Prediction of survival benefit using an automated bone scan index in patients with castration-resistant prostate cancer


Yozo Mitsui, Department of Urology, Shimane University School of Medicine, 89 1 Enya cho 693-8501, Izumo, Japan. e-mail:


Study Type – Prognosis (case series)

Level of Evidence 4

What's known on the subject? and What does the study add?

A bone scan index (BSI) can quantify the extent of bone involvement and response to treatment, but it has not been widely accepted, because of its time-consuming nature.

The study is the first to demonstrate that automated BSI calculated with a computer-assisted diagnosis system is effective in judging the chemotherapeutic response of bone metastatic lesions in patients with castration-resistant prostate cancer.


  • • To evaluate the value of an automated bone scan index (aBSI), calculated using a computer-assisted diagnosis system, to indicate chemotherapy response and to predict prognosis in patients with castration-resistant prostate cancer (CRPC) with bone metastasis.


  • • Forty-two consecutive CRPC patients underwent taxane-based chemotherapy between November 2004 and March 2011 at our institution.
  • • The aBSIs were retrospectively calculated at the diagnosis of CRPC and 16 weeks after starting chemotherapy.
  • • Cox proportional hazards regression models were applied to multivariate analyses with and without aBSI response in addition to the basic model.
  • • Based on the difference in the concordance index (c-index) between each model, the prognostic relevance of adding the aBSI response was determined.


  • • A decrease in aBSI was found in 28 patients (66.7%), whereas a response was shown by bone scan in only 23.8% of patients.
  • • Patients with a reduction in aBSI had longer overall survival (OS) in comparison with the other patients (P= 0.0157).
  • • Multivariate analysis without aBSI response showed that performance status (P= 0.0182) and PSA response (P= 0.0375) were significant prognosticators.
  • • By adding the aBSI response to this basic model, the prognostic relevance of the model was improved with an increase in the c-index from 0.621 to 0.660.


  • • The aBSI reflected the chemotherapy response in bone metastasis.
  • • The index detected small changes of bone metastasis response as quantified values and was a strong prognostic indicator for patients with CRPC.

automated bone scan index


prostate cancer


castration-resistant PCa


androgen deprivation therapy




docetaxel, estramustine and carboplatin


bone scan


BS index


computer-assisted diagnosis


automated BSI


performance status


upper limit of normal


artificial neural network


response evaluation criteria in solid tumours


extent of disease


overall survival


concordance index


Prostate cancer (PCa) has become the second leading cause of cancer death in the majority of western countries [1], and there is also a trend towards an increasing number of PCa deaths in Japan. Among patients who die from PCa, the incidence of skeletal involvement appears to be >85% [2,3]. The standard treatment for patients with PCa with metastatic spread to the bone is androgen deprivation therapy (ADT); however, the vast majority of patients with PCa with bone spread finally become resistant to androgen deprivation and progress to castration-resistant PCa (CRPC). We have previously shown that a docetaxel-based chemotherapeutic strategy with a combination of docetaxel, estramustine (EM) and carboplatin, known as DEC therapy, confers significant clinical benefits as well as an acceptable drug toxic profile in patients with CRPC [4,5]. In a phase II study [5] on DEC therapy, partial response was achieved in 66.7% of patients with extra-osseous measurable disease, but in only 8.3% of patients with bone metastasis.

In patients with PCa, bone scan (BS) is the most frequently used imaging technique for detecting or identifying bone metastasis, and it is also used to evaluate changes in metastatic spread involving bone tissues [6]. However, the changes detected on BS images are essentially determined by subjective evaluation focusing on the intensity and/or size of osseous lesions, which makes it difficult to compare images in a longitudinal fashion. Scher et al. [7] showed that BS is more likely than other variables to identify bone lesion as stable disease, even when those variables indicate a beneficial response.

The bone scan index (BSI), which makes it possible to quantify bone involvement as well as therapeutic response, was originally developed by Imbriaco et al. [8], but BSI has not become widely accepted because it is time-consuming to analyse the data and because of the special training necessary to apply it to routine clinical work [9].

