| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Original Studies |
Departments of Internal Medicine I (T.S., B.S., H.B, G.S., R.Z., J.P.), Gynecology (I.D.), and Pathology (B.K.) University of Heidelberg; Rehabilitation Center Berchtesgadener Land (S.S.), Schönau am. Königssee; and Lilly Deutschland GmbH (W.F.B.), Bad Homburg, Germany
Address all correspondence and requests for reprints to: T. Seck, Department of Internal Medicine I, University of Heidelberg, Luisenstraße 5, D-69115 Heidelberg, Germany.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In summary, our study suggests the following. 1) The concentration of IGF-I in cancellous bone undergoes age-related decreases that are similar to those of circulating IGF-I. 2) Menopause has no effect on this age-related decline. 3) Physiological differences in bone matrix IGF-I are associated with differences in iliac crest cancellous bone volume. 4) Bone matrix IGF-I is a better predictor of cancellous bone volume than circulating IGF-I. 5) The role of IGF-II in human bone tissue is clearly distinct from that of IGF-I.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Between 1991 and 1996, 533 Jamshidi bone biopsies from the anterior-superior iliac crest were obtained during surgery for breast cancer at the Gynecology Department of the University of Heidelberg. Cortical bone was separated from cancellous bone with a scalpel blade under a dissection microscope. The biopsies were part of a prospective study that examines the impact of local bone growth factors on the future development of bone metastases. For the present study we only analyzed bone samples from women who had an early tumor stage (T1, T2), no metastases at the time of surgery, normal serum calcium concentrations, serum concentrations of PTH-related peptide below the detection limit of 5 pmol/L by RIA (23), and no other evidence of systemic tumor disease. Of the 533 women, 192 women were premenopausal [mean age, 42.8 ± 6.4 (SD) yr; range, 25–57 yr], 235 were postmenopausal and did not receive hormonal replacement therapy (mean age, 61.9 ± 8.4 yr; range, 37–83 yr), and 76 had received hormonal replacement therapy before surgery (mean age, 55.8 ± 5.9 yr; range, 39–72 yr). In the remaining 30 women, menopausal status could not be reliably determined. Blood samples were obtained in a subset of 164 women after overnight fasting before surgery for the measurement of osteocalcin, bone-specific alkaline phosphatase (BAP), and circulating IGF components. The study was reviewed and approved by the institutional review board. Bone and serum samples were stored at -80 C until further analysis.
Bone matrix extraction
The skeletal concentrations of IGF-I, IGF-II, and IGF-binding protein-3 (IGFBP-3) were determined in extracts from a single bone biopsy from the left iliac crest. The biopsies were immersed in liquid nitrogen and mechanically crushed into small fragments of several millimeters in diameter. The fragments were repeatedly washed in cold distilled water until the washings were free of blood and defatted in cold isopro-pylether. The defatted bone fragments were then ground into smaller particles (40–60 µm) in a liquid nitrogen-cooled freezer mill (Raetsch, Haan, Germany). The yield of mineralized bone powder varied between 20 mg and 50 mg, depending on biopsy size.
Duplicate extractions of the IGFs were performed from 15 mg of
lyophilized bone powder each, as described previously (24), with the
exception that the bone extracts were redialyzed against 10
mM acetic acid instead of PBS. In 34% and 7% of the
samples, less than 15 mg of bone powder were available for one or both
extractions, respectively. The minimum amount of bone powder for a
single extraction was 6 mg. The lyophylized bone powder was placed in
microcentrifuge tubes (Twist-lock, Eppendorf, Hamburg, Germany), and
1.7 mL extraction solution (see below) was added. Spectrapor 3 dialysis
tubing (3.5-kDa cutoff; Spectrum Medical Industries, Houston, TX) was
then placed over the tube opening and secured by a melted out tube cap,
as described by Overall et al. (25). The tubes were then
inverted and firmly fixed upside down in a circular rack floating on
top of the extraction solution (600 mL for 30 tubes). Extraction was
achieved by dialysis against 0.05 M tetrasodium EDTA
(Serva, Heidelberg, Germany), 4 M guanidin-HCL (Sigma,
Deisenhofen, Germany), 30 mM Tris (Merck, Darmstadt,
Germany), and 1 mg/mL BSA (Sigma; RIA grade) at a pH of 7.4. The
following protease inhibitors were added to the extraction solution: 5
mM benzamidine-HCl, 1 mM phenylmethylsulfonyl
fluorid, and 0.1 M 
-aminocaproic acid (Sigma). Dialysis
was carried out at 4 C under constant stirring for 24 h. After
extraction, the samples were redialyzed against 10 mM
acetic acid (pH 7.4) for 72 h. Dialysis medium (600 mL for 30
tubes) was replaced every 24 h. The supernatant extracts were
recovered after centrifugation at 5000 x g for 10 min
and stored at -80 C until assayed for growth factor activity.
