Dehydroepiandrosterone Secretion in Healthy Older Men and Women: Effects of Testosterone and Growth Hormone Administration in Older Men

Abstract

Context

Aging is associated with diminished gonadal steroid and GH/IGF-I axis activity; whether these changes contribute to the parallel declines of dehydroepiandrosterone (DHEA) and DHEA sulfate (DHEAS) production is unknown, as are the effects of sex steroid and/or GH administration on DHEA and DHEAS production.

Objective

Our objective was to evaluate morning DHEAS concentrations and nocturnal DHEA secretory dynamics in healthy older men and women, before and after chronic administration of sex steroid(s) alone, GH alone, sex steroid(s) combined with GH, or placebo alone.

Design

We compared nocturnal DHEA secretory dynamics (2000 h to 0800 h, sampling every 20 min, analyzed by multiparameter deconvolution and approximate entropy algorithms) in healthy older (65–88 yr) men (n = 68) and women (n = 36), both before and after 26 wk of administration of sex steroid(s) alone [testosterone (T) in men or estrogen/progesterone in women], GH alone, sex steroid(s) combined with GH, or placebo alone.

Results

Morning concentrations of DHEAS were lower; nocturnal DHEA pulsatile production rate, burst frequency, and amplitude were higher; and half-life was shorter in women (P < 0.05). Nocturnal integrated DHEA concentrations, total production rate, and approximate entropy did not differ significantly by sex. Because of small treatment group sizes in women, only hormone intervention results in men are presented. In men, T and T plus GH administration significantly decreased nocturnal integrated DHEA but not morning DHEAS concentrations. GH alone exerted no significant effects on nocturnal DHEA secretion or morning DHEAS.

Conclusions

Spontaneous nocturnal DHEA secretion is sexually dimorphic in healthy older individuals, and T administration decreases nocturnal DHEA secretion in older men. The clinical significance of sex steroid modulation of DHEA secretion in older persons remains to be elucidated.

DEHYDROEPIANDROSTERONE (DHEA) and its sulfated conjugate DHEAS are the most abundant circulating adrenal steroids, with significantly higher concentrations in men than in women (1–3). Considered biomarkers of aging, DHEA and DHEAS concentrations peak in early adulthood and decline steadily thereafter. By age 80 yr, DHEA(S) concentrations diminish to 10–20% of peak values (1, 2). Epidemiological studies suggest an association between low DHEA(S) concentrations and certain adverse effects of aging (4). In the United States, DHEA is widely available as an over-the-counter dietary supplement. Although DHEA has been promoted in the lay press, its effect on a wide variety of conditions, such as aging, diabetes, obesity, immune dysfunction, arthritis, and depression, as well as its long-term safety, remain uncertain (5). Moreover, the mechanisms regulating the production of DHEA(S), the basis of its age-dependent sexual dimorphism, and the biological significance of DHEA remain unclear.

Adrenal DHEA secretion is episodic and follows a diurnal rhythm similar to that of cortisol, under the stimulus of CRH and ACTH (6, 7). In contrast to the age-associated decline in adrenal androgens, basal and ACTH-stimulated concentrations of cortisol change minimally with aging, suggesting an ACTH-independent, age-related diminution in adrenal androgen secretion (8). Factors other than ACTH, such as estrogens, GH, IGF-I, and TGF-β, have been proposed to modulate DHEA production (9). Age-related declines in DHEA and DHEAS are thought to be related to a selective decline in zona reticularis (ZR) mass of the adrenal gland, reduced ZR expression and activity of 17α-hydroxylase and 17,20-lyase (P450c17), and attenuated responsiveness of ZR cells to ACTH (10). Aging is associated with diminished activities of the gonadal steroid and GH/IGF-I axes (11). However, the possible contributions of these hormonal changes to the parallel decline of DHEA production and the effects of sex steroid without or with GH administration on adrenal DHEA(S) secretion in healthy older individuals have not been reported previously. This analysis reports the sexually dimorphic nature of DHEA secretion in healthy older adults and the effects of 26 wk of administration of T, without or with recombinant human GH, on DHEA secretory dynamics in healthy older men.

Subjects and Methods

The study design and participants have been described previously (12). Of the total of 125 subjects completing the study, 68 men and 36 women underwent the overnight frequent sampling procedure as described below. They did not take any medications known to interfere with the gonadal or adrenal steroid or GH axes. None of the men had previously taken T replacement. All women were postmenopausal, and none had used exogenous estrogen or progestin for at least 3 months before evaluation. The study protocol was approved by the Institutional Review Board of the Johns Hopkins Bayview Medical Center. Written informed consent was obtained from each participant.

