Comparison of 10 serum bone turnover markers in prostate carcinoma patients with bone metastatic spread: Diagnostic and prognostic implications



Our aim was to assess the diagnostic accuracy of bone markers in serum of patients with prostate cancer (PCa) for early detection of bone metastases and their usefulness as predictors of PCa-caused mortality. In sera of 117 PCa patients (pN0M0, n = 39; pN1M0, n = 34; M1, n = 44), 35 healthy men and 35 patients with benign prostatic hyperplasia, bone formation markers [total and bone-specific alkaline phosphatase (tALP, bALP), amino-terminal procollagen propeptides of type I collagen (P1NP), osteocalcin (OC)], bone resorption markers [bone sialoprotein (BSP), cross-linked C-terminal (CTX) and cross-linked N-terminal (NTX) telopeptides of type I collagen, tartrate-resistant acid phosphatase isoenzyme 5b (TRAP)] and osteoclastogenesis markers [osteoprotegerin (OPG), receptor activator of nuclear factor κB ligand (RANKL)] were measured. tALP, bALP, BSP, P1NP, TRAP, NTX and OPG were significantly increased in PCa patients with bone metastases compared to patients without metastases. OPG showed the best discriminatory power to differentiate between these patients. Logistic regression analysis resulted in a model with OPG and TRAP as variables that predicted bone metastasis with an overall correct classification of 93%. Patients with concentrations of OPG, P1NP, tALP, bALP, BSP, NTX, TRAP and CTX above cut-off levels showed significantly shorter survival than patients with low marker concentrations. Multivariate Cox proportional hazards regression revealed that only OPG and BSP were independent prognostic factors for PCa-related death. Thus, the importance of serum OPG in detecting bone metastatic spread, alone or in combination with other bone markers, and predicting survival in PCa patients has been clearly demonstrated. © 2004 Wiley-Liss, Inc.

The most frequent cancer in men, PCa, is characterized by the occurrence of skeletal metastases in about 65–75% of patients with advanced disease.1 Bone metastases alter the balance between bone formation and bone resorption by influencing the involved bone cells (osteoblasts and osteoclasts) through local release of cytokines and growth factors. To detect and monitor this metastatic bone involvement, bone scintigraphy is the widely applied standard method. Altered bone remodeling activity can also be assessed either directly by measuring components of the affected bone cells (osteoblasts and osteoclasts) or indirectly by analyzing metabolic products released from the bone matrix following the changed bone formation or resorption. Bone turnover markers that reflect either bone formation in consequence of osteoblast proliferation or analytes indicating bone resorption as the opposite bone remodeling activity have been recommended as tools in the assessment of bone metastasis in PCa.2, 3, 4, 5, 6, 7 The balance between osteoblastic and osteoclastic activity in bone is essentially determined by osteoclastogenesis, which is regulated by 3 proteins: RANK, RANKL and OPG. OPG and RANKL could be, in addition to bone formation and resorption markers, biomarkers to detect bone metastases.8, 9 It was therefore interesting that these 2 proteins were overexpressed in bone metastases of PCa patients.10, 11

Several studies have assessed the diagnostic efficacy of both bone formation and resorption markers for the detection of bone metastases in PCa.6 However, they often included only a few bone markers or a limited number of patients.7, 12 While little comparative data are available on the diagnostic accuracy of the newly developed assays for the various analytes including OPG and RANKL, the conclusions regarding the diagnostic usefulness need to be reconsidered.9, 13 In addition, the recommendation of diagnostic tests was often substantiated by the univariate evaluation of data without taking into account the usefulness of multivariate analysis. Therefore, our aims were (i) to measure serum markers of bone formation, bone resorption and osteoclastogenesis as noninvasive, easy-to-determine analytes of bone metastases in the same serum samples; (ii) to evaluate and compare the diagnostic validity of all analytes concerning their differential efficiency between patients with and without metastases; (iii) to recommend, based on univariate and multivariate analyses of data, one analyte or an analyte combination as a useful tool for detecting bone metastases in PCa patients; and (iv) to verify whether bone markers could be used for predicting prognosis in PCa patients. We studied the bone formation markers tALP, bALP, OC and P1NP; the bone resorption markers BSP, NTX, CTX and TRAP; and the osteoclastogenesis markers OPG and RANKL.


