Effects of androgen deprivation therapy and bisphosphonate treatment on bone in patients with metastatic castration-resistant prostate cancer: Results from the University of Washington Rapid Autopsy Series
Department of Urology, University of Washington, Seattle, WA, USA
Nearly all patients who die of advanced prostate cancer (PCa) have bone metastases.1, 2 Bone metastases are associated with bone pain, fracture, nerve and spinal cord compression, and myelophthisis.3–7 Androgen-deprivation therapy (ADT), mainstay treatment for metastatic disease in PCa patients, adversely induces osteoporosis and fragility fractures.8 Since 2004, bisphosphonates (BP) have been considered standard care for castration-resistant prostate cancer (CRPC) patients with bone metastases as they reduce the risk for skeletal related events (SREs). Zoledronic acid is the most potent of the chemical antiresorptives, and to date, no histomorphometric study evaluating the effects of an antiresorptive agent on human PCa bone metastases has been published.
To better characterize the bone changes induced by metastatic CRPC, ADT, and BP therapy on both tumor involved and uninvolved bone, we assessed the qualitative and quantitative bone parameters in 20 predetermined sites in 44 patients who died of metastatic CRPC. Each bone specimen from these 20 sites were divided into two halves; one was decalcified and paraffin embedded for qualitative assessment, and the other, nondecalcified, was directly embedded in methylmethacrylate (MMA) for quantitative bone histomorphometry. Our results confirm bone loss in uninvolved bone specimens from androgen-deprived patients and the existence of a spectrum of osteoblastic to mixed to osteolytic metastases, frequently in the same patient. Furthermore, BP increased bone volume (BV) in the osteoblastic metastases as expected but surprisingly did not decrease the number of osteoclasts in metastatic or uninvolved bone biopsy specimens. Giant osteoclasts were observed rarely in bone specimens of patients treated with alendronate and zoledronic acid. In addition, increased osteoid was observed in both high BV and low BV metastases, and osteoclast numbers were significantly higher in low BV BP-treated metastases versus high BV BP-treated metastases.
Materials and Methods
Samples were obtained from patients who died of metastatic CRPC and who signed written informed consent for a rapid autopsy to be performed ideally within 2 hours of death, under the aegis of the Prostate Cancer Donor Program at the University of Washington.9 The study was approved by the institutional review board at the University of Washington. Bone biopsy specimens of all bone sites, except ribs (a dyke cutter was used for rib biopsies), were obtained using a drill with an attached trephine (11-mm-diameter coring drill bit). These specimens were obtained from 20 predetermined anatomical sites.2 Additional cores were obtained from either additional obvious sites of bone metastases or as indicated by the last bone scintigraphy survey.
Of the 44 patients in the study, 34 were treated with bisphosphonates. Based on notes from the medical charts of multiple patients, the usual BP dosing was as follows: For zoledronic acid (27 patients, 28 if you include one patient who received alendronate and zoledronic acid), the usual regimen was 4 mg iv monthly. This was varied based on patient factors, such as renal function. Some men were in a clinical trial and received zoledronic acid every 3 weeks. For pamidronate, (5 patients), the usual regimen was 80 mg iv monthly. One patient received alendronate 70 mg orally weekly.
Bone samples for qualitative histology were fixed in formalin, decalcified in 10% formic acid, and embedded in paraffin. Bone samples for quantitative histomorphometric analysis were fixed in 70% ethanol and embedded in a methylmethacrylate (MMA) medium without decalcification.10
Histology and histomorphometry
Five-micron sections of decalcified paraffin-embedded bone samples were stained with hematoxylin and eosin and assessed by two pathologists for bone and tumor content (MR and LT). Five-micron sections of nondecalcified MMA-embedded bone samples were stained with Goldner's method for histomorphometric analysis.11 The reagent NBT/BCIP was used to evaluate alkaline phosphatase (ALP) activity. An aqueous coupling azo reactive dye mixture was used to evaluate osteoclast number (acid phosphatase [TRAcP] positivity). All histomorphometric measurements are expressed in accordance with the American Society for Bone and Mineral Research.12 Histomorphometry was performed on a 13.55-mm2 area at 20x magnification. Image analysis was conducted with an Olympus BX41 microscope using Bioquant Nova Prime software (Bioquant Image Analysis Corp., Nashville, TN, USA). In each biopsy specimen where the number of osteoclasts was >1, osteoclasts were counted and separated into <9 nuclei, 9 to 12 nuclei, 13 to 20 nuclei, or >20 nuclei groups. Woven bone area was measured using two polarizing filters on Goldner's and TRAcP-stained bone sections where BV/TV was equal or more than 45%. Mineralization was not studied in this cohort of patients.