Study design

The study was conducted as a randomized, double-masked, double-dummy, placebo-controlled, non-crossover clinical trial that followed a 2 × 2 factorial design (factors: sex steroid and GH). Subjects received sex steroid alone, GH alone, sex steroid with GH, or placebo alone for 26 wk, as previously described (12). Recombinant human GH (Nutropin; Genentech, Inc., South San Francisco, CA) was administered in a dose of 20 μg/kg body weight, self-injected sc three times per week in the evening. Sex steroid was given as a 100 μg/d estradiol patch (Estraderm; Novartis Pharmaceuticals, New York, NY) plus 2.5 mg medroxyprogesterone acetate (Provera; Pharmacia & Upjohn, Kalamazoo, MI) for the first 10 d of each month. Testosterone was administered as an im injection of 100 mg testosterone enanthate (T) (Delatestryl Injection; Bio-Technology General, East Brunswick, NJ) once every 2 wk.

Study participants were admitted to the General Clinical Research Center of the Johns Hopkins Bayview Medical Center at approximately 0800 h (clock time) on d 1. Body weight and height were obtained, from which body mass index (BMI, kg/m²) was computed. Meals were served between 1200 and 1300 h and between 1700 and 1800 h. At 1900 h of d 1, an iv catheter was inserted into a forearm vein and kept open with heparinized (1000 U/liter) 0.9% sodium chloride. From 2000 to 0800 h, blood samples (2 ml) were collected at 20-min intervals, from which DHEA concentrations were measured. Subjects were encouraged to sleep beginning at 2300 h, and room lights were turned off from 2400 to 0700 h. After an overnight fast, at 0800 h on the morning of d 2, blood was collected for measurements of DHEAS, T, free T, SHBG, estradiol, LH, and IGF-I. All sera were stored at −80 C until assayed. Participants were discharged and readmitted at wk 26, at which time the aforementioned protocol was repeated.

Assays

Total T (TT) concentrations were measured in men and women, using a sensitive RIA (13) that has been validated against liquid chromatography tandem mass spectrometry, widely considered the gold standard (14). The sensitivity, defined as hormone concentration corresponding to the 90% B/B₀ point, was 0.008 nmol/liter (0.22 ng/dl). Intra- and interassay coefficients of variation (CV) were 13.2 and 8.2%, respectively. Free T (FT) concentrations were measured by a sensitive equilibrium dialysis method (15), optimized to measure low concentrations with precision and accuracy. The sensitivity of the FT assay is 0.6 pg/ml (2.0 pmol/liter), with intra- and interassay CV of 4.2 and 12.3%, respectively. SHBG was measured by coated-tube immunoradiometric assay from Diagnostic Systems Laboratories (DSL) (Webster, TX) with a sensitivity of 5 nmol/liter (ng/ml), with intra- and interassay CV of 6.7 and 5.0%, respectively. Serum concentrations of DHEAS were measured in duplicate by coated-tube RIA by Covance Laboratories (previously Hazleton Laboratories, Vienna, VA) using commercial kits (DSL). The DHEAS assay sensitivity was 141 nmol/liter (5.2 μg/dl), with intra- and inter-assay CV of 4.3 and 6.6%, respectively. Serum DHEA concentrations were measured by ELISA at the National Center for Complementary and Alternative Medicine (NCCAM) and at DSL using commercial (DSL) kits. A correction factor was applied to the samples measured at NCCAM to account for lot-to-lot variation. The sensitivity of the DHEA assay was 0.347 nmol/liter (0.1 ng/ml), with intra- and interassay CV of 10.7 and 5.7%, respectively. DHEA ELISA measurements correlated strongly (r² = 0.87) with values quantified by tandem gas chromatography-mass spectrometry. Serum LH was measured by immunoradio-metric assay using commercial kits (Nichols Institute Diagnostics, San Juan Capistrano, CA). Sensitivity and intra- and interassay CV of the LH assay were 1 mIU/ml and 3.2 and 4.5%, respectively. Serum IGF-I and E2 were measured as previously described (12).

Analysis of hormone secretion

Serum DHEA concentrations reflect the combination of an underlying secretory event and endogenous (subject-specific) metabolic clearance. Multiparameter deconvolution analysis was used, as it allows for simultaneous determination of duration, maximal secretory rate (amplitude), and temporal locations of all significant DHEA secretory bursts and endogenous hormone half-life (16). In addition, basal hormone secretion was determined by the deconvolution analysis. To assess the performance of deconvolution analysis in identifying DHEA bursts, we compared deconvolution-derived burst positions with the known (true-positive) burst positions with simulated data sets. The simulated DHEA concentrations closely mimic the pulsatile nature and measurement uncertainty characteristics of the observed physiological data. We classified our derived bursts as either true positives [derived burst position occurs at the same time as the known burst position time-half delta time (10 min), where δ is the width of the sampling interval] or false positives (inferred burst center removed from known position by more than one-half δ time). The true-positive, false-positive, and false-negative rates of the deconvolution were 91.4, 8.6, and 17.3%, respectively.

