Cortical Remodeling Following Suppression of Endogenous Estrogen with Analogs of Gonadotrophin Releasing Hormone


  • Part of this work was presented at the 1996 meeting of the ASBMR, Seattle, Washington, U.S.A. (Bell et al., J Bone Miner Res 11:S334).


The effects of estrogen suppression on osteonal remodeling in young women was investigated using transiliac biopsies (eight paired biopsies + four single pre; three single post biopsies) taken before and after treatment for endometriosis (6 months) with analogs of gonadotrophin releasing hormone (GnRH). Estrogen withdrawal increased the proportion of Haversian canals with an eroded surface (106%, p = 0.047), a double label (238%, p = 0.004), osteoid (71%, p = 0.002), and alkaline phosphatase (ALP) (116%, p = 0.043) but not those showing tartrate-resistant acid phosphatase (TRAP) activity (p = 0.25) or a single label (p = 0.30). Estrogen withdrawal increased TRAP activity in individual osteoclasts in canals with diameters greater than 50 μm (p = 0.0089) and also the number of osteons with diameters over 250 μm (p = 0.049). ALP activity in individual osteoblasts was increased but not significantly following treatment (p = 0.051). Wall thickness was significantly correlated with osteon diameter (p < 0.001). In a separate group of patients (four pairs + one post biopsy) on concurrent treatment with tibolone, there was no significant increase in the osteon density, cortical porosity, median canal diameter, or the markers of bone formation and resorption. Enzyme activities and numbers of active canals were also not increased with the concurrent treatment, but there was still an increase in the osteon diameter. As previously shown for cancellous bone, estrogen withdrawal increased cortical bone turnover. We have now shown that resorption depth within Haversian systems was also increased with treatment. The enhanced TRAP activity in individual osteoclasts supports the concept that osteoclasts are more active following estrogen withdrawal in agreement with theoretical arguments advanced previously. Understanding the cellular and biochemical mechanisms responsible for increased depth of osteoclast resorption when estrogen is withdrawn may allow the development of new strategies for preventing postmenopausal bone loss.


Cortical bone comprises about 80% of the skeleton and has been shown to contribute considerably to the strength of bone.1 However, despite the structural and physiological importance of the cortices, many studies of remodeling have only investigated cancellous bone.

Cortical bone is remodeled through discrete basic multicellular units (BMUs)2,3 known as osteons. Within each osteon, it has been shown that remodeling occurs through a fixed sequence of events termed activation, resorption, and formation.4,5 When the osteon is formed, osteoclasts create a longitudinal cutting cone. Behind this advancing resorptive front, osteoblast activity begins filling in the cavity to form the closing cone. This centrifugal resorption and centripetal bone formation at a given site has been termed the remodeling cycle.3

In normal healthy adults, the amount of cortical or cancellous bone that is resorbed and formed is in balance. However, with increasing age or the onset of disease there is evidence that this balance gets disrupted, resulting in bone loss. In cancellous bone the mechanisms of loss remain controversial. Many studies have shown that with increasing age there is a decrease in wall thickness indicating reduced bone formation6–8 but there remains disagreement about the possible role of increased erosion depth in either age-related bone changes7,9 or in postmenopausal osteoporosis.10,11 Comparatively few studies have examined the changes in cortical bone with either increasing age or the onset of disease.12–16 With increasing age, cortical bone loss appears to be associated with increasing Haversian canal size12–14 and canal number.14 Brockstedt et al.15 found a negative cortical bone balance with aging in both sexes which in females was mediated through a decrease in wall thickness. In addition, following the menopause, an increase in activation frequency led to an enlargement of the remodeling space, resulting in an increase in cortical porosity.15