Significance of differences was evaluated using a two-sample equal variance Student's t test with paired observations when appropriate and a two-tailed distribution. p values ≤ 0.05 indicated statistical significance.
We processed bone samples from 44 rapid autopsies of patients who died with metastatic CRPC. Patients' demographics and treatments are presented in Supplemental Table S1. The mean age at autopsy was 72.8 years. All patients had received ADT with a median treatment duration of 65.9 months. Thirty-four patients (77%) received BP therapy. Zoledronic acid was the most common BP treatment; among those treated with BP, the median treatment duration was 17.2 months (range 1 month to 70 months). One patient received alendronate and zoledronic acid, and one patient received only alendronate for 42 months. Five patients received pamidronate for 6 to 22 months. In addition, 29 patients (66%) received secondary hormonal manipulation, and 31 patients (71%) received at least one form of chemotherapy.
From a potential of 894 decalcified paraffin-embedded samples, 873 were suitable for histological analysis. Five hundred eighty-seven had tumor involvement (67%), accounting for 38 of 44 patients (82%) having at least one bone metastasis. Qualitatively, of the decalcified paraffin-embedded metastatic samples, 30% were predominantly osteoblastic, 38% were osteoblastic with a small osteolytic component, 9% were osteolytic with a small osteoblastic component, 14% were predominantly osteolytic, and 9% had no bone response to tumor (Fig. 1A). Two patients were predominantly osteolytic in all bone metastatic sites.
For nondecalcified MMA-embedded samples, of 894 potential samples, 233 (26%) of the bone samples could not be evaluated because of technical reasons. The remaining 661 MMA-embedded samples from 44 patients were analyzed. The main histomorphometric parameters that we measured were: BV, new unmineralized bone formation (ie, osteoid volume), osteoclast numbers, and ALP. Four hundred forty-three of the 661 MMA-embedded samples had tumor involvement (67%), accounting for 37 of 44 patients (84%) with at least one bone metastasis (Fig. 1B, showing bone volume in all specimens, metastatic or not). This is similar to but not identical to the 64% tumor involvement noted in the paraffin-embedded samples because of different portions of each bone core being used for these two processing methods.
We calculated a weighted mean for BV based on the most comprehensive published analysis of BV in the aging male.13 This weighted mean was based on the age profile of the patients. The weighted mean average BV for our cohort based on age was 14.7%. Of the 661 available MMA-embedded samples, 3 were excluded as statistical outliers, 215 (33%) had no tumor with a mean BV of 10.5% (range 2.33% to 25.42%). Of those, 176 (82%) had <14.7% BV and 39 (18%) had >14.7% BV, indicating that ADT induces bone loss. In addition, of the 661 available samples, 443 (67%) had tumor as described above, with 147 (33%) containing tumor with <14.7% BV and 296 (67%) containing tumor with >14.7% BV, indicating a predominantly osteoblastic response in the PCa bone metastases (Supplemental Table S2).