As part of routine clinical work, some computer-assisted diagnosis (CAD) systems have been developed and are now available, which have contributed to an improvement in breast cancer and colon polyp diagnosis [10–13]. Sadik et al. [14–17] used a CAD system to identify bone metastasis on BS and reported it had a sensitivity of 90% and a specificity of 74%. Since a CAD system can easily quantify BS image findings, those can be converted into a BSI in a more comprehensive and objective way to compare images obtained at different time points during the clinical course.

At present, PSA level has been considered an essential and practical marker for management of patients with PCa, but an elevated PSA level itself cannot always determine whether the region of bone, lymph nodes or visceral organ is metastatic spread or local recurrence. In addition, although PSA is the best marker for PCa monitoring, it is not PSA that represents or reflects the comprehensive status of PCa aggressiveness. Considering the bone as the most prevalent site of metastatic spread in PCa, we think that a more detailed approach to analysing bone metastasis would contribute to improvements in prognosis and even confer a survival benefit on the affected patients. In the present study, we hypothesize that BS images evaluated by a CAD system could identify small but significant alterations involving the bone that might be closely associated with the biological behaviour of PCa. In patients with CRPC who underwent docetaxel-based combination chemotherapy or DEC therapy, we validated the superiority of automated BSI (aBSI) to conventional and subjective BS evaluation, when determining the chemotherapeutic effect of DEC on the target bone lesions.



Forty-two Japanese patients with CRPC with bone metastasis, who underwent DEC therapy between November 2004 and March 2011 at our institution, were included in the present study. The eligibility criteria for DEC chemotherapy were as follows: (i) Eastern Cooperative Oncology Group performance status (PS) score of 0–3; (ii) baseline leukocyte count >3000/µL; (iii) haemoglobin ≥8.0 g/dL; (iv) platelet count >100 000/µLl; (v) adequate renal function defined as serum creatinine ≤1.5 times the upper limit of normal (ULN); (vi) adequate liver function defined as bilirubin < ULN and aspartate transaminase <1.5 times ULN; (vii) adequate cardiac function; (viii) life expectancy of >3 months; and (ix) >8 weeks must also have elapsed since any major surgery, radiotherapy or any previous chemotherapy. The DEC therapy comprised weekly administration of i.v. docetaxel at 30 mg/m2, daily oral EM 10 mg/m2, and i.v. carboplatin every 28 days to reach an area under the curve value of 6 on day 1 of every 4-week cycle [4,5]. All patients had histologically confirmed PCa and showed progression despite discontinuation of anti-androgen (androgen withdrawal). During DEC therapy, ongoing ADT was also applied. Pretreatment evaluation included medical history, a physical examination, a complete blood count, a chemistry profile, serum PSA, alkaline phosphatase and lactate dehydrogenase levels, 24-h creatinine clearance, a 12-lead electrocardiogram, a chest X-ray, CT, BS and MRI. Treatment was continued until disease progression, an unacceptable adverse event or patient refusal occurred. Institutional review board approval and written informed consent from all patients were obtained. All study protocols were approved by the ethics committee of the Shimane University Faculty of Medicine in accordance with the 1975 Declaration of Helsinki.


Bone scan examinations were performed ∼ 3.5 h after i.v. injection of 740 MBq Tc 99m methylene diphosphonate (FUJIFILM RI Pharma, Tokyo, Japan). Whole-body images, at anterior and posterior view scan speeds of 15 cm/min (matrix 256×1024), were obtained with a gamma camera equipped with low-energy high-resolution parallel hole collimators (Siemens, Milton Keynes, UK). Energy discrimination was provided by a 7.5% window centred on the 140 keV of the Tc 99m.


