Although median levels of bone turnover are increased in postmenopausal women, it is unclear whether the low circulating levels of endogenous estrogen exert a regulatory role on these levels. This issue was evaluated by assessing the effect of a blockade of estrogen synthesis on bone turnover markers in 42 normal women (mean age ± SD, 69 ± 5 years) randomly assigned to groups receiving the potent aromatase inhibitor letrozole or placebo for 6 months. Letrozole treatment reduced serum estrone (E1) and estradiol (E2) to near undetectable levels (p < 0.0001). This treatment did not affect bone formation markers but, as compared with the placebo group, increased bone resorption markers (urine 24-h pyridinoline [PYD] by 13.3% [p < 0.05] and 24-h urine deoxypyridinoline [DPD] by 14.2% [p < 0.05]) and decreased serum parathyroid hormone (PTH) by 22% (p = 0.002). These data indicate that in late postmenopausal women even the low serum estrogen levels present exert a restraining effect on bone turnover and support the concept that variations in these low levels may contribute to differences in their rate of bone loss.
Women undergo two phases of involutional bone loss: an early rapid phase beginning at menopause that lasts for 6-10 years and a subsequent slow phase that lasts indefinitely.(1) The early phase can be prevented by estrogen replacement and is clearly the result of the cessation of ovarian function. The late slow phase traditionally has been attributed to age-related processes, especially to the development of progressive secondary hyperparathyroidism. However, two independent studies(2, 3) recently have shown that both the increased bone resorption and the secondary hyperparathyroidism that are associated with the slow phase also are consequences of continuing estrogen deficiency. These changes probably result from loss of estrogen action on increasing calcium absorption and renal calcium conservation.(1)
After the cessation of ovarian function at menopause, serum estrogen levels fall rapidly over a 2- to 3-year interval to reach low levels that then remain relatively constant through the remainder of life.(3) In postmenopausal women, circulating estrogen (E) levels result from aromatization of testosterone and weak adrenal androgens to estrogens in peripheral tissues.(4) Although the aromatase enzyme is present in many tissues including bone cells,(5, 6) most of the circulating estrogens in postmenopausal women are derived from aromatase activity in adipose tissue.
Earlier attempts to relate these low endogenous concentrations of serum estrogen to bone mineral density (BMD) or to bone loss in elderly women generally have given negative or inconsistent results. More recently, several studies using more sensitive immunoassays for serum E have been more successful. In 73 postmenopausal women, Slemenda et al.(7) related serum levels of estrone (E1), estradiol (E2), and sex hormone-binding globulin (SHBG) to BMD measurements at three scanning sites. They found weak direct correlations with serum E1 or E2 and SHBG in about one-half of the comparisons. Also, in three nested case-cohort studies in women ≥65 years from the Study of Osteoporotic Fractures, levels of serum E1, E2, and SHBG were related to the occurrence of hip and spine fractures,(8) BMD and fractures,(9) and changes in BMD at the calcaneus and hip.(10) In all three studies, direct correlations were found with serum E1 or E2, and inverse correlations were found with SHBG, for at least some of the dependent variables. Moreover, the data from all three of these studies suggested a threshold level for serum E2. Serum E2 levels of ≤5 pg/ml correlated with indices of osteoporosis, whereas values above this threshold did not. There were weak inverse correlations of serum E1 with hip and spine fractures.(8) Serum SHBG correlated directly with fractures in one study(8) and had a weak direct correlation with the rate of bone loss in another.(10) Thus, it is unclear from observational data whether the failure to find a clear relationship across the entire range of values was related to a true threshold effect or was because of statistical noise in the measurement of these low residual levels of sex steroids. Moreover, in another nested case-cohort study from the EPIDOS study of French women ≥75 years,(11) no relationship was observed between serum E2 or E2/SHBG and the occurrence of hip fractures for those subjects in the lower three quartiles. Subjects in the highest quartile were at lower risk than those in the lowest quartile, but this protection was lost after adjustment for body weight. However, because the data were corrected for differences in body weight, they do not necessarily exclude the possibility that higher levels of serum estrogen produced in adipose tissue were protective.
