Department of Epidemiology, Graduate School of Public Health, University of Pittsburgh (J.V.Z., J.A.C., E.O.T.), Pittsburgh, Pennsylvania 15261; Department of Obstetrics and Gynecology, University of Rochester School of Medicine (D.S.G.), Rochester, New York 14642; and Division of Endocrinology and Metabolism, University of Louisville Health Sciences Center (S.J.W.), Louisville, Kentucky 40292
Address all correspondence and requests for reprints to: Jeanne V. Zborowski, Ph.D., Department of Epidemiology, University of Pittsburgh, Graduate School of Public Health, 507 Parran Hall, 130 DeSoto Street, Pittsburgh, Pennsylvania 15261. E-mail: JVZST{at}pitt.edu
![]() |
Introduction |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Women with PCOS typically have acyclic production of 17ß-estradiol, with circulating concentrations similar to those seen in the follicular phase of cycling women, but considerably lower than the mean 17ß-estradiol concentration observed across the normal menstrual cycle (1). This static level of 17ß-estradiol, with the absence of the estradiol surge associated with ovulation, would intuitively be expected to negatively affect bone density. Testosterone and androstenedione, however, are produced in excess by the PCOS ovary and perhaps by the adrenals of some affected women. This chronic elevation in androgens may exert a positive influence on bone in PCOS women, either directly through androgen receptors on bone-related cells or indirectly after conversion to 17ß-estradiol and estrone, respectively, in peripheral tissues. Moreover, elevated circulating insulin levels, also associated with PCOS, may offer some additional protection against a reduction in bone mass in these women. Hyperinsulinemia has been shown to positively affect bone density (5, 6) through direct stimulatory effects on osteoblastic activity (7) and by the suppressive influences exerted by insulin on the production of sex hormone-binding globulin (SHBG) (8) and insulin-like growth factor (IGF)-binding proteins (IGFBPs) (9). This insulininduced suppression of these binding proteins may result in increased exposure of target tissues, including bone, to the stimulatory effects of elevated concentrations of free sex steroids and IGFs.
Both thin and obese women develop PCOS, a presentation that allows for evaluation of the effects of life-long obesity, alterations in body composition (e.g. increased central adiposity), and related metabolic abnormalities (e.g. hyperandrogenemia and hyperinsulinemia) on the skeleton. The relatively high prevalence of PCOS and its early-life presentation render this disorder of particular importance as a model for assessing the effects of androgen and potentially insulin on the attainment of maximal bone mass.
The objectives of this review article are to 1) summarize the literature exploring the androgen-bone mass relationship in normal women; 2) assess the effects of PCOS, androgen excess, and androgen-estrogen balance on bone mass in women; 3) examine study design issues related to PCOS, androgen excess, and bone mass; and 4) suggest areas for future research related to the influence of PCOS, androgen excess, and insulin on bone in women.
![]() |
Osteoporosis, peak skeletal mass, and androgens |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Osteoporosis in women is a reflection not only of aging and estrogen-dependent bone loss during the menopause, but also of peak skeletal mass attained and maintained in young adulthood (2030 yr of age) (14, 15). The natural history of maximal skeletal mass has been linked to a variety of influences, including genetic and familial factors, race and ethnicity, nutrient intake, mechanical loading, and, in particular, reproductive hormones (16). The influence of estrogens (both endogenous and exogenous) on bone among women has been examined extensively. The influence of androgenic hormones on bone mass among women is less certain. If the protective effect of estrogens on female bone also extends to androgens, then androgenic influences during the adrenarche and the pubertal transition may be of particular importance in achieving maximum bone density in women. Evaluation of the androgen-estrogen equilibrium in PCOS women and its relationship to menstrual cyclicity and bone density may offer important insights into the potential synergistic relationship between estrogens and androgens in the protection against the development of osteoporosis and its debilitating sequelae.