The CAD system used was one that is able to perform completely automated detection and analysis of hot spots and also determines complete classification based on hot-spot analysis findings. The method used for interpretation of BS findings consists of image-processing techniques and artificial neural networks (ANNs). The program analysed anterior and posterior images in digital format, and no manual steps were required. The first step included image segmentation, hot-spot detection, and feature extraction, then the resulting highlighted features of the images were used as input to ANNs for classifying hot-spot networks [14–17]. Hot spots classified as possible metastasis were indicated as red, while those classified as benign (e.g. degenerative changes, fractures, symmetric hot spots) were indicated in blue (Fig. 1). For the pressent study, we used the BONENAVI® system developed by FUJIFILM RI Pharma, which was constructed from data obtained from 904 Japanese cases with BS findings.

Figure 1.

Representative images obtained from a 62-year-old patient with CRPC (cT4N0M1). The automated bone scan index (aBSI) value was 7.435% at the time of diagnosis of CRPC. Suggestive bone metastases are marked in red, while symmetric or benign radiotracer uptake is shown in blue. At 8 weeks after starting the chemotherapy, the flare-up phenomenon was seen on radionuclide BS images (aBSI, 8.418%). A remarkable decrease in aBSI was found 16 weeks after starting the chemotherapy.

Bone scan examinations were performed for all the patients and the results converted to the aBSI at diagnosis of CRPC, and 8 and 16 weeks after starting DEC therapy. With regard to the flare-up phenomenon in radionuclide BS findings after treatments targeting PCa (Fig. 1), we defined the BS images at 16 weeks after starting DEC therapy as reliable data to show the therapeutic response of bone metastatic lesions [18,19].


For patients with measurable disease, response was determined according to the Response Evaluation Criteria in Solid Tumours (RECIST) on the basis of imaging studies, including chest X-ray, CT scans, and BS at least every 8 weeks for four cycles. PSA levels were determined every 4 weeks and PSA response was defined as a reduction of ≥90% from the baseline at 16 weeks after initial chemotherapy. Absolute changes in the intensity or size of osseous lesions by BS were objectively difficult to interpret; therefore, for bone metastasis response, complete response was defined as the disappearance of all positive areas in BS, partial response as a decrease in the extent of disease (EOD) grade, and progressive disease as an increased number of positive sites [20]. PSA progression was defined as three consecutive increases in PSA level of at least 50% over the nadir with a minimum of 4 ng/mL. Overall survival (OS) was determined from the initiation of DEC therapy to the day of death or last follow-up examination.


Parametric data were analysed using an anova test followed by a post-hoc test. Correlation analysis was performed using Pearson's coefficient correlation. Survival curves were conducted using the Kaplan–Meier method and the differences between two curves were analysed using a log-rank test. Univariate and multivariate analyses for OS were performed using a Cox proportional hazards regression model. Harrell's concordance index (c-index) was used to determine whether adding the aBSI response meaningfully helps risk stratification. A two-tailed P value of <0.05 was considered to indicate statistical significance. Statistical analysis was performed using the R project (



To exclude false-positive and false-negative findings, all hot spots indicated as positive for metastasis by the BONENAVI® system were re-evaluated by focusing on whether those spots were compatible with metastatic bone regions found by the combination of MRI, CT, and plain radiographic findings. The clinicopathological characteristics of the 42 patients in the study are shown in Table 1. Twenty-one (50.0%) patients had EOD grade 1, 15 (35.7%) had EOD grade 2, and the remaining six (14.3%) had EOD grade 3. Thirty (71.4%) and 12 (28.6%) patients were treated with ADT alone and ADT plus EM, respectively, before introduction of DEC therapy. At the diagnosis of CRPC, 32 (68.7%) patients had an aBSI value <3.0% and the remaining 10 (31.3%) had a value of ≥3.0%. As shown in Fig. 2A, aBSI was strongly correlated with EOD grade (P < 0.001). The aBSI threshold value was determined as 3.0% according to the previous study [8].

