Alendronate in the Prevention of Bone Loss After a Fracture of the Lower Leg


  • Dr. Lips has been a consultant and received research funding by Merck & Company and Eli Lilly & Company. Dr. van der Poest Clement served on the regional advisory council on “raloxifene” for Eli Lilly & Company and received funding by Merck & Company. All other authors have no conflict of interest.


Fracture of a leg and the consequent absence from weight-bearing lead to local bone loss. A 1-year, single-center, prospective, randomized, double-blind study was conducted, to determine whether bone loss would occur in the proximal femur and the calcaneus after a fracture of the lower leg and whether this loss could be prevented by the antiresorptive drug bisphosphonate alendronate. Twenty-three men and 18 women with a recent unstable fracture of the lower leg were randomized to receive either 10 mg of alendronate daily or placebo. Bone mineral density (BMD) of both hips and the lumbar spine was measured at baseline and 6 weeks and 3, 6, and 12 months after start of the treatment. Quantitative ultrasound (QUS) measurements of the calcaneus were performed at baseline on the noninjured side and at 6 weeks and 3, 6, and 12 months after start of treatment on both sides. After 1 year, in the placebo group, there was a significant decrease from baseline in BMD of the hip on the side of the fracture. In the alendronate group, there was no significant change from baseline. The differences in BMD between the two treatment groups on the side of the fracture were significant in all sites of the hip: 4.4% (p = 0.016) in the trochanter, 4.6% (p = 0.016) in the femoral neck, and 3.9% (p = 0.009) in the total hip. In the hip on the contralateral side, there were no significant changes from baseline in either treatment group and there was no difference between the two treatment groups. BMD in the lumbar spine increased in the alendronate group, and after 1 year there was a significant difference between the active treatment and placebo group of 3.4% (p = 0.04). One year after fracture, ultrasound parameters of the calcaneus in the placebo group were significantly lower on the fractured side compared with the contralateral side (p < 0.01). In the alendronate group, no significant difference between the two sides was observed. In conclusion, BMD of the proximal femur was still decreased 1 year after a fracture of the lower leg. Alendronate prevented this bone loss.


IMMOBILIZATION LEADS to bone loss, whether related to bed rest, space flight, or after trauma.(1–3) Fractures of the leg lead to bone loss at several sites along the leg and in the hip, as was shown in cross-sectional and prospective studies.(4–11) This bone loss occurs both close to the fracture and at more distant sites in the same extremity. Bone loss starts immediately after fracture but still can be detected several years later.(7, 10)

In a previous study, during the 2 months of no weight bearing after a tibia fracture, bone mineral density (BMD) of the trochanteric region of the hip decreased 9% on the side of the fracture. One year later, this loss had increased to 15%.(7) Five years after fracture some recovery was observed, but not to baseline BMD values.(10) Bone loss of the same magnitude was found in the trochanteric region after tibial shaft fractures.(9) After less severe ankle fractures with earlier weight bearing, a smaller but significant decrease of trochanter BMD was encountered.(11) At sites distal to a fracture of the lower leg substantial bone losses, especially of trabecular bone, were observed.(8, 11) These changes in BMD after fracture may reflect an increased risk for later fragility fractures.(12, 13)

Quantitative ultrasound (QUS) measurement of the os calcis has had considerable interest in the past years for its ability to predict the risk of hip fractures.(14–16) Especially, changes in broadband ultrasound attenuation (BUA) may indicate changes in bone structure besides lower bone mass. Up to now only one study examined changes in ultrasound measurement of the calcaneus, distal to ankle fractures. In this study, an early decrease of BUA was found without return to baseline after 1 year.(11)

The metabolic condition of bone after a fracture is characterized by increased bone turnover.(17) In animal studies skeletal unloading stimulated bone resorption and decreased bone formation.(18) In humans the biochemical markers of bone turnover increased after fracture, probably reflecting the effect of unloading and fracture healing.(11,19–22)

Therefore, it appears logical to study the effect of an antiresorptive drug on the bone loss after a fracture of the lower leg. Bisphosphonate alendronate has been effective in several conditions characterized by increased bone turnover and bone loss, including postmenopausal osteoporosis and local bone loss after wrist fracture.(23–25)

This report deals with a double-blind study on the effect of alendronate on postfracture bone loss in patients with a fracture of the lower leg.