Thus, the relationship between low residual levels of serum E and abnormal bone and calcium metabolism in elderly postmenopausal women is still unclear. In elderly women, all circulating estrogen is produced by aromatization of androgens in peripheral tissues.(12) If the low levels of serum E in late postmenopausal women do have a restraining effect, we reasoned that a reduction in estrogen synthesis would result in further increases in bone turnover. Thus, we conducted a randomized, placebo-controlled, 6-month prospective trial in normal elderly women treated with letrozole, a selective and potent aromatase inhibitor, or placebo to test this hypothesis.
MATERIALS AND METHODS
After approval by the Mayo Institutional Review Board, 42 normal women, with a median age of 69 years (range, 60-76 years) who were at least 10 years postmenopausal were recruited by announcements and gave informed consent for the study. Subjects were interviewed for medical history; underwent a physical examination; had a prestudy laboratory assessment that included a complete blood count (CBC), chemistry group, platelet count, and urinalysis. Subjects with significant medical diseases such as renal failure (creatinine >1.5 mg/dl), malabsorption, active malignancy, congestive heart failure, or obesity (>20% of normal body mass index [BMI]) were excluded. Nutritional status was assessed using a food frequency questionnaire by a trained dietitian. No subject was taking any medication known to affect calcium and bone metabolism including calcium supplements of >500 mg/day, vitamin D supplements of >1000 U/day, sodium fluoride, bisphosphonates, calcitonin, estrogen, suppressive dose of thyroid hormones, glucocorticoids, calcium channel blockers, thiazides, anticonvulsants, or lithium. All subjects had BMD measured at the lumbar spine (L2-L4). Patients with osteoporosis were excluded based on previous fractures of the vertebrae, hip, or a value for BMD that was ≤ −2.5 SD of the premenopausal mean.
We performed a prospective, double-blind, placebo-controlled clinical trial. Subjects were assigned randomly to one of two groups: those receiving the aromatase inhibitor letrozole (Novartis Pharmaceuticals, Inc., Basel, Switzerland), 2.5 mg orally daily, and those receiving a matching placebo. They were studied as outpatients in the General Clinical Research Center of the Mayo Clinic and Mayo Foundation. Throughout the study period, subjects were maintained on their habitual calcium intake. Fasting blood samples and 24-h collections of urine were obtained at baseline and after 6 months of treatment. Because the bone remodeling cycle is completed in 4-6 months,(13) we felt that this interval would be sufficient to allow all remodeling foci to reach new steady-state conditions. Measurements of the blood samples included serum calcium, phosphorus, creatinine, E1, E2, osteocalcin (OC), bone-specific alkaline phosphatase (bone ALP), both markers of bone formation, and intact parathyroid hormone (PTH). We also measured total pyridinoline (PYD) and total deoxypyridinoline (DPD), both markers of bone resorption, and for urinary calcium in the 24-h urine samples. Dietary calcium intake was estimated from a 5-day diet diary, which has been shown to provide a good estimate of habitual intake.(14) For the markers of bone turnover (OC, bone ALP, PYD, and DPD), collections were made on 2 consecutive days to minimize intraindividual variability. Compliance was assessed by pill counts.