Although it is known that in remodeling, formation follows resorption, the timing of these processes is less well determined. Frost17,18 proposed that bone formation and mineralization may not be a continuous process but that there may be interruptions and termed this the ON-OFF phenomenon. In describing resorption, Jaworski et al.19 reported the presence of a bimodal distribution of small and large resorption cavities in cortical bone. Parfitt20 has suggested that this non-normal distribution may be due to aborted or interrupted resorption in some cavities. Evidence has also been presented indicating skewness in the distribution of erosion depth in individual cancellous bone resorption cavities so that there appears to be a greater proportion of smaller cavities. One explanation advanced was that there may be interruptions in the resorption process of cancellous bone.21,22

The detrimental effects on bone mass resulting from gonadotrophin-releasing hormone (GnRH) analog treatment of endometriosis are well documented.23–27 Recently, a histomorphometric study of iliac crest cancellous bone following GnRH analog therapy indicated that this treatment may be associated with adverse structural effects that may be irreversible.28 Concomitant treatment with tibolone appeared to prevent these changes. In the current study, we have used the same transiliac biopsies to investigate the effect of GnRH analog or GnRH analog and tibolone therapy on cortical bone using both histomorphometric and, in some cases, in situ biochemical approaches.



Twenty-one subjects, women with endometriosis, were recruited from the Royal Free Hospital Obstetrics and Gynaecology out-patient department.28 Each gave informed written consent to the following procedure in the way that had been specified by the district ethics committee for the hospital. Eight of the subjects had suitable bone biopsies before and 6 months after treatment with GnRH analogs. Four patients had a single biopsy before GnRH analog treatment and another four patients were only biopsied after the GnRH analog treatment. One of these patients was subsequently found to have been on treatment for 20 months and was therefore excluded from the analysis. The GnRH analog therapy consisted of either subcutaneous Goserelin (3.6 mg) at four weekly intervals or Triptorelin (3.75 mg) intramuscularly at four weekly intervals. In addition, paired biopsies were obtained from four patients given tibolone (Org OD14, Livial, Organon Technika, Boxtel, The Netherlands), 2.5 mg daily in conjunction with GnRH analogs. A further single biopsy was taken from one patient after treatment with the GnRH analog and tibolone. Prior to each biopsy, each patient received two in vivo tetracycline labels, given as 300 mg demethylchlorotetracycline for 2 days, followed by a 12 day gap, then a further 2 day course of treatment ending 4 days before the biopsy.

None of the subjects had a history of any other complicating medical condition or any other condition at any previous time relevant to the skeleton. None had taken any medication known to affect bone metabolism.

The transiliac bone biopsies were taken with a 7.5 mm internal diameter modified Bordier trephine with the repeat biopsies taken from the contralateral iliac crest. In some cases biopsies were halved with one half being embedded in LR White medium resin (London Resin Co., Berkshire, U.K.)28 and the other being briefly dipped in a 5% solution (w/v) of polyvinyl alcohol (PVA) then chilled in n-hexane and stored at −70°C prior to sectioning on a cryostat.29

Sections from some of the chilled specimens have been used to investigate the relationship between alkaline phosphatase (ALP) and trabecular bone mineralization29 and the effect of estrogen withdrawal on osteocyte apoptosis.30 A previous study based on sections from the embedded material examined the effects of estrogen withdrawal on the remodeling and structural changes in cancellous bone and included some of the data on estradiol levels and BMD at the lumbar spine.28

Biochemical markers and bone mineral density

Estradiol was measured using a DPC immunoradiometric assay kit with a normal postmenopausal range of <70 pmol/l. Plasma ALP levels were measured following an International Federation of Clinical Chemists method using a Hitachi 717 autoanalyzer (Tokyo, Japan) with a normal range of 35–150 U/l. Intact parathyroid hormone was measured using an N-tact PTH (Incstar Corp., Stillwater, MN, U.S.A.; normal range, 10–55 pg/ml) kit and osteocalcin with an125I RIA kit (Incstar; normal range, 1.8–6.6 ng/ml). Urinary calcium was measured using an automated spectrophotometric method,31 urinary hydroxyproline by the method of Prockop and Udenfriend,32 and creatinine by enzymatic determination.33 The normal ranges for the hydroxyproline/creatinine and the calcium/creatinine ratio were 0.01–0.08 μmol/l and 0.02–0.16 mg/mg, respectively.