Approximate entropy (ApEn)

We calculated the ApEn of the individual subject's DHEA concentration-time series. The ApEn statistic measures the regularity or orderliness of hormone release, with a higher ApEn reflecting a more random or disordered pattern of hormone secretion (17). Three hundred Monte Carlo simulations were used for each series to estimate the sd of ApEn in each subject by perturbing each original DHEA concentration-time series with random experimental variation defined by the dose-dependent within-sample sd.

Statistical analysis

Although the results presented in this manuscript were not part of the analyses planned when the study was originally designed, they are a natural extension of the original protocol design and dovetail with our previously published data (12). Data were analyzed with the SAS statistical software package, version 9.1 (SAS Institute, Inc., Cary, NC). The distribution of the data was explored by visual inspection of Q-Q plots, stem and leaf plots, and box plots. Variables were log-transformed as required before any statistical analyses. Means, sd, and confidence intervals (CI) were calculated from the log-transformed data, as appropriate, and were subsequently back transformed (by taking the antilog of the values) to yield geometric means and corresponding 95% CI. Sex differences in baseline measures were examined by unpaired Student's t tests. Analysis of covariance (ANCOVA) was performed to adjust morning DHEAS concentrations and deconvolution-derived DHEA secretion parameters for age and BMI; sex differences were determined by the LSMEANS procedure. Relationships between variables were assessed by linear regression analyses using Pearson's correlation coefficient. To clarify the determinants of DHEA secretion and DHEAS concentrations, multiple regression analyses, adjusted for age, BMI, and morning LH, FT, and IGF-I levels, were performed, and standardized regression coefficients of multivariate-adjusted analyses are presented. Standardized regression coefficients are interpreted as the sd change in dependent variable for a change of one sd in the independent variable. Standardized regression coefficients compare the relative explanatory power of different independent variables; the largest coefficient indicating the independent variable with the greatest influence on the dependent variable. Only those relationships that are statistically significant are reported.

Differences between treatment groups at 26 wk in men were assessed by ANCOVA. These analyses were not reported in women because of modest sample size in the various treatment groups. The dependent variables in the ANCOVA were the changes (post – pre) in values of the outcome variable being studied. Independent variables included the subject's age, BMI, change in BMI, the initial value of the outcome variable, and two variables indicating treatment group (T, GH, T plus GH, or neither T nor GH). Each treatment group was compared with the placebo group; the method of Dunnett was used to adjust for multiple comparisons. A P value of <0.05 (two-tailed) was considered significant.

Results

Subject characteristics and baseline serum hormone concentrations

We studied the 104 healthy older men (n = 68) and women (n = 36) from the original study (12) who had undergone overnight frequent blood sampling before and after hormone or placebo intervention. Table 1 summarizes the mean values for age, BMI, and a single 0800-h sampling of LH, TT, FT, and IGF-I before hormone treatment. Mean age did not differ by sex, whereas BMI, TT, FT, and IGF-I concentrations were higher in men than in women. LH concentrations were higher in women. Within-sex group characteristics did not differ significantly.

TABLE 1.

Baseline age, BMI, and serum levels of LH, TT, FT, and IGF-I in healthy older men and women