Serum E1 and E2 were measured by high-sensitivity radioimmunoassays: E1 was measured by a kit from Diagnostic Systems Laboratories (Webster, TX, USA) and E2 by Diagnostic Products Corp. (Los Angeles, CA, USA). Inter- and intra-assay CVs were <10%. The detection limits for these assays were 1 pg/ml for serum E1 and 2 pg/ml for serum E2. Serum OC was measured by immunoradiometric assay (inter- and intra-assay CVs, 13.6% and 3.9%, respectively; CIS US, Bedford, MA, USA). Serum bone ALP was measured by antibody capture assay (inter- and intra-assay CVs, 7.6% and 2.8%, respectively; Metra Biosystems, Mountain View, CA, USA). Total PYD and DPD were measured in acid hydrolysates of urine by high-performance liquid chromatography with fluorometric detection after preliminary fractionation on cellulose columns using an automated system.(15) Inter- and intra-assay CVs were 5.1% and 3.8% and 6.2% and 4.2%, respectively.(15) Serum intact PTH was measured by ELISA (inter- and intra-assay CVs, 6.3% and 6.2%, respectively; Diagnostic Systems Laboratories). Chemistry group was determined by routine automated methods (Hitachi 911 Analyzer; Boehringer Mannheim, Indianapolis, IN, USA). CBC and platelet count were performed by electronic method (Coulter STKS; Coulter, Miami, FL, USA).
Because the data were found to not have a Gaussian distribution, nonparametric statistics were used. The Wilcoxon rank-sum test was used to compare the baseline and the 6-month baseline differences between the letrozole and placebo treatment groups. The Spearman correlation coefficient was used to relate baseline characteristics with each other. A value of p < 0.05 was considered significant. Values are expressed as medians and ranges. Undetectable samples were assigned an arbitrary value below the limit of detection. All analyses were performed on the Statistical Analysis System Software Program (SAS Institute, Inc., Cary, NC, USA).
The clinical and biochemical characteristics of the two study groups before treatment were similar (Table 1), indicating that the randomization process had been successful. For merged values of both groups, there were the following correlations at baseline: a positive correlation between E1 and E2 (r = 0.89; p = 0.0001), a trend for negative correlation between E1 and OC (r = −0.25; p = 0.10), a negative correlation between E1 and DPD (r = −0.37; p = 0.02), a positive correlation between E1 and lumbar spine BMD (r = 0.32; p = 0.04), and a positive correlation between E2 and BMI (r = 0.30; p = 0.05). Serum E1 and BMI levels were correlated positively (r = 0.14; p = 0.36). No other correlations were significant.
Table Table 1. Median and Range for Baseline Values in the Two Treatment Groups
The effects of treatment were assessed by comparing letrozole-treated and the placebo-treated groups (Table 2). At baseline, all samples of serum E1 and all but one sample of serum E2 were above the detection limit. Serum E1 (p < 0.001) and serum E2 (p < 0.001) levels were significantly reduced in the subjects receiving letrozole. After letrozole treatment, 5 of the 22 patients had undetectable values for serum E1 and all patients had undetectable values for serum E2. Serum PTH decreased by 22% in the letrozole group and increased by 3% in the placebo group; the percentage change between groups was significant (p = 0.001). For bone formation markers, there were no significant changes in serum OC or bone ALP. For bone resorption markers, urinary PYD increased by 13.3% (p = 0.03) and urinary DPD increased by 14.2% (p = 0.03 both as compared with the placebo group). Although the two groups did not differ with respect to BMI, total fat mass is known to affect estrogen production in postmenopausal women. Thus, we reanalyzed the significance of changes in serum PTH and bone resorption markers after adjusting for BMI. After reanalysis, the changes in serum PTH (p < 0.001), urine PYD (p = 0.028), and urine DPD (p = 0.031) remained significant. Changes in individual values for urine PYD as a result of placebo or letrozole treatment are given in Fig. 1 and corresponding changes for urine DPD are given in Fig. 2. For each figure, the upper panel gives absolute values at baseline and after 6 months of treatment and the lower panel gives the percentage change between the two visits. None of the components of the serum lipids changed significantly with treatment. There was no significant correlation between values for the changes in serum PTH, E1, or E2 and the changes in bone resorption markers. Treatment was well tolerated and there were no symptoms reported in either the letrozole or the placebo group.