Bone mineral density (BMD, g/cm2) at the lumbar spine (L2–L4) and proximal femur was measured with dual X-ray absorptiometry on a Hologic 1000W bone densitometer (Hologic Inc., Waltham, MA, U.S.A.) and at the radius using a peripheral quantitative computerized tomography (pQCT) device (Densiscan, Scanco, Zurich, Switzerland) in g/cm3.

Embedded biopsies


Unstained sections (15 μm) were cut from each biopsy for fluorescence microscopy for the assessment of the proportion of osteons with single or double labels and the presence of a crenated surface. Although various sets of sections had been prepared from each biopsy, the order of the sections that were available for analysis could not be determined. Hence to ensure that multiple measurements were not made on the same osteonal system, only one section from each set was measured. Preliminary studies indicated that the two cortices of the biopsy had similar characteristics, and because not all biopsies had both cortices, the thickest cortex (pre, 1.00 ± 0.09; post therapy, 0.97 ± 0.14 mm, mean ± SEM) from each biopsy was used to map each osteon and the presence of label or resorption surface.

The presence of osteoid was investigated using Von Kossa and eosin stained sections (8 μm) in a similar manner. Total cortical area was measured and used for the determination of Haversian canal density (canals/mm2) and percent cortical porosity ([total canal area/total area]100).

Chilled material

Alkaline phosphatase

Cryostat (Brights, Huntingdon, U.K.), sections (10 μm) were reacted for 10 minutes at 25°C in 2 mmol/l alpha-naphthyl acid phosphate (monosodium salt) (Sigma-Aldrich, Dorset, U.K.), magnesium chloride (2 mmol/l), and Fast Red TR (1 mg/ml, Sigma) in 0.1 M barbitone buffer at pH 9.2. The sections were rinsed in 1% acetic acid and washed in distilled water. Enzyme activity was measured using a Vickers M85a scanning and integrating microdensitometer (Vickers Instruments, York, U.K.) with the following settings: ×40 objective, mask size A1 (5 μm diameter), wavelength 525 nm. Those biopsies for which cryostat sections were not available8 had supplied sections for two other studies,29,30 and all the material had been used up.

Tartrate-resistant acid phosphatase

Cryostat sections (10 μm) were reacted for 5 minutes at 37°C in 0.1 M citrate buffer at pH 4.5 containing naphthol AS-BI phosphate (sodium salt) (Sigma) and (10 mmol/l) sodium tartrate. The sections were then washed in cold distilled water containing 50 mmol/l sodium fluoride prior to a 1 minute of coupling in 0.1 M acetate buffer pH 6.2 containing 2.2 mmol/l Fast Garnet GBC (Sigma). Sections were rinsed in distilled water. Activity was measured by microdensitometry with the following settings: ×40 objective, mask size A3 (15 μm diameter) wavelength 540 nm.

Haversian canal and osteon diameters

Measurement of the Haversian canal diameters was performed using a semiautomatic image analysis package34 with a digitizing tablet (Summasketch II, Summagraphics, Fairfield, CT, U.S.A.) and cursor with a Polyvar microscope (Reichert-Jung, Vienna, Austria) and drawing tube attachment. At ×250 magnification all of the Haversian canals were mapped and only those affected by artefacts, e.g., arising at the cut edge of the cortex, were excluded. The edge of each canal was traced, and the minimum diameter and area were obtained (Fig. 1). Osteons are approximately cylindrical so their dimensions are only estimable by measuring along the shortest axis due to the likelihood of each osteon being transected obliquely to a greater or lesser extent.35 The use of the minimum diameter allowed us to include those canals that had been transected longitudinally or obliquely. Such canals may have both resorption and formation surfaces, but in this study only 2.9% of canals with a fluorochrome (single or double) label had an eroded surface.

Figure FIG. 1.

Diagrammatic representation of the measurements of minimum dimensions within osteons. On.Dm, osteon diameter; H.Ca.Dm, Haversian canal diameter; W.Th., wall thickness.