Forty-one hospitalized patients were recruited, shortly after occurrence of an unstable fracture of the lower leg. All fractures necessitated immobilization and absence from weight bearing for several weeks. Fractures involved the tibia, including 10 tibial and fibular shaft fractures, 5 distal tibial fractures (compression or pilon tibiale fractures), and 26 ankle fractures, 15 bimalleolar and 11 trimalleolar. Only 2 patients were treated conservatively with fracture reduction and a plaster cast, 2 patients with an open fracture were treated with external fixation, the other patients were operated on and had osteosynthesis with a nail, plate, screws, or wire. Twenty-six patients were operated on within 24 h after the fracture occurred, 6 patients were operated on 5–7 days after fracture, and 7 patients were operated on ≥10 days after fracture. Bed rest was prescribed for 5–7 days after the treatment. Then, mobilization without weight bearing on the fractured leg was allowed. Four to six weeks later, initial weight bearing was allowed, with a gradual increase over the subsequent 4–6 weeks. Polytrauma patients and all patients with fractures at more than one site were excluded. Twenty-three men and 18 women, aged 30–75 years old were recruited. Eleven of the women were postmenopausal, with a mean of 12.6 years since last menstruation (range, 3.7–28.3 years), 7 women were pre- or perimenopausal. Of the postmenopausal women, one used tibolone for climacterial complaints and one used a progestin for 6 years because of endometriosis. Of the premenopausal women, two used oral contraceptives as contraception, two used oral contraceptives because of perimenopausal complaints, and one used a progestative for 3 years to postpone menstruation. Two perimenopausal women, 47 and 48 years old, respectively, used no hormonal medication. Apart from hormonal therapy, no other medication was used that might have influenced bone and mineral metabolism. In all patients hyperthyroidism, vitamin D deficiency, and other diseases influencing bone metabolism were excluded before randomization. All patients were in good general health and had performed normal daily activities before the occurrence of the fracture. The hospital Ethical Review Committee approved the protocol and written informed consent was obtained from all patients.

Study design and treatment

After a baseline period, starting as soon as possible after tibial fracture, eligible patients were randomized to treatment either with alendronate or placebo. Randomization was planned to occur up to 2 weeks after fracture. The study duration was 1 year, with clinic visits 6 weeks and 3, 6, 9, and 12 months after the start of treatment. Patients were instructed to take one tablet, containing 10 mg of alendronate or placebo, each morning, at least ½ h before breakfast. All patients also took a calcium supplement containing 500 mg of elemental calcium and 400 IU of vitamin D3 in the evening at mealtime. The study drug was supplied by the manufacturer (Merck & Co., Inc., Whitehouse Station, NJ, USA) in bottles containing 110 tablets of alendronate or placebo for each 3-month treatment period. Compliance to the study drug was monitored by counting the medication returned every 3 months. Noncompliant patients, missing >20% of the scheduled doses were reminded of the importance of good compliance for a successful outcome of the study.

All patients answered a previously validated questionnaire(26) that assessed the daily calcium intake of dairy products in a semiquantitative way. To account for the daily calcium intake from vegetables and cereals, 250 mg were added for each subject.

Efficacy criteria

BMD measurements

BMD was measured by DXA using a Hologic QDR 2000 densitometer (Hologic Corp., Waltham, MA, USA). The bone density of the lumbar spine and both hips was obtained twice during the baseline period, the mean providing the baseline value. Follow-up measurements of lumbar spine and both hips were done 6 weeks and 3, 6, and 12 months after the start of treatment. Regions of interest in the hip were the femoral neck, trochanter, and total hip. Apart from the nuclear medicine physician, the clinical investigators were blinded to the bone densitometry results obtained after randomization. All BMD scans were reviewed at a quality assurance center.

Bone markers

First-morning voided urine samples were collected twice pretreatment and once at 6 weeks and 3, 6, and 12 months after start of treatment for determination of the biochemical markers for bone resorption N-telopeptide cross-links of type I collagen (NTX) and deoxypyridinoline (DPD), both corrected for the creatinine concentration. Serum samples were obtained in the morning of the same days for measuring the bone formation markers bone-specific ALP (BAP) and osteocalcin (OC). All samples were kept frozen until analysis, which was done at the biochemical and endocrine laboratories of the study center, after the last patient had finished the study. NTX was measured by ELISA (Osteomark; Ostex, Inc., Seattle, WA, USA), DPD was measured by competitive immunoassay (Chiron Diagnostics, East Walpole, MA, USA), BAP was measured by immunoradiometric assay (IRMA; Tandem-R Ostase; Hybritech, Inc., San Diego, CA, USA), and OC was measured by ELISA (Biosource Diagnostics, Fleurus, Belgium).