AUC, area under the curve; bALP, bone-specific alkaline phosphatase; BPH, benign prostatic hyperplasia; BSP, bone sialoprotein; CTX, cross-linked C-terminal telopeptides of type I collagen; NTX, cross-linked N-terminal teleopeptides of type I collagen; OC, osteocalcin; OPG, osteoprotegerin; PCa, prostate carcinoma; P1NP, amino-terminal procollagen propeptides of type I collagen; PSA, prostate-specific antigen; RANK, receptor activator of nuclear factor κB; RANKL, RANK ligand; ROC, receiver operating characteristic; tALP, total alkaline phosphatase; TRAIL, tumor necrosis factor–related apoptosis-inducing ligand; TRAP, tartrate-resistant acid phosphatase isoenzyme 5b.


Study population

Our retrospective study included 187 men (Table I): 175 were investigated at the Department of Urology, Charité University Hospital, and 12 at the Department of Urology, University Hospital (Munster). Archived (at –80°C) and unthawed serum samples collected between May 1998 and December 2001 were used. The study was performed in accordance with the ethical standards of the Helsinki Declaration and approved by the local ethical boards of the hospitals.

Table I. Characteristics of the Study Groups1
Stage pN0M0Stage pN1M0Stage M1
  • 1

    Values are medians, with lower and upper quartiles in parentheses. n.d. not determined; IPSS, International Prostate Symptom Score.

Number of patients3535393444
Race, white (%)100100100100100
Age (years)50 (42, 60)68 (61, 71)64 (59, 69)68 (63, 71)66 (61, 72)
PSA (μg/l)2.1 (0.8, 3.1)4.6 (1.7, 11.6)8.6 (4.9, 12.8)19.4 (9.8, 34.2)51 (11, 205)
Prostate volume (cm3)n.d.50 (42, 60)29 (22, 36)25 (21, 35)n.d.
Uroflow (Qmax, ml/sec)n.d.9.0 (5.8, 14.7)n.d.n.d.n.d.
IPSSn.d.18 (13, 21)n.d.n.d.n.d.
Tumor stage     
 T2  22119
 T3  172329
 T4    6
Tumor grade     
 G1  31 
 G2  232019
 G3  131325

The control group consisted of 35 men with no history of prostate disease or metabolic bone disease. None of the subjects received any medication known to interfere with bone metabolism (calcium supplementation, bisphosphonate) or had signs of infection; gastrointestinal, hepatic, cardiac, or renal disease; tumors; or immunologic disease. Additionally, liver and kidney diseases were excluded as far as possible since the subjects had values of alanine aminotransferase and creatinine within the reference intervals.

Another group of 35 men were classified as BPH patients who had received no treatment for prostatic disease at the time of serum sampling. The clinical diagnosis of BPH was histologically confirmed by examination of prostatic specimens obtained by ultrasound-guided sextant or octant biopsies or after transurethral resection.

There were 117 men (median age 66 years, range, 47–83) with PCa. PCa was diagnosed histopathologically by microscopic examination of prostatic specimens after biopsy or additionally at radical prostatectomy. Cancer stage was assigned according to the TNM system, and histologic grade was classified as grade 1, 2 or 3.14 Bone scintigraphy and, in special cases, X-ray, computerized tomography or magnetic resonance imaging were used to diagnose bone metastases. There were 44 carcinoma patients with bone metastases (indicated as subgroup M1). The 73 patients without distant metastases received surgical lymph node staging (pelvic lymphadenectomy) with histologic examination and were therefore subdivided into groups without (pN0M0, n = 39) and with (pN1M0, n = 34) lymph node metastasis.

