts into 6 prospectively-planned 4 h treatment-timing blocks (referenced to the baseline acrophases), and adding age group and gender as additional factors to produce 6 × 2 × 2 analyses. A criterion for significance of p < 0.05 was selected. No correction for multiple testing seemed appropriate, since most tests were significant, and correction would be problematic with tests which were intercorrelated.
Comments
View archived comments (2)