Changes in bone density and structure of the calcaneus were evaluated using two ultrasound parameters: BUA and velocity of sound (VOS). The contact ultrasound system CUBA Clinical (McCue Ultrasound, Winchester, UK) was used. As previously described, its long-term CV is 4.9% for BUA and 1.3% for VOS.(27) At baseline, that is, pretreatment, only the calcaneus on the nonfractured (contralateral) side could be measured. When feasible, 6 weeks after treatment started, both calcanei were measured, supplying the first result on the fractured side. In 7 patients, the first measurement on the fractured side had to be postponed to 3 months after the start of treatment, because of the use of a cast, or sustained edema. In 1 patient, no measurement on the fractured side could be performed at all because of sustained excessive edema. All measurements were at least performed twice, with repositioning of the foot. When the results differed >8%, a third measurement was done. Therefore, the definitive result for a visit was the mean of two or three measurements.


Adverse events were recorded at each visit using nonleading questions. A complete physical examination was performed before randomization, and limited physical examinations were performed at subsequent visits. Clinical laboratory tests to monitor safety were performed at each visit.

Fracture healing was monitored by the surgeon in charge of the patient. The timing of X-rays and necessary procedures for the management of the fracture were at his discretion. Decisions regarding mobilization and the amount of weight bearing on the fractured leg were made by the surgeon in collaboration with a physiotherapist. The diagnosis of delayed healing or malunion was made by the surgeon. The diagnosis of reflex sympathetic dystrophy (RSD) was made by the surgeon in collaboration with the anesthesiologist, experienced in diagnosing and treating this disorder.


The primary evaluation of the efficacy data were based on the intention-to-treat principle. Three patients without any follow-up measurement after baseline were excluded from statistical evaluation. In case of missing values, the last measurement was carried forward. Percent changes from baseline BMD values were calculated and ANOVA models were used to analyze them. A factor “treatment” was used to asses the between group differences. Apart from the intention-to-treat analysis, an active treatment analysis was performed on the bone densitometry results of those patients who were at least 80% compliant for the study medication over 1 year. The study had 95% power to detect a 4.3% difference in trochanter BMD.

For biochemical markers of bone turnover, the log-transformed percent of baseline value was analyzed using ANOVA.

For analysis of ultrasound results, the data of 33 patients who completed the full year of the study protocol were used. When measurements had to be postponed because plaster or swelling made them impossible, results were carried backward from the first possible measurement. Because no baseline measurements were available of the fractured side, no percent change from baseline could be calculated. The measured values for BUA and VOS were plotted and ANOVA was used to analyze the data over time on the nonfractured side and to compare the fractured versus the contralateral side. Paired sample t-tests were used to compare the values after 1 year on the fractured versus the nonfractured side. All treatment comparisons were two-sided and a value of p < 0.05 was considered statistically significant.


After randomization, 3 patients, 2 on placebo and 1 on alendronate, withdrew from the study before the first follow-up visit, 6 weeks after the start of the treatment. Reasons for this very early withdrawal were because 1 patient did not receive the trial medication in time, 1 patient experienced gastrointestinal problems on nonsteroidal anti-inflammatory drugs (NSAIDs) and did not start trial medication, and 1 patient, once released from the hospital, considered the trial too much trouble. During the year, 3 more patients were lost to follow-up and 1 patient discontinued the study because he was not able to come to the scheduled visits. These four patients had at least one follow-up visit and are included in the analysis. Baseline data of all randomized patients are presented in Table 1.

Table Table 1.. Baseline Characteristics of the 41 Randomized Patients
original image



In all three regions of interest of the hip, no difference in BMD between the hip on the side of the fracture and the contralateral hip existed at baseline.

Fracture side

In the placebo group, there was significant bone loss from the hip on the side of the fracture. The bone loss occurred in the trochanter, femoral neck, and total hip (Fig. 1). Six weeks after the start of the treatment (for most patients at the end of total absence from weight bearing), a significant loss had occurred in all three regions of the hip, progressing thereafter, with maximal loss from baseline after 1 year (Table 2).