![]() |
Androgens and bone |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Similarly, osteoblast-like cells in bone in both men and women are able to aromatize androgens to estrogen (23, 24). Moreover, aromatase deficiency and estrogen receptor polymorphisms in men have been shown to lead to osteopenia and sustained linear growth, with failure of epiphyseal closure (25, 26). Estrogen replacement in aromatase-deficient males has also been shown to increase bone mass (27). These observations suggest that estrogen plays a crucial role in mediating bone mass in both men and women, a theory recently articulated by Riggs and colleagues in the bone literature (28).
However, the presence of androgen receptors on osteoblasts also
suggests a potential direct effect of androgens on bone regardless of
gender. Colvard et al. first described the expression of the
androgen receptor in human bone cells in vitro (29). Kasperk
et al. demonstrated that androgens stimulate the
proliferation of osteoblasts and osteoblast-like cells in
vitro (30); dihydrotestosterone, a potent nonaromatizable
androgen, was shown to increase the formation of human osteoblasts and
to promote their differentiation and maturation. Human osteoblasts also
exhibit 5-reductase activity and are able to convert testosterone
into dihydrotestosterone (31). Furthermore, androgenic hormones appear
to stimulate bone formation in human osteosarcoma cells by acting as a
transcription factor for the synthesis of
1(I)-procollagen and transforming growth
factor-ß (32). The protein analogue to the
1(I) messenger ribonucleic acid is
representative of two of the three chains that form the triple helix of
the type I collagen molecule, the major extracellular matrix protein in
bone (32). This finding implies that androgenic hormones may stimulate
bone formation by the induction of type 1 collagen synthesis (32).
Androgens, similarly to estrogens, have also been shown to influence bone metabolism indirectly by several potentially interrelated mechanisms. These pathways include inhibition of bone resorption through a reduction in interleukin-6 production by osteoblasts (33), inhibition of the production of PGE (34), and inhibition of the effect of PTH on osteoblasts (35). In addition, androgens may improve calcium balance by increasing intestinal absorption (36), decreasing renal excretion (37), and increasing circulating levels of free 1,25-dihydroxyvitamin D3 (38).
![]() |
Androgens and BMD in women |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Cross-sectional epidemiological studies assessing the specific effects of endogenous androgens on BMD or bone development in female youth and in adult premenopausal women have generally demonstrated a positive association between androgen levels and BMD. Evaluations of androgens and bone mass in prepubertal and adolescent girls are, however, relatively sparse. Dhuper et al. evaluated the independent effects of androgens and estrogens on bone density in 43 healthy nulliparous adolescent females (age, 1320 yr) (49). BMD of the spine, wrist, and foot was moderately correlated (r = 0.300.54) at all sites with testosterone levels and with a composite score of integrated estrogen exposure, but not with serum 17ß-estradiol, DHEAS, PRL, LH, or FSH. Lloyd et al. suggested bone mass accretion during puberty was best predicted by body weight, height, and androgen-dependent growth characteristics, as measured by pubic hair staging, in 112 premenarchal Caucasian females (mean age, 11.9 yr) (50). Estrogen-dependent development (breast staging), on the other hand, was less closely associated with whole body bone density. Porcu et al. observed that skeletal maturity after the menarche (defined by closed vs. open epiphyses) in 82 adolescent (12- to 19-yr-old) females was unrelated to testosterone, dihydrotestosterone, or androstenedione (51). However, DHEA and DHEAS levels were higher in the closed epiphyses group, suggesting a possible association of this weak androgen with bone maturity in young females. Interestingly, adolescent females with complete androgen insensitivity syndrome have reduced bone density, whereas bone turnover and fracture risk appear to be unaffected (52). These observations imply that both androgens and estrogens may be required for acquisition of normal bone density in females during childhood; whether androgens influence bone remodeling and fracture risk, however, is less clear. Conversely, long-term androgen receptor blockade with flutamide in a group of adolescent women with idiopathic hirsutism had no apparent detrimental effect on BMD (53).