The recovery of the IGF components was determined by spiking bone powder from the human iliac crest with known amounts of purified IGF-I, -II, or IGFBP-3 before extraction. The mean recovery of IGF-I, -II, and IGFBP-3 was 98%, 92%, and 95%, respectively. Coefficients of variance of the extraction and measurement procedure were determined by extracting three different pools of human bone matrix containing low, medium, and high concentrations of the three IGF components over a period of time and assaying the concentration of these growth factors individually by duplicated measurements for IGF-I, -II, and IGFBP-3. The combined interassay variance for repeated extractions was less than 15% for all three IGF components. Preceding experiments had shown that the yield of extractable growth factors from human bone could not be improved by reextraction. Furthermore, no additional growth factors were released when the residual bone matrix was digested with highly purified collagenase, suggesting a complete recovery of the endogenous growth factors.
Measurement of serum and bone matrix IGF components
IGF-I in the bone matrix extracts and in serum was measured using a polyclonal rabbit antibody against human IGF-I (Mediagnost, Tübingen, Germany) and recombinant human IGF-I (GroPep, Adelaide, Australia) as a tracer and standard (11, 24). IGFBP artifacts were avoided by initial dissociation of IGF from the binding proteins present using an acidic buffer. Reassociation was then blocked by inducing binding protein saturation through the addition of excess IGF-II (Mediagnost) (26). This assay method has been extensively validated and shows good correlations with measurements obtained by acidic size-exclusion chromatography (27). Moreover, previous experiments had shown that with this procedure unlabeled IGF-I is completely recovered after preincubation with the extracts, and that there is no change in the measurement of endogenous IGF-I in the presence of up to 10 ng/mL IGFBP-3 (24). IGF-II cross-reactivity in this RIA was < 0.05%. Interassay variation was 6%. IGF-II was measured by RIA using a polyclonal rabbit antibody against human IGF-II (Mediagnost). Measurements were calibrated against recombinant human IGF-II (GroPep). Assay conditions were similar to those described for the IGF-I RIA, except that excess IGF-I was used to block interferences caused by IGFBPs. Validation was performed as described for IGF-I. In particular, there was no change in the measurement of endogenous IGF-II in the presence of up to 10 ng/mL IGFBP-3. Cross-reactivity of the IGF-II antibody with IGF-I was less than 0.05%. Interassay coefficient of variation was 6.5%. IGFBP-3 was measured using a commercial enzyme-linked immunosorbent assay (ELISA) from Diagnostic Systems Lab. (Webster, TX). Interassay coefficient was 7%. Cross-reactivity with IGF-I, -II, and IGFBP-1 was less than 0.05%. Extracts from two independent extractions of bone powder from the same iliac crest biopsy were assayed in duplicate for all three IGF components.
The residual matrix after extraction was resuspended twice in distilled water, lyophilized, and weighted. The concentration of the IGFs in the bone matrix was expressed per milligram mineralized bone or per milligram dry weight of the residual bone matrix, as described previously (24). The normalized values of the two separate extractions were then averaged. Both normalized indices were highly correlated (r > 0.9 for all IGF components).