The 443 tumor-containing samples were compared with the 215 samples without tumor for various histomorphometric characteristics (Table 1). The mean tumor volume as a percent of tissue volume in the metastatic sites was 24.7%. Overall, there was a considerable amount of bone present in the metastatic sites (range 2.8% to 76.6%; mean 25.7%; %BV/TV), compared with uninvolved bone (range 2.3% to 25.4%; mean 10.5%; %BV/TV) (Table 1). The significant increases in bone volume (p < 0.001), bone surface (p < 0.001), trabecular thickness (p < 0.001), and trabecular number (p < 0.001) in metastatic bone sites were the result of the addition of woven bone and osteoid onto preexisting trabeculae and in between a nest of cancer cells within the bone marrow microenvironment. Osteoid but no woven bone was observed in purely osteolytic lesions. As expected, BV (p < 0.001), osteoid area (p < 0.001), and ALP (p < 0.001) activity were also higher in the metastatic sites versus uninvolved sites, with 10-fold more osteoid area and more than double the percentage of metastatic sites being positive for ALP compared with the uninvolved bone sites. The osteoid volume compared with total BV was fivefold higher in the metastatic sites (Table 1). No difference was found in osteoid volume between BP-treated and untreated bone specimens. As expected, osteoclast numbers were significantly higher (p = 0.005) in metastases than in uninvolved bone (Table 1). However, the osteoclast number per bone surface unit was no different (p = 0.660) in metastases and uninvolved bone, revealing that both involved and uninvolved bone had similar osteoclast coverage on the bone surface (Table 1).
Table 1. Histomorphometric Analysis of Metastatic and Nonmetastatic Bone Specimens From 44 Patients From the CRPC Rapid Autopsy Series
Bone sites with tumor (n = 443) and no tumor (n = 215) were assessed for quantitative tumor volume and bone response. The percentage of bone volume in tissue volume (%BV/TV), bone surface (BS; mm), percentage of bone surface in tissue volume (%BS/TV), trabecular thickness (Tb.Th; µm), trabecular number (Tb.N; 1/nm), trabecular separation (Tb.Sp; µm), percentage of tumor volume in tissue volume (%TuV/TV), osteoid volume (OV; mm2), total bone volume (Tt.BV; mm2), ratio of osteoid volume to total bone volume (OV/Tt.BV; %), alkaline phosphatase activity (ALP), osteoclast number (N.Oc), and ratio of osteoclast number to bone surface (N.Oc/BS) were measured in each sample. BV/TV demonstrates that metastatic bone biopsy specimens have an overall osteoblastic phenotype (25.7% versus 14.7%, which is considered as normal BV for age-matched patients). Nonmetastatic bone biopsy specimens have an osteoporotic phenotype with a BV/TV of 10.5%. Greater osteoid but similar osteoclast numbers on the bone surface in involved bone as in uninvolved bone demonstrates higher bone accumulation is associated with metastasis when compared with uninvolved bone.
No. of samples
Tt. BV (mm2)
OV/Tt. BV (%)
ALP (% positive)
Our patients had an extremely wide range of BP treatment times (0.8 to 70 months). An analysis was completed for all BP versus nonbisphosphonate-treated (NBP) samples. Metastatic sites in patients receiving BP had a similar tumor volume as equivalent metastatic sites in patients not receiving BP treatment (Table 2). BV as a percentage of tissue volume (%BV/TV) was significantly greater in BP-treated patient metastatic sites (p < 0.001) when compared with metastatic sites of NBP patients (Supplemental Fig. S1A—D). There was a significant increase in mean osteoid volume in BP-treated metastatic lesions (p = 0.005), with fewer ALP-positive sites present in patients receiving BP. There was no significant difference in the number of TRAcP-positive osteoclasts in NBP and BP metastatic specimens, but there was a significant difference in osteoclast coverage of bone surface (p = 0.048) in metastatic sites from patients receiving BP treatment versus metastatic sites from NBP patients (Table 2). In uninvolved samples, BP treatment had no effect on BV (p = 0.209), with no significant effect on osteoid volume, ALP-positive sites, and osteoclast numbers (Table 2). In addition, we also elected to confine this analysis to those patients who had been treated for a minimum of 12 months. The results were similar (Supplemental Table S3).
