Changes in Bone Turnover Markers and Menstrual Function After Short-term Oral DHEA in Young Women with Anorexia Nervosa


  • Catherine M. Gordon M.D.,

    Corresponding author
    1. Division of Adolescent/Young Adult Medicine, Children's Hospital, Harvard Medical School, Boston, Massachusetts
    2. Division of Endocrinology, Children's, Hospital, Department of Pediatrics, Harvard Medical School, Boston, Massachusetts
    • Children's Hospital 300 Longwood Avenue Boston, MA 02115 U.S.A.
    Search for more papers by this author
  • Estherann Grace,

    1. Division of Adolescent/Young Adult Medicine, Children's Hospital, Harvard Medical School, Boston, Massachusetts
    Search for more papers by this author
  • S. Jean Emans,

    1. Division of Adolescent/Young Adult Medicine, Children's Hospital, Harvard Medical School, Boston, Massachusetts
    Search for more papers by this author
  • Elizabeth Goodman,

    1. Division of Adolescent/Young Adult Medicine, Children's Hospital, Harvard Medical School, Boston, Massachusetts
    Search for more papers by this author
  • Margaret H. Crawford,

    1. Division of Adolescent/Young Adult Medicine, Children's Hospital, Harvard Medical School, Boston, Massachusetts
    Search for more papers by this author
  • Meryl S. Leboff

    1. Division of Endocrine/Hypertension, Brigham and Women's Hospital, Department of Medicine, Harvard Medical School, Boston, Massachusetts
    Search for more papers by this author


Bone loss is a serious consequence of anorexia nervosa (AN). Subnormal levels of serum dehydroepiandrosterone (DHEA) are seen in patients with AN and may be causally linked to their low bone density. We hypothesized that oral DHEA would decrease markers of bone resorption (urinary N-telopeptides [NTx]), and increase markers of bone formation (serum bone-specific alkaline phosphatase and osteocalcin [OC]). Fifteen young women (age 15–22 years) with AN were enrolled in a 3-month, randomized, double-blinded trial of 50, 100, or 200 mg of daily micronized DHEA. Blood and urinary levels of adrenal and gonadal steroids and bone turnover markers were measured. No adverse clinical side effects of DHEA were noted, and a 50 mg daily dose restored physiologic hormonal levels. At 3 months, NTx levels had decreased significantly in both the 50 mg (p = 0.018) and the 200 mg (p = 0.016) subgroups. OC levels simultaneously increased within treatment groups over time (p = 0.002). Eight out of 15 (53%) subjects had at least one menstrual cycle while on therapy. Short-term DHEA was well-tolerated and appears to normalize bone turnover in young women with AN. Resumption of menses in over half of subjects suggests that DHEA therapy may also lead to estradiol levels sufficient to stimulate the endometrium in this group of patients.


A serious consequence of anorexia nervosa (AN) is the compromise of bone density. Patients with chronic AN have a 7-fold increased incidence of fractures and often develop early osteoporosis.(1) Poor nutrition deprives these young women of calcium and other macronutrients that strengthen bone.(2) While physical activity is important for maintaining bone mass, patients with AN often exercise excessively, leading to weight loss with subsequent hormonal alterations that predispose these patients to skeletal losses. (3-8) Because adolescence is a critical period for the acquisition of bone mineral,(9,10) the identification of safe and effective strategies to preserve bone density in these young women is an important public health issue.

Subnormal levels of the adrenal steroid dehydroepiandrosterone (DHEA) have been observed in patients with AN. Adrenocorticotropic hormone stimulation tests have suggested decreased adrenal 17–20 lyase activity in AN, with a predominance of glucocorticoid over androgenic pathways.(6,7) This enzymatic block results in increased cortisol and decreased DHEA production.(6,7) In adolescents with AN, DHEA and gonadal steroid levels revert back to the range of a prepubertal child. Subnormal androgen levels, paradoxically accompanied by normal to increased levels of cortisol, have been speculated to represent a regression of hormonal function, with levels normalizing after weight restoration.(6) Normally, the secretion of DHEA rises sharply during adolescence, when bone mass is increasing, and reaches its peak during the third decade. DHEA levels subsequently decline with advancing age.(11-14) In patients with AN, declines in both DHEA and insulin-like growth factor I (IGF-I) can occur.(6-8) In some adult studies, DHEA levels are positively correlated with bone mineral density (BMD), suggesting that DHEA may play an important role in bone accretion and the prevention of bone loss associated with low DHEA states (e.g., AN and aging).(15,16)

