Among the applicants for medical examination in a CHA health-promoting center, 359 city dwelling volunteers (192 women and 167 men) were recruited from July to September, 2004. Their saliva and blood samples were collected between 10 and 11 AM. Four saliva samples throughout the day (20-30 minutes after awakening, 11 AM, 4 PM, and bedtime) were collected from 78 participants (42 women and 36 men; 20 to 40 years of age) who had a normal sleep and wake-up cycle. None of the participants selected in both groups presented with diabetes, hypo- or hypertension. Those who were undergoing hormone replacement therapy, taking hormones for birth control, or using sleeping pills were excluded. General information about volunteer health is summarized in Table. 1

Participants were asked to rinse their mouth with water before collecting saliva. A minimum volume of 1 mL saliva was obtained directly by expectorating into a collecting tube. Food and beverages such as tea, soft drinks, and coffee were not permitted 30 minutes prior to any sample collection. Because salivary cortisol concentration is independent of flow rate and sugarless gum does not interfere with the salivary assay, 21 participants were permitted to chew sugarless gum if needed to stimulate saliva flow. Participants were also instructed not to brush their teeth 30 minutes before saliva collection and to refrain from wearing lipstick on the day of saliva collection. When saliva and blood were simultaneously collected from a volunteer, saliva was always collected first. Saliva samples which were contaminated with blood were excluded from this study. The collected saliva was subjected to two freezing-thawing cycles in order to precipitate mucins, and then centrifuged (10,000 × g, 15 min, 4 ℃).22 The supernatant was collected and stored at -70 ℃ until the day of assay. Blood samples collected between 10 to 11 AM were kept at room temperature and allowed to clot. The samples were then centrifuged to obtain serum, which was stored at -70 ℃ until the day of assay.

Total cortisol and DHEA levels in serum were measured with a commercial RIA kit (Diagnostic Products Cooperation, DPC). The intra-assay variation coefficient for cortisol was 3.5% (n = 10) and 3.7% for that of DHEA (n = 10). The inter-assay coefficient of variation for cortisol was 4.5% (n = 6) and that for DHEA was 4.3% (n = 6)

Commercially supplied kits were not adequate for the measurement of cortisol levels in saliva.23 Therefore, sophisticated assay techniques were required for the analysis of salivary steroids. We developed a hormone assay system based on a liquid phased-double antibody method. For standards, pure cortisol and DHEA were obtained from Sigma-Aldrich Chemical Co. (St. Louis, MO, USA). These steroids were dissolved in methanol (1.0mg/mL) and further diluted with gelatin containing 0.1M phosphate buffered saline (GPBS, pH = 7.0, containing 0.015M NaN₃ and 0.1% gelatin). Iodine-125, labeled cortisol [cortisol-3-(O-caboxymethyl oximino)-2-¹²⁵I iodohistamine], was obtained from Amersham Biosciences (Buckinghamshire, UK) and ¹²⁵I-labeled DHEA, a component of the DHEA RIA kit obtained from DPC, was used for the salivary DHEA assay. The cortisol and DHEA antiserums were purchased from Biogenesis (Oxford, UK). The cortisol antiserum cross-reacts 0.001% with aldosterone, 4.1% with 11-deoxycorticosterone, 5.7% with 11-deoxycortisol, 0.5% with 21-deoxycortisol, 1.2% with corticosterone, and less than 0.01% with other steroids. The DHEA antiserum cross-reacts 6.3% with 5α-andorstane-3β, 17β-diol, 1.3% with androstenedione, 0.1% with testosterone, and less than 0.05% with other related compounds. General assay procedures were adapted from those described by our previous reports.24,25 The assay procedure is briefly described below. The volume of standards added to assay tubes was 0.1mL. A standard with no added cortisol or DHEA was also included. Tubes in which GPBS was substituted for the antiserum were prepared to estimate nonspecific binding (NSB). An equal volume of charcoal-stripped saliva was added to the standard tubes as in sample tubes. The total radioactivity of ¹²⁵I-cortisol or ¹²⁵I-DHEA per tube was approximately 20,000 cpm. The antiserum was diluted to give approximately 30-35% binding. The total reaction volume was 0.4 mL/tube. Routinely, two sets of standards were included in each assay. Standards and samples were incubated overnight at 4℃. Then 0.5 mL of an antirabbit IgG coated (developed in goat) cellulose bead solution was applied to the assay tubes. The tubes were incubated at room temperature for 30 minutes, and then centrifuged (1,500 × g, 20 minutes). The supernatant was aspirated, and the radioactivity in the pellet was counted using a gamma counter (Cobra 5005, Perkin Elmer Life and Analytical Sciences). The steroid levels for individual assay were calculated by interpolation from the standard curves using RiaSmart software (PerkinElmer Life and Analytical Sciences). Multiple tubes of low and high level quality control (QC) samples were prepared and kept frozen for use in each assay. The intra-assay coefficient of variation for cortisol was 5.4% (n = 10) and 5.1% for DHEA (n = 10). The inter-assay coefficient of variation for cortisol was 6.5% (n = 6) and 6.5% for DHEA (n = 6). The analytical sensitivity for cortisol was 1 pg/mL and that for DHEA was 0.1pg/mL.