	n	Age (yr)	BMI (kg/m2)	LH (IU/liter)	TT (nmol/liter)	TT (nmol/liter)	IGF-I (μg/liter)
Men
Placebo	15	70.7 ± 4.8	27.3 ± 1.7	3.0 (2.4–3.7)	14.7 (12.1–17.7)	155 (138–175)	144 ± 50
T	18	70.4 ± 2.9	26.8 ± 3.2	3.2 (2.1–4.8)	16.2 (14.6–17.9)	154 (138–174)	124 ± 46
GH	15	70.3 ± 4.8	27.2 ± 2.4	3.1 (2.2–4.3)	16.6 (14.7–18.6)	166 (139–197)	149 ± 49
GH + T	20	73.2 ± 6.3	27.0 ± 3.2	3.3 (2.4–4.5)	15.1 (12.8–17.8)	142 (125–162)	116 ± 41
All	68	71.2 ± 4.9	27.1 ± 2.7	3.1 (2.7–3.7)	15.6 (14.5–16.7)	153 (143–163)	131 ± 47
Women
Placebo	9	71.4 ± 4.7	25.5 ± 2.4	19.1 (14.7–24.6)	1.1 (0.7–1.6)	27 (22–33)	102 ± 34
E/P	10	71.9 ± 3.0	25.1 ± 1.9	17.9 (13.7–23.3)	1.0 (0.6–1.9)	29 (25–34)	111 ± 40
GH	9	70.5 ± 3.8	26.2 ± 3.4	18.8 (13.7–25.5)	1.2 (1.0–1.4)	30 (27–34)	110 ± 42
GH + E/P	8	71.5 ± 4.9	24.3 ± 3.7	20.3 (12.8–32.4)	0.9 (0.6–1.5)	25 (21–30)	134 ± 51
All	36	71.0 ± 4.0	25.3 ± 2.8	18.9 (16.5–21.7)	1.1 (0.9–1.3)	28 (25–29)	114 ± 42
P value		NS	0.002	0.0001	<0.0001	<0.0001	0.05

Data are presented as arithmetic mean ± SD (age, BMI, IGF-I) or as geometric means (95% CI) (LH, TT, FT); n, no. of subjects. P values indicate significance for comparisons between men and women. NS, Not significant.

Characteristics of DHEA secretion

Serum DHEAS concentrations were higher in men than in women (Table 2). Basal DHEA secretion rate, mass secreted per burst, total production rate, mean and integrated concentrations, and ApEn did not differ significantly by sex. However, DHEA burst frequency, amplitude, and pulsatile production rate were higher, and half-life, assessed by deconvolution analysis, was lower in women compared with men (Table 2). Although the pulsatile production rate was higher in women, the total production rate was similar in both sexes.

TABLE 2.

Baseline DHEAS, selected deconvolution, and overnight DHEA rhythm parameters in healthy older men and women

Variables	Men (n = 68)	Women (n = 36)	P value
DHEAS (0800 h, μmol/liter)	2.07 (1.83–2.32)	1.40 (1.17–1.68)	0.001
DHEA (deconvolution parameters)
Basal secretion rate (nmol·liter−1·min−1)	0.24 (0.19–0.30)	0.27 (0.19–0.37)	NS
Mass/burst (nmol·liter−1)	13.0 (10.9–15.5)	15.8 (12.3–20.0)	NS
Burst frequency (no./12 h)	6.26 ± 2.79	7.71 ± 2.79	0.01
Amplitude (nmol·liter−1·min−1)	1.0 (0.8–1.3)	1.6 (1.1–2.2)	0.03
Pulsatile production rate (nmol·liter−1·12 h−1)	76 (63–92)	107 (82–140)	0.04
Total production rate (nmol·liter−1·12 h−1)	261 (213–320)	317 (240–421)	NS
Mean (nmol/liter)	12.98 (11.38–14.80)	12.69 (10.58–15.23)	NS
Integrated (nmol·liter−1·min−1)	9452 (8240–10842)	9302 (7936–10902)	NS
Half-life (min)	32 ± 3.2	20 ± 1.5	0.02
ApEn	0.86 (0.82–0.89)	0.83 (0.78–0.89)	NS

Values are adjusted for age and BMI (ANCOVA) and presented as arithmetic mean ± sd (burst frequency, half-life) or geometric mean (95% CI) (DHEAS, DHEA basal secretion, mass/burst, amplitude, pulsatile and total production rates, mean, integrated, and ApEn); n, no. of subjects. P values indicate significance for comparisons between men and women. NS, Not significant.

Multiple regression analyses of DHEA deconvolution parameters vs. age, BMI, LH, FT, and IGF-I

Multiple regression analyses adjusted for age, BMI, and morning LH, FT, and IGF-I levels revealed that, in men, FT concentrations, and, in women, age, BMI, and FT were the only significant predictors of mean and integrated DHEA concentrations (Table 3). In women, BMI was inversely and FT directly related to DHEA secretory burst mass, BMI directly and IGF-I inversely related to ApEn, and FT and IGF-I directly related to DHEAS (0800 h).

TABLE 3.