Although many clinicians use estrogen replacement therapy (ERT) for adolescents with AN, results of this treatment have been conflicting.(17,18) A recent short-term study of combined androgen and estrogen replacement in postmenopausal patients noted that although ERT inhibits bone resorption, it also decreases bone formation, while androgen replacement stimulates bone formation.(19) As DHEA is converted into both estrogens and androgens, inhibiting bone resorption and stimulating bone formation, respectively, restoration of DHEA levels may counteract several of the factors that contribute to bone loss in this population. There are no prior data on the effect of DHEA repletion in patients with AN.

In the present study, we examined the effects of short-term supplementation of DHEA on levels of bone turnover, estradiol (E2), androgens, and IGF-I in young patients with AN. Because DHEA is metabolized into estrogens and androgens, we hypothesized that supplemental DHEA would increase serum levels to a physiologic range in these patients, simultaneously increasing markers of bone formation (serum osteocalcin [OC] and bone-specific alkaline phosphatase [BAP]) and decreasing markers of bone resorption [cross-linked N-telopeptide, NTx]).



Fifteen young Caucasian women aged 15–22 years (mean age 17.3 ± 2.7 years) with AN by Diagnostic and Statistical Manual of Mental Disorders, Revised Fourth Edition criteria participated in the study. The subjects were recruited from the Eating Disorders Program at Children's Hospital, Boston and a local suburban adolescent medicine practice. All patients were hemodynamically stable, free of any acute or other chronic disease, and were taking no anticonvulsants, glucocorticoids, or sex steroids, medications known to affect BMD. All patients gave informed consent according to the guidelines of the Committee for Clinical Investigation at Children's Hospital, Boston.

Study design, treatment, and measurements

Subjects were randomized to receive a total daily dose of micronized DHEA of either 50, 100, or 200 mg. DHEA was taken twice daily. Micronized DHEA was used because increased lymphatic absorption has been documented with this preparation, thereby decreasing hepatic first-pass effects.(13) The doses of DHEA tested were intended to approximate the high levels of this steroid that are found during adolescence, doses higher than would be generally recommended for replacement therapy in adults. At baseline and 1, 2, and 3 months, subjects had venous blood drawn from the antecubital vein and a second morning urine collected in the outpatient division of the General Clinical Research Center, Children's Hospital, Boston. Samples were obtained between 7:00 a.m. and 10:00 a.m. after an overnight fast. At baseline and at 3 months, BMD (total body, femoral neck, and lumbar spine) and body composition were measured by dual-energy X-ray absorptiometry (DXA). Nutritional and activity questionnaires, including a detailed assessment of calcium intake (both dietary and supplemental calcium), were completed at baseline and 3 months. (20) Psychological questionnaires, including the Beck Depression Inventory, Speilberger State Inventory (an anxiety assessment), and the Eating Attitudes Test (a tool for evaluating body image and anorexic behavior),(21-24) were completed at baseline and at 3 months. Compliance was assessed using monthly interviews, pill counts, and serum DHEA levels.

At monthly visits, subjects' weight in kilograms and height in centimeters were determined in a hospital gown after voiding. All weights were obtained on the same scale at each visit. Height was obtained using the same stadiometer (Perspective Enterprises, Kalamazoo, MI, U.S.A.). Body mass index was calculated from these measurements in kilograms per square meter. Percentage of ideal body weight (%IBW) was estimated using standard percentile tables from the National Center for Health Statistics.(25)

Each participant had a BMD measurement of the total body, lumbar spine, and femoral neck at baseline and at 3 months by DXA with a Hologic 2000 machine (Hologic, Inc., Waltham, MA, U.S.A.). Body composition by DXA was also obtained at these time points. With this instrument, the precision error (CV% ± SEM) for BMD of the spine was 0.53 ± 0.075% and 0.77 ± 0.14% at the femoral neck for premenopausal females.(26) Bone density of the spine was compared with age- and gender-matched controls. BMD of the hip was extrapolated from adult normal bone density data (age 20 years).