Blood samples were collected before any treatment except in the groups pN1M0 and M1. In all other cases, blood samples were taken before any diagnostic procedure, transurethral resection of the prostate or prostatectomy or 4 weeks after digital rectal examination, prostatic biopsy or transrectal ultrasound. In the pN1M0 group, 21 patients were untreated and 13 received hormonal therapy (orchiectomy, luteinizing hormone–releasing hormone analogs and antiandrogens). In the M1 group, 17 patients were untreated and 27 received hormonal therapy or had this treatment after radical prostatectomy or radiotherapy.


Blood samples were collected in evacuated Monovette plastic tubes (Sarstedt, Nümbrecht, Germany; catalog 04.1904.001) between 7:00 and 9:00 A.M. and centrifuged at 2,000g for 10 min at 4°C within 2 hr after venipuncture. Blood samples for OC determination were collected on ice. Supernatants were frozen at −80°C and not thawed before analysis.

Bone formation markers

tALP was measured with a standard enzyme assay on the Hitachi 717 analyzer (Roche, Mannheim, Germany). bALP was determined by the Tandem-MP Ostase Immunoenzymetric Assay (Beckman Coulter, Fullerton, CA), which specifically quantifies skeletal ALP with low immunoreactivity for the liver/kidney isoforms.15 OC (N-Mid-Osteocalcin Assay, Roche) and P1NP (Total P1NP-Assay, Roche) were measured on the Elecsys 2010 analyzer (Roche). The P1NP assay is a new electrochemiluminescent assay that detects both tri- and monomeric P1NP forms.

Bone resorption markers

BSP was measured by radioimmunoassay (Immundiagnostik, Bensheim, Germany) to avoid the instability of radioiodinated reagents.16 The assay characteristics (linearity, recovery, detection limit) corresponded to data previously published.16 CTX was determined by the β-CrossLaps Assay (Roche) on the Elecsys 2010 analyzer17 and NTX by the ELISA Osteomark NTX Assay (Ostex, Seattle, WA).18 NTX could be measured in only 18 of the 44 patients of the M1 group because the tests could not be supplied by the German distributor anymore. TRAP was determined by the Bone TRAP ELISA (Medac Diagnostika, Wedel, Germany), which quantifies the active isoform 5b.19

Osteoclastogenesis markers

OPG was measured as previously described by ELISA (Immundiagnostik), which equally detects both monomeric and homodimeric forms.20 RANKL was measured by another ELISA of the same company. This assay quantifies the soluble, uncomplexed RANKL (isoform 2, 27.7 kDa) and not the membrane-type form (isoform 1, 35.5 kDa).20

Imprecision control

Using control material generally containing analytes at the upper reference limit, the percentage coefficient of variation as indicator of interassay imprecision was 2.8% for tALP (n = 9), 5.2% for bALP (n = 9), 2.9% for OC (n = 19), 3.1% for P1NP (n = 19), 7.7% for CTX (n = 19), 13.1% for NTX (n = 6), 9.8% for TRAP (n = 9), 6.7% for OPG (n = 10) and 10.5% for RANKL (n = 10). BSP was measured in one series, with intra-assay and interassay variations of 6.1% and 9.2%, respectively.

Routine clinical chemistry determinations

Alanine aminotransferase (upper reference limit 41 U/l), C-reactive protein (5 mg/l) and creatinine (105 μmol/l) were measured by standard assays on the Hitachi 717 analyzer. Total and free PSA were measured with the Immulite PSA and Free PSA kits (Diagnostic Products, Los Angeles, CA).

Statistical analysis

Statistical calculations were performed with SPSS 11.5 for Windows (SPSS, Munich, Germany) and GraphPad Prism 4.1 (GraphPad, San Diego, CA). We used the nonparametric Kruskal-Wallis ANOVA with Dunn's post-test, the Mann-Whitney U-test, Spearman's rank correlation coefficients (rs), the distribution fitting procedure of Kolmogorov-Smirnov and the logistic regression approach. The Kaplan-Meier product-limit method was used to determine survival probability for subgroups. Univariate and multivariate analyses of risk factors predicting PCa-specific death were performed using the Cox proportional hazards regression model. Diagnostic accuracy was evaluated by ROC curve analysis using the software MedCalc 7.2 (MedCalc, Mariakerke, Belgium). Reference intervals were calculated according to the International Federation of Clinical Chemistry (IFCC)-recommended procedure.21p < 0.05 was considered statistically significant.


