Bone remodeling in postmenopausal women who discontinued denosumab treatment: Off-treatment biopsy study


  • The results of this study were submitted in abstract form to the IOF WCO/ECCEO (European Congress on Clinical and Economic Aspects of Osteoporosis and Osteoarthritis) (2010, poster presentation), ASBMR (American Society of Bone and Mineral Research) (2010, oral presentation), ACR (American College of Rheumatology) (2010, oral presentation), and EULAR (European League Against Rheumatism) (2011, oral presentation).


Denosumab is a fully human monoclonal antibody that neutralizes the activity of RANKL, leading to the inhibition of osteoclast maturation, bone-resorbing activity, and survival. Evaluation of trans–iliac crest bone biopsy specimens in the phase 3 pivotal fracture study with denosumab in postmenopausal women with osteoporosis showed evidence of reduced bone turnover at the tissue level in subjects receiving denosumab, and up to one-third of subjects did not have evidence of tetracycline labeling in trabecular or cortical bone. Discontinuation of denosumab therapy has demonstrated that the effects of denosumab are reversible, as assessed by biochemical markers of bone turnover (BTM) and BMD. The precise nature of changes that occur at the tissue level with denosumab discontinuation have not been explored. Fifteen subjects were enrolled in a cohort study to evaluate the effects of denosumab discontinuation at the tissue level. Subjects had discontinued osteoporosis treatment for a mean time of 25.1 months (range 21 to 29 months). Bone histomorphometry results were compared with results from placebo-treated women with osteoporosis in the denosumab phase 3 pivotal fracture bone biopsy substudy, and BTMs were compared with subjects' pretreatment values. The results of this study showed normal histology and bone remodeling similar to those observed in untreated postmenopausal women with osteoporosis. With treatment cessation, 100% of biopsy specimens had evidence of tetracycline labels. Biochemical markers were comparable to and highly correlated with pretreatment levels. These data confirm that the effects of denosumab on bone turnover at the tissue level are fully reversible. © 2011 American Society for Bone and Mineral Research


The rate of bone resorption increases with age and with estrogen loss at menopause, potentially leading to bone loss and increased risk of fracture. Osteoporosis treatments have been developed to inhibit bone resorption. Denosumab is a fully human monoclonal antibody with high affinity and specificity for RANK ligand (RANKL), an essential mediator of the formation, activity, and survival of osteoclasts. By neutralizing the activity of RANKL, denosumab inhibits osteoclast-mediated bone resorption with subsequent increases in BMD. Denosumab significantly reduces the risk of new vertebral fractures, nonvertebral fractures, and hip fractures, as demonstrated in a 3-year randomized, double-blind, placebo-controlled phase 3 trial in women with postmenopausal osteoporosis (FREEDOM study).1 The study also showed that denosumab reduced bone turnover and increased BMD at the lumbar spine and total hip compared with placebo.

The effects of denosumab on bone resorption and BMD are reversible on discontinuation of treatment. Denosumab discontinuation was associated with initial increases in markers of bone turnover (BTM) to levels above baseline, but BTM levels remained within the range for placebo-treated postmenopausal women and approached pretreatment levels by 48 months after therapy cessation.2, 3 Gains in BMD achieved with denosumab treatment generally returned to pretreatment levels at all measured sites by 36 months following discontinuation but remained above levels seen in subjects in the placebo group, who continued to lose bone during the study.2, 3

Evaluation of trans–iliac crest bone biopsy specimens from subjects receiving denosumab in the FREEDOM study revealed expected reductions in bone turnover at the tissue level, confirming the global reductions in bone turnover observed through BTM measurements.4 Up to one-third of subjects treated with denosumab did not have evidence of tetracycline labeling in trabecular or cortical bone.4 However, there have not been any histologic and histomorphometric studies to show tissue-level changes following denosumab discontinuation. The objective of this study was to characterize the effects of denosumab discontinuation on bone remodeling, as assessed by bone histology and quantitative histomorphometric analysis as well as BTM measurements.


Study participants

Subjects who participated in this study had completed denosumab treatment in one of two clinical trials, study 15 and study 26 (noted as parent study 1 and parent study 2, respectively), for which the primary results have been reported previously.5, 6 Both these studies enrolled postmenopausal women with low bone mass. The BMD requirement for eligibility varied slightly; in parent study 1, eligible subjects were required to have a baseline BMD T-score at the spine or total hip of −2.0 to −3.0, whereas in parent study 2, eligible subjects had a baseline BMD T-score at the spine or total hip of −2.0 or below. All subjects provided written informed consent before participating in this off-treatment biopsy study, which was performed in compliance with the International Conference on Harmonization (ICH) guidelines for good clinical practice standards.