Table 2. Histomorphometric Analysis of Bone Specimens From Bisphosphonate- and Nonbisphosphonate-Treated Patients
Metastatic and nonmetastatic specimens from bisphosphonate (BP)- and nonbisphosphonate (NBP)-treated patients were assessed by histomorphometry. The histomorphometric parameters are described in Table 1. Bone volume (BV) is greater in metastatic specimens treated with BP than in metastatic NBP specimens with no change in the number of osteoclasts. In uninvolved specimens, there is a trend to greater BV in BP-treated specimens than in NBP specimens with no change in the number of osteoclasts.
No. of samples
Tt. BV (mm2)
OV/Tt. BV (%)
ALP (% positive)
To determine if the site of skeletal metastasis influenced the relationship between the tumor and the corresponding bone response, bone histomorphometric parameters in the long bones, vertebrae, sacrum, and other sites, ie, sternum, ribs, and clavicle, were compared (Supplemental Table S4). Fifty-four percent of the long bones had tumor compared with 73%, 74%, and 69% in vertebrae, sacrum, and other sites, respectively. Larger numbers of osteoclasts were present in the sternum, ribs, and clavicle specimens versus the other bone specimens, reaching 67 per specimen in a 7-month zoledronic acid–treated patient (Supplemental Table S4). The high osteoclast numbers may be associated with increased bone remodeling in these more flexible bone sites (Supplemental Table S4).
Histomorphometric features of metastatic sites from patients treated with BP with >14.7% or <14.7% BV were compared to determine if features other than BV could explain differences in bone formation (Supplemental Table S5). Besides a decrease in ALP activity in the <14.7% BV samples, no obvious factors were observed. A similar analysis mentioned above was done for patients who received BP or NBP with <14.7% or >14.7% BV, and again, no significant differences were observed (Supplemental Table S2). Based on these data, the challenge to identify histomorphometric parameters that are significantly different between the >14.7% and <14.7% BV samples reaches an unanticipated level of complexity.
Therefore, to better define the role of the osteoblasts and osteoclasts in the bone remodeling process, extreme ends of the spectrum were examined, ie, highly osteoblastic (>35.3% BV) and highly osteolytic (<9.1% BV) metastases in BP-treated patients (Table 3). We selected these cutoffs of 35.3% and 9.1% because they were the mean BV measurements in metastatic sites from patients who received BP with >14.7% and <14.7% BV (Supplemental Table S5). The mean %BV/TV was 49.9% in extremely osteoblastic samples and 6.5% in extremely osteolytic sites (Table 3). Remarkably, even with this significant difference in %BV/TV, osteoid volume and ALP positivity were similar between these two groups. However, the osteoid bone coverage was nearly fourfold higher in the osteolytic metastatic sites, indicating an important bone-forming process in BP-treated bone.
Table 3. Histomorphometric Analysis of Highly Osteoblastic and Highly Osteolytic Metastatic Bone Specimens From Bisphosphonate-Treated Patients
This subgroup of 16 patients received BP treatment for a minimum of 1 year. Bone volume >35.3% versus <9.1% assesses the significant differences between highly osteoblastic and highly osteolytic castrate-resistant BP-treated metastatic bone sites. Histomorphometric parameters are described in Table 1. Two extremes of the bone phenotype spectrum in prostate cancer metastasis are shown with no difference in osteoid volume between the two extremes; however, significantly more osteoclasts relative to bone surface were present in osteolytic lesions.
Tt. BV (mm2)
OV/Tt. BV (%)
ALP (% positive)
A marked increase in osteoclast number (p = 0.057) and osteoclast coverage of the bone surface (p = 0.003) was observed in the extremely osteolytic sites together with a high osteoid bone coverage, indicating a higher bone turnover in osteolytic lesions compared with osteoblastic lesions (Table 3; Supplemental Fig. S1E, F; Supplemental Fig. S2). Similar results were observed when we assessed patient-paired osteoblastic and osteolytic bone biopsy specimens from a much smaller subset of 5 BP-treated patients with metastatic bone sites that had both extremely osteoblastic (>35.3% BV) and extremely osteolytic (<9.1% BV) metastases (Supplemental Table S6).
