Department of Family and Preventive Medicine University of CaliforniaSan Diego La Jolla, California 92093-0607
To the editor:
Drs. Davis and Tran raise important issues that can be summarized to three main areas: 1) the effect of adjusting results for body mass index (BMI); 2) the validity of the testosterone measurements and their relation to premenopausal levels; and 3) a plausible biological explanation for the findings.
1) The effect of adjusting results for BMI.
The relation between hormone levels and chronological and menopausal age was analyzed 1) without adjusting for covariates, 2) adjusting for BMI, and 3) adjusting for BMI, smoking, exercise, and alcohol consumption. In the interest of conserving space, only the BMI-adjusted results were presented. The association (or absent association) with age did not differ in unadjusted analyses, and the total and bioavailable testosterone levels were virtually identical to the BMI-adjusted values shown in the figures. The minimal impact of BMI on bioavailable testosterone in this study may reflect the narrow range of BMI among the Rancho Bernardo cohort (95% confidence interval, 23.824.5 kg/m2), in whom BMI did not vary across age groups.
2) The validity of the testosterone measurements and their relation to premenopausal levels.
Total testosterone levels were measured by a RIA developed in the steroid research laboratory of Samuel Yen, M.D., during the 1970s using an antibody produced at the University of CaliforniaSan Diego. An important feature of steroid hormone measurements in this laboratory is sample purification by solvent extraction and celite column chromatography before RIA. The testosterone assay sensitivity of 0.07 nmol/L is comparable with published values (0.030.05 nmol/L) for assays using similar sample purification steps (1, 2, 3). Because of the high sensitivity and specificity of this method, measured values are lower than direct serum measurements with commercial kits. However, the testosterone levels in this study were similar to levels in other large studies of postmenopausal women using comparable assay methods (1, 2, 3).
The authors also question the validity of the testosterone level cited for premenopausal women. Levels for postmenopausal women were compared with a recently published value (0.63 ± 0.23 nmol/L) from the Yen laboratory for 32 women sampled in the early follicular phase of the menstrual cycle (4). A 1979 study (5) using the same assay reported a 24-h mean level of 0.69 ± 0.04 nmol/L for early follicular phase women, therefore, the assay/laboratory has good internal consistency. (The 1973 Judd and Yen article cited by Davis and Tran used an earlier, less specific assay.) Although the sensitivity of the Sinha-Hikim total testosterone assay is very high, total testosterone levels were apparently measured directly in the dialysate. Equilibrium dialysis would not eliminate cross-reacting substances, which may account for higher (1.04 ± 0.76 nmol/L) early follicular phase testosterone levels with this assay compared with the Yen assay. In our view, steroid hormone assays may yield differing results for a variety of reasons. Direct comparisons are valid only when all samples are measured with the same assay techniques, as was the case in our paper.
Drs. Davis and Tran correctly point out that the finding of an increase in testosterone levels with postmenopausal aging was, in large part, based on lower levels among only 29 intact women in the 50- to 59-yr age range. As stated in the article, "although age-specific levels were based on relatively few women for the youngest decade, the similarity of hormone patterns after stratification by decade and by years since menopause supports the validity of the age-related associations." Among the intact women, 125 were less than 20, 136 were 2030, and 119 were more than 30 yr menopausal. Testosterone levels were 25% higher among intact women more than 20 yr postmenopause compared with intact women less than 20 yr postmenopause.
3) A plausible biological explanation for the findings.
The increase in circulating testosterone levels with postmenopausal aging is likely to be ovarian in origin. Testosterone levels did not increase with age among the oophorectomized women, and the metabolism and interconversion of androgens in women is not altered after menopause (6). In recent studies, ovarian vein testosterone levels were related to the degree of stromal hyperplasia in 18 postmenopausal women (7) and correlated positively with age in 52 women aged 4269 yr (8). Postmenopausal ovarian androgen production is thought to be at least partially gonadotropin driven. Gonadotropin receptors have been identified in the stroma of menopausal ovaries (9), and peripheral steroid levels in postmenopausal women increase after hCG stimulation (10) and decline after administration of long-acting GnRH agonists (11, 12). Thus, stimulation by elevated gonadotropins could account for increased ovarian testosterone synthesis in older women. Although postmenopausal gonadotropin levels tend to decline with advancing age (13), they remain well above premenopausal levels. It is possible that prolonged exposure to high gonadotropin stimulation may up-regulate ovarian responses to LH. To our knowledge, no published studies address this possibility.
The absence of a postmenopausal increase in androstenedione levels with age similar to the increase in testosterone is not surprising. As shown in previous studies and confirmed in this one, the postmenopausal ovary contributes 40% of circulating testosterone, but only 1015% of circulating androstenedione. Thus, an increase in ovarian androstenedione production would be masked by the many-fold greater adrenal contribution.
Our conclusion that testosterone levels return to the premenopausal range in older, intact postmenopausal women needs clarification. The internal (unpublished) normal laboratory values for this assay are 0.72 ± 0.34, 1.02 ± 0.48, and 0.93 ± 0.38 nmol/L for the early follicular, mid-cycle, and luteal phases, respectively. Because testosterone levels do not undergo cyclic changes in postmenopausal women, our results suggest that older, as well as younger, postmenopausal women are relatively testosterone deficient when compared with younger cycling women, with levels similar to the early follicular phase. We appreciate this opportunity to expand and clarify our findings. Clearly, interpretation of cross-sectional data are subject to error. Prospective studies are needed to define with certainty the pattern of changing testosterone levels in older women.
References
HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Endocrinology | Endocrine Reviews | J. Clin. End. & Metab. |
Molecular Endocrinology | Recent Prog. Horm. Res. | All Endocrine Journals |