In the cryostat sections the presence or absence of enzyme activity was noted and only canals in which cellular material was present were included in the analysis to ensure that there were no false negatives for the enzyme activity. Where an intact cement line, uninterrupted by subsequent bone resorption occurring during the development of a newer osteon, could be identified using polarized light, this boundary was also traced to obtain the minimum osteon diameter and wall thickness measurements (Fig. 1).

Statistical analysis

To analyze the effect of estrogen withdrawal on the indices of bone remodeling activity, the data are presented as the mean ± SEM of individual patients within each treatment group and analyzed by paired and unpaired Student's t-test.

Canal diameters within individual cortices were not normally distributed so that medians instead of mean diameters were used in the statistical analysis. In addition, to make these distributions more normal, the canal diameters were log-transformed. In contrast, the osteon diameters in the untreated population were normally distributed. To examine the relationship between individual Haversian canal or osteon size and activity, the results were analyzed using the JMP statistical package (SAS Institute, Cary, NC, U.S.A.) with subject, treatment, the presence of alkaline phosphatase (ALP), tartrate-resistant acid phosphatase (TRAP), osteoid, single labels and double labels as nominal categorical variables and log canal diameter, wall width, osteon diameter, ALP, and TRAP activity as continuous variables. The modeling strategy tested the relationships between the presence of labels, osteoid, ALP, or TRAP activity (as dependent variables) and log canal diameter, subject, and treatment (as independent variables) using nominal logistic regression analysis.

Because it has been suggested that the majority of canals with diameters below 50 μm are either aborted or mature Haversian remodeling systems,19,20,35 the canals were subsequently divided into those below and above 50 μm in diameter to investigate this hypothesis.


Effect of GnRH analog therapy given alone

Biochemistry and bone mineral density

Hypoestrogenism following treatment was confirmed as mean serum estradiol fell from 302.9 ± 177.8 p/mol to 38.1 ± 19.9 p/mol (mean ± SEM) post-treatment (Table 1). Serum osteocalcin increased but serum PTH and total ALP were unchanged by the GnRH analog treatment. Both the urinary calcium/creatinine and the hydroxyproline/creatinine ratio were higher following therapy, but only the increase in the urinary calcium/creatinine ratio reached statistical significance (Table 1). The lumbar spine showed a statistically significant decrease in BMD (Table 1). The BMD was lower at most other sites following GnRH analog therapy, but these changes were not statistically significant.

Table Table 1. Biochemical Markers and Bone Mineral Density
original image

Embedded biopsies


Treatment with the GnRH analog did not affect either the osteon density or cortical porosity to an extent that was statistically significant, although these were both higher post-treatment (Table 2). While there was no change in the proportion of Haversian systems which had a single label, the proportion with a double label was significantly (p = 0.004) elevated post-treatment (Table 2). Analysis of the results from the eight paired samples also showed no significant increase in single labels (pre, 4.5 ± 2.1; GnRH analog, 9.6 ± 3.3, p = 0.30) but a significant increase in double labels (pre, 2.7 ± 1.7; GnRH analog, 7.6 ± 1.7, p = 0.0045).

Table Table 2. Assessment of Osteon Activity in Embedded Sections
original image

The presence of osteoid was also significantly higher after treatment both for all samples (Table 2) and for the paired samples (pre, 12.4 ± 2.3; GnRH analog, 23.4 ± 1.8, p = 0.01). While the proportion of canals showing the presence of an eroded surface rose after treatment (Table 2) and reached significance for all subjects (p = 0.047), the change was not significant for the paired samples (pre, 3.3 ± 1.4; GnRH analog, 7.8 ± 2.1, p = 0.12).

Haversian canal diameter

The overall median canal diameters were not affected by the treatment (Table 3). Although the median diameter of unlabeled canals was lower than those with either a single or double fluorochrome label, this was only significant (p = 0.031; unpaired t-test) after GnRH analog treatment. There were no differences, either before or after treatment, in the median canal diameters between those canals lined with osteoid and those without (Table 3).