At monthly intervals, subjects had venous blood samples obtained for OC, DHEA, and dehydroepiandrosterone-sulfate (DHEAS) measurements by double-antibody radioimmunoassay (RIA), and BAP levels by immunoradiometric assay. At monthly intervals, urinary levels of NTx were also determined using enzyme-linked immunosorbent assay on a second morning void. At baseline, levels of follicle-stimulating hormone, luteinizing hormone, thyroxine, thyroid-stimulating hormone, and prolactin were also measured by RIA. At baseline, 1 month, and 3 months, calcium, phosphorus, liver function tests, total cholesterol, and high-density lipoprotein (HDL) levels were measured by Ektachem methodology (cholesterol oxidase). At baseline and 3 months, parathyroid hormone, total estrogen and testosterone, sex hormone binding globulin (SHBG), IGF-I, and a fasting glucose and insulin were analyzed by RIA, and levels of free and total testosterone were determined by dialysis at Endocrine Sciences (Calabasas Hills, CA, U.S.A.). Blood glucose was determined by glucose hexokinase methodology (Boehringer-Mannheim, Indianapolis, IN, U.S.A.). Serum insulin levels were determined by a microparticle enzyme immunoassay (Abbott Laboratories, Abbott Park, IL, U.S.A.).

Statistical analysis

Data were evaluated statistically with paired or unpaired Student's t-tests in the case of a normal distribution. In the case of a skewed distribution, Wilcoxon signed-rank or rank-sum tests were used, as appropriate. A one-way analysis of variance (ANOVA) was used to compare baseline means (bone marker and hormonal levels, and demographic variables) among the three treatment groups. A two-way repeated measures ANOVA was used to evaluate changes in BAP, OC, and NTx levels within and among the three treatment groups over time. To examine relationships among continuous variables, a Spearman's correlation analysis was performed. Data are presented as mean ± SD with two-tailed significance levels reported. Bone marker and hormonal and demographic variables were not significantly different among groups at baseline, and a dose-response relationship was not seen in response to therapy. Therefore, data are presented as a pooled sample for baseline means and correlation analyses, and as individual dosage subgroups for selected variables. Statistical analyses were performed using SPSS software (SPSS, Inc., Chicago, IL, U.S.A.). A p-value < 0.05 was considered statistically significant. Given the small sample size, all p values ≤ 0.10 are reported.


Demographic characteristics

Baseline demographic variables, presented in Table 1, were not significantly different across treatment groups. The median duration of AN at study entry was 20 months, with a range of 3–99 months. None of the demographic variables changed significantly over the course of the study, including calcium intake and duration of weekly exercise. All subjects had Tanner stage 5 breast and pubic hair development. Six patients (40%) had a family history of osteoporosis and one patient had previously been on ERT (more than 1 year prior to the study).

Table Table 1.. Demographic Characteristics of All Subjects at Baseline (Means ± SD)
original image

Hormonal levels

Baseline and 3-month hormone levels for the 50, 100, and 200 mg subgroups are presented in Table 2. There were no mean baseline differences in hormonal levels among groups.

Table Table 2.. Hormonal Parameters of Individual Subgroups of DHEA Treatment: Baseline and 3 Months
original image

For the total sample (n = 15), the mean baseline DHEA level was 359.7 ± 204.5 ng/dl and the median 290 ng/dl (range of 92–849 ng/dl). Nine of the 15 (60%) subjects were below the normal physiologic range of 215–850 ng/dl. The mean baseline DHEAS level for the sample was 170.6 ± 70.5 μg/dl (range 77–289 μg/dl), with a normal range of 183–283 μg/dl. DHEA levels increased at each time point following therapy, with significant increases from baseline to 1 and 2 months in all dosage subgroups. Increases in DHEA and DHEAS from baseline to 3 months were directly correlated with increases in testosterone over the same interval (r = 0.62, p = 0.016 and r = 0.66, p = 0.011, respectively).