Table Table 2. Comparison of 6-Month Minus Baseline Values for Both Groups
Estrogen acts directly on bone through high-affinity receptors in osteoblasts and osteoclasts.(1) Its major effect is to decrease bone turnover and to maintain a balance between bone formation and bone resorption. The loss of this balance because of estrogen deficiency is now believed to be the major cause of osteoporosis in postmenopausal women and may contribute to bone loss in aging men.(1)
Previous tests of the hypothesis that low levels of estrogens actively regulate bone turnover in postmenopausal women have been made using correlation data from observational studies. However, correlation cannot establish causality. For this, intervention studies are necessary. The recent development of safe and effective aromatase inhibitors has made it possible to test the hypothesis directly by assessing the effect of blocking the peripheral conversion of androgens to estrogens on levels of bone turnover markers. In the current studies, we used letrozole, a highly selective, potent aromatase inhibitor.(16) We confirmed the potency of this agent by showing that its administration could reduce the residual production of estrogen virtually to undetectable levels. We measured serum estradiol using a new ultrasensitive immunoassay that was able to detect concentrations as low as 1 pg/ml for serum E1 and 2 pg/ml for serum E2. Before letrozole administration, serum values for E1 and E2 were detectable in almost all subjects. We found that letrozole treatment was well tolerated and produced no side effects.
In late postmenopausal women, we found that letrozole treatment significantly increased bone resorption markers without affecting bone formation markers. The failure to increase bone formation in response to an increase in bone resorption is consistent with the age-related impairment in osteoblast work capacity that has been shown previously with bone histomorphometry(17) and biochemical marker(18) measurements. In this study, presumably osteoblast activity was already maximal because bone formation markers did not increase further in response to increases in resorption. However, we cannot exclude the possibility that these markers would have increased if the period of observation had been longer than 6 months.
Although bone resorption markers increased to only 14% above baseline levels, this was clearly physiologically significant as evidenced by the compensatory decrease of 22% in serum PTH. Presumably, the increase in bone resorption was caused by loss of the well-established direct restraining effect of estrogen on bone cell function. This mechanism should be distinguished from that of secondary hyperparathyroidism associated with an intake of calcium that is insufficient to offset loss of the extraskeletal effects of estrogen on intestinal calcium absorption and renal calcium conservation.(1) In this study, there was no change in urinary calcium excretion in response to letrozole treatment.
We can only speculate on the reason for the paradox that serum PTH decreased rather than increased after letrozole treatment in late postmenopausal women. One possible explanation is suggested by the hypothesis of Frost(19) that an innate sensing system for biomechanical signals exists in bone (“the mechanostat”) and that estrogen deficiency reduces this sensing. Recently, this hypothesis has been supported by in vitro studies showing that the response of osteoblastic cells to mechanical strain and to estrogen administration share at least part of the same signaling system.(20) Thus, immediately after menopause, bone mass may be sensed by the mechanostat as inappropriately high. This leads to rapid bone loss that continues until a new steady state is reached where strain signals again are sensed as appropriate and bone loss ceases. If so, the exacerbation of the preexisting estrogen deficiency by letrozole therapy again could reset the mechanostat and induce a new phase of rapid bone loss that would continue until a new steady state appropriate to the new serum E level is once again reached. In contrast, the late slow phase of bone loss would continue indefinitely because it is mediated by increased PTH secretion rather than by loss of direct skeletal effects of estrogen that reset the mechanostat.
The significance of the observed increase in bone resorption is better appreciated when it is expressed in relation to the remodeling imbalance that is present in late postmenopausal women. This can be estimated from the proportional increases in levels of resorption markers minus formation markers present in late postmenopausal women over levels present in premenopausal women. Based on previously published data,(3, 18) the remodeling imbalance induced by menopause is ∼25%. Thus, the letrozole-induced increase in bone resorption of 14% with no change in bone formation would be roughly equivalent to about one-half of the postmenopausal remodeling imbalance (i.e., the gap between levels of bone resoption and levels of bone formation). If so, virtual elimination of the low residual levels of serum E1 and E2 postmenopausal, which represent only 40% and 7%, respectively, of the premenopausal medians,(21) had a disproportionately large effect of bone remodeling. Although serum E1 levels had declined less than serum E2, the latter steroid is 3- to 4-fold more potent. However, the osteoblast also contains aromatase.(5, 6) Thus, the effect of letrozole on sharply reducing circulating estrogen levels could have been enhanced by blockade of synthesis of estrogen within the osteoblast.