Table Table 2.. Mean Percent Changes From Baseline in BMD of Both Hips and the Lumbar Spine After 1 Year of Treatment With 10 mg of Alendronate or Placebo, According to Intention-to-Treat Analysis
original image
Figure FIG. 1..

Mean percent changes (±SE) from baseline in the proximal femur on the fractured side. (A) Trochanter, (B) femoral neck, and (C) total hip. Intention to treat analysis. Significance between treatment groups are *p < 0.05, **p < 0.01, and ***p < 0.001.

In the alendronate group, after 1 year, no significant loss from baseline was observed. In the trochanter, the difference between the two treatment groups was significant after 6 weeks and 1 year (Fig. 1A). In the femoral neck and total hip, the difference was significant throughout the year (Figs. 1B and 1C).

Contralateral hip

Significant changes from baseline in the contralateral hip did not occur in either treatment group, and significant differences between the two treatment groups were not observed (Table 2).

Lumbar spine

A significant difference in BMD between the treatment groups was found after 6 months and after 1 year (Table 2).

In addition to the intention-to-treat analysis, an active treatment analysis was performed on the data of the 33 patients who completed the full year of the protocol with a compliance of at least 80% of the study medication. The results of two noncompliant patients who discontinued the study after 3 months and 6 months, respectively, were excluded from the analysis. This had an effect on the outcome, making the treatment differences larger. The difference between placebo and alendronate increased from 3.9% to 5.5% in the total hip, from 4.4% to 6.5% in the trochanter, and from 3.4% to 4.1% in the lumbar spine.



On the contralateral side, no significant change from baseline was observed in either treatment group. On the fractured side, in the placebo group, BUA increased between 3 months and 1 year after fracture, but at 1 year a significant difference between contralateral and fractured side still existed (Fig. 2B). In the alendronate group, no difference was observed between the contralateral and the fractured side (Fig. 2A).

Figure FIG. 2..

Mean BUA values (±SE) in the calcaneus on the fractured side versus the contralateral side. Patients completing 1 year of treatment. (A) Placebo group, (B) alendronate group. Significant difference of **p < 0.01) in a paired sample t-test 1 year after fracture.


A pattern similar to BUA was observed. On the contralateral side, there was no change from baseline in either treatment group. On the fractured side, in the placebo group VOS increased from 3 months to 1 year, but was still significantly lower than on the contralateral side (p = 0.003). In the alendronate group, no difference between the two sides was observed.

Bone markers

Figure 3 shows the response of the biochemical markers of bone resorption and formation, urine NTX and DPD, and serum BAP and OC to fracture and treatment. At baseline, there were no differences between either treatment groups with respect to any marker.

Figure FIG. 3..

Mean percent changes from baseline (±SE) of the urine bone resorption markers (A) NTX/creatinine, and (B) DPD/creatinine, and the serum bone formation markers (C) BAP and (D) OC, according to intention-to-treat analysis. Significance between treatment groups are *p < 0.05, **p < 0.01, and ***p < 0.001.


In the placebo group both resorption markers initially seemed to increase after fracture, but after 1 year both were below baseline: NTX, 16% (p < 0.01); and DPD, 33% (p < 0.001), respectively. In the alendronate group NTX significantly decreased from baseline throughout the year. After 1 year this decrease amounted to 48% from baseline (p < 0.001). DPD did not decrease significantly from baseline until 6 months after fracture. After 1 year it had decreased 41% (p < 0.001). The differences between the two treatment groups were significant at all time points for NTX but not significant for DPD.


The bone formation markers showed an initial increase after fracture in both treatment groups. In the placebo group BAP and OC increased significantly from baseline and reached their maximum 6 months after fracture, 59% and 29%, respectively, above baseline (p < 0.001). After 1 year, both had partially returned to baseline. OC differed not significantly from it, but BAP remained 30% elevated above baseline (p = 0.001). In the alendronate group BAP increased less, maximally at 33% after 3 months (p < 0.05), and then returned to baseline. In the alendronate group, OC only was significantly above baseline after 6 weeks: +13% (p < 0.05) and it returned to baseline afterward. For BAP, the differences between the two treatment groups were significant after 6 months and 1 year and for OC only after 6 months.