The baseline E2 level was 25.1 ± 12.1 pg/ml for the entire sample (range 10–44 pg/ml), with a normal range of 30–300 pg/ml. Mean levels within treatment subgroups did not change significantly over the course of the study. Increases in E2 over the study were moderately correlated with increases in body fat as measured by DXA (r = 0.55, p = 0.050).

The mean baseline total testosterone for the 15 subjects was 20.4 ± 10.0 ng/dl (range 5.2–41.0 ng/dl) with a normal range of 10–55 ng/dl. Although no adverse clinical side effects were noted, total testosterone levels in the 100 and 200 mg groups at 3 months approached or were in a supraphysiologic range. As is seen in Table 2, total testosterone levels for the 50 mg subgroup remained in a physiologic range, despite a significant increase over the 3 months.

The mean baseline free testosterone for the total sample was 1.7 ± 1.2 pg/ml (range 0.5–4.6 pg/ml), with a normal range of 1.1–6.3 pg/ml. As is seen in Table 2, final free testosterone levels in the 100 mg and 200 mg subgroups were elevated, while those in the 50 mg subgroup were normal.

The mean baseline SHBG level for the entire sample was 1.6 ± 0.6 ng/dl (range 0.8–3.8 ng/dl, normal range of 1.0–3.0 ng/dl) with levels significantly decreased after 3 months of DHEA (p = 0.008). However, subgroup analyses revealed no statistically significant changes after 3 months of therapy within any of the treatment groups.

The mean fasting baseline IGF-I level for the entire sample was 272.9 ± 101.6 ng/ml (range 166–566 ng/ml) with a normal range of 240–660 ng/ml. IGF-I levels did not change significantly over the 3-month study period within any of the treatment groups. However, IGF-I levels increased after 3 months of DHEA in 9 out of 15 (60%) subjects. Within this subgroup, increases in IGF-I correlated with increases in BAP (r = 0.67, p = 0.049). The mean 3-month IGF-I level for the total sample was also strongly correlated with the final mean E2 level (r = 0.65, p = 0.011).

The baseline luteinizing hormone level for the entire sample was 2.1 ± 2.1 mIU/ml (range of 0.15–7.0 mIU/ml) with a normal range of 0.4–11.7 mIU/ml. The mean baseline follicle-stimulating hormone level was 4.28 ± 2.16 (range of 0.24–7.04 mIU/ml) with a normal range of 1.0–9.2 mIU/ml.

The mean baseline insulin level was 4.9 ± 2.1 μU/ml, accompanied by a mean glucose of 76.2 ± 8.7 mg/dl before DHEA. After therapy, levels of both insulin and glucose did not change significantly within any of the treatment groups.

Bone resorption markers

Urinary NTx levels at baseline and 3 months for the 50, 100, and 200 mg subgroups are shown in Table 2. There were no differences in mean NTx levels at baseline across the three treatment groups.

The mean urinary NTx level was increased for the 15 subjects at baseline: 108.8 ± 47.9 nmol/mmol creatinine (normal 10–65). As is shown in Fig. 1, significant decrease in NTx levels was seen in the 50 (p = 0.018) and 200 mg (p = 0.016) dosage subgroups. At baseline, levels of NTx were inversely correlated with levels of DHEAS (r = −0.74, p = 0.002). There was also a strong inverse correlation between 1-month levels of both DHEA and DHEAS and the 1-month NTx levels (r = −0.60, p = 0.023 and r = −0.92, p = 0.001, respectively); a similarly strong inverse correlation was seen at 2 months for the same variables (r = −0.56, p = 0.031 and r = −0.78, p = 0.001, respectively). Increases in DHEAS from baseline to 3 months also weakly correlated with decreases in NTx over the same interval (r = −0.45, p = 0.09).

Figure FIG. 1..

Urinary NTx levels among treatment subgroups. Baseline and 3-month urinary NTx levels are depicted for the three DHEA subgroups: 50, 100, and 200 mg dosage groups. A significant decrease was seen comparing baseline to 3-month levels for the sample using Student's t-tests for paired data (n = 15, p = 0.002).

Bone formation markers

Baseline and 3-month BAP and OC levels for the three dosage subgroups are shown in Table 2. OC levels over the course of the study are shown in Fig. 2. There were no differences in levels of bone formation markers at baseline across treatment groups.