Buchanan et al. first demonstrated the apparent independent and potentially additive effects of androgens and estrogen on peak trabecular bone density in adult premenopausal Caucasian women (n = 30; all nulliparous; age, 1822 yr) (54). Trabecular bone density, measured by quantitative computed tomography, was correlated with bioavailable (free and albumin-bound) testosterone (r = 0.48; P = 0.007), total testosterone (r = 0.49; P = 0.007), and androstenedione (r = 0.40; P = 0.03). In a multiple regression analysis, the overall variance in peak bone density was best explained by the combination of bioavailable testosterone and 17ß-estradiol (r2 = 0.38; P = 0.002). No adjustment was made, however, for weight or relative obesity [e.g. body mass index (BMI)]. Body composition parameters are highly correlated with the sex steroids; in addition, body weight and weight distribution exert mechanical stresses on bone that may influence overall and regional densities (55).
Circulating SHBG may also play an important role in the actions of androgens in bone among healthy women. The concentration of circulating SHBG influences the association between testosterone and bone density by regulating the availability of unbound, biologically active sex hormones. SHBG binds dihydrotestosterone with the highest affinity, followed by testosterone and 17ß-estradiol (56). Steinberg et al. observed an inverse relationship between SHBG and BMD in 292 healthy, normally menstruating, pre- and perimenopausal Caucasian women (aged 3550 yr) (57). After adjusting for weight, lower SHBG was associated with higher bone density at all sites in the premenopausal women. Free testosterone was positively correlated with BMD in the lumbar spine and femoral neck; the percentage of both free testosterone and free estradiol was correlated with BMD at the femoral neck and greater trochanter. The DHEAS-BMD association was statistically significant only at Wards triangle and the femoral neck. However, controlling for SHBG in the multivariate model eliminated the apparent association between bone density and the sex steroids, suggesting that SHBG levels mediated the association between BMD and sex hormones. Johnston et al. observed similar associations in a group of healthy premenopausal women (58). Free testosterone was the androgen most consistently and significantly associated with BMD at all measured sites. DHEAS was correlated with BMD at the femoral neck and radius only. Again, SHBG was negatively correlated with BMD at the both the spine and femoral neck.
These data suggest that free endogenous androgens, potentially modified
by circulating SHBG levels, might play a role in the determination of
bone mass in premenopausal women. Specific androgens may, however,
exert influences differentially across the various bone sites. DHEAS
has little androgenic activity, and direct stimulatory effects on bone,
although possible, are unlikely. Therefore, DHEAS may serve rather as a
surrogate marker for the effects of DHEA, which can
directly stimulate both osteoblast proliferation and differentiation
[similar to testosterone and DHT (59)], or for
5-androstenediol (60). Conversely, DHEAS may
be metabolized to estrogens and exert its influence through the
estrogen receptor (61). Alternatively, the association may reflect the
effects of an unmeasured substance, influencing both DHEAS levels and
bone density (e.g. interleukin-6 and transforming growth
factor-ß) (62, 63).
The generally positive correlations observed between androgens and BMD
in premenopausal women contrast with those reported in the
postmenopausal population, in which the androgen-bone density
relationship is less clear. Although ovarian function, adrenal androgen
production, and BMD all decrease with age, early studies in elderly,
primarily Caucasian, populations (64, 65, 66) demonstrated no apparent
association of androgens with bone density. Cauley et al.
also later investigated the contribution of differences in sex steroid
hormone levels to racial differences in BMD in older women (65 yr of
age) (67). No association between androgens and bone density was
evident in either black or white older women. Also, in a prospective
study in an elderly ambulatory population, baseline DHEA
levels were not associated with later life BMD (68).
In contrast, Johnston et al. found positive correlations between serum testosterone and vertebral bone mass in 84 peri- and postmenopausal women categorized by menstrual status (58). It was not known, however, if these associations were independent of estradiol or estrone. Deutsch et al. also found a significant correlation between adrenal androgens and lumbar spine BMD in elderly women (69). In other retrospective studies in postmenopausal osteoporotic women, lower androgen levels have been associated with both a lower BMD and an increase in hip fractures (70). Jassal et al. more recently observed that both estimated and measured bioavailable testosterone levels predicted future height loss in elderly women, independent of age, obesity, estrogen, smoking, alcohol intake, and use of thiazide diuretics (71). They concluded that a reduction in height loss, as a surrogate for osteoporotic vertebral fractures, is compatible with a direct effect of testosterone on BMD or bone remodeling. Cummings et al., however, found no independent association of baseline free testosterone, after adjustment for 17ß-estradiol levels, with subsequent vertebral fractures among older women in the Study of Osteoporotic Fractures (72).