Baseline values were calculated as the mean of two pretreatment samples, taken 7.9 days (range, 3–17 days) and 9.8 days (range, 4–18 days) after fracture. A significant difference between the two samples was found only for NTX, the second being 14% higher than the first (p = 0.015).


For a full safety analysis, there were not enough participants in this study. It lacked the power to detect small differences and draw conclusions about the treatment with alendronate. Generally, the study medication was well tolerated. No patient discontinued before study completion because of a clinical or laboratory adverse event. One patient in the alendronate group had a serious adverse event 1 day after randomization with a suspected pulmonary embolism. This was soon after complete immobilization and operation and while wearing a plaster cast.

Fracture healing

Two patients had a complicated fracture healing. Nonunion was diagnosed. One patient was on placebo and the other patient was on alendronate. Both had open fractures (of tibial shaft and ankle, respectively) and were primarily treated with external fixation. The patient in the placebo group was operated on after 9 months with bone grafting and plate fixation. The patient in the alendronate group was not completely mobile at the end of study and was awaiting an operation. Although a group of two patients does not allow statistical analysis, some differences in the outcome of study parameters were observed. In the patient on placebo, BMD of the trochanter and total hip had decreased 9.4% and 8.5%, respectively, after 1 year, BUA of the calcaneus on the fractured side was 43% below the BUA on the contralateral side, and there was an increased bone turnover; after 6 months NTX was 77% and BAP was 112% above baseline. In the patient on alendronate, BMD had increased 1.6% and 1.2%, respectively, after 1 year. Swelling of the ankle did not allow BUA measurements on the fractured side throughout the year. NTX had fallen 31% below baseline after 1 year, but BAP still was 30% above baseline.

In 2 other patients RSD was diagnosed and treated. Both had ankle fractures and were in the alendronate group. After 1 year, both still experienced pain with walking. Their study results did not differ remarkably from the means in the alendronate group: total hip BMD was +2.0% and −0.8% compared with baseline, respectively and the BUA on the fractured side was 95.7% and 105.7% of the contralateral side after 1 year.

All other patients had an uneventful recovery, were allowed full weight bearing 3 months after fracture, and were completely mobile at the end of the study.


This study shows that bone loss, occurring after a fracture of the tibia, can be prevented by the use of bisphosphonate alendronate. Earlier observations of bone loss in the hip after a fracture of the tibia were confirmed in the placebo group of this study.(7–11) In this group, there was a significant decrease of BMD in the hip at the end of the period of no weight bearing. The bone loss continued after weight bearing was resumed by the patients. As in our earlier study,(7, 10) maximum loss occurred after 1 year. Alendronate prevented this bone loss, with a significant suppression of the markers of bone turnover.

In this study, the decrease of BMD of the hip was less than that found in our earlier study, which included tibial plateau fractures, with prolonged immobilization.(7) Greater losses after tibial shaft fractures were also observed.(9) Smaller losses in the proximal femur were observed after stable ankle fractures that were treated conservatively, allowing partial weight bearing 1 week after fracture.(11) Therefore, severity of the fracture and duration of immobilization seem to play an important role in the amount of bone lost in the proximal femur. Analyzing other factors, which might explain the differences in bone loss, we did not observe a significant influence of sex or age on the bone loss in the placebo group. After 6 months, the patients with unstable ankle fractures had lost more bone than the patients with tibial shaft fractures. Although a nutritional calcium deficiency did not exist, all patients in this study were given a calcium supplement in the evening, possibly influencing the bone turnover in the placebo group.

In this study, two cases of malunion occurred, one in the placebo-group and one on alendronate. In both cases, the nature of the fracture can explain the complicated fracture healing. The question whether it is safe to give a bisphosphonate to patients with a recent fracture, because it might influence fracture healing, was addressed in several animal studies.(28, 29) In the study using alendronate, callus formation was increased and callus remodeling was slower, but mineralization was normal and bone strength at the fracture site was not influenced.(28) In the clinical trials in which alendronate was used, fractures occurred and there were no reports of delayed fracture healing. In a previous study, in which alendronate was given after a recent Colles' fracture, good fracture healing was observed.(25) Generally, bisphosphonates do not appear to influence fracture healing unfavorably, with the exception of high doses of etidronate.(30) As more potent bisphosphonates are being developed, it is prudent to be aware of this potentially harmful effect and to investigate it.(31)