Study design and procedures

This was an international single-arm, noninterventional cohort study. The trial included a screening visit, two cycles of tetracycline dosing, and a trans–iliac crest bone biopsy procedure, with a follow-up visit 7 days after the biopsy procedure (Fig. 1). No study drug was administered.

Figure 1.

Study schema.

Subjects scheduled for a biopsy followed a double tetracycline labeling procedure as follows: tetracycline hydrochloride for the first 3-day cycle, a 14-day interval without tetracycline, tetracycline for a second 3-day cycle, followed by the biopsy procedure 5 to 14 days after the last dose of tetracycline. The dose of tetracycline was 250 mg four times daily for 3 days. Subjects were instructed to take each dose 1 to 2 hours before a meal or 2 hours after a meal. Urine was collected on days 3 and 20 of the study for tetracycline derivative measurements (Esoterix Laboratories, Austin, TX, USA) to ensure compliance with the tetracycline regimen.

Trans–iliac crest bone biopsy specimens were obtained from the anterior iliac crest using a Rochester bone biopsy trephine (diameter 7 to 8 mm) (Medical Innovations International Inc., Rochester, MN, USA). Specimens were stored and shipped in 70% ethanol and dehydrated and embedded in glycol methacrylate at the Bone Histomorphometry Laboratory of the Mayo Clinic (Rochester, MN, USA), where histomorphometric analyses were performed, as described previously.4

Blood samples were drawn for hematology and serum chemistry analysis. All blood samples were collected in a nonfasting state except for samples collected for evaluation of serum C-telopeptide (sCTX) and procollagen type 1 amino-terminal propeptide (P1NP), which were collected from subjects in the fasting state before noon on day 3 or day 20 and within 24 hours of the last dose of the respective tetracycline cycle. Serum CTX and P1NP levels were assessed using the same assay methodology used in parent studies 1 and 2 (sCTX, Serum Crosslaps ELISA; Nordic Bioscience Diagnostics A/S, Herlev, Denmark; assay conducted by Covance Laboratories, Princeton, NJ, USA; P1NP, Orion radioimmunoassay, Orion Diagnostica Oy, Espoo, Finland; assay conducted by Covance Laboratories), but samples from the off-treatment biopsy study and parent studies 1 and 2 were not batched within the same assay kit for evaluation.


Adverse events or new conditions that occurred during the study after signing of the informed consent, including those related or unrelated to study procedures (ie, tetracycline administration and bone biopsy procedures) were monitored and collected.

Statistical analysis

All analyses for this study were descriptive. Since this cohort study did not enroll a control group, a descriptive comparison with placebo-treated subjects enrolled in the biopsy substudy for FREEDOM was tabulated (comparator group).4 The analyses of all histomorphometric indices (for subjects in this study and for the comparator group) were performed by the same laboratory (Mayo Clinic). Since the 2.5 and 97.5 percentiles for histomorphometric measurements were either the same or very similar to the minimum and maximum values observed in the comparator group, the observed maximum and minimum values were used as the comparison instead of the 95% CI.

Agreement between BTM levels before denosumab treatment initiation and after denosumab discontinuation (in individual subjects) was assessed using Lin's concordance correlation.7, 8 The Lin coefficient measure combines measures of accuracy (the proximity of the regression line through the data points to the line of perfect concordance, which is at 45 degrees) and precision (the tightness of the data about the regression line through the data) to determine whether observed data significantly diverge from the line of perfect concordance. The value of Lin's coefficient increases in relation to the accuracy and precision of the observed data.

Categorical variables are presented as frequencies and percentages. Continuous variables are summarized as means, SDs, medians, and minimum, maximum, and interquartile ranges (IQRs).


A total of 15 subjects were enrolled: 10 subjects from parent study 1 and 5 subjects from parent study 2. The mean (SD) age for subjects was 62.1 (5.9) years, with a mean (SD) time of 15.8 (7.9) years since menopause (Table 1). Mean time since denosumab discontinuation, defined as 6 months after the last denosumab injection prior to the start of the first tetracycline labeling cycle, was 25.1 months (range 21 to 29 months). Mean serum 25-hydroxyvitamin D [25(OH)D] concentration was 79.5 nmol/L (SD 17.3 nmol/L; range 56.2 to 104.1 nmol/L) . None of the 15 subjects had received bisphosphonate treatment since discontinuing the parent study. Demographics for the comparator group of placebo-treated subjects from the FREEDOM study are also shown in Table 1. Of note, the comparator group was older (70.0 years) and had a longer mean time since menopause (23 years).

Table 1. Subject Baseline Demographics
 Off-treatment biopsy subjects (n = 15)Comparator groupa (n = 45)
  • N/A = not applicable.