On the other hand, investigators tend to agree that exogenous androgens may positively influence bone density in older women, independent of estrogen. Androgenic hormones have recently been employed as a supplemental agent in the prevention and treatment of osteoporosis in postmenopausal women (73). The assessment of bone density and bone remodeling parameters in this population suggests that the addition of androgens to the treatment regimen results in significant increases in BMD (74) and bone formation (75) compared with the effect of either estrogen or estrogen-progestin therapy alone.
![]() |
Endogenous androgen excess, polycystic ovaries, and bone |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Buchanan et al. compared mean BMD among regularly menstruating women (aged 2148 yr) with androgen excess [bioavailable testosterone levels >15 ng/dL (0.52 nmol/L)] with that of control women of similar age, body size, and gynecological history (78). Nineteen women were classified by biochemical evaluation as having androgen excess, and 27 women (11 hirsute and 16 nonhirsute) were classified as normal. Trabecular bone density, measured again by three-dimensional quantitative computed tomography, was significantly higher in the androgen excess group compared to the control group (172 vs. 153 mg/mL; P = 0.03). Adjusting for androstenedione abolished the difference between groups, whereas adjusting for other androgens, including testosterone, did not significantly alter the difference. Trabecular bone density in the total group was directly correlated with androstenedione (r = 0.31; P = 0.03), but not with other androgens. Radial density was correlated with both total testosterone (r = 0.53; P = 0.004) and bioavailable testosterone (r = 0.47; P = 0.01), but only in the 27 women was it associated with normal bioavailable testosterone. Controlling for height and weight in this subgroup did not alter the androgen-bone density relationship.
The investigations by Buchanan were unique in that computed tomography, unlike two-dimensional absorptiometry techniques, allowed the investigators to assess the volumetric density, specifically of the trabecular component in the lumbar spine and independently of bone size. The higher trabecular density noted in the androgen excess group is therefore unlikely to be an artifact of increased bone thickness. The observed between-group difference in mean bone densities, however, may actually be an underestimate of the true differences among women with androgen excess compared to normal controls due to the inclusion of hirsute women in the control group. Bardin and Lipsett demonstrated that the production and clearance rates for androstenedione and testosterone were significantly higher in women with hirsutism, androgen excess, and/or PCOS than in controls (79). The presence of hirsutism suggests, if not an increase in free androgens, potentially an increase in androgen sensitivity. It is not known whether this potential increase in sensitivity also extends to bone. Therefore, increased bone tissue exposure or sensitivity to androgens is possible in the hirsute controls even in the apparent absence of elevated androgen levels.
Dixon et al. also reported an apparent compensatory effect of androgens on BMD in hirsute women (80). In 32 nulliparous women with androgen excess, BMD at the lumbar spine and femoral neck was not significantly different compared with that in 32 nonhirsute nulliparous women with regular menstrual cycles despite significantly lower estradiol levels in the hirsute women. In the 5 hirsute women with the lowest estradiol levels (<70 pmol/L), bone density was also comparable to that in controls. However, the hirsute women were significantly more obese than the control women. In a subgroup analysis, the investigators addressed the difference in weight and body fat between hirsute cases and controls by examining BMD among women with a normal BMI (2025 kg/m2). Similar to the results observed in the total sample, no significant difference in BMD between the hirsute and nonhirsute groups was noted. These findings suggest that hyperandrogenemia or related metabolic processes may preserve bone mass independently of obesity and low estradiol levels. However, no measures of weight distribution (e.g. waist circumference) or body composition (e.g. fat mass and lean mass) were available to assess the contribution of differences in regional body composition and/or fat distribution to the observed differences in BMD between cases and controls.