Multiple linear regression analysis: standardized regression coefficients between baseline age, BMI, LH, FT, IGF-I, selected DHEA deconvolution parameters, and DHEAS

Dependent variables	Independent variables	Men		Women
		β	P value	β	P value
Log mean DHEA (nmol/liter)	Age	−0.07	ns	−0.27	0.04
	BMI	−0.02	ns	−0.32	0.04
	LogFT	0.43	0.001	0.73	0.0001
Log integrated DHEA (nmol·liter−1·min−1)	Age	−0.07	ns	−0.27	0.05
	BMI	−0.03	ns	−0.31	0.04
	LogFT	0.43	0.001	0.73	0.0001
Log DHEA mass/burst (nmol/liter)	BMI	ns	ns	−0.41	0.02
	LogFT	ns	ns	0.54	0.003
DHEA ApEn	BMI	ns	ns	0.46	0.012
	IGF-I	ns	ns	−0.42	0.009
Log DHEAS (0800 h, μmol/liter)	LogFT	ns	ns	0.55	0.008
	IGF-I	ns	ns	0.33	0.03

ns, Not significant.

Effects of T and/or GH on serum LH, T, E2, and IGF-I concentrations

The effects of sex steroids, without and with concomitant GH administration, on T and IGF-I levels (0800 h) in the total study population have been published previously (12). In this subset of men in the study cohort, T administration alone increased TT by 25% (P = 0.02) and decreased serum LH by 94% (P = 0.0001) and SHBG by 17% (P = 0.003) (Table 4). Similarly, GH plus T increased TT by 25% (P = 0.02) and decreased serum LH by 87% (P = 0.01) and SHBG by 16% (P = 0.007) (Table 4). As expected, administration of GH alone or GH plus T significantly increased IGF-I levels in men. Treatment with GH alone did not significantly change sex steroid levels, and administration of T alone did not alter IGF-I concentrations (Table 4).

TABLE 4.

Effects of hormone administration on serum T, SHBG, LH, and IGF-I in healthy older men

	Placebo	T	GH	T + GH
TT (nmol/liter)
Baseline	14.7 (12.1–17.7)	16.2 (14.6–17.9)	16.6 (14.7–18.6)	15.1 (12.8–17.8)
26 wk	14.9 (13.1–17.1)	19.6 (16.0–24.0)	15.3 (13.3–17.5)	19.2 (17.1–21.3)
Ratio2		1.28 (1.10–1.54)a	0.9 (0.81–1.19)	1.29 (1.04–1.55)a
FT (pmol/liter)
Baseline	155 (138–175)	154 (138–174)	166 (139–197)	142 (125–162)
26 wk	158 (138–181)	192 (154–239)	147 (132–166)	198 (174–221)
Ratio		1.21 (0.99–1.48)	0.89 (0.72–0.91)	1.27 (1.04–1.55)
SHBG (nmol/liter)
Baseline	96 (78–116)	91 (72–114)	102 (81–128)	93 (60–144)
26 wk	98 (77–128)	76 (62–96)	91 (71–117)	77 (51–117)
Ratio		0.82 (0.73–0.92)a	0.88 (0.78–0.99)	0.82 (0.74–0.93)a
LH (IU/liter)
Baseline	3.0 (2.4–3.7)	3.2 (2.1–4.8)	3.1 (2.2–4.3)	3.3 (2.4–4.5)
26 wk	2.4 (1.6–3.7)	0.2 (0.06–0.8)	3.1 (2.0–4.8)	0.3 (0.1–1.0)
Ratio		0.07 (0.02–0.23)a	1.14 (0.32–4.04)	0.16 (0.05–0.55)a
IGF-I (μg/liter)
Baseline	144 ± 50	124 ± 46	149 ± 49	116 ± 41
26 wk	121 ± 38	144 ± 35	266 ± 92	219 ± 53
Change1		31 (−5–68)	140 (102–177)a	119 (82–156)a

Baseline and wk 26 data are unadjusted values expressed as arithmetic mean ± sd (IGF-I) or geometric means (95% CI) (TT, FT, SHBG, LH); Δ, difference in post- and pretreatment values are adjusted for age, BMI, change in BMI, baseline value, and specific group. Differences in mean Δ between the treatment groups (T, GH, or GH + T) and placebo are expressed as changes (adjusted arithmetic mean, 95% CI)¹ or the ratio² (95% CI) of the adjusted geometric means of Δ.

^aP < 0.05 as compared to placebo (Dunnett's test).

Effects of T and/or GH on overnight DHEA secretion and morning DHEAS concentrations

In men, T or GH plus T administration significantly decreased integrated DHEA concentrations (Table 5). T decreased DHEA mass per burst, but this effect did not reach statistical significance (P = 0.07). In contrast, morning DHEAS concentrations, total and pulsatile nocturnal DHEA production rates, secretory burst frequency, basal secretion (time invariant) (Table 5), burst duration, half-life, and ApEn (data not shown) were not significantly altered in men receiving T or GH plus T, and GH alone did not significantly alter DHEAS concentrations or any DHEA secretory parameter.

TABLE 5.