Table Table 3. Size Distribution of Haversian Canals in Sections of Embedded Biopsies
original image

Statistical modeling of Haversian canal characteristics

In individual Haversian canals, the presence of either single or double labels was independent of subject. However, the presence of a single label, although not affected by treatment, was positively associated with log canal diameter (p = 0.0081). The presence of double labeled canal surfaces was positively associated with both log canal diameter (p < 0.0001) and treatment (p = 0.0054). The proportion of canals bearing a double label was significantly increased by GnRH analog treatment in those with diameters below (p = 0.0008) and above (p = 0.0181) 50 μm in diameter (Fig. 2).

Figure FIG. 2.

Distribution of single and double tetracycline labels and osteoid in individual Haversian canals from subjects before and after treatment with GnRH analogs. The numbers of canals within each group were as follows. Single labels: under 50 μm, pre 19 with label/284 without label, GnRH analog 18/220; over 50 μm, pre 16/147, GnRH analog 16/127. Double Labels: under 50 μm, pre 4 with labels/299 without labels, GnRH analog 16/222; over 50 μm: pre 9/154, GnRH analog 19/124. Osteoid: under 50 μm, pre 59 with osteoid/342 without osteoid, GnRH analog 79/277; over 50 μm: pre 22/122, GnRH analog 27/72. ***p < 0.001; *p < 0.05. Likelihood ratio Chi square.

In all Haversian systems, the frequency of the presence of osteoid was not associated with canal size but was increased after estrogen withdrawal (p = 0.0003) and there was a weak effect of subject (p = 0.03). The treatment effect on the presence of osteoid was seen both in canals under 50 μm in diameter and canals exceeding 50 μm in diameter (p = 0.023) (Fig. 2).

Chilled biopsies

For assessment of ALP and TRAP activity in the chilled biopsies, four biopsies from each of the pretreatment and the GnRH analog groups were available. There were three paired samples for ALP activity and two paired samples for TRAP activity. There were no significant differences in the overall median canal diameters compared with the plastic sections or between those before and after GnRH analog treatment (Tables 3 and 4).

Table Table 4. Size Distribution of Haversian Canals in Sections of Cryostat Biopsies
original image

Alkaline phosphatase activity

The overall percentage of ALP positive canals was increased by treatment with the GnRH analog (pre, 19.8 ± 6.6%; GnRH analog, 42.8 ± 8.1%; p = 0.043, unpaired t-test). In the analysis of individual Haversian canals, there was no relationship between the presence of ALP and the overall log canal diameter. The proportion of canals with ALP activity was significantly increased by the therapy in those canals with diameters below (p = 0.027) as well as those above (p < 0.0001) 50 μm (Fig. 3).

Figure FIG. 3.

Distribution of ALP and TRAP in individual Haversian canals from subjects before (n = 4) and after (n = 4) treatment with GnRH analogs. The numbers of canals within each group were as follows. ALP: under 50 μm, pre 27 with ALP/85 without ALP, GnRH analog 60/104; over 50 μm, pre 7/32, GnRH analog 38/26. TRAP: under 50, pre 31 with TRAP/115 without TRAP, GnRH analog 31/122; over 50 μm, pre 14/22, GnRH analog 30/18. ****p < 0.0001; *p < 0.05. Likelihood ratio Chi square.

The level of ALP activity in individual canals was not significantly increased by GnRH analog treatment either when grouped according to subject (14.4 ± 1.5 vs. 18.3 ± 2.3 MIA × 100, p = 0.18) or when individual canals were grouped according to treatment (15.92 ± 1.9 vs. 19.9 ± 1.1, p = 0.051). Overall, ALP activity was weakly associated with log canal diameter (r = 0.182; p = 0.041) but this was not affected by treatment (Fig. 4).

Figure FIG. 4.

Relationship between ALP and TRAP activity in individual Haversian canals and the log canal diameter before (solid line) and after (broken line) treatment with GnRH analogs.