In a recent study of lean women with PCOS, Good et al. suggested that androgens might significantly influence bone mass in hyperandrogenic PCOS women through androgen-mediated changes in body composition. A significantly higher BMD was noted in the left arm, right arm, and left ribs of the PCOS group compared to controls (81). Total body BMD was not, however, different between the two groups. A strong correlation was apparent between total and bioavailable testosterone levels and the observed increase in upper body bone mass. A trend toward a decreased percentage of body fat in the lean PCOS group was also suggested. The regional differences in bone mass, with significantly increased BMD in the upper skeleton, may reflect differential androgenic influences on lean mass accretion in PCOS (82, 83, 84), with subsequent, localized stimulatory effects (mechanical or otherwise) on bone (84, 85, 86, 87). The sample size in this study was small, however, and further evaluation is required.
Interventional studies in women with androgen excess also provide insight into the possible role of androgens and androgen-estrogen equilibrium in the maintenance of bone mass. Androgen suppression by androgen antagonists or GnRH agonists is a common treatment option in women with PCOS that results in the suppression of androgens or both androgens and estrogens, respectively. Prezelj and Kocijancic observed that BMD was significantly reduced (mean decrease, 0.032 g/cm2; 1.095 to 1.063 g/cm2; P < 0.001) in 15 of 17 eumenorrheic subjects with androgen excess, after a 12-month treatment with a combination of spironolactone and linestrenol (88). A decrease in androstenedione was the only hormonal variable significantly associated with the decline in bone density (r = 0.50; P = 0.037). Simberg et al. studied the effect of a GnRH agonist on bone density in 20 hirsute women with hyperandrogenemia compared with 19 age-matched nonhirsute, normoandrogenic controls (89). At baseline, after controlling for BMI, hirsute patients had higher BMD at all measured sites (lumbar spine, proximal femur, femoral neck, and trochanter major). BMD in the lumbar spine and proximal femur correlated with free testosterone (r = 0.62; P = 0.003) and total testosterone (r = 0.46; P = 0.045) in the whole study group. GnRH analog treatment significantly reduced BMD (2.07.9%) in the lumbar spine and greater trochanter region of the hirsute women. However, the higher baseline BMD observed in the hirsute females prevented the posttreatment BMD from falling to the level in the normoandrogenic controls. The significant correlation of androstenedione and total and free testosterone with BMD again suggests some contribution of androgens to bone density in women.
Prepubertal children and adolescents with congenital adrenal hyperplasia (CAH), also characterized by androgen excess, appear to be protected from the well described (90, 91), detrimental influence on bone density of long-term glucocorticoid replacement therapy. Peak bone density attainment does not appear to be impaired in young, glucocorticoid-treated CAH subjects (92, 93, 94), and bone mass may be higher compared to that in controls of similar age. Given the replacement nature of the therapy, it is not clear, however, that this protective effect is androgen mediated. In addition, it appears that adults with CAH, particularly postmenopausal adult women, may not be immune to the development of osteopenia (95). Osteoporosis, therefore, remains a consideration in certain hyperandrogenic populations.
![]() |
Androgen-estrogen synergy, menstrual function, and BMD |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
DiCarlo et al. compared BMD of the lumbar spine in 266
amenorrheic women who were either amenorrheic due to PCOS or
amenorrheic related to other disparate diagnoses (96). For the
analyses, the non-PCOS amenorrheic group was further subdivided into
those with polycystic-appearing ovaries on ultrasound and those with
normal-appearing ovaries. Patients with PCOS-related amenorrhea had
significantly higher BMD compared to non-PCOS amenorrheic patients
(Fig. 1). Non-PCOS amenorrheic subjects
with polycystic ovaries on ultrasound examination, but not diagnosed
with the endocrine syndrome (n = 65), had higher BMD than non-PCOS
subjects with normal ovaries (n = 142). Any reduction in BMD,
often associated with diverse causes of amenorrhea, appeared to be
attenuated in amenorrheic PCOS women. The presence of
polycystic-appearing ovaries on ultrasound was associated with a higher
BMD compared with normal-appearing ovaries regardless of whether the
endocrine syndrome itself was overtly manifested. However, there are
several considerations in interpreting the results of this study. No
attempt was made to control in the analysis for age, duration of
amenorrhea, or other factors, specifically differences in body weight,
that may have contributed to the observed differences in bone density
between the groups. The relative contributions of androgens and
estrogens to BMD in this population were also not estimated. Moreover,
the non-PCOS amenorrheic group consisted of a heterogeneous
composite of women with amenorrhea from various causes, including
premature ovarian failure, hyperprolactinemia, hypogonadatropic
hypogonadism, excessive weight loss, excessive exercise, and Turners
syndrome. These disorders have distinctively different hormonal
profiles and very different effects on BMD, and are, as an aggregate,
not a classic control group for comparative purposes.