Characteristics of study groups

Table I summarizes the demographic and essential clinical data of the groups studied. There were no age differences between BPH and PCa patients or within the PCa groups (p > 0.05). Controls had a somewhat lower mean age.

Concentrations of serum bone markers

Figure 1 shows the scatter plots and medians of all markers in controls, BPH patients and PCa patients subdivided in the 3 groups N0M0, N1M0, and M1. Statistical evaluation of the data can be summarized as follows. (i) There was no significant difference in the concentrations of all markers except CTX between controls and BPH patients (p < 0.05). Thus, the data of both groups could be compiled for the calculation of provisional reference limits and showed a gaussian distribution (Kolmogorov-Smirnov test). The parametric upper 95% reference limits are indicated as dotted lines in the figures. (ii) Analytes OC, RANKL and CTX were not significantly different between the groups and could not be used as tools to differentiate between benign and malignant prostatic diseases. (iii) In all cases, 3 of the 4 bone formation markers (tALP, bALP and P1NP) and bone resorption markers (BSP, TRAP and NTX) as well as OPG were significantly increased in PCa patients with bone metastases compared to patients without metastases as well as controls and BPH patients. (iv) PCa patients without bone metastases showed no differences for all analytes compared to BPH patients, except for BSP, which was increased in the lymph node-positive group.

Figure 1.

Scatter plots of bone formation (a–d), bone resorption (e–h) and osteoclastogenesis markers (i,j) in serum of controls and patients with BPH or PCa. Median values of the groups are shown as horizontal lines with corresponding figures; dotted lines indicate the upper parametric 95th percentiles of the combined data from controls and BPH patients. Significant differences (Kruskal-Wallis nonparametric ANOVA with Dunn's post-test, p < 0.05 at least) are shown by the following symbols; a, compared to controls; b, compared to BPH patients; c, compared to PCa patients without lymph node metastases (group pN0M0); d, compared to PCa patients with lymph node metastases (group pN1M0); e, compared to PCa patients with bone metastases (group M1).

Association between bone markers and clinical data

Significant correlations (Spearman's rank correlation coefficients) existed not only between the various bone formation and resorption markers (rs = 0.195–0.730, at least p < 0.05) but also, except for OC and CTX, to OPG (rs = 0.172–0.384, at least p < 0.05), while no associations were found between RANKL and the other bone markers. Only tALP, bALP and P1NP were weakly correlated with tumor stage (rs = 0.192–0.209); and no correlations were found with histologic grade of tumor. PSA significantly correlated with all bone markers (rs = 0.223–0.391, at least p < 0.05) except OC, CTX and NTX.

To evaluate the effect of hormonal therapy on bone markers, we compared their concentrations in patients with and without treatment in the groups pN1M0 and M1. In the pN1M0 group, 21 patients were untreated and 13 received hormonal therapy before sample collection (median 1.7 months, range 0.5–2.8). In the M1 group, 17 patients were untreated and 27 received hormonal therapy or had this treatment after radical prostatectomy or radiotherapy before sample collection (median 25 months, range 8–97). The concentrations of all markers in the M1 group did not differ between patients with and without hormonal treatment (Mann-Whitney U-test, p = 0.509–0.990). Similarly, no significant differences (p = 0.163–0.901) were observed in the pN1M0 group except for TRAP and CTX. Patients who received hormonal therapy showed slightly increased median values of both analytes (TRAP, medians 4.41 vs. 3.96 U/l, p = 0.033; CTX, 0.48 vs. 0.31 μg/l, p = 0.009). However, separate calculations showed that the differences in these 2 analytes between the groups did not change whether the calculations were made only with untreated patients or with the combined group of patients with and without hormonal treatment. Thus, all further calculations were performed with the data of all patients in the respective groups.