Bone histomorphometry
Histomorphometric measurements at the iliac crest were independently performed in two Jamshidi biopsies from the left and right iliac crest. The biopsies were 2 mm in diameter and had an average length of 3–4 cm. The means of the histomorphometric indices from the left and right iliac crest were averaged. All bone samples that had been assigned to histomorphometry were fixed in 70% ethanol and embedded undecalcified in methyl methacrylate. Three-micrometer sections were cut with a Jung Supercut microtome (Cambridge Instruments, Nussloch, Germany), and stained according to a modified Goldner method. Cancellous bone volume [cancellous bone volume in percent of tissue volume (BV/TV%)] was determined by image analysis (Digithurst, Nürnberg, Germany) at a 25-fold magnification. The coefficient of variance for BV/TV% between the left and the right iliac crest was 14.6%. The percentages of the cancellous bone surface covered with osteoblasts (OB.S/BS) and undergoing active bone resorption (OC.S/BS) were evaluated with a Merz grid at a 100-fold magnification. All measurements were blindly performed without knowledge of the corresponding IGF concentrations.
Measurement of serum osteocalcin and BAP
Osteocalcin was measured by chemoluminescence (BRAHMS, Berlin, Germany) using an antibody directed against the N-terminal region of the osteocalcin molecule. The assay had an intraassay precision of 4.7% and an interassay precision of 7.5%, respectively. BAP was determined by ELISA (Metra, Mountain View, CA). Intra- and interassay coefficients of variance were 3.9% and 7.6%, respectively. All samples were measured in duplicates and the means calculated.
Statistical analysis
Differences between groups were analyzed by one-way ANOVA using Fisher’s test for multiple comparisons. Simple and partial Pearson’s correlations were used to describe the associations between two continuous variables. All analyses were performed with the Statistical Analysis System software program (SAS Institute, Cary, NC).
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The concentration of IGF-I in iliac crest bone was inversely
related to age. IGF-I concentrations in the eighth decade were 35%
lower than in the third decade (Fig. 1A
and Table 1). Age-related differences in
IGF-I of a similar magnitude were observed in serum (r = -0.18,
P = 0.02, n = 164; r = -0.30
P < 0.0001 when three outliers were excluded). The
slope of the inverse association between bone matrix IGF-I and age was
most pronounced in premenopausal women, but a weak inverse association
between bone matrix IGF-I and age was still present in postmenopausal
women (Fig. 1A). This difference in the slope of the regression of
IGF-I on age was also present when the women were arbitrarily
stratified into women less than 50 yr of age (r = -0.21,
P = 0.003, n = 200) and more than 50 yr of age
(r = -0.13, P = 0.02, n = 333). There were
no significant differences in bone matrix IGF-I concentrations between
age-matched 45-to 55-yr-old premenopausal women, postmenopausal women,
and postmenopausal women on hormonal replacement therapy (HRT) (Table 2), suggesting that the age-related
decline in bone matrix IGF-I is independent from menopause.
|
|
|
). As opposed to IGF-I, serum IGF-II concentrations
failed to be associated with age (r = +0.04, P =
0.64). Bone matrix IGF-II concentrations were even slightly positively
correlated with age (Fig. 1B). Menopause had no effect on bone matrix
IGF-II concentrations (Table 2). The concentration of bone matrix IGFBP-3 was weakly inversely related to age in premenopausal women (r = -0.16; P = 0.03), but failed to be associated with age in postmenopausal women with HRT (r = -0.05, P = 0.65) and without HRT (r = +0.05, P = 0.41).
Association between bone matrix IGFs and bone turnover
Bone matrix IGF-I was positively associated with the percentage of
the osteoblast-covered cancellous bone surface, the percentage of the
bone surface undergoing active bone resorption, and with serum
osteocalcin levels (Table 3). With the
exception of the osteoclast surface, these associations were
independently observed both for pre- and postmenopausal women (data not
shown). Interestingly, bone matrix IGF-I failed to be associated with
serum BAP, despite the fact that BAP was significantly correlated with
serum osteocalcin (n = 166, r = +0.54, P <
0.0001) and with the osteoblast-covered surface (n = 96, r =
+0.28, P = 0.006). In contrast, the strength of the
relationship of intact serum PTH (r = +0.34, P <
0.0001) and free serum T3 (fT3) (r =
+0.20, P = 0.001) with BAP in the same sample
collective even exceeded that of the relationship of PTH (r =
+0.25, P < 0.0001) and fT3 (r =
+0.12, P = 0.06) with osteocalcin (n = 246
including samples for which no bone matrix IGF-I measurements were
available).