|
|
|
|
|
![]() |
Discussion |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Various lines of evidence in both normal women and women with androgen excess (PCOS) tend to support an independent association of androgens with peak bone mass attainment and maintenance in premenopausal women. PCOS is, however, a complex disorder, and several components of the syndrome may be related to bone density through common metabolic events not necessarily mediated specifically by androgenic steroids. Notably, women with PCOS tend to be overweight; body weight is a strong predictor of bone mass and is highly collinear with sex steroids. Moreover, differences in regional body composition (e.g. increased central adiposity and truncal mass) may contribute to site-specific increases in BMD in PCOS. This important aspect of the PCOS-BMD association has not been adequately explored.
Generally, the association of androgens and BMD in the androgen excess populations appears to be stronger in bone sites with a greater percentage of trabecular bone (lumbar spine and femoral neck) compared to cortical bone sites (radius and hip), suggesting potentially more pronounced androgenic effects, whether direct or indirect, in more metabolically active bone. This finding is of interest given the documented effects of androgens on cortical bone during development. As noted in rat models, however, skeletal growth (bone modeling) as a finite event and skeletal maintenance (bone remodeling) as a lifelong process reflect distinctly different biological systems. It is significant to note that in growing rats, androgens stimulate periosteal bone formation, whereas in aged rats, androgens prevent endosteal bone resorption (101, 102).
The higher bone density generally observed in PCOS/hirsute women with regular menses compared with that in oligomenorrheic/amenorrheic PCOS/hirsute women or normal controls suggests that estrogens and androgens may interact to determine bone density in women. In the presence of adequate estrogen, androgens may have a positive effect on bone However, this is a highly controversial area in PCOS research. In PCOS, bone density appears to be maintained at levels comparable to those observed in control women, implying that the deleterious effects associated with anovulation may be attenuated by bone exposure to androgens. However, no net positive effect on bone was observed without the midcycle estradiol peak and/or luteal phase progesterone production associated with the normal menstrual cycle (100), suggesting again that estradiol plays a critical role. Therefore, some oligo- and amenorrheic women with PCOS may be at risk for osteopenia, if estrogen levels, integrated across the menstrual cycle, are particularly low. However, the estrogen profile across the menstrual cycle in PCOS-affected women has not been completely characterized. If PCOS is actually a disorder of hyperestrogenism, as suggested by Barnes, the higher bone density seen in eumenorrheic, hyperandrogenic women compared to that in both oligo- and amenorrheic PCOS women and normal controls may be estrogen mediated.
Alternatively, androgen-dependent changes in body composition (e.g. increased central adiposity and/or lean mass), rather than a direct effect of androgens, may be the primary influence in maintaining bone density in these women. The absence of any notable differences in bone remodeling markers between women with PCOS and healthy controls suggests that any direct effects of androgens on bone formation may be limited. As observed by Good et al., the association between upper body BMD and increased testosterone levels in women with PCOS suggests that increased truncal mass may certainly be a consideration. To date, however, no larger epidemiological studies have examined this important aspect of the BMD-androgen excess relationship.