TRAP activity

Overall, GnRH analog therapy did not significantly increase the proportion of TRAP positive canals (pre, 24.5 ± 4.8; GnRH analog, 30.4 ± 2.7, p = 0.245). In the analysis of all the individual Haversian canals, the presence of TRAP was independent of canal diameter. However, the presence of TRAP was increased by GnRH analog therapy in canals with diameters above 50 μm but not those below (Fig. 3).

In the analysis of all individual Haversian canals, the level of TRAP activity was associated with log canal diameter (r = 0.504; p < 0.0001) but was independent of treatment and subject. TRAP activity was stimulated by GnRH analog treatment in canals above 50 μm diameter (pre, 20.66 ± 2.34; post, 33.25 ± 2.78; p = 0.0089) (Fig. 4) but not in those with smaller diameters (pre, 18.17 ± 1.38; post, 20.35 ± 2.08; p = 0.387).

Osteon diameter and wall thickness

Following GnRH analog therapy there was a significant (p = 0.0149) increase in the mean osteon diameter (pre, 147.3 ± 7.2 μm; GnRH analog, 175.7 ± 8.7 μm). This was due to an increase in the number of uninterrupted osteons of diameter greater than 250 μm with the GnRH analog (pre, 5/73; GnRH analogue, 16/94; p = 0.049). Osteonal wall thickness was correlated with osteon diameter, and although the mean wall thickness was significantly increased (pre, 105.3 ± 5.8 μm; GnRH analog, 126.8 ± 6.5 μm: p = 0.0166) after GnRH analog treatment this was due to the increase in osteon diameter (Fig. 5).

Figure FIG. 5.

Relationship between osteonal wall thickness and osteon diameter before and after GnRH analog therapy. Pretreatment: y = 0.354x + 0.5545, r2 = 0.776. GnRH analog: y = 0.3005x + 9.8125, r2 = 0.663. (Corrected per Publication Correction dated August, 1999)

Effect of combined GnRH analog and tibolone therapy

Some of the subjects (five patients; four paired) received concomitant tibolone treatment with the GnRH analog.


This combined treatment had no significant effect on canal density or cortical porosity (Table 2) or on the median canal diameters before or after treatment (Table 3). The proportion of osteons with fluorochrome labels or osteoid was higher after GnRH analog and tibolone treatment but did not reach significance (Table 2).

Enzyme activities

Even before treatment, these patients had a higher proportion of ALP positive canals (57.8 ± 21.5) than the subjects on GnRH analog therapy alone so that although there was no increase after treatment with the GnRH analog and tibolone (64.8 ± 7.7) the values following treatment were not different from those of subjects treated with the analog alone (see above). Similarly, the level of ALP activity of these canals prior to treatment was higher (22.8 ± 2.5) but did not increase following the GnRH analog and tibolone therapy (24.5 ± 1.4). The proportion of TRAP positive Haversian canals (pre, 20.4 ± 12.8; GnRH analog + tibolone, 11.7 ± 3.4; p = 0.931) and TRAP activity at individual sites (pre, 24.3 ± 1.5; GnRH analog + tibolone, 23.6 ± 5.9; p = 0.902) did not change significantly following the GnRH analog and tibolone therapy.

Osteon diameter and wall thickness

Following GnRH analog and tibolone therapy there was a significant (p = 0.0121) increase in the mean osteon diameter (pre, 153.0 ± 6.2 μm; GnRH analog + tibolone, 176.1 ± 8.7 μm). This was due to an increase in the number of uninterrupted osteons of diameter greater than 250 μm with the GnRH analog and tibolone (pre, 1/62; GnRH + tibolone, 14/116; p = 0.016). Although the mean osteonal wall thickness was also significantly increased (pre, 104.8 ± 4.6 μm; GnRH analog + tibolone, 121.1 ± 4.8 μm; p = 0.0004) after GnRH analog and tibolone treatment this was due to the increase in osteon diameter.