Steroid-free saliva was prepared as follows. Approximately 50,000 cpm of tritium labeled cortisol (1,2,6,7-³H-hydrocortisone, PerkinElmer Life and Analytical Sciences) or DHEA (1,2,6,7-³H-DHEA, PerkinElmer Life and Analytical Sciences) was added to 20mL of saliva sample. Twenty percent (w/v) of activated charcoal powder (Sigma-Aldrich Chemical Co.) was added to each tube and stirred at 4℃ for overnight. The tubes were then centrifuged (15,000 × g, 30 minutes, 4℃) and the supernatant was collected. The radioactivity remaining in the supernatant was determined using a liquid scintillation analyzer (Tricarb 2900, PerkinElmer Life and Analytical Sciences). The charcoal stripping procedures were repeated until there were not more than 100 cpm remaining in a 1 mL sample. The charcoal stripped sample was filtered through a 0.45µm filter and stored at -70 ℃ until use. To determine the RIA accuracy, unlabeled cortisol or DHEA was added to the charcoal stripped saliva, and the steroid level was determined using RIA as described above. Non-detectable levels of cortisol or DHEA were observed in the charcoal stripped saliva. Exogenously added cortisol (250pg/0.1mL) was detected with a range from 236 to 277pg/0.1mL (255.2 ± 14.7 pg/0.1mL, n = 10), while exogenous DHEA (50 pg/0.1mL) was detected with a range from 48 to 61pg/0.1mL (53.9 ± 3.5pg/0.1mL, n = 10).

Conventional methods (mean and SD, Student's t-test, and ANOVA) were used to analyze the diurnal variation in cortisol and DHEA levels with the level of significance set at p < 0.01. Linear regression analysis was carried out to determine the relationship between hormone levels and age by using STATISTICA version 5.1 for Windows (Tulsa, OK, USA). The Pearson's correlation (r, P) was calculated for correlations of hormone levels between saliva and serum with GraphPad Prism version 4 for Windows (San Diego, CA, USA)

Fig. 1 shows the relationship of salivary (A) and serum (B) cortisol levels (sample collection, 10-11 AM) with age. The age-related decline pattern and concentration of cortisol appeared to be very similar in both genders. Therefore, cortisol levels of men and women were combined together in the plot. Salivary cortisol levels were similar at all ages examined (9.6 ± 0.68, p = 0.62 by ANOVA) and the results of linear regression analysis showed that the salivary cortisol levels (Fig. 1A) did not significantly change with age (slope = -0.078, t = -0.389, p = 0.697). However, serum cortisol levels (Fig. 1B) declined significantly with age (slope = -2.29, t = -4.43, p < 0.001). The serum cortisol levels in the 20s age group were relatively higher (288.2 ± 14.5nmol/L) than those in other age groups (30s; 258.1 ± 11.7, 40s; 228.4 ± 9.3, 50s; 216.0 ± 13.6, and 60s; 221.4 ± 21.7nmol/L). In addition, serum cortisol levels were significantly lower in the 40s age group when compared with the 20s age group (p < 0.01, by Duncan's multiple range test). Saliva represented 3.4% of cortisol levels found in serum for the 20s age group, but this ratio increased with age (3.8%, 4.5%, 4.9%, and 4.9%; 30s, 40s, 50s, and 60s, respectively). The results of linear regression analysis also showed that the saliva: serum cortisol ratio increased with age (Fig. 1C, slope = 0.051, t = 3.61, p < 0.001).