The association of polycystic-appearing ovaries on ultrasound with higher BMD, without clinically apparent androgen excess, has implications in the general population. In a recent study in an unselected female population, the prevalence of ultrasound-diagnosed polycystic ovaries was 14.2% (103). Therefore, some of the variability observed in BMD in apparently normal premenopausal women might be attributed to this often silent and rather common form of mild androgen excess.
Despite some compelling evidence, previous studies have not conclusively established that the higher bone density observed in some hyperandrogenic women is the result of the action of androgens. Instead, these findings may be at least partially the result of altered body composition, hyperestrogenism (primarily estrone), peripheral conversion of androgens, or an increase in free estradiol resulting from the suppression of SHBG by androgen excess and/or by hyperinsulinemia (8). DiCarlo et al. observed that patients with PCOS were apparently more estrogenized compared to other amenorrheic women, as evidenced by both a thicker endometrium and a greater uterine cross-sectional area (96).
It has also not been established, as sometimes articulated in the literature, that all women with PCOS are protected from osteopenia. PCOS is a complex disorder that manifests with different clinical presentations. Obesity, hyperandrogenemia, and hyperinsulinemia, the critical dimensions through which PCOS may positively influence bone mass, are not present in every PCOS subject.
Study design issues
Several study design issues are also of concern in the evaluation of the androgen-bone density relationship in normal and hyperandrogenic women. In a study seeking to establish an etiological relationship between an agent (androgens) and an outcome (BMD), a relatively homogeneous population with a documented exposure to the agent of interest is desirable. Clinical criteria such as hirsutism, rather than biochemical measurements, are often usd in the definition of case status in these studies. Hirsutism, however, may be present in women with both elevated and normal circulating androgens and may demonstrate considerable familial and ethnic variability (104). It is also not unique to any one particular hyperandrogenic disorder. Hirsutism can be observed in Cushings syndrome, acromegaly, congenital adrenal hyperplasia, and adrenal and ovarian tumors or with the administration of certain medications in addition to PCOS (105). These disorders present with very different endocrine profiles; the androgen-bone density relationship may therefore be highly confounded in these heterogeneous populations.
In certain studies women with hirsutism but normal androgen levels were included in the control group. As previously noted, hirsute women may have higher levels of androgen production with increased clearance rates and/or increased sensitivity to androgens. This presentation could represent either increased tissue exposure or responsiveness to androgens with apparently normal total circulating levels. Inclusion of these women in the control group, with reliance on a single, free testosterone measurement for classification, may inadvertently underestimate the actual androgen excess-bone density association in females. In the subject selection process, recruitment of case subjects with a specific hyperandrogenic disorder and control women with no clinical or biochemical evidence of androgen excess would address this limitation.
Moreover, in studies of hyperandrogenic disorders, the bone effects
attributed to androgens may be partially mediated by unmeasured
cytokines (interleukin-6), growth factors (IGF-I and transforming
growth factor-ß), or other hormones (insulin and free estrogens).
DHEAS, only weakly androgenic, may exert a direct influence on bone or,
perhaps more likely, an indirect influence, as a substrate in the
formation of more potent androgens
(5-androstenediol) or estrogens. DHEAS may
also stimulate other unmeasured bone-remodeling hormones such as IGF-I.
An increase in free circulating sex hormones and IGFs may result from a
reduction in SHBG and IGFBPs, respectively. An increase in free
estradiol (with normal or low total estradiol), particularly in obese,
hyperinsulinemic PCOS subjects, may also contribute to the effects
often ascribed to excess androgens. As previously noted,
hyperinsulinemia is a common finding among women with PCOS and has been
linked by Barrett-Conner and others to increased BMD. Aberrations in
IGF-I, IGF-II, or IGFBPs in PCOS (106, 107, 108, 109) and increases in free IGF-I
(110) may also mediate bone metabolism. Follistatin has recently been
linked to both PCOS (111) and bone remodeling. Therefore, in the
polycystic ovary model, the roles of insulin, the GH-IGF axis,
follistatin, and activin in mediating the potential relationship
between BMD and hyperandrogenism should be thoroughly investigated.
