Sex-Specific Effects of Dehydroepiandrosterone (DHEA) on Bone Mineral Density and Body Composition

A Pooled Analysis of Four Clinical Trials

Catherine M. Jankowski; Pamela Wolfe; Sarah J. Schmiege; K. Sreekumaran Nair; Sundeep Khosla; Michael Jensen; Denise von Muhlen; Gail A. Laughlin; Donna Kritz-Silverstein; Jaclyn Bergstrom; Richele Bettencourt; Edward P. Weiss; Dennis T. Villareal and Wendy M. Kohrt


Clin Endocrinol. 2019;90(2):293-300. 

In This Article

Materials and Methods

Sources of Data

Individual-level data from four RCTs of DHEA therapy in older adult[7–10] were merged into a central database by investigators at the University of Colorado Anschutz Medical Campus (CU-AMC).[7] Although each of the RCTs included additional outcomes and, in one RCT an additional treatment arm,[8] the focus of the pooled analysis was on the 12-month changes in BMD, body composition, and circulating hormones and growth factors in response to oral DHEA therapy as compared to placebo. The main characteristics of the studies are briefly described below and in Table 1. Investigators obtained local institutional review board approval for their RCT.[7–10]


The RCTs included women and men aged 55–85 years,[9] 60 years or older[7,8] or 65–75 years,[10] who were not using sex hormone therapy. Low serum DHEAS concentration was an inclusion criterion in two studies.[7,8]


Participants were randomized to oral daily DHEA or placebo for 12 months. The period of intervention was 24 months in one RCT,[8] but only data through the 12-month end-point were included in the pooled analyses. Similarly, data from a second year of open-label DHEA therapy in one RCT were excluded from the pooled analyses.[10] The dose of DHEA was 50 mg/d[7,9,10] except in one RCT8 in which the dose was 75 mg/d for men. Pills or capsules were produced by different companies across trials, but all were administered in a double-blinded manner in all trials. Compliance was monitored by pill counts at 1- to 3-month intervals.[7,9,10] Calcium and vitamin D supplements were provided to participants in two of the RCTs.[7,10]

Tests and Procedures

Bone mineral density and body composition were measured, and blood samples collected before and after the 12-month intervention. Data from interim measurements of study outcomes[7,9,10] were not included in the pooled analysis.

Proximal femur (total hip, neck, trochanter and subtrochanter regions) and lumbar spine (L2-L4) BMD were measured by dual-energy x-ray absorptiometry (DXA) using a Lunar DPX-IQ,[8] Hologic GDR 20009 or Hologic Delphi 4500-W10 instrument. One RCT7 used both Lunar DPX-IQ and Hologic Delphi-W units because of a transition in instrumentation, but each participant had baseline and 12-months measurements on the same instrument. Total body DXA scans were performed for the determination of FM and FFM.

Serum DHEAS, (17)estradiol (E2), testosterone (T), sex hormone-binding globulin (SHBG) and insulin-like growth factor-1 (IGF-1) concentrations were measure during chemiluminescense, radio-immunoassay, immunoradiometric assay or enzyme-linked immunosorbent assay, as previously reported.[7–10] Blood samples were collected after an overnight fast.

Data Management

A data dictionary was sent to each RCT investigator, and corresponding data files containing individual data for the cases that had been included in intention-to-treat analyses were returned to CU-AMC. Units of measurement were verified, and raw data converted as needed. For three RCTs,[7,9,10] undetectable serum testosterone concentrations in women were replaced with 16.90,[7] 2.88[9] or 20.0 ng/dL[10] (0.59, 0.10, 0.69 nmol/L, respectively), values representing the lower limit of detection for the assay used at the site. In the fourth RCT,[8] all serum testosterone values were within the detection limits of a high-sensitivity chemiluminescense assay.

No substitution rules were applied for missing data. Thus, the number of cases for each outcome varied and is provided in Table 1, Table 2 and Table 3 and Supporting Information Table S1.

Statistical Analyses

Differences in the baseline characteristics of women and men across the four sites were evaluated by one-way ANOVAs. Sex allocation to treatment arm was balanced by design; the balance was verified in the ITT population by a chi-square test for equal proportions. The primary outcomes were the per cent changes in BMD and the absolute changes in FM and FFM in response to the intervention. Between-group differences (change in DHEA minus change in placebo) were evaluated in an ANCOVA model controlling for baseline measures and performance site. Separate regressions were completed for women and men combined (controlling for sex), women only and men only. Differences in the 12-month changes in sex hormones and growth factors were evaluated in the same manner. Changes were expressed in absolute and relative terms.

We performed a sensitivity analysis for all outcomes to address the impact of baseline BMD using BMD T-scores from 3 of the 4 RCTs. A variable for normal BMD (T-score >−1.5) or low BMD (T-score ≤−1.5) was created. Due to the few cases of osteoporosis, we collapsed low bone mass and osteoporosis into a single category. The BMD indicator was first examined as a covariate in all regression models to test group differences adjusting for normal vs low BMD. Additional analyses tested the possibility of a differential effect of condition on outcomes by baseline BMD, through the inclusion of a statistical interaction term between condition and the BMD indicator. All conclusions drawn from these sensitivity analyses were the same as the primary approach, and there were no significant interactions between condition and BMD indicator. We therefore report the primary analyses only.

All analyses were performed using SAS software (version 9.3; SAS Institute, Inc, Cary, NC, USA). Results are reported as means and standard deviations (SD) or standard error of the mean (SE). To account for multiple comparisons, the alpha level was set at 0.01.