Fig. 2 shows the relationship of AM salivary (A) and serum (B) DHEA levels with age. Similar levels of DHEA were detected in saliva collected from the 20s, 30s, and 40s age groups (0.79 ± 0.03 nmol/L, p = 0.9 by ANOVA), but DHEA levels decreased substantially in the 50s age group (0.62 ± 0.0.5nmol/L, p < 0.01, when compared with the 40s age group), and decreased further in the 60s age group (0.42 ± 0.07nmol/L, p < 0.05, when compared with the 50s age group). Results of linear regression analysis also showed that salivary DHEA levels declined significantly with age (Fig. 2A, slope = -0.007, t = -3.761, p < 0.001). The declining pattern of serum DHEA levels with age was similar to that observed in saliva. Serum DHEA levels were similar between the 20s, 30s, and 40s age groups (21.2 ± 0.2nmol/L), but the levels were significantly less in the 50s age group (16.9 ± 1.3nmol/L, p < 0.01), and decreased further in the 60s age group (10.9 ± 1.8 nmol/L, p < 0.01). Results of linear regression analysis also showed that serum DHEA levels declined significantly with age (Fig. 2B, slope = -0.197, t = -4.9, p < 0.001). Saliva represented 3.5-3.7% of the DHEA concentrations found in serum, and this ratio did not change significantly with age (Fig. 2C, slope = 0.0016, t = 0.344, p = 0.73).

Fig. 3 shows the correlation of cortisol (Fig. 3A) and DHEA (Fig. 3B) concentrations between saliva and serum. Data obtained from 358 saliva and serum samples were used to determine the correlation coefficients. The data presented here was obtained from only paired saliva and serum samples; the results show a direct comparison of hormone levels in saliva with simultaneously obtained serum samples in the absence of hormonal stimulation or suppression. A Pearson's correlation (r, P) was calculated for the comparison of hormone levels between saliva and serum samples. Cortisol levels in saliva were well correlated with those in serum (Fig. 3A, r = 0.59, p < 0.0001). Furthermore, DHEA levels in saliva were also highly correlated with those in serum (Fig. 3B, r = 0.864, p < 0.001).

Fig. 4 shows diurnal changes in salivary cortisol and DHEA levels. Saliva samples were collected at four designated time points throughout a day from 74 volunteers (ages ranging from 20 to 40 years). Cortisol and DHEA levels were measured and are depicted in Fig. 4A and B, respectively. Salivary cortisol diurnal rhythm peak levels occurred first just after awakening (14.1 ± 0.58 nmol/L), followed by a marked decline thereafter, and finally reaching its lowest level at midnight (11 AM; 6.5 ± 0.37, 4 PM; 4.9 ± 0.3, midnight; 1.5 ± 0.18nmol/L)(Fig. 4A). The diurnal pattern of salivary DHEA followed that of cortisol, with the highest levels observed following awakening (1.3 ± 0.2nmol/L) and gradually decreasing thereafter (0.9 ± 0.06, 0.9 ± 0.09, and 0.7 ± 0.06nmol/L, 11AM, 4PM, and midnight, respectively). There was a significant decrease in DHEA concentration between post- awakening and 11AM (p < 0.001, by t-test), but a slight decrease from 11AM to midnight was observed (p = 0.11 by ANOVA) (Fig. 4B).