  • a

    Baseline results from placebo-treated subjects in FREEDOM study.4

  • b

    Based on results at baseline in parent studies (Seeman and colleagues5 and Brown and colleagues6).

  • c

    Based on month 12 results in parent studies.

Age, years, mean (SD)62.1 (5.9)70.0 (6.3)
Age groups, n (%)
 ≥65 years5 (33.3)37 (82.2)
 <65 years10 (66.7)8 (17.8)
Years since menopause, mean (SD)15.8 (7.9)23.0 (8.1)
25(OH)D, nmol/L, mean (SD)79.5 (17.3)63.4 (26.2)
Time since denosumab treatment discontinuation, months, mean (min, max)25 (21, 29)N/A
BMD T-score, mean (SD)
 Lumbar spine−2.2 (0.4)b/−1.8 (0.5)c−2.9 (0.5)
 Total hip−1.3 (0.8)b/−1.0 (0.8)c−1.8 (0.8)
 Femoral neck−1.6 (0.7)b/−1.5 (0.6)c−2.0 (0.7)

All biopsy specimens were evaluable for histology and showed normal architecture and quality without evidence of mineralization defects, woven bone, or marrow fibrosis (Table 2). Fourteen of the 15 biopsy specimens were evaluable for histomorphometry; one specimen had a crush artifact and was not evaluable for all histomorphometric variables.

Table 2. Histology and Tetracycline Evaluation
 Off-treatment biopsy subjects (n = 15)
Histology, n (%)
 Normal lamelar bone15 (100)
 Normal mineralization15 (100)
 Marrow fibrosis0 (0)
 Woven bone0 (0)
 Osteomalacia0 (0)
Tetracycline label status, trabecular or cortical bone, n (%)
 Any label15 (100)
 Any double label14 (93)
 Any double label in trabecular region13 (87)
 Any double label in cortical region14 (93)
 Only single label1 (7)
 No label0 (0)

The presence of tetracycline labels was assessed to calculate dynamic histomorphometric variables. Urinanalysis confirmed that all 15 subjects were compliant with tetracycline administration. Thirteen specimens had double label in trabecular and cortical bone, and one specimen had single label in trabecular bone and double label in cortical bone (Table 2). The specimen with crush artifact and fragmented cortex had evidence of single label in trabecular bone and no label in cortical bone. An example of a tetracycline-labeled biopsy specimen is provided in Fig. 2.

Figure 2.

Tetracycline-labeled bone biopsy specimen (65-year-old female 21 months off treatment).

The median and IQR values for static and dynamic histomorphometry variables are shown in Figs. 3 through 5, along with ranges for bone histomorphometric parameters for the comparator group.4 Values for all static and dynamic histomorphometry variables for biopsy specimens obtained in this study generally were within the range of the comparator group. Median levels of sCTX and P1NP—evaluated in these 15 study subjects approximately 2 years after discontinuation of denosumab—had returned to levels comparable with predenosumab treatment levels (ie, baseline levels in studies 1 and 2; Fig. 6).

Figure 3.

Static histomorphometric indices of bone formation and resorption. Data are shown as box and whisker plots with the interquartile range as the box, the median as a line, and the range of the nonoutlier observations as whiskers. Dashed reference lines show median, minimum, and maximum in the comparator group.4

Figure 4.

Dynamic histomorphometric indices of bone formation. Data are shown as box and whisker plots with the interquartile range as the box, the median as a line, and the range of the nonoutlier observations as whiskers. Dashed reference lines show median, minimum, and maximum in the comparator group.4

Figure 5.

Dynamic histomorphometric indices of bone formation and activation frequency. Data are shown as box and whisker plots with the interquartile range as the box, the median as a line, and the range of the nonoutlier observations as whiskers. Dashed reference lines show median, minimum, and maximum in the comparator group.4

Figure 6.

sCTX (A) and P1NP (B) levels of subjects at the beginning and end of the 12-month period of the parent studies and for the off-treatment biopsy study.

The relationship between pretreatment bone remodeling levels (ie, baseline values for studies 1 and 2) and bone remodeling levels in the current cohort was evaluated. The analysis showed a moderate correlation between pretreatment and postdiscontinuation BTM values; for sCTX and P1NP, the Lin correlations were 0.80 (0.48, 0.93) and 0.59 (0.15, 0.81), respectively (Fig. 7).

Figure 7.

Correlation between parent study baseline and off-treatment BTMs for the 15 subjects.


No serious adverse events were reported. All subjects experienced biopsy-related adverse events, with procedural pain (12 subjects, 80%) and postprocedural hematoma (6 subjects, 40%) being the most commonly reported. Other infrequently reported procedure-related adverse events included dyspepsia, hot flush, myalgia, and vomiting.