To determine BP effects on osteoclasts, we compared the number of osteoclasts found in metastatic biopsy samples from patients untreated and treated with pamidronate, alendronate, and zoledronic acid for <1 year, <2 years, <3 years, and >3 years. There was no significant difference between the number of osteoclasts in biopsy samples from each of the groups (Fig. 2A). Pamidronate-treated metastases had the lowest number of osteoclasts. Alendronate had the highest number of osteoclasts, but could not be statistically compared with the other groups because only one patient was treated with alendronate. It is noteworthy that all metastases of one patient in each BP and NBP group had no osteoclasts, suggesting that individual PCa's may induce no osteoclasts at all.
To define giant osteoclasts, we sorted the osteoclasts by the number of nuclei visible on the metastatic sections. When we compared the BP groups defined above, we found that 8-or-less-nuclei and 9-to-12-nuclei osteoclasts were observed in all groups representing the bulk of the osteoclasts present in the metastases (Fig. 2B). Giant osteoclasts, defined as either containing 13 to 20 nuclei and more than 20 nuclei, were observed in the patient treated with alendronate, as well as in all zoledronic acid–treated patient groups (Fig. 3C, E–G), they were absent in the pamidronate-treated and in NBP-treated groups. Concerning osteoclast apoptosis, we found only 3 osteoclasts (out of 1971 reviewed) with pyknotic nuclei in a 4-year zoledronic acid–treated metastasis (Fig. 3H). However, we noticed that many osteoclasts from the zoledronic acid group had atypical nuclei with angulated nuclear membrane, clear open chromatin, and one large irregular nucleolus in each nucleus (Fig. 3E).
To further analyze bone formation in the metastatic specimens from CRPC patients, we measured the woven bone component in 78 biopsies that had more than 45% BV/TV. This was chosen as a cut-off because 45% BV/TV and above is without a discussion an increase of bone mass. Woven bone was found occupying from 15.6% to 98.5% of total bone in osteoblastic biopsies with an average of 88.1%. No difference was observed in woven bone between metastatic specimens from BP- and NBP-treated patients. Woven bone was formed de novo from the tumor stroma in the bone marrow or deposited directly on lamellar trabeculae (Fig. 3I, J).
At least four processes are known to influence bone structure in PCa bone metastasis: (1) normal bone physiology, (2) systemically and locally secreted tumor factors, (3) androgen deprivation therapy, and (4) BP treatment.
The physiological and perpetual systemic bone remodeling involves balanced and coupled actions of osteoblasts and osteoclasts. All 44 patients in our study were on ADT, and the results show that ADT induces serious bone loss similar to postmenopausal osteoporosis in women by systemic activation of osteoclasts. As expected, we observed a huge increase in the number of osteoclasts per uninvolved biopsy specimen.13 BPs, because of their high affinity for skeletal mineralized bone matrix, induce a detrimental effect on osteoclast biology and a systemic and local loss of osteoclast activity. Thus BP treatment has been shown to prevent bone loss in men receiving ADT.14 However, in our study, in uninvolved bone biopsy specimens, BPs were not able to significantly decrease osteoclast number. Consequently, %BV/TV averaged 11.0% in those patients with BP treatment, which was not significantly different from %BV/TV (10.1%) measured in bone specimens from NBP-treated patients. These low numbers contribute to the fragility fractures observed in ADT patients (Table 2).
Regarding the elevated osteoclast numbers in BP-treated patients (Table 2 and Fig. 2), we found that a small percentage of those osteoclasts were giant osteoclasts, defined here as hypernucleated osteoclasts with greater than 13 nuclei. Previously, it was reported that long-term alendronate treatment is associated with an increase in the number of osteoclasts, which include distinctive giant hypernucleated detached osteoclasts that are undergoing protracted apoptosis.15 Similarly, we found in a single alendronate-treated patient that the number of osteoclasts per biopsy specimen was the highest of all BP-treated bone metastases, with 157/mm2 osteoclasts in a metastatic sample. We did not morphologically observe pyknotic nuclei in those alendronate-treated osteoclasts but did not use immunohistochemistry to test DNA fragmentation in our plastic embedded sections. In addition, we did not often observe giant osteoclasts with pyknotic nuclei in bone biopsies from patients treated long term with zoledronic acid. One explanation for the presence of the giant osteoclasts in BP-treated specimens is that they may have lost their basic functions by perturbation of the mevalonate pathway but have not as yet disappeared from the bone surface.15, 16 The study of apoptosis markers in osteoclasts in paraffin sections of metastasis specimens of this series is warranted.