Bone markers as diagnostic indicators of bone metastatic spread

ROC analyses were performed to characterize the diagnostic usefulness of all the bone markers investigated to differentiate between PCa patients with and without bone metastases (Fig. 2). Both bone formation (Fig. 2a) and resorption (Fig. 2b) markers as well as the osteoclastogenesis marker OPG (Fig. 2c) were helpful in this respect. However, OPG showed the largest AUC. The AUC difference between OPG and bALP or tALP, respectively, was significant (p = 0.004) or showed a corresponding tendency (p = 0.072). Also, some formation (e.g., OC) as well as resorption (e.g., CTX) markers were rather ineffective for this purpose. Table II shows the diagnostic sensitivities and specificities at the cut-off level with the highest diagnostic accuracy in the ROC analysis and at the conventional upper reference limit of the respective marker. OPG demonstrated the best discriminatory power. The better diagnostic accuracy of OPG compared to tALP became evident when the method-specific upper limit of 200 U/l for tALP was used in the same way as for routine diagnostics (Table II).

Figure 2.

ROC curves for bone formation (a), bone resorption (b) and osteoclastogenesis markers (c) to distinguish between PCa patients with and without bone metastases. AUC values ± SE were as follows: (a) tALP, 0.91 ± 0.03; bALP, 0.84 ± 0.04; P1NP, 0.84 ± 0.044; OC, 0.56 ± 0.06; (b) BSP, 0.82 ± 0.05; TRAP, 0.82 ± 0.07; NTX, 075 ± 0.07; CTX, 0.59 ± 0.08; (c) OPG, 0.97 ± 0.02; RANKL, 0.63 ± 0.05.

Table II. Diagnostic Sensitivity and Specificity of Bone Formation, Bone Restoration and Osteoclastogenesis Markaers to Distinguish PCa Patients with and without Bone Metastases1
VariableSensitivity (%)Specificity (%)
  • 1

    Sensitivity and specificity (with 95% confidence intervals in parentheses) of the various markers were calculated using either

  • 2

    the cut-off level with the highest diagnostic accuracy obtained from ROC analysis performed with 73 patients without bone metastases and 44 patients with bone metastases or

  • 3

    the conventional upper reference limits provided by our hospital laboratory, by the producers of the test kits or established in this study (OPG, RANKL).

Bone formation marker  
 tALP (U/l)  
  129281 (66–91)93 (85–98)
  200345 (30–61)97 (90–100)
 bALP (ng/l)  
  15.2275 (60–87)93 (85–98)
  20.1366 (50–80)95 (87–98)
 OC (μg/l)  
  11.3224 (12–40)95 (87–99)
  46314 (5–27)96 (88–99)
 P1NP (μg/l)  
  50.6276 (60–88)90 (81–96)
  80361 (46–76)96 (88–99)
Bone resorption marker  
 BSP (μg/l)  
  17.2273 (57–86)83 (72–91)
  21.6364 (48–78)89 (79–95)
 CTX (μg/l)  
  0.627230 (17–47)93 (85–98)
  0.854319 (8–33)95 (87–98)
 NTX (nmol/l BCE)  
  26.9261 (36–83)96 (88–99)
  24.4361 (36–83)89 (80–95)
 TRAP (U/l)  
  4.62277 (60–90)85 (75–92)
  5.73354 (37–71)92 (83–97)
Osteoclastogenesis marker  
 OPG (pmol/l)  
  3.44293 (81–99)94 (87–99)
  3.36393 (81–99)93 (85–98)
 RANKL (pmol/l)  
  0.65246 (30–61)84 (73–91)
  1.55393 (81–99)12 (6–22)

We also used logistic regression analysis to predict bone metastasis. Using single variables, the overall correct classification was <70% for RANKL (62%), OC (67%) and CTX (69%); 70–90% for BSP (78%), TRAP (81%), bALP (84%), NTX (84%), P1NP (84%) and tALP (88%); and >90% for only one analyte, OPG (92%). Using all variables together with a stepwise selection method, only OPG and TRAP remained as independent variables for differentiation, resulting in an overall correct classification of 94%.