Despite the importance of the cortex in bone strength, most recent studies of bone remodeling have concentrated on cancellous bone.22,28,36–38 The few investigations on cortical bone have been on normal subjects15,35,39 to show changes with aging and gender or those with diseases known to affect bone metabolism.16,35

In the present study, the effects of estrogen withdrawal, due to GnRH analog therapy for endometriosis, on cortical bone in young premenopausal women has been investigated. Unlike previous histomorphometric studies, in which only a set number of Haversian systems within the middle two quarters of the cortex were analyzed,15,16,35,39 we have examined the whole area of cortical bone because in the femoral cortex it has been reported that there are differences in the characteristics of osteons within the various regions of the cortical bone.12,40 In addition to the histomorphometric approach, we have also used in situ biochemical techniques to study ALP and TRAP activity and distribution within a subset of patients. ALP has been shown to be an essential enzyme for bone formation, and similarly TRAP has a key role in osteoclast function as suggested by recent studies on a gene knock-out animal model.41

Before treatment, the proportions of canals showing resorption, osteoid, and labeling with tetracycline were comparable to those reported from a similar age group by Agerbæk et al.39 The withdrawal of estrogen resulted in a higher proportion of active Haversian systems within the cortex. Although the proportion of active osteons varied for individual parameters related to resorption (eroded surface and TRAP activity) and formation (osteoid, double label, and ALP), the response to GnRH analog therapy was comparable in them all with an increase of approximately 2-fold. This is in accord with a previous study of postmenopausal women which reported a significant doubling in the activation frequency in Haversian systems in comparison with premenopausal women.15 The increase in double, but not single, labeled canals also supports the view that the bone formation rate within Haversian canals is increased in postmenopausal women.15

Following estrogen withdrawal, there was no increase in the median Haversian canal diameter or cortical porosity. This is in contrast to the study of Brockstedt et al.15 who observed an increase in the average Haversian canal diameter and cortical porosity between pre- and postmenopausal women. This lack of a significant change in the canal diameter or porosity in our study may be due to the small sample size or a reflection of the length of the treatment time (6 months) in comparison with the lifespan of a BMU which is thought to be 6–12 months.42 To develop a significant increase in porosity may require a prolonged negative balance between resorption and formation. If this imbalance were relatively small, it might require several remodeling cycles at individual sites for an imbalance in the remodeling process in the whole cortex to be identified.

In cortical bone, the minimum diameter of the uninterrupted cement lines of secondary osteons is a record of the depth of bone resorbed by the most recent generations of osteoclasts. TRAP activity was increased following estrogen withdrawal in canals with large diameters. Furthermore, after GnRH analog treatment, there was an increase in the number of osteons of large diameter (>250 μm) compared with biopsies taken prior to treatment. These results suggest that estrogen withdrawal results in enhanced osteoclastic activity which leads to an increase in osteonal size. In previous studies of normal premenopausal women aged 19–38 years, the mean resorption period was 44 ± 9 days,15 thus in the current study the treatment period of 6 months is sufficient to have allowed ongoing and new sites of resorption to go to completion.

The minimum wall thickness within each osteon reflects the amount of bone formed by osteoblasts since the reversal phase and thus is dependent on time since reversal, rate of bone formation, and the availability of space for newly mineralized bone to be laid down. In the current study, GnRH analog therapy was associated with increases in the wall width of the large osteons which may be compensatory for the increase in erosion depth and does not seem to reflect an increase in the rate of bone formation that would result in a positive bone balance. This is in contrast to the study of Brockstedt et al.15 on postmenopausal women which found no significant change in overall mean wall thickness or the osteon diameter.

In previous studies, on femoral cortical bone, a gradation in osteon diameter from the periosteal to endosteal region40,43 has been identified. However, in the ilium we found no pattern in the size distribution of osteons and the large osteons that developed with treatment were evenly dispersed across the cortical bone from the periosteal to endosteal margin.