Local tumor-secreted factors, still to be identified, induce the characteristic osteoblastic phenotype of PCa metastasis. Our study, confirming a pilot study previously reported from 12 patients,2 show that the osteoblastic phenotype can manifest as a spectrum of osteoblastic to osteoblastic/osteolytic (mixed) and osteolytic bone responses observed in the same patient at different biopsy sites. Not surprisingly, this study clearly demonstrates that PCa in the bone is predominantly osteoblastic, with an osteolytic component (Fig. 1). However, this spectrum of lesions has not been previously reported in such detail. Bone scans preferentially expose osteoblastic lesions, thereby underestimating osteolytic lesions. The existence of predominantly osteolytic lesions in a bone site together with osteoblastic lesions in another bone site can be explained either by (1) a maturation of lesions, ie, osteolytic first and then osteoblastic, or osteoblastic and then osteolytic, or (2) the presence of multiple tumor clones secreting different bone acting factors. This study lacks the knowledge from sampling different bone metastases in each patient at different times as the sampling was made for all patients at death.
In addition to the osteoblastic response of the bone to PCa cells, a high bone turnover is observed in tumor-involved bone biopsy specimens as demonstrated by the presence of a high volume of osteoid and a high number of osteoclasts, both in BP-treated and NBP-treated patients. Osteoid, defined as collagen 1 before mineralization, is fourfold higher in metastases than in uninvolved bone, indicating the ongoing osteoblastic process in response to the local presence of prostate tumor cells. This suggests that bone response is a localized event. If physiologic bone coupling was maintained, then we would expect that an increase in new bone formation would be associated with an increase in osteoclast number. However, in low BV metastases, osteoid bone coverage and osteoclast number are much higher than in high BV metastases, suggesting that normal physiologic coupling is perturbed by the tumor (Table 3; Supplemental Fig. S2).
Results of this study clearly demonstrate the complex biology taking place in CRPC bone metastases, including the dramatic loss of bone induced by ADT, the dramatic addition of woven bone, the limited efficacy of BPs to change the bone responses, and the need to adjust BP treatment to individual tumor-related resorption. Finally, this study highlights the persistent need to identify the factor(s) that promote osteoblast activity so that better targeted therapies can be delivered clinically to improve the quality of life of patients suffering from CRPC skeletal metastases.17, 18–20
MPR was employed by Amgen Inc. within the past 36 months before submission of this manuscript. All other authors state that they have no conflicts of interest.
We thank the patients and their families who were willing to participate in the Prostate Cancer Donor Program, for without them, research of this nature would not be possible. We also acknowledge Bruce Montgomery, MD; Peter Nelson, MD; Susan Ott, MD; the rapid autopsy teams; and students in the Urology Department at the University of Washington. This material is the result of work supported by resources from the VA Puget Sound Health Care System, Seattle, WA (RLV is a Research Career Scientist; PHL is a staff physician), the Pacific Northwest Prostate Cancer SPORE (P50CA97186), the NIH grant PO1CA085859, and the Richard M LUCAS Foundation. CM is a recipient of the Career Development Award from the Pacific Northwest Prostate Cancer SPORE (P50CA097186).
Authors' roles: CM accepts responsibility for the integrity of the data. MPR, LDT, PHL, CSH, and RLV were responsible for the conception of the experiments. CM, AD, and MPR were responsible for conducting the experiments. CM, AD, MPR, MK, CW, EC, and RLV were responsible for analyzing data. CM, MPR, AD, LDT, MK, CW, EC, PHL, CSH, and RLV all participated in the drafting of the manuscript.