|
). Association between bone matrix IGFs and cancellous bone volume at iliac crest
Cancellous bone volume at the iliac crest was inversely related to
age (n = 466, r = -0.15, P = 0.001). The
concentration of bone matrix IGF-I was positively associated with the
cancellous bone volume (Fig. 2 and Table 4). This association was independently
observed both for pre- and postmenopausal women. Adjustment for age had
no substantial effect on the strength of the association (Table 4). On
the other hand, adjustment for IGF-I partially explained the
age-associated bone loss (r = -0.11, P = 0.02
after adjustment for IGF-I). Based on the estimates of the linear
regression equation, women with an IGF-I concentration 2 SD
above the mean (0.6 ng/mg) had a 20% higher average cancellous bone
volume than women with an IGF-I concentration 2 SD below
the mean (0.13 ng/mg).
|
|
). Association between bone matrix IGF-I and serum IGF-I
The relationship between the concentrations of IGF-I in the bone
matrix and in the circulation was examined in a subset of 164 women for
whom both serum and bone matrix IGF measurements were available. In
both compartments, IGF-I concentrations were strongly correlated with
the respective concentration of IGFBP-3 in bone matrix or serum (Table 5). There was no significant association
between the serum levels of IGF-I and the concentration of IGF-I in
bone matrix (Table 5). However, a weak positive relationship between
the two variables became apparent after adjustment of the serum IGF-I
concentration for the concentration of serum IGFBP-3. Similar to our
findings in the entire sample collective, bone matrix IGF-I in this
subset of women was significantly associated with the percentage of the
osteoblast-covered bone surface, serum osteocalcin levels, and
cancellous bone volume. In contrast, neither serum IGF-I (Table 5) nor
serum IGF-II and IGFBP-3 were correlated with any parameter of bone
turnover or with bone volume, regardless whether serum IGF-I or IGF-II
were adjusted for circulating IGFBP-3 or not (data not shown).
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
There are some limitations to our study because of the fact that the biopsies were obtained from women with breast cancer and therefore may not be entirely representative for the normal female population. Although our selection criteria made it unlikely that the tumor itself had any influence on bone metabolism at the time of surgery; breast cancer has been shown to be associated with a higher bone mass, presumably because of the linked association with higher estrogen exposure (28, 29). However, because this association is weak, and our sample contained a sufficiently large number of women with both low and high bone volume, this bias is unlikely to invalidate our results. Moreover, because 12% of all women will have breast cancer diagnosed in their lifetime (30), this is a rather common risk factor.
In accordance with previously reported findings at different skeletal sites (19, 21, 22), we observed an inverse relationship between bone matrix IGF-I at the iliac crest and donor age. This decline in bone matrix IGF-I with age may be related to the decline in the general activity of the GH/IGF axis, but direct age-related decreases in IGF-I production by bone cells may also be involved (31). Interestingly, the slope of the association between bone matrix IGF-I and age was most pronounced in premenopause, suggesting that the loss of skeletal IGF-I to a large part already occurs during the first half of life. Nevertheless, it should be pointed out that based on the estimates of the linear regression analysis, age explained only 4.4% of the total variability of bone matrix IGF-I, indicating that at least for this particular sample collective, age was only a modest contributor to the overall variability in skeletal IGF-I concentration.