Bone markers as predictors of survival

Incomplete follow-up of 2 patients (subgroup pN0M0) resulted in 115 eligible patients of the 117 with PCa for survival analyses. Mean follow-up time was 36.1 ± 17.2 months (range 2.3–77.8). The primary end point of our analyses was cancer-related survival, as measured from the date of surgery or visit at our institution to the last follow-up or cancer-related death. According to the death certificates and the information of general practitioners, 21 patients died from PCa. To determine whether the serum concentrations of bone markers correlated with disease outcome, patients were stratified into 2 groups using the cut-off points of the analytes with the maximal diagnostic accuracy to differentiate between patients with and without bone metastases. To identify the significant prognostic factors associated with PCa-specific death, univariate and multivariate risk factor analyses were performed using the Cox proportional hazards regression model with the stratified groups (Table III). Levels of OPG, P1NP, tALP, bALP, BSP, NTX, TRAP and CTX as well as the presence of bone metastases were significant univariate predictors of death from PCa, whereas age, tumor stage, tumor grade, the presence of lymph node metastases, PSA and the markers OC and RANKL were not. Similar results were obtained with Kaplan-Meier survival analysis. Patients with concentrations of the above-mentioned 8 markers higher than the cut-off levels had significantly shorter overall survival time than patients with low concentrations (Fig. 3). Multivariate analysis of the significant predictors, however, showed that only OPG and BSP were independent predictors of cancer-related death (Table III). According to the forward or backward stepwise calculation model, both markers remained significant variables, whereas the other variables were excluded.

Table III. Univariate and Multivariate Analysis of Serum Bone Markers and Clinicopathologic Factors in Relation to PCa Survival1
VariableRelative risk95% Confidence intervalp
  • 1

    Because of 2 incomplete follow-ups, 115 of 117 patients were available for analysis of independent prognostic significance. Using the Cox proportional hazards regression model, calculations were performed with dichotomized data of bone markers according to the concentrations below and above the cut-off point with the maximal diagnostic accuracy as given in Table II. Criteria for dichotomous classification of the other variables were, for age, <60 and ≥60 years; for PSA, <20 and ≥20 μg/l; for tumor stage, T1–2 and T3–4; for tumor grade, G1–2 and G3; for bone metastasis and lymph node metastasis, presence or not. For further details, see text.

Univariate analysis   
 Tumor stage3.350.97–11.60.451
 Tumor grade1.470.51–4.210.473
 Bone metastasis8.512.73–28.6<0.0001
 Lymph node metastasis2.430.22–26.80.469
Multivariate analysis   
 Bone metastasis0.230.01–3.830.303
Figure 3.

Cumulative cancer-related survival in PCa patients with bone marker concentrations below and above the cut-off points for maximal diagnostic accuracy to differentiate between patients with and without bone metastases. Cut-off points were taken from the data in Table II. Survival distributions were calculated using the Kaplan-Meier method and compared by the log-rank test. Of 117 patients, 115 were eligible for survival analysis due to exclusion of 2 patients with incomplete follow-up.


Bone markers as diagnostic tools

Bone scintigraphy is considered to be the gold standard for monitoring metastatic bone involvement.22 However, since scintigraphy is expensive, lacks specificity and is not particularly suitable for the follow-up of patients, bone markers as indicators for bone metastasis in PCa patients have been studied.2, 3, 4, 5, 6 There are no clear recommendations which markers or marker combinations should be used.9 Analyses were performed in both serum and urine, but consistent results were not always described. The preanalytic variability of bone markers, their different stability in vitro especially in urine and use of assays with different antibodies recognizing different epitopes may explain many of the discrepant results.23, 24, 25, 26, 27 To avoid these discrepancies as far as possible, we measured bone markers exclusively in serum collected during a defined time period, between 7:00 and 9:00 A.M. In elderly patients, blood sampling is always more reliable and more practicable than urine sampling. As commercial assays for measuring bone formation and resorption in serum are available, measurements in urine could be replaced by those in serum/plasma.28