Figure FIG. 2..

Serum OC levels among treatment subgroups. Serum OC levels are shown at baseline, and after 1, 2, and 3 months of DHEA therapy for each of the three dosage groups: (A) 50 mg, (B) 100 mg, and (C) 200 mg. Over the 3-month study period, significant changes were noted within groups (p = 0.002) by repeated measures ANOVA. Significant changes from baseline are noted with an asterisk.

The mean baseline OC level was 2.2 ± 1.3 ng/ml for the entire sample, low normal compared with an adult normal range of 2–24 ng/ml, and subnormal compared with a normal range for adolescents of 42–225 ng/ml. As is shown in Fig. 2, OC levels increased significantly over time within all subgroups, particularly at the 1-month time point (p = 0.002). Increases in OC were strongly correlated with increases in BAP (r = 0.70, p = 0.004). The 3-month OC level was also strongly correlated with the final IGF-I level (r = 0.63, p = 0.012), and the final E2 level (r = 0.67, p = 0.009).

The mean baseline BAP level for the sample was normal at 17.1 ± 8.8 ng/ml compared with a normal young adult range of 2–22 ng/ml, but was subnormal against expected levels of 28–38 ng/ml reported for healthy 15- to 18-year-old females.(27) There were no statistically significant changes in BAP over time or within subgroups. BAP increased in 9 out of 15 (60%) subjects from baseline to 3 months. Within this subgroup, the increase in BAP after DHEA was strongly correlated with both an increase in IGF-I (r = 0.74, p = 0.034) and an increase in E2 levels (r = 0.77, p = 0.010). Four of the patients whose BAP levels increased were in the 50-mg treatment group.

Lipid panels

There was no significant difference between HDL levels at baseline and at 3 months; the mean HDL level was 48.6 ± 14.0 mg/dl before DHEA and did not change significantly in any subgroup after therapy. There was also no significant difference in total cholesterol with a level of 158.1 ± 41.9 mg/dl at baseline without significant changes at 3 months.

Bone density and body composition measurements

The mean baseline lumbar BMD for all subjects was 0.92 ± 0.12 g/cm2 (range 0.58–1.024 g/cm2). Expressed as a Z score, the mean bone density was −0.73 ± 1.14 SD (range −4.02 to 0.64 SD) below that expected for age and gender. The mean whole body BMD was 1.1 ± 0.6 g/cm2 (range 0.80–1.8 g/cm2) with a Z score of −1.3 ± 0.8 SD (range −3.8 to −0.75 SD). The mean BMD of the femoral neck was 0.89 ± 0.10 g/cm2 (range 0.54–0.99 g/cm2) with a Z score of −0.71 ± 0.87 SD (range −3.5 to 0.19 SD). The mean baseline percentage body fat by DXA was 15.6 ± 6.9% (range 5.7–24.0%). There were no significant changes in either mean BMD at any site or body composition after 3 months of DHEA. Baseline BMD (g/cm2) was inversely correlated with both duration of amenorrhea (r = − 0.52, p = 0.046) and duration of AN (r = −0.57, p = 0.028). Percentage body fat was inversely correlated with the amount of weekly exercise at baseline (r = −0.57, p = 0.027).

Psychological parameters

There were no significant changes in psychological measures over the study. The baseline mean Eating Attitudes Test score was elevated for the sample at baseline (48.3 ± 23.4 [normal < 15]) and did not change significantly at 3 months in any group. The baseline mean Beck Depression Inventory score was also increased at baseline at 18.5 ± 13.0 (normal < 9) and was unchanged at 3 months. Lastly, the Speilberger State Inventory score was 49.1 ± 11.8 (range of test, 20–80) at baseline and did not change significantly after DHEA.