The interaction between estrogens and the GH/IGF-axis is complex. Estrogens augment the pituitary secretion of GH but induce a relative resistance to the stimulatory effect of GH on IGF-I production (32). The net effect of estrogens on serum IGF-I concentrations, however, depends on the route of administration. Whereas oral estrogens decrease systemic IGF-I levels, presumably via a direct hepatic effect, transdermal delivery of estrogens has no significant effect on circulating IGF-I levels. With respect to bone, estrogens have been shown to augment IGF-I production in rodent osteoblast cell cultures (33, 34). In contrast, ovariectomy has no effect on (35, 36) or actually increases bone IGF-I (37, 38) in the rat. In the present study, we failed to observe any significant difference in the bone matrix concentration of IGF-I between age-matched 45-to 55-yr-old early postmenopausal women without estrogen replacement, as compared to postmenopausal women on HRT and premenopausal women. Our study therefore suggests that the decline in bone matrix IGF-I with age in women is independent of human menopause, and that physiological concentrations of estrogens have no major impact on the skeletal concentration of bone matrix IGF-I.
The concentration of IGF-I in the bone matrix was significantly positively associated with histomorphometrical parameters of bone formation and bone resorption. Interestingly, the matrix IGF-I concentration was also positively related to circulating osteocalcin levels, but not to serum BAP. We can only speculate on the reason for these differential relationships. Osteocalcin measurements may provide a better reflection of the bone forming surface than BAP measurements, as suggested by a stronger association of the osteoblast-covered bone surface with serum osteocalcin (r = 0.57, P < 0.001) than with serum BAP (r = 0.28, P = 0.006). However, in the same sample collective, the strength of the association of intact serum PTH and fT3 with BAP even exceeded that of the association between these two hormones and serum osteocalcin. IGF-I is known to promote osteoblast differentiation (39) and osteocalcin is a marker for a more advanced osteoblastic phenotype than BAP. It is therefore also possible that the better association of bone matrix IGF-I with serum osteocalcin is in part related to this differentiation-promoting action of IGF-I.
The concentration of bone matrix IGF-I was positively associated with iliac crest cancellous bone volume. Our study suggests that physiological variations in the local IGF-I concentration in human bone tissue may only explain a small percentage of the overall variability of normal bone mass. Nevertheless, based on the estimates of a linear regression analysis, a woman with an IGF-I concentration at the upper limit of the normal distribution (+2 SD) will have a 20% higher cancellous bone volume than a woman at the lower limit (-2 SD), indicating that local IGF-I concentrations outside of the normal range may be associated with clinically relevant differences in bone mass.
The bone matrix concentration of IGF-II was 8-fold higher than that of IGF-I. However, despite similar anabolic effects of the two IGF species on bone cells in vitro (40), only IGF-I was associated with bone volume in the present study. Together with the finding that there is no age-related decline in bone matrix IGF-II, this points to a very different role of this IGF species in human bone tissue as compared with that of IGF-I. Recent experiments in primary cultures of rat osteoblasts suggest that IGF-II may be involved in the early strain-related response of osteoblasts in loading-related modeling (41).
The concentration of IGF-I and -II in human bone matrix is potentially determined by a number of different variables. These include the circulating concentration of IGF-I and -II, the local production of IGF-I and IGF-II by cells within the bone tissue, and the local production of IGFBPs and binding protein proteases. Evidence for the involvement of the IGFBPs as determinants of the skeletal IGF concentration is provided by the finding that the concentration of IGFBP-3 in the bone matrix explained 19% of the bone matrix IGF-I concentration. In a recent study by Nicolas et al. (20), the skeletal concentration of IGFBP-5, another IGFBP that is abundant in bone matrix and binds to hydroxyapatite, even explained 39% of the variability of IGF-I in human cortical bone matrix. Unfortunately, sensitive assays for IGFBP-5 or other IGFBPs apart from IGFBP-3 were not available to us in the present study.
Circulating IGF-I was only weakly positively associated with bone matrix IGF-I, and this association became only apparent after adjustment of serum IGF-I for the concentration of its major binding protein in circulation, IGFBP-3. Thus, the amount of IGF-I in bone tissue that is provided by circulating IGF-I may depend on the ratio of IGF-I to IGFBP-3 rather than on the total concentration of serum IGF-I. Indeed, only free IGF-I is thought to be able to cross the capillary endothelial barrier (42). However, even the ratio of IGF-I to IGFBP-3 explained only 6% of the variability of bone matrix IGF-I, suggesting that circulating IGF-I levels are only a poor predictor of skeletal IGF-I concentrations at the iliac crest.