DHEAS and selected DHEA deconvolution parameters at baseline and after 26 wk of hormone administration in healthy older men

DHEA (deconvolution parameters)	Placebo	T	GH	T + GH
Integrated (nmol·liter−1·min−1)
Baseline	9.79 (6.67–14.40)	11.33 (8.30–15.45)	8.39 (6.62–10.64)	7.92 (6.10–10.27)
26 wk	10.02 (7.09–14.15)	7.85 (5.72–10.76)	7.59 (5.86–9.28)	4.99 (3.67–6.77)
Ratio2		0.69 (0.59–0.81)a	0.88 (0.74–1.04)	0.62 (0.53–0.74)a
Total production rate (nmol·liter−1·12 h−1)
Baseline	327 (225–475)	334 (196–570)	228 (136–380)	227 (149–346)
26 wk	215 (117–395)	170 (116–249)	162 (91–290)	135 (78–232)
Ratio		0.72 (0.35–1.50)	0.79 (0.37–1.68)	0.67 (0.31–1.41)
Pulsatile production rate (nmol·liter−1·12 h−1)
Baseline	103 (65–164)	85 (54–134)	59 (35–97)	70 (50–99)
26 wk	69 (38–126)	55 (34–87)	47 (25–91)	48 (26–88)
Ratio		0.66 (0.31–1.38)	0.72 (0.32–1.56)	0.71 (0.33–1.52)
Amplitude (nmol·liter−1·min−1)
Baseline	1.38 (0.94–2.02)	0.99 (0.70–1.38)	0.81 (0.37–1.79)	1.04 (0.66–1.64)
26 wk	1.41 (0.75–2.64)	0.97 (0.69–1.35)	0.70 (0.39–1.27)	0.66 (0.36–1.22)
Ratio		0.71 (0.35–1.42)	0.57 (0.28–1.19)	0.51 (0.25–1.05)
Mass/burst (nmol/liter)
Baseline	15.01 (9.79–23.01)	15.14 (10.24–22.39)	11.32 (7.77–16.49)	11.83 (8.93–15.67)
26 wk	13.69 (8.59–21.82)	8.78 (6.54–11.79)	8.07 (4.88–13.32)	8.63 (5.40–13.77)
Ratio		0.55 (0.32–0.94)b	0.61 (0.35–1.07)	0.69 (0.40–1.20)
Burst frequency (number/12 h)
Baseline	7.07 ± 1.59	6.22 ± 3.11	5.80 ± 2.36	6.30 ± 2.10
26 wk	5.86 ± 2.64	6.83 ± 3.11	6.46 ± 2.79	6.15 ± 2.71
Change1		0.54 (−1.54–2.62)	0.57 (−1.61–2.76)	−0.29 (−2.38–1.79)
Basal secretion rate (nmol·liter−1·min−1)
Baseline	0.30 (0.20–0.44)	0.32 (0.17–0.57)	0.22 (0.12–0.39)	0.21 (0.13–0.34)
26 wk	0.21 (0.11–0.43)	0.15 (0.09–0.22)	0.15 (0.08–0.27)	0.12 (0.07–0.20)
Ratio		0.61 (0.28–1.34)	0.73 (0.32–1.62)	0.61 (0.27–1.36)
DHEAS (0800 h, μmol/liter)
Baseline	1.89 (1.29–2.77)	2.46 (1.96–3.09)	2.04 (1.55–2.67)	1.86 (1.47–2.37)
26 wk	2.02 (1.31–3.12)	2.02 (1.67–2.44)	2.22 (1.70–2.89)	1.83 (1.44–2.33)
Ratio		0.84 (0.69–1.01)	1.07 (0.88–1.30)	0.91 (0.75–1.10)

Baseline and wk 26 data are unadjusted values expressed as arithmetic mean ± SD (IGF-I) or geometric means (95% CI) (TT, FT, SHBG, LH); Δ, difference in post- and pretreatment values are adjusted for age, BMI, change in BMI, baseline value, and specific group. Differences in mean Δ between the treatment groups (T, GH, or GH + T) and placebo are expressed as changes (adjusted arithmetic mean, 95% CI)¹ or the ratio² (95% CI) of the adjusted geometric means of Δ.

^aP < 0.0001 as compared to placebo (Dunnett's test).

^bP = 0.07 as compared to placebo (Dunnett's test).