In this study, as part of the analysis the Haversian canals were split into those below and above 50 μm in diameter because it has been suggested that canals with diameters below 50 μm are predominantly either aborted or mature Haversian remodeling systems.19,20,35 In this study, some of the canals with diameters below 50 μm showed the presence of TRAP activity (27%), single or double labels (4%), osteoid (17%), or ALP activity (32%) prior to treatment. Furthermore, although the effect of estrogen withdrawal was to increase the levels of resorption and formation activity in the cortex, this was not limited to canals of diameter above 50 μm. These findings would suggest that both resorption and formation occur in a significant proportion of osteons with small canal diameters.

Cohen and Harris,40 using a three-dimensional reconstruction of the Haversian systems of the midfemoral cortex of adult dogs, demonstrated that there was spatial and temporal intermittency in the resorption and formation of the osteons. In the current study, ALP activity and the presence of fluorochrome labels were correlated with canal diameter indirectly supporting the view that the rate of bone formation decreases during the closure of the Haversian canal.39 However, our data also indicate that the size of the canal cannot be used to predict whether TRAP activity, ALP activity, osteoid, or even fluorochrome labels will be present or absent. Furthermore, the changes in these indicators of bone formation associated with estrogen withdrawal were not limited to cohorts of canals with similar diameters suggesting that temporal intermittency may occur at various stages of bone formation. While our data do not allow us to determine the periodicity of interruptions of bone formation, it has been reported44 that pauses in bone mineralization of between 3 and 34 days do not occur in cancellous bone of women with severe (two vertebral fractures) postmenopausal osteoporosis.

In this study, we also investigated the effects of dual therapy of tibolone and GnRH analog. Tibolone is a synthetic steroid with estrogenic, progestrogenic, and androgenic properties which may reduce some of the adverse effects on cancellous bone microstructure induced by GnRH analog therapy.28 Some of the changes that we observed following GnRH analog therapy, such as the enhanced development of very large osteons were also present in the subjects receiving concomitant tibolone. However, tibolone did appear to prevent the rise in the number of TRAP active sites that was associated with GnRH analog therapy, although it did not affect the level of TRAP activity in individual Haversian canals.

There are some limitations to this study. First, the study only included a few women so the power of the study to demonstrate significant changes was therefore limited. Similarly, the addition of tibolone to the GnRH analog therapy only involved five patients so it is not possible to draw precise conclusions as to any possible beneficial effects. The number of biopsies for in situ biochemistry was reduced because some of the chilled samples did not have a suitable cortex while others had been used up in previous studies.29,30 Nonetheless, based on the sizes of the Haversian canals, the biopsies that were available were representative of all the samples. Finally, as a model of the early changes in cortical bone turnover leading to postmenopausal osteoporosis, the use of GnRH analogs does not mimic the slow decline in hormone levels associated with the menopause but is more pertinent to a surgically induced hypoestrogenic state.

In conclusion, we have confirmed that estrogen suppression in young premenopausal women is associated with increased bone turnover in the iliac cortex. The increase in the size-related proportion and activity of TRAP positive canals together with the increased numbers of very large osteons supports the concept that osteoclasts in some newly forming osteons resorb significantly more deeply in response to estrogen withdrawal than did their predecessors before the start of treatment. As expected, overall osteoblast activity also increased following estrogen suppression with an increased number of active formation sites and a compensatory increase in bone formation in large osteons. The fact that both the proportion of double fluorochrome labeled (but not single labeled) sites and the presence of ALP activity within individual canals increase after estrogen suppression suggests that the rate of bone formation at any one site also increases following estrogen suppression.


This work was supported by MRC Programme Grant 9321536 (to J.R.) and by MRC project grant G892787OSA (to J.E.C. and R.S.). The authors thank Angela Mitchell and Richard Hesp (Clinical Research Centre, Harrow) for the measurements of bone density, Z. Varghese (Royal Free Hospital) and Jeffrey Green (Clinical Research Centre, Harrow) for the measurement of serum and urinary markers of bone turnover, Jeremy Bradbeer and Peter Croucher for the preparation of the biopsy material, and K. Yamaguchi for the preparation of the sections of the embedded samples.