The lack of correlation between circulating IGF components and trabecular bone volume in the present study is in apparent contradiction to findings from several previous studies reporting significant positive associations between circulating IGF components and bone mass or bone turnover. Reed et al. (10) showed positive associations between circulating IGF-I and several histomorphometric parameters of bone formation in patients with idiopathic osteoporosis. Although these findings were observed in a small number of patients, it is possible that in particular subsets of patients with bone diseases, circulating IGF-I levels may explain variability in bone turnover to a larger extent than in individuals without known bone disease.
Sugimoto et al. (16) recently reported striking associations between circulating IGF-I levels and bone mineral density (BMD) measurements at all skeletal sites in postmenopausal women. However, in a separate study from the same group, only weak associations were observed between serum IGF-I and forearm BMD in perimenopausal women (17). In the few studies that have examined the relationship between circulating IGF components and parameters of bone metabolism in population-based settings, only weak associations were observed that could be further attenuated by adjustment for age and/or other confounding variables (11, 12, 13). These weak positive associations may have been missed in the present study because of the larger variability of the histomorphometric measurement as compared with BMD measurements.
We should mention that the changes in circulating IGF-I levels with age were rather modest. In particular, IGF-I serum levels in the younger women were somewhat lower that those reported in previous studies (43, 44, 45, 46). The reason for this is unclear. There was no obvious difference in fasting or physical condition compared with the older women. The association with age became considerably stronger when three IGF-I-outliers were excluded from analysis. Nevertheless, this did not result in any improvement of the associations between circulating IGF-I levels and bone turnover or volume. Our study therefore clearly suggests that, at least for trabecular bone and the iliac crest, the bone matrix IGF-I concentration is a better predictor of local bone turnover and mass than circulating IGF. Whether this may also apply to cortical bone or different skeletal sites remains to be determined.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
Received December 2, 1997.
Revised March 4, 1998.
Accepted March 18, 1998.
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  F. Rivadeneira, J. J. Houwing-Duistermaat, N. Vaessen, J. M. Vergeer-Drop, A. Hofman, H. A. P. Pols, C. M. van Duijn, and A. G. Uitterlinden Association between an Insulin-Like Growth Factor I Gene Promoter Polymorphism and Bone Mineral Density in the Elderly: The Rotterdam Study J. Clin. Endocrinol. Metab., August 1, 2003; 88(8): 3878 - 3884. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. Ueland, K. Brixen, L. Mosekilde, L. Mosekilde, A. Flyvbjerg, and J. Bollerslev Age-Related Changes in Cortical Bone Content of Insulin-Like Growth Factor Binding Protein (IGFBP)-3, IGFBP-5, Osteoprotegerin, and Calcium in Postmenopausal Osteoporosis: A Cross-Sectional Study J. Clin. Endocrinol. Metab., March 1, 2003; 88(3): 1014 - 1018. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Sowers, T. Scholl, J. Grewal, X. Chen, and M. Jannausch IGF-I, Osteocalcin, and Bone Change in Pregnant Normotensive and Pre-Eclamptic Women J. Clin. Endocrinol. Metab., December 1, 2001; 86(12): 5898 - 5903. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B. C. Nindl, C. R. Scoville, K. M. Sheehan, C. D. Leone, and R. P. Mello Gender differences in regional body composition and somatotrophic influences of IGF-I and leptin J Appl Physiol, April 1, 2002; 92(4): 1611 - 1618. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 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 |
) (r = -0.25, P = 0.0005, n = 192);
postmenopausal women without HRT (•) (r = -0.17,
P = 0.01, n = 235). Postmenopausal women on
HRT 
) (r = -0.30, P = 0.009, n = 76).
Menopausal status could not be reliably determined in remaining 30
women. B, Mean values and 95% prediction interval of linear
regressions of bone matrix IGF-II on age. Premenopausal women (