For this study, 74 full-time employed individuals working in the hospital were recruited (35 male and 39 female subjects). The mean age of the study subjects was 41.39 ± 10.22 years. Due to age-related decreases in DHEA levels,26 no participant over the age of 56 years of (ages ranged between 22-55 years old) was included. The subjects worked Monday through Friday from 9 a.m. to 6 p.m. and on Saturday from 9 a.m. to 4 p.m. All of the subjects had worked more than three months at the position they held at the time of this study, and none of the subjects were taking any type of medication. None of the study subjects had any pathological conditions such as diabetes, cancer, hypotension, or hypertension. The samples were collected between May and July of 2008. Among the 74 participants, more than one batch of samples was collected from some of the participants. Saliva samples were collected from 31 subjects for the preliminary experiment, and from 46 participants for the main objectives of this study.

All participants were instructed to abstain from smoking, eating certain kinds of food items, and consuming caffeine-containing beverages. They were instructed to neither brush their teeth nor rinse their mouth with water before collecting the saliva samples. The participants were instructed to label the samples by providing information regarding the sampling time, date, and duration of sleep on the provided labels. A minimum volume of 2 mL of saliva was collected directly into a collecting tube by expectoration. We selected this collection method because cotton-based saliva collection is not adequate for the salivary immunoassay of DHEA.27 This study was conducted in accordance with the principles laid down by the Declaration of Helsinki. All participants provided their informed consent and were provided with information about their hormone levels. Ethical approval for this study was obtained from the local research ethics committee.

A batch of samples collected from a participant was simultaneously excluded when a saliva sample was missed or was collected outside of the designated time point (s) if discovered after the examination of the labels on each tube. In addition, a batch of samples was rejected if a sample was useless, reddish in color, or if the sample contained an insufficient volume to perform the steroid assays. In conclusion, we accepted saliva samples collected from 26 subjects for the preliminary experiment, and 28 subjects for the main objectives of the present study. General information on the saliva collection time and basic health of the 28 subjects who collected saliva samples for 7 consecutive days is summarized in Table 1. All saliva samples were analyzed at the Hormone Research Center at Chonnam National University.

Steroid levels in the saliva were determined using a radioimmunoassay (RIA), employing a previously described liquid-phase double-antibody method.26 Iodine-125-labeled cortisol [cortisol-3-(O-carboxymethyl) oximino-(2-[¹²⁵I]iodohistamine)] was obtained from PerkinElmer life and analytical sciences (MA, USA). Since iodine-125-labeled DHEA [DHEA-7-(O-carboxymethyl) oximino-(2-[¹²⁵I] iodohistamine)] is not commercially available in Korea, it was prepared by modifying a previously described protocol for the radioiodination of steroids.28 Cortisol and DHEA antisera were purchased from Biogenesis (Oxford, UK). Cortisol antiserum cross-reacts with aldosterone (0.001% cross reactivity), 11-deoxycorticosterone (4.1%), 11-deoxycortisol (5.7%), 21-deoxycortisol (0.5%), corticosterone (1.2%), and other steroids (< 0.01%). DHEA antiserum cross-reacts with 5α-androstane-β, 17β-diol (6.3% cross reactivity), androstenedione (1.3%), testosterone (0.1%), and other steroids (< 0.05%). Multiple quality control samples were prepared, kept frozen, and used in each assay. The intra- and inter-assay coefficients of variation for cortisol were 7.34% (n = 20) and 8.28% (n = 9), respectively. The intra- and inter-assay coefficients of variation for DHEA were 6.90% (n = 20) and 8.28% (n = 7), respectively. The analytical sensitivities of the assay for cortisol and DHEA were 10 pg/mL and 1 pg/mL, respectively.

The results are presented as means ± standard deviations (SDs) and as medians and percentiles. Since cortisol and DHEA levels were repeatedly measured for seven consecutive days, the inter-subject variations were eliminated in determination of the day-to-day differences by using an analysis of variance (ANOVA) for a randomized complete block design (RCBD). The seven daily measurements of cortisol levels, DHEA levels, or the molar C/D ratios from each participant were considered to be a complete block. This complete block was used to determine the day-to-day differences,29 and a post-hoc test (Duncan's multiple-range test) was performed to determine the significance of the day-to-day differences. Pearson's correlation coefficients for the cortisol and DHEA levels, the molar C/D ratios for each day of the sample collection week, and the area under the curve (AUC) were calculated using GraphPad Prism version 5.01 for Windows (San Diego, CA, USA), and statistical calculations were performed using SAS version 9.1 (SAS Institute Inc., Cary, NC, USA).

The cortisol levels at different periods after awakening (immediately upon awakening, 15, 30, and 45 minutes after awakening) were measured from saliva samples collected on workdays from 26 subjects (16 male and 12 female, aged 28-55 years old) (Fig. 1). Cortisol levels immediately upon awakening, 15, 30, and 45 minutes after awakening in males were 7.18 ± 3.06 nmol/L, 14.09 ± 6.68 nmol/L, 14.77 ± 7.89 nmol/L and 8.46 ± 4.88 nmol/L, respectively, and those in females were 8.49 ± 2.42 nmol/L, 14.08 ± 6.35 nmol/L, 15.69 ± 4.45 nmol/L and 10.45 ± 5.37 nmol/L, respectively. Because there were no differences in cortisol levels at each collection time between male and female (p > 0.05 in all analyses), the data was pooled together.