None of the previous epidemiological studies conducted in
hyperandrogenic women addressed these abnormalities and their potential
bone-related effects in specific subgroups (e.g. PCOS).
To infer causality, higher levels of circulating androgens should be clearly associated with higher BMD, independent of BMI and other factors associated with peak bone mass. In most studies, no attempt was made to identify a doseresponse relationship between the levels of circulating androgens and BMD. The demonstration of such a linear trend in the bone density-androgen relationship would strengthen the argument for an androgen effect on bone in women independent of estrogen.
Control of the confounding effects of age, body weight, body composition, and lifestyle factors (dietary calcium intake, caffeine and alcohol consumption, smoking, and physical activity) is essential to adequately assess the androgen-BMD association. Multicollinearity in the underlying variables, specifically between anthropometric characteristics (BMI, waist circumference, and intraabdominal fat) and sex hormones, is of concern and might be addressed through specific data reduction techniques, such as principal components or factor analysis. Case-control matching of weight/BMI and age (the two variables most significantly and consistently associated with BMD) would aid in minimizing biased comparisons between cases and controls and enhance evaluation of the independent effect of PCOS and androgens levels on bone density. In such a matched pair design, where BMI and age in cases and controls were balanced effectively across all subgroups, the assessment of the influence menstrual pattern on the BMD-androgen association would result in more precise estimates of subgroup-specific associations (112).
In attributing increases in bone mass to androgen excess, methodological issues in the assessment of bone density are also of concern. For example, dual energy x-ray absorptiometry is a two-dimensional estimate of volumetric bone density. Differences in bone size may, therefore, confound the PCOS-androgen-BMD association assessed by this technique (113). In addition, absorptiometry offers an integrated measure of all bone at a given site; the trabecular and cortical compartments cannot be separately distinguished (113). Therefore, the differential influence of androgen excess on these distinct bone compartments cannot be assessed by dual energy x-ray absortiometry. However, Buchanan, as previously noted, employed a true, three-dimensional technique (computed tomography) and observed a difference in trabecular BMD among women with androgen excess compared with controls, suggesting that 1) bone size may not be a deterministic factor in the observed differences; and 2) the trabecular compartment is apparently influenced in some fashion by the presence of androgen excess.
Finally, the powerful influence of estrogen in bone among premenopausal women renders assessment of the independent effects of androgens in these women difficult at best. Menstrual acyclicity may attenuate the relationship between androgens and bone density in premenopausal PCOS and control women, leading one to hypothesize a possible interaction between estrogen and androgen in the determination of BMD in younger affected females. If this is the case, although certain women with PCOS may be protected against the development of osteopenia, treatment interventions targeted at reducing androgen, estrogen, and insulin levels, without attention to their individual effects on bone mass and without subsequent restoration of normal cycling, may place a subgroup of these women potentially at risk for osteoporosis later in life.
Directions for future research
Regional body composition may have a profound influence on bone mass at specific sites in women. The relationship among regional BMD, androgens, and increased truncal mass should be more thoroughly examined. Moreover, little is known about the influence of androgens on bone size or bone quality in healthy women or among women with PCOS and androgen excess. Quantitated computed tomography and ultrasound evaluations of bone in PCOS may provide insights into the roles of androgens in influencing bone size and in preserving or improving the structural organization of bone and its material properties in women. Prospective studies of BMD in PCOS and androgen excess disorders are also needed to determine the relative importance among these women of androgens, estrogens, and insulin in bone mass attainment, maintenance, and loss. Both epidemiological and clinical investigations of bone density in preadolescent females with premature adrenarche/pubarche and in PCOS women during the menopausal transition would be of value in further defining the importance of the androgen-estrogen equilibrium in the modeling and remodeling of bone. Ultimately, comparisons of bone density and fracture rates among normal postmenopausal women and postmenopausal women with androgen excess will determine the relative protective influence of PCOS and androgens on later-life risk of osteoporosis and its deleterious health outcomes.
Received April 24, 2000.
Revised June 7, 2000.
Accepted June 27, 2000.
![]() |
References |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|