In the placebo group, the biochemical markers of bone turnover followed a pattern already described(11,19–22): an early increase after fracture, probably reflecting increased bone turnover caused by operation, fracture repair, and immobilization. After a maximum at 8 weeks after the fracture, the bone resorption markers decreased to baseline and below baseline values. Because the baseline value in this study was the mean of two samples taken in the first 2 weeks after fracture (mean, first sample, 8 days after fracture; second sample, 10 days after fracture), a very early increase of the resorption markers may have been missed. The bone formation markers rose to a later maximum and remained increased above baseline after 1 year. This may reflect ongoing bone formation to compensate for early bone loss. In the alendronate group, an increase of bone resorption markers was not observed, but instead a decrease was seen from the first sample after treatment start, continuing to a 50% decrease after 1 year for NTX. Despite this remarkable suppression of bone resorption, bone formation showed an increase in the first 3 months after fracture and then returned to baseline, probably reflecting the phase of fracture repair. Although the study was not designed to show differences at the lumbar spine, a significant increase in BMD of the lumbar spine in the alendronate group, compared with the placebo group, was observed. Repetitive QUS measurements after a fracture were performed only in one other study(11) showing different patterns of change for BUA and VOS after an ankle fracture, with VOS, after an initial decrease, making a complete recovery to the contralateral baseline after 1 year, and BUA continuing to decrease. In this study, in the placebo group, BUA and VOS showed largely parallel patterns after fracture. Both increased from 3 months after fracture but remained below the values on the contralateral side 1 year after fracture. In the alendronate group, no difference between the fractured side and the contralateral side was observed.

Patients in the placebo group showed significant bone loss in the proximal femur on the side of the fracture. In the alendronate group, no significant bone loss occurred in the proximal femur. The difference between the alendronate and placebo groups was ∼5%. One may speculate on the clinical significance of these observations. Several studies suggest that bone loss after a fracture is a risk factor for a subsequent new fracture. In a cross-sectional study, patients who had a previous fracture in a lower limb had a greater risk of a subsequent fracture on that side than patients without a previous fracture.(12) In another study, a previous ankle fracture was associated with a twofold increase in the occurrence of new fractures in the following 30 years, and there was a 24% higher prevalence of new fractures in the injured limb than in the contralateral limb. The corresponding figures for a previous tibial fracture were a 2.5-fold increase of new fractures and a 35% higher prevalence of new fractures in the previously injured lower limb.(13) The mean difference of 5% in BMD between active treatment and placebo approximately corresponds to a ½ SD difference in T score. From the literature it is known that each SD bone loss in the proximal femur in postmenopausal women increases the risk for hip fracture by a factor of (relative risk [RR]) 2.6 (femoral neck) or 2.7 (trochanter).(32) The 5% reduction in BMD of the hip in our study can be translated to an RR of 1.6 for a hip fracture 1 year after a fracture of the lower leg. Several studies have shown that a history of any fracture (since the age of 50 years) independently increases the risk of hip fracture.(33) In a recent meta-analysis, the relative risk for hip fracture after a previous other fracture was 2.1 (95% CI, 1.2–3.5).(34) Several possible pathways might explain why previous fractures are associated with an increased risk for new fractures, independently of bone mass. A prevalent fracture may be an indicator of impaired bone quality, with a different bone structure. Another explanation is the higher bone turnover, which may be an additional risk factor for new fractures. Treatment with bisphosphonates decreases the risk of fractures early after the start of treatment, when only modest effects on BMD have occurred.(24,35,36) This suggests that mechanisms other than increase of BMD, such as rapid decrease in bone remodeling rate, play a role in the reduction of fracture risk.(35)

It is tentative to speculate that increased bone turnover and bone loss after a fracture of the lower leg contribute to a higher risk for subsequent fractures in the same limb or elsewhere in the skeleton and that the prevention of this bone loss could prevent future hip and other fractures. The exact influence of a tibial fracture on the RR for a hip fracture and the number of patients with a tibial fracture to treat to prevent one subsequent hip fracture remains a subject for additional study.


The authors thank the personnel of the Department of Traumatology for their cooperation, technicians and staff of the Department of Nuclear Medicine for the performance of the BMD measurements, staff and personnel of the Endocrinological and Central Chemical Laboratories for carrying out the measurements of the markers of bone turnover, and Greetje Asma for assistance in the organization of this study. The study was supported by a grant from Merck & Co., Inc.