Menstrual function

Eight of the 15 subjects (53%) reported having at least one menstrual period while on DHEA. Each episode of vaginal bleeding lasted a minimum of 3 days. This finding was unexpected because weight gain was significant in only 2 out of 8 (25%) subjects in this subgroup, with weight loss also seen in 2 out of 8 (25%). As is shown in Table 3, there was no significant change in weight, percentage body fat by DXA or %IBW within this subgroup after 3 months of DHEA. Compared with subjects who remained amenorrheic, there were no significant differences either at baseline or 3 months in these same variables, or duration of AN or amenorrhea. Significant increases in testosterone were seen in this subgroup comparing baseline to 3-month levels. Also shown in Table 3 are significant changes in bone turnover markers from a post hoc analysis of those subjects whose menses returned. This subgroup also had significant increases in BAP at 3 months (p = 0.027) compared with the amenorrheic group. Three-month E2 levels were significantly increased in this subgroup, compared with those subjects who remained amenorrheic (mean 51.4 ± 43.6 vs. 17.7 ± 6.4 pg/ml, respectively; p = 0.010), although the baseline E2 levels were not significantly different (29.7 ± 13.5 vs. 20.6 ± 9.2 pg/ml, p = NS). Within 3 months after discontinuing DHEA, seven of the above eight patients showed cessation of menses.

Table Table 3.. Clinical Characteristics of Subjects Who Resumed Menses
original image

Adverse effects

One patient in the 200-mg subgroup had an elevation of transaminases after 1 month of DHEA, in association with mild upper respiratory symptoms. Without stopping therapy, laboratory values decreased within 48 h and were back within the normal range within 2 weeks. This episode was attributed to a viral illness rather than the DHEA, since parameters rapidly normalized despite continuation of therapy. No other abnormalities in other biochemical parameters and no signs of hirsutism or acne were noted.


The current study documents changes in several biochemical and clinical parameters after DHEA therapy in a cohort of adolescents and young adults with AN. To our knowledge, this is the first report to describe the effects of DHEA on bone turnover markers and menstrual function in young women with this disease. Our data suggest that DHEA may both decrease bone resorption and increase bone formation markers significantly in subjects with AN and may be associated with the resumption of menses unrelated to weight gain or changes in percentage of body fat. DHEA levels were subnormal in 60% of subjects studied, and a 50-mg dose of DHEA restored DHEA, E2, testosterone, and IGF-I levels to a physiologic range. Although the 100-mg and 200-mg doses resulted in supraphysiologic-free testosterone levels, this therapy was well-tolerated without adverse clinical effects.

Recent research has affirmed that deficiencies of androgens jeopardize skeletal health.(29,30) For example, patients with complete androgen resistance have been documented to have a lower BMD, despite increased E2 levels.(30) The androgen DHEA is a steroid precursor of estrogens through aromatization in peripheral tissues.(31-34) DHEA stimulates human osteoblastic cell proliferation through the androgen receptor, with alkaline phosphatase production through transforming growth factor-β.(35) A recent study(33) demonstrated that the stimulatory effect of DHEA on BMD and bone mineral content is primarily androgenic in nature. Our results corroborate with this report since the level of aromatization was not significantly increased at any of the DHEA doses studied. Our preliminary findings with oral DHEA, and those of Labrie et al. with topical DHEA, (36) suggest that this androgen has both anabolic and antiosteolytic properties. In both studies, urinary bone resorption markers decreased, suggesting suppressed osteoclast production and bone breakdown. Serum bone formation markers also increased, implying a stimulation of osteoblast function and bone formation. Similar changes in bone turnover markers were also seen after 1 year of DHEA in the animal study of Martel et al.(33) However, larger prospective studies are necessary to determine the long-term effects on both bone turnover markers and bone mass.

Accompanying the increased bone resorption in AN, a decrease in bone formation has been previously described.(5,17) Our data support this observation because low BAP and OC levels and elevated NTx levels were observed at baseline in our study subjects. Our subjects also had a reduced mean body mass index, which has been identified as an independent risk factor for the development of osteoporosis.(37) The decreased bone formation in AN is particularly problematic in adolescent patients who normally acquire 45–60% of their bone mass during the teenage years and reach their peak bone mass during this period.(2,3) Bone loss associated with this disease has also been characterized as being of rapid-onset and often irreversible.(1,5) Furthermore, a compromise of peak bone mass increases these individuals' risk of developing subsequent osteoporosis and fractures. (37) To prevent bone loss, stimulation of bone formation with minimization of bone resorption is needed. These pilot data suggest that DHEA fulfills this need as it normalizes levels of endogenous hormones that should be present abundantly during adolescence, stimulating the critical bone formation of this developmental period.