The mean cortisol levels were increased and reached a peak 30 minutes after awakening. The mean cortisol level, at its peak, was increased by 113.7 ± 76.96% (7.4 ± 5.6 nmol/L) from those samples that were collected immediately upon awakening.

The results of the preliminary study showed that the cortisol levels reached a peak 30 minutes after awakening. Thus, participants were asked to collect their saliva samples between 25-35 minutes after awakening and to continue collecting samples at the same time for 7 consecutive days.

Cortisol levels were measured in saliva samples collected from 28 subjects (15 male and 13 women, aged 24-53 year) for seven consecutive days at the designed time point (i.e., 30 ± 5 minutes after awakening from a normal nocturnal sleep) (Fig. 2).

The mean cortisol levels on Monday through Saturday were in the range of 13.67-14.94 nmol/L, and those on Sunday were 10.39 nmol/L (Fig. 2). Since cortisol levels from each participant were repeatedly measured for seven consecutive days, the inter-subject variations were eliminated in determination of the day-to-day within subject differences by using ANOVA for a RCBD. Seven consecutive measurements of cortisol levels from each participant were considered a block. The complete block was used to determine the day-to-day differences in the cortisol levels. RCBD-ANOVA tests revealed that there were significant day-to-day differences in the mean cortisol levels between days of the week (F_{6, 168} = 5.77, p < 0.01). Duncan's post-hoc test revealed that the cortisol levels on Sunday were significantly lower than those on the six workdays (p < 0.05). There were no differences between males and females in cortisol levels on workdays (15.7 ± 10.48 nmol/L and 15.04 ± 5.78 nmol/L, respectively; p > 0.05) and on Sunday (11.57 ± 7.78 nmol/L and 10.75 ± 5.25 nmol/L, respectively; p > 0.05).

There was a moderate to strong correlation in the cortisol levels between workdays (Pearson's r = 0.53-0.77, p < 0.01), except between Monday and Friday (Pearson's r = 0.44, p < 0.05). The cortisol levels on Sundays were also correlated with those on the six workdays, however, the correlation was found to be relatively weaker (Pearson's r = 0.30-0.65, p < 0.01) than the correlation between the cortisol levels on the six workdays.

DHEA levels were determined from the same saliva samples that were used for determining the cortisol levels (Fig. 3). The mean DHEA levels on Monday (0.93 ± 0.58 nmol/L) and Tuesday (0.90 ± 0.54 nmol/L) were lower than the other days (ranged 1.02 nmol/L to 1.18 nmol/L) (Fig. 3). The RCBD-ANOVA test revealed significant day-to-day differences in the mean DHEA levels between the days of the week (F_{6, 168} = 2.23, p > 0.05). Duncan's post-hoc test revealed that the DHEA levels on Monday and Tuesday were significantly lower than those on Thursday or Friday (p < 0.05), but the DHEA levels on Saturday, Sunday, and Wednesday were not different from each other (Fig. 3). There were no differences between males and females in the overall DHEA levels on the seven consecutive days (1.0 ± 0.48 nmol/L and 1.13 ± 0.7 nmol/L, respectively; p > 0.05). The DHEA levels between workdays correlated with each other (Pearson's r = 0.62-0.78, p < 0.01). The levels measured from Monday and Friday were weakly correlated (Pearson's r = 0.39, p < 0.05). The DHEA levels on Sunday were weakly correlated with those on the six workdays (Pearson's r = 0.38-0.57, p < 0.05), but the levels measured from the other workdays were well correlated with those from Monday (Pearson's r = 0.63, p < 0.01).

The molar C/D ratios were calculated for each subject by using the data from the cortisol and DHEA levels (Fig. 4). The mean molar C/D ratios for the week were able to be divided into three subgroups, Monday and Tuesday (20.70 and 21.42, respectively), Sunday (12.12), and other workdays (range, 16.63-17.72) (Fig. 4). The RCBD-ANOVA test revealed that there were significant day-to-day differences in the mean molar C/D ratios between the days of the week (F_{6, 168} = 8.10, p < 0.01). Duncan's post-hoc test revealed that the molar C/D ratios on Monday and Tuesday were significantly higher than those on the other four weekdays (p < 0.05), except Wednesday. Interestingly, the molar C/D ratios on Sunday were significantly lower than those on the other six workdays (p < 0.05). There were no significant differences in the molar C/D ratios among Wednesday, Thursday, Friday, and Saturday (Fig. 4). The overall molar C/D ratios on the six workdays were not different between the male and female subjects (19.81 ± 2.26 and 18.22 ± 11.31, respectively; p > 0.05), and the molar C/D ratios on Sunday were not different between the male and female subjects (13.82 ± 6.97 and 10.29 ± 5.78, respectively; p > 0.05). There was a moderate to strong correlation in the molar C/D ratios between the workdays (Pearson's r = 0.62-0.80, p < 0.01), as well as between Sunday and the workdays (Pearson's r = 0.48-0.65, p < 0.01).