Denosumab is a fully human monoclonal antibody that binds RANKL, preventing RANKL activation of RANK and thereby inhibiting the formation, activity, and survival of osteoclasts. This activity is demonstrated biochemically in bone resorption, as reflected by sCTX,5, 6 and treatment cessation has shown that the effect of denosumab therapy is fully reversible.2, 3 Results from our cross-sectional bone biopsy study demonstrated that the histomorphometric indices in biopsy specimens of subjects who discontinued denosumab were consistent with those derived from a postmenopausal population with osteoporosis, and BTMs returned to pretreatment baseline levels.

In this evaluation, we used as our reference group the placebo-treated subjects who participated in the biopsy substudy of the denosumab pivotal phase 3 fracture trial.4 Demographics from this comparator group indicated some differences from those of the current study, notably older age and lower BMD at study entry. This being acknowledged, the chosen comparator group provided a useful and relevant reference point from which to interpret our findings. We found that all static and dynamic parameters were within range of the group of untreated postmenopausal women with osteoporosis, and there was no evidence of decay in structural parameters. Denosumab potently reduces bone resorption, as shown by BTMs and histomorphometry,4 and has been associated with large increases in BMD and a significant reduction in osteoporosis-related fractures. With treatment cessation, assessments of bone remodeling at the tissue level in this study showed a fully reversible effect that was not associated with deleterious effects on bone microstructure.

Since pretreatment biopsy specimens are not available for these subjects, the small, cross-sectional nature of this cohort study does not provide an in-depth evaluation into baseline level of bone remodeling at the tissue level. This being said, the reference population of placebo-treated subjects from the pivotal phase 3 fracture trial4 provides some insights into level of remodeling in this study because both groups involved untreated postmenopausal women. The coupling between bone resorption and bone formation was preserved in the postdiscontinuation period, patients with high resorption parameters having high bone formation parameters, and vice versa. All patients, including those with the highest osteoclast number, had normal osteoclast morphology and size/nuclear number. Furthermore, there was a moderate correlation between pretreatment and postdiscontinuation BTM values for sCTX and P1NP consistent with a return to the pretreatment baseline bone remodeling levels.

Subjects enrolled in this cohort study had discontinued denosumab for approximately 2 years. Evaluation of global remodeling, as reflected by sCTX and P1NP, demonstrated that the level of turnover observed in these individuals was similar to that of the pretreatment evaluation. While serum samples were batch run at different time points, which may have introduced some variability, there was evidence of a modest correlation between off-treatment and baseline levels. This observation suggests that the skeletal mechanosensory system remained responsive during and after denosumab treatment, an observation with denosumab that has been noted elsewhere.2, 3 The effects of pharmacologic agents often are not sustained once treatment is discontinued, and this has been observed in the management of many diseases, including hypertension and diabetes mellitus. Reversibility with other antiresorptive therapies has been observed with hormone-replacement therapy and selective estrogen receptor modulators9–19 and, more recently, with odanacatib, a selective cathespin K inhibitor.20 Reversibility with bisphosphonate treatment is a function of adsorption affinity for hydroxyapatite. Prompt increases in bone turnover and declines in BMD have been observed 12 months after discontinuation of risedronate21 compared with more gradual changes over a few years after discontinuation of alendronate.22 Denosumab, a soluble inhibitor of RANKL, does not incorporate into bone matrix, and the observed reversibility of effect is similar to that of therapies that do not have binding affinity for hydroxyapatite in matrix, including hormone therapy.

In conclusion, the effects of denosumab on bone remodeling are fully reversible, as assessed by biochemical markers of bone turnover and histomorphometry. These data suggest that subjects who discontinue denosumab return to their pretreatment level of bone remodeling structural parameters at the tissue level.


JPB has received research grants, consulting fees, or speakers' bureau fees from Abbott, Amgen, Bristol-Myers Squibb, Eli Lilly, Merck, Novartis, Pfizer, Roche, Sanofi-Aventis, Servier, and Warner-Chilcott. DWD has received research grants, consulting fees, and speakers' bureau fees from Amgen, Eli Lilly, Merck, and Genentech. were BD, RDA, JSM, AG, and RBW are employed by Amgen and may own Amgen stock or stock options. JZ has received consulting fees from Amgen, Eli Lilly, Merck, Pfizer, and Servier.


This study was sponsored by Amgen, Inc. Yeshi Mikyas of Amgen, Inc., provided writing assistance for this article.

Authors' roles: Drs. Brown and Zanchetta were investigators in the study and enrolled patients. Dr. Dempster provided expert interpretation of the bone biopsy data. Drs. Brown, Dempster, Dent-Acosta, San Martin, Grauer, and Wagman were involved in the design and conduct of the study. Dr. Ding did the statistical analysis for the study. All were involved in the data interpretation and made substantial contribution in the writing of this manuscript.