The mechanisms behind the decrease in bone resorption observed in this study are unclear. NTx levels were inversely correlated with DHEAS levels at baseline, and significant increases in DHEA and DHEAS in all subgroups at 1 and 2 months were inversely correlated with decreases in NTx. DHEA may induce the release of certain mediators that inhibit bone resorption. For example, DHEA itself and/or DHEAS may suppress secretion of proresorptive cytokines. This suggestion is only speculative since cytokines were not measured in this study. Previous research has established that DHEA is converted into estrogen, (12) and supplemental estrogen has been well validated as a therapy for postmenopausal bone loss.(28) Additionally, 17β-E2 in vitro inhibits the production of proresorptive cytokines.(38,39) Interestingly, NTx levels were not inversely correlated with E2 levels in the current study. A report by Labrie et al. showed that changes in intracellular sex steroids, as may occur after DHEA administration, are not necessarily translated into parallel changes in circulating sex steroid levels.(40) Data from the study of Martel et al.(33) also corroborate our results, indicating that both the antiosteolytic and anabolic effects of DHEA on bone are due mainly to local formation of androgens in bone cells, rather than the estrogenic effects of DHEA. Other potential mechanisms behind the antiosteolytic actions of DHEA deserve further study.

IGFs play an important role in the maintenance of bone mass and may mediate the anabolic actions of DHEA on the skeleton.(41,42) Previous work has shown that IGF-I levels are low in patients with AN.(8,43) In vitro studies have shown that IGF-I has effects on osteoblast function and collagen formation.(44,45) One previous clinical study showed that DHEA produced a rise in IGF-I and free IGF-I as it decreased IGF binding protein-1 levels.(46) In the current study, it was hypothesized that IGF-I was an anabolic mediator of DHEA's actions, stimulating bone formation, although no significant increases were seen within the small sample studied. However, increases in the bone formation marker, BAP, were strongly correlated with increases in IGF-I, and the 3-month IGF-I levels were directly correlated with final levels of a second formation marker, OC. As only future protocols can explore, this therapy may increase secretion of anabolic mediators other than IGF-I, unidentified to date, that were not measured in the current study.

Direct correlations between levels of E2 and bone formation markers were found in this study. We identified a strong correlation between increases in BAP and E2, and the final OC and E2 levels. DHEA is converted into E2(12,34); and previous studies have documented important effects of E2 on bone formation and growth. Estrogen has been shown to stimulate transforming growth factor-β and IGF-I, each stimulating bone formation in vivo.(47) Although not examined in this study, estrogen also inhibits prostaglandin E2 and interleukins 1 and 6,(48) thereby blocking bone resorption. Although an animal study has suggested that DHEA's effects on bone formation are primarily androgenic rather than estrogenic,(33) the beneficial effects of this steroid's conversion to estrogen in humans may be an important aspect of this hormonal agent and merits further study.

Androgenic hormones are receiving increased acceptance for use in female patients, but careful surveillance for androgenic side effects is warranted. Using pharmacological doses (1600 mg/day) for 28 days in six menopausal women, Mortola and Yen showed that DHEA produced increases in levels of testosterone, androstenedione, and E2.(12) While no adverse clinical effects were noted on these doses for a short-term course, cholesterol, HDL, and SHBG levels decreased. Two other reports showed a trend of beneficial effects on serum lipid profiles after 12 months of percutaneous DHEA,(36,49) although in one of these studies(36) two women developed slightly increased facial hair and two others noted mild acne. Insulin resistance is another potential side-effect of DHEA therapy, although one report demonstrated evidence of decreased insulin resistance after 12 months of the topical form of this therapy.(49) In the current study, no significant changes in fasting insulin or lipid levels, acne, or hirsutism were seen after 3 months of replacement doses of DHEA. Given our data showing restoration of physiologic hormonal levels with the lower DHEA dose, and the potential for adverse side effects during longer treatment periods with higher doses, 50 mg appears to be an appropriate dose for long-term trials of DHEA in adolescent females with AN.