Discussion

The current study, to our knowledge, is the first to report the sexually dimorphic nature of spontaneous DHEA secretion in healthy older individuals and the effects of administration of T and GH, alone and in combination, on DHEA secretion in healthy older men. At baseline, in subjects without hormone treatment, we identified gender disparities in nocturnal DHEA pulse frequency, half-life, and pulsatile production rate and sex-dependent associations of specific features of DHEA secretion with morning concentrations of FT. In addition, in men, after administration of T alone or T combined with GH, but not after GH alone, we observed significantly decreased integrated DHEA concentrations without change in nocturnal DHEA pulsatility, pattern regularity, or half-life.

In this study, older women exhibited lower morning DHEAS concentrations than did men, in conjunction with higher nocturnal DHEA pulse frequencies and pulsatile production rates and a shorter DHEA half-life. The seeming contradiction of women exhibiting lower DHEAS concentrations and higher DHEA production rates is explained, at least in part, by the more rapid DHEA clearance in women. In contrast, DHEA total production rates and mean and integrated DHEA concentrations did not differ by sex. Few studies have objectively quantified ultradian patterns of DHEA secretion. In one report, baseline 24-h mean DHEA concentrations and DHEA pulse amplitude were lower in postmenopausal vs. premenopausal women (7). We are unaware of similar studies in men or detailed comparisons of ultradian DHEA secretion in healthy older men and women. We found that mean nocturnal DHEA concentrations were similar in older men and women, as were values for integrated DHEA. Consistent with this, in another study, 24-h mean plasma DHEA concentrations were similar in women and men 50 yr of age or older (18). Pharmacokinetic studies have reported the circulating half-life of DHEA to vary from 17–44 min (19, 20). Assuming single-elimination kinetics for DHEA, the current deconvolution analysis yielded DHEA half-life estimates well within the published range. The finding of a shorter circulating DHEA half-life in women suggests a gender-related difference in the elimination kinetics of endogenous DHEA in older individuals, which requires confirmation by isotopic infusion techniques. The mechanism(s) underlying this putative difference in DHEA half-life in older men and women is (are) unknown.

In our study, FT was positively associated with nocturnal mean and integrated DHEA concentrations in men and women. As was evident from the multiple regression analyses, this relationship was considerably stronger in women, in agreement with other studies in which FT was directly related to DHEAS concentrations in women but not men (21). This finding may reflect the substantial peripheral plasma conversion of adrenal DHEA, DHEAS, and concurrently released androstenedione to T in older women. In contrast, the contributions of the aforementioned steroids to plasma T concentrations are negligible in older men (<1%). As in other studies (22), we found that DHEA but not DHEAS concentrations were negatively related to BMI. Of note, we found that DHEA ApEn was positively related to BMI but negatively related to serum IGF-I concentrations in older women. This suggests that increased fatness in older women is associated with more irregular DHEA production. ApEn reflects the degree of coordination and stability in integrated network systems. Although the feedback and feed-forward pathways regulating DHEA secretion are not completely characterized, bidirectional interactions between the GH/IGF-I and hypothalamo-pituitary-adrenal axes are well recognized. Similarly, the magnitude of adiposity is known to modulate other endocrine axes such as the somatotropic, corticotropic, and insulin axes. Additionally, higher BMI has been associated with irregular ACTH release and with lower IGF-I levels (23, 24); ACTH and IGF-I modulate DHEA secretion, and DHEA supplementation up-regulates the GH/IGF-I axis (25, 26). Thus, we speculate that in older women, adiposity and lower IGF-I levels modulate the feed-forward pathways regulating DHEA secretion. The clinical significance, if any, and the basis for the gender-specific nature of these findings remain to be elucidated.

Gonadal steroids can modulate adrenal androgen production in humans, suggesting that they may underlie the observed sexual dimorphism in adrenal androgen production. In gonadectomized men with advanced prostate cancer, androgen receptor blockade for 2 months significantly decreased basal and ACTH-stimulated serum DHEA and DHEAS concentrations (27). Similarly, in postmenopausal women, flutamide decreased serum DHEAS concentrations by 50% (28). Conversely, T administration in women increases ACTH-stimulated DHEA release (29). These studies suggest that androgens can increase DHEA production, as do in vitro studies demonstrating a direct stimulatory effect of T on human adrenal DHEA production (30).

In contrast, T administration to ovariectomized healthy women for 3 wk decreased circulating DHEA concentrations (31), and T administration to normal men for 16 wk suppressed plasma DHEAS and endogenous testicular androgens (32). The above noted, apparently conflicting effects of gonadal steroids on adrenal androgens may result from varying study designs, age and comorbidity of study subjects, duration and dosage of T administration, and frequency of sampling for DHEA or DHEAS measurements. We are unaware of other studies examining the effects of exogenous T on DHEA secretion in healthy older men. In the present study, T administration alone or in combination with GH significantly decreased nocturnal integrated DHEA concentrations but did not significantly affect other measured indices of nocturnal DHEA secretion. T administration tended to suppress DHEA secretory burst mass, which may have contributed to the lower total production rate. There were no additive effects of T plus GH coadministration, compared with T alone, on DHEA secretion.