Our present study, to our knowledge, involves the largest number of bone turnover markers in serum of PCa patients for both diagnostic and prognostic evaluation to date. Of the 10 bone turnover markers measured, 3 of the 4 bone formation (tALP, bALP and P1NP) and resorption (BSP, TRAP and NTX) markers as well as OPG were significantly increased in PCa patients with bone metastases. These results correspond to histomorphologic evidence that not only osteoblastic but also resorptive processes occur.29 Differences between the markers could reflect the occurrence of a marker in either an early or a late period of bone formation or resorption.2 Hormonal therapy could similarly give rise to differences. Nonmetastatic PCa patients receiving antiandrogen therapy had moderately increased serum bALP,6, 30 and the type of therapy (orchiectomy, luteinizing hormone–releasing hormone analogs and/or antiandrogens) is likely to influence bone turnover.31 In our study, bone formation markers were not different between patients with and without hormonal therapy in the N1M0 or M1 groups. Only slightly increased values of the resorption markers CTX and TRAP were observed in lymph node-positive patients who received hormonal therapy. These changes did not affect the accuracy of the markers for diagnosing bone metastasis. However, since not only “fresh” but also treated patients were investigated, this limitation of our study should be considered in characterizing the diagnostic accuracy of the bone turnover markers.

Bone formation markers.

Bone ALP is a well-established bone marker. It can be considered together with tALP as a “standard” analyte when the utility of other bone markers is to be assessed and compared.2 In our study, the advantage of bALP compared to tALP became evident in a higher diagnostic sensitivity when the assay-specific upper cut-off of tALP in our hospital was used as the decision point (Table II). In contrast to bALP as an enzyme directly secreted by osteoblasts, P1NP is a metabolic indicator of bone formation. P1NP measured with a radioimmunoassay (Orion, Espoo, Finland) has been recommended as a promising bone marker in PCa patients.5, 32, 33 Our data confirm these results using a new assay that measures both tri- and monomeric P1NP forms, while the Orion radioimmunoassay detects only the trimeric form. The diagnostic sensitivity of OC was low compared to the other bone formation markers. Its appearance in the late bone formation phase in contrast to ALP or P1NP as typical markers of the early phases of osteoblastic proliferation and matrix maturation could explain the divergent diagnostic efficacy.2 In addition to the clinical validity, both assays are uncomplicated and reliable with good analytic performance; thus, they fulfill the essential preconditions for routine measurements. The P1NP assay is an automated assay on a general-purpose analyzer (Elecsys). Meanwhile, bALP can also be measured on an analyzer (Access, Beckman-Coulter), which gives results almost identical to those of the Tandem Ostase test used in our study. All of these arguments support the view that bALP and P1NP are the most suitable single bone formation markers to detect bone metastases or to confirm their absence in PCa patients.

Bone resorption markers.

Dissociation was also observed between bone resorption markers. CTX showed lower diagnostic sensitivity than NTX, TRAP and BSP (Table II). NTX had the lowest analytic performance, and the test kits were not always available. The TRAP assay used in our study is based on a combined approach of immunoassay and activity measurement. In contrast to other TRAP immunoassays, this new ELISA with reliable analytic performance detects the active isoform 5b without any interference from inactive enzyme forms or fragments.34, 35, 36 BSP is generally considered to be a bone resorption marker.37 It is overexpressed in PCa tissue and bone metastases.38, 39 As yet, only few data exist on the behavior of serum BSP in PCa patients. Sensitivities and specificities of about 90% for the diagnosis of bone metastases have been described.40 Our data were similar, though BSP was determined without disrupting its interaction with complement factor H. However, BSP was the only marker with increased concentrations already in the pN1M0 group, and numerous patients of the pN0M0 group showed increased values. As there was no further evidence for the occurrence of bone metastases in those patients, we assumed that overexpression of BSP in PCa tissue already raises the BSP concentration in serum before bone metastasis develops. Therefore, BSP should not be considered a specific bone metastasis marker in PCa as suggested for breast cancer.16 In addition, the instability of reagents and the disadvantage of a radioimmunoassay hindered determination of BSP as a routine parameter up to now. Taking these data of bone resorption markers together, we recommend TRAP as a single bone resorption marker for PCa patients.