The resumption of menstrual bleeding in 53% of subjects was surprising and may contribute to our understanding of factors affecting return of menses in patients with AN. Studies have noted that anorexic patients resume menses at varying body weights.(50-52) Data from a study by Golden et al. suggest that an E2 level of 30 pg/ml may be a marker for the onset of menses in these patients.(51) We found that baseline E2 levels were moderately correlated with percentage of body fat. We also documented significant increases in E2 at 3 months in subjects who resumed menses with a mean level exceeding 50 pg/ml. Increases in E2, combined with a significant increase in testosterone after DHEA, may explain why favorable, significant changes in bone turnover markers occurred within this subgroup. DHEA is a significant precursor of ovarian estrogen secretion.(34) Its conversion to estrogen in the ovary may explain the recurrence of menses seen in the current study. From the significant decrease in SHBG levels among the total sample, one can postulate that DHEA, and the testosterone to which it is converted, suppress SHBG, possibly providing increased bioavailable E2 to stimulate the endometrium. However, Labrie et al.(36) found that 12 months of DHEA therapy did not induce estrogenic changes in the endometrium of postmenopausal patients. Whether DHEA administration stimulates the endometrium of young adolescent patients is currently unknown and deserves investigation. The fact that cyclic vaginal bleeding subsided within 3 months after discontinuing the medication further suggests that DHEA promoted the return of menses.

Our study has several limitations. First, given the small sample size of this pilot study, results must be considered preliminary. Second, the amount of activity and nutritional intake was determined by self-report, and patients with AN are known to under-report exercise and over-report dietary intake.(52) Furthermore, our study subjects were adolescents who may not have been compliant with therapy. Despite appropriate pill counts and histories of adhering to the protocol, varying DHEA and DHEAS levels throughout the study suggest intermittent compliance which may have impacted on our data. We also did not see a clear dose-response relationship among the 50, 100, and 200 mg dose groups, which could have been related to either noncompliance or other factors. Although not addressed specifically in this pilot study, factors relating to absorption and bioavailability may exist that are unique to patients with AN. There may be a threshold effect such that higher doses inhibit formation and/or increase resorption, or the maximal effect may be reached with the 50 mg dose. In addition, the small sample size (five subjects per group) gave us little power to determine definitive dose-response relationships and may have caused us to miss other significant associations. Last, only changes in markers of bone turnover were documented in the current study. Although increasingly sensitive,(53) they are not exact measures of changes in bone mass. As only a longitudinal protocol can address, the effect of DHEA on bone mass requires further study.

This preliminary pilot study was designed to evaluate the effect of a new therapy on bone turnover markers in order to begin to address its effects on reversing the skeletal loss associated with AN. This bone marker data, and the unexpected finding regarding menstrual function, indicate that DHEA may be an effective means of hormonal replacement for young women with this disease. Our data suggest that DHEA may be one of only a few agents which have shown promise in preventing the irreversible bone loss associated with AN. We have shown that short-term DHEA significantly decreased levels of bone resorption and increased markers of bone formation. DHEA was well tolerated and, as an oral therapy, was convenient for adolescent patients. These pilot data emphasize the therapeutic potential of DHEA. Longitudinal studies are needed to determine the long-term biological actions of this agent, including its effects on bone mass.


The authors gratefully acknowledge the dedicated patient care of the nursing staff of the outpatient division of the General Clinical Research Center; Joseph A. Majzoub, M.D. for his support of this project and critical review of the final manuscript; Mr. Charles Hakala of Belmar Pharmacy, Lakewood, Colorado for supplying the micronized DHEA; Cara Campobasso and Jennifer Franklin for expert DXA technical assistance; David Zurakowski, Ph.D. for biostatistical advice; and Rebecca Lamm for help with preparation of the manuscript. This work was supported by the Clinical Investigator Training Program: Harvard/Massachusetts Institute of Technology Health Sciences and Technology-Beth Israel Deaconess Medical Center, in collaboration with Pfizer Inc. (to C.M.G.); The National Osteoporosis Foundation/Mazess Research Program (to C.M.G.); Grant MO1 RR02172, General Clinical Research Resources, National Institutes of Health (NIH); Program Grant MCJ-MA 259195 from the Maternal and Child Health Bureau (to S.J.E.); and NIH Grants RO1 AG2271–03 and RO1 AG13519-02 (both to M.S.L.).