Few studies have examined the regulation of DHEA secretion in humans and the mechanisms underlying sex steroid regulation of adrenal DHEA(S) production. This study was not designed to explore mechanisms underlying the potential effects of sex steroid and/or GH treatment on DHEA secretion. Nevertheless, possible mechanisms by which T can decrease DHEA concentrations include direct or indirect suppression of adrenal and/or testicular DHEA production, altered catabolism and peripheral conversion of DHEA to T, and change in the elimination kinetics of DHEA.

Our finding of reduced DHEA burst mass, with unchanged burst frequency, basal secretion, or clearance suggests that a decline in steroidogenesis partly explains the suppressive effects of T on DHEA secretion. Androgen receptors have been demonstrated in human adrenal glands and the inhibitory effects of T could be androgen receptor mediated (33). A high expression of 17α-hydroxylase, 17,20-lyase, cytochrome P450 (P450c17), cytochrome_b5 (cyt_b5) and DHEA-sulfotransferase (SULT2A1) combined with low expression of competing 3β-hydroxysteroid dehydrogenase is necessary to synthesize DHEA and DHEAS from pregnenolone and is characteristic of ZR of the adult adrenal gland. T administration has been reported to decrease aldosterone and CRH-stimulated cortisol production in men (34, 35). However, T administration did not affect basal production of cortisol in men in the current study (data not shown) or others (35). Thus, T may differentially modulate steroid production in the adrenals. In regard to DHEA secretion, T may inhibit 17α-hydroxylase/17,20-lyase activity in addition to modulating 3β-hydroxysteroid dehydrogenase activity to decrease DHEA synthesis. T is known to influence multiple components of the hypothalamo-pituitary-adrenal axis. T-mediated inhibition of ACTH secretion and/or action could be one explanation for reduced DHEA production. However, T administration in men has been reported to augment, rather than reduce, CRH-stimulated ACTH levels without modulating basal cortisol levels (35). T may inhibit testicular DHEA production either directly or by suppression of LH. However, plasma DHEA and DHEAS concentrations are determined almost exclusively by the adrenals, with minor contribution from the gonads (~10%). Thus, T effects in this study are most likely a result of its suppressive effects on adrenal DHEA production.

In vivo and in vitro studies suggest that GH and IGF-I can influence adrenal steroidogenesis (36–41). DHEAS concentrations are lower in GH-deficient patients; however, GH replacement does not consistently increase DHEAS concentrations in these patients (37, 39). Most of these studies were performed in children, and data in adult patients are limited. In adult hypopituitary patients, serum DHEAS increased after 6 months of GH treatment only in patients with an adequate ACTH reserve (37). These data suggest that the GH effect on adrenal steroid production is partly ACTH dependent but that GH/IGF-I may also exert direct adrenal effects. In our study, GH administration alone did not significantly alter DHEAS concentrations or DHEA secretory parameters. Although we did not measure ACTH concentrations, we have no reason to suspect impaired ACTH reserve in these men. Taken together, these data suggest that GH and/or IGF-I levels do not modulate DHEA secretion in older men and that reduced GH/IGF-I activity may not be the proximate cause for diminished DHEA secretion associated with aging.

This study has several limitations. The number of participants, particularly women, was relatively small, and women were not treated with T, making it difficult to determine definitively whether there is gender dimorphism in T-modulated nocturnal DHEA secretion and morning DHEAS concentrations in older individuals. The doses and duration of hormone intervention may have been inadequate to evaluate effects on DHEA and DHEAS fully. Finally, we did not assess the clinical or physiological significance of T-modulated changes in DHEA. Although age-associated decline in DHEA is well established, the biological role of DHEA in humans is incompletely understood. In that context, the clinical significance of T-induced reduction in DHEA concentrations is unclear. Taken together, the current data demonstrate that spontaneous nocturnal DHEA secretion is sexually dimorphic in healthy older individuals and that testosterone decreases nocturnal DHEA secretion in men. These results provide additional insights into the regulation of DHEA secretion in older individuals.

Abbreviations

ANCOVA

Analysis of covariance

ApEn

approximate entropy

BMI

body mass index

CI

confidence interval

CV

coefficient of variation

DHEA

dehydroepiandrosterone

DHEAS

dehydroepiandrosterone sulfate

FT

free testosterone

T

testosterone

TT

total testosterone

ZR

zona reticularis.