Osteoclastogenesis markers.

As osteoclastogenesis is the connecting link between bone formation and bone resorption, determination of OPG and RANKL was of special interest. Both OPG and RANKL were overexpressed in bone metastases of PCa patients.10, 11 OPG has been postulated to be a survival factor in PCa cells.41, 42 We and another group found increased serum OPG concentrations with advanced PCa,8, 13 while elevations were not observed in breast, colorectal, pancreatic and renal tumor patients with bone metastases.43 The present study confirms the discriminatory power of OPG in distinguishing between PCa patients with and without bone metastases. However, despite this clinical significance, it has to be considered that OPG is also produced by a variety of other organs.44 Thus, changed serum concentrations of OPG caused by other diseases, such as vascular diseases or rheumatoid arthritis, need to be taken into account when interpreting serum OPG values.44, 45 Based on the role of OPG and RANKL,11, 41, 42 it could be assumed that the ratio of RANKL to OPG would facilitate the diagnosis of bone metastasis. However, neither serum RANKL concentrations nor the ratio of RANKL to OPG was related to metastatic stage (Fig. 1). Thus, measurement of RANKL does not appear to be helpful in this respect. Similar findings have been reported in patients with Paget's disease.46

Multivariate analysis.

As the univariate evaluation of data showed dissociated changes of serum marker concentrations for both bone formation and resorption as well as for those linked with osteoclastogenesis, multivariate analysis was appropriate. Logistic regression analysis of all variables showed that the combination of OPG and TRAP is the best way to differentiate between bone metastasis and nonmetastasis. The overall correct classification of 93% achieved by the combination of TRAP and OPG corresponds to the results described when the ratio of OC to P1NP or OC to bALP as a marker characteristic of different bone formation phases is used.33 As OPG alone resulted in a sensitivity and specificity >90%, a rather small improvement of diagnostic accuracy could be expected by TRAP.

Bone markers as prognostic tools

The extent of bone metastasis monitored by bone scans is a strong predictor of survival in PCa patients.47, 48 Some studies have also demonstrated this association with ALP, P1NP and CTX.12, 49, 50 PSA was not directly associated with bone progression in all studies.12, 49, 50 Univariate analyses confirmed the strong predictive value of bone markers concerning survival in our study (Table III, Fig. 3). Not only did OPG yield the highest relative risk value in the univariate analysis but OPG and BSP were the only independent predictors of PCa-specific death in the multivariate Cox model (Table III). Thus, OPG really appears more as a marker of tumor burden or activity than a simple indicator of bone turnover. The value of serum BSP as an independent prognostic factor for subsequent bone metastasis was previously demonstrated in patients with primary breast cancer.51

The current study provides data on the association between serum OPG, bone metastases and survival in PCa patients. Our results are consistent with the functional role of OPG in facilitating the survival of PCa cells by protecting them from apoptosis induced by TRAIL.41, 52 The design of our study, with the limitation that patients with hormonal pretreatment were included, did not allow an exact answer as to whether the increase of serum OPG occurs in the early phase, during or in the progression phase of metastasis.

In conclusion, our results suggest that, despite the limitations, measurement of serum OPG concentration is a powerful test alone or in combination with other bone turnover markers to detect bone metastatic spread and to predict survival probability in PCa patients. In addition, the association of OPG with the survival of PCa patients could be used for stratifying patients with advanced PCa for clinical trials in the current treatment options with bisphosphonates or other chemotherapeutic agents to individualize treatment.53, 54, 55 However, increased serum OPG caused by other organ alterations must be considered.


This report includes parts of the doctoral thesis of K.v.H. and was kindly supported in part by the following sources and grants: Funds of the German Chemical Industry (to K.J., grant 400770) and Sonnenfeld-Stiftung (to K.J. and P.S.). We thank Ms. S. Klotzek and Ms. J. Reiche for excellent technical assistance.