Our findings are consistent with the recent observations from the Study of Osteoporotic Fractures. In 269 women ≥65 years of age, Stone et al.(10) found that women with levels of serum E2 of <5 pg/ml had an 8-fold greater rate of bone loss from the hip than did women with levels ≥10 pg/ml. Also, in another study from the same cohort, Cummings et al.(8) found that women with serum E2 values below the detection limit of 5 pg/ml had a 2.5-fold greater risk than did those with detectable values. This value is similar to the median pretreatment value for serum E2 of 6.5 pg/ml for the postmenopausal women in our study. Thus, our data, when combined with those from the two publications from the Study of Osteoporotic Fracture,(8, 10) strongly suggest that the effect on decreasing serum estrogen levels on increases in bone resorption is not linear. Further decreases of serum E2 below 10 pg/ml and particularly below 5 pg/ml appear to have a disproportionately larger effect on increasing bone resorption and possibly also on increasing bone loss and fracture risk.
Although the optimal dosage of estrogen for prevention of bone loss in early postmenopausal women has been believed to be 0.625 mg/day of conjugated estrogen or its equivalent,(22) recent trials of about one-half of this dosage plus calcium supplements have resulted in comparable efficacy without inducing endometrial hyperplasia.(23, 24) Our finding that low levels of serum estrogens have a disproportionately larger regulatory effect on bone resorption provides theoretical support for the use of this regimen. Because there appears to be a dose relationship between estrogen exposure and risk of breast cancer(25) and because substantial side effects still occur on standard dosage regimens,(26) low-dose estrogen may improve the tolerability and acceptance of estrogen-replacement therapy. Although the low dosages of estrogen replacement that prevent bone loss will result in higher levels of circulating estrogen than the low levels observed in these untreated women, it is obvious that these low levels are exerting a restraining effect on bone loss because reducing them to undetectable levels increased bone resorption.
Letrozole administration is approved for the treatment of advanced breast cancer in postmenopausal women. For such women, our current findings suggest that it would be prudent to monitor bone density during treatment and to begin antiosteoporosis therapy if decreases in bone density are detected. Also, estrogen administration has been shown to have a beneficial effect on serum lipids,(27) and, in principle, letrozole therapy could have adverse effects. Therefore, it was reassuring that we found no significant effects of treatment on serum lipids over the 6-month duration of our study.
In conclusion, we have found that experimental reduction of the low residual levels of serum E1 and E2 present in late postmenopausal women to virtually undetectable levels is associated with further increases in the already high levels of bone turnover. Thus, serum E1 and E2 levels that were 40% and 7%, respectively, of median levels of premenopausal women appear to inhibit actively bone resorption. Moreover, there appears to be no lower levels of serum E1 and E2 that do not regulate bone resorption. Our data also are consistent with the hypothesis that individual variability in the residual levels of estrogen may play a contributing role to the pathogenesis of osteoporosis, even in elderly postmenopausal women.
We thank K. Egan and E. Atkinson for assistance in statistical analysis; J.M. Peterson for assistance in data management and graphics; J.M. Muhs for recruiting the patients; the nurses of the General Clinical Research Center, Mayo Clinic; the Mayo Foundation for performing the samplings; L. Oenning, H.M. O'Connor, and S. Nayar for nutritional assessment; P.S. Helwig for bone densitometry; and C. McAlister, R.A. Soderberg, D.M. Hanson, S.H. Showalter, and D.W. Heser for technical assistance. This study was supported by grants AG04875 and M01 RR00585 from the National Institutes of Health (NIH) and the U.S. Public Health Service (USPHS).