Table 1. Clinical characteristics of the 42 study patients
Median (range) age, years73 (52–86)
PS, n (%) 
 <233 (78.6)
 ≥29 (21.4)
Median (range) PSA at CRPC diagnosis, ng/mL65.3 (0.1–3584.1)
Median Gleason sum8
EOD classification, n (%) 
 121 (50.0)
 215 (35.7)
 36 (14.3)
Lymph node involvement, n (%) 
 Neg.22 (52.4)
 Pos.20 (47.6)
Liver metastases, n (%) 
 Negative39 (92.9)
 Positive3 (7.1)
Previous treatment, n (%) 
 Hormonal therapy alone30 (71.4)
 Hormonal therapy + EM12 (28.6)
aBSI at CRPC diagnosis, n (%) 
 <3.0%32 (76.2)
 ≥3.0%10 (23.8)
aBSI 16 weeks after starting chemotherapy, n (%) 
 Increase14 (33.3)
 Decrease28 (66.7)
PSA reduction rate (16 weeks after starting chemotherapy), n (%) 
 <50%4 (9.5)
 ≥50%, <90%10 (23.8)
 ≥90%28 (66.7)
Effect of bone metastasis, n (%) 
 Complete response3 (7.1)
 Partial response7 (16.7)
 Stable disease32 (76.2)
Effect of lymph node involvement, n (%) 
 Partial response13 (75.0)
 Stable disease5 (25.0)
Effect of liver metastases, n (%) 
 Partial response3 (100)
Figure 2.

Correlation between aBSI and EOD, and aBSI and PSA changes. A, aBSI value was significantly associated with EOD grade. B, aBSI change was weakly but significantly correlated with PSA change.


Clinical outcomes of DEC therapy are shown in Table 1. Sixteen weeks after starting DEC therapy, a PSA reduction rate of >50% from the baseline was observed in 38 patients (90.5% [Table 1]). Of 20 patients with overt lymphadenopathy, 13 (75.0%) showed partial response, and all of three patients with measurable lung metastasis showed partial response. The beneficial effect of DEC therapy on the bone metastasis (complete and partial response) was only found in 10 patients (23.8%). By contrast, 28 patients (66.7%) showed a reduction in aBSI score 16 weeks after starting DEC therapy. A weak but significant correlation was found between the reduction rate of aBSI and PSA (Pearson's coefficient correlation 0.157; P= 0.023, Fig. 2B). A representative case with stable disease shown by BS but a 98.7% decrease in aBSI is shown in Fig. 3. In this case, consecutive levels of aBSI and PSA were measured during the progression, and the patient showed parallel increases and decreases in aBSI and PSA levels along with a rise in aBSI value preceding a rise in PSA level.

Figure 3.

Representative case. A 75-year-old patient with CRPC (cT4N1M1) was treated with DEC therapy. At 8 weeks after starting the chemotherapy, the effect of treatment on bone metastasis was SD, and PSA level and aBSI value were decreased by 89.0% and 98.7%, respectively. The patient refused to continue chemotherapy, and thereafter showed a parallel increase of aBSI and PSA levels. PSA failure occurred ∼ 9 months after the initial treatment, while a decrease in aBSI was seen 2 months before PSA failure. Although changes in the intensity of osseous lesions were recognized on BS images, the BONENAVI® CAD system used with BS more clearly showed the changes in osseous lesions during progression.


In the present series, the median OS was 26.6 months. The 42 patients who underwent DEC therapy were classified into aBSI response (+) (decrease in aBSI) and aBSI response (−) groups (increase in aBSI) at 16 weeks after starting DEC therapy. OS was significantly longer in the aBSI response (+) group than the aBSI response (−) group (P= 0.016; Fig. 4B). Likewise, at the time of CRPC diagnosis, the difference in OS between the aBSI value of > and <3.0% was close to significance (P= 0.061; Fig. 4A). In addition to aBSI response, PS and PSA response were identified as significant factors contributing to OS in univariate analysis, while both EOD grade and bone response after chemotherapy evaluated by EOD grade were not significant contributors. The basic multivariate model without aBSI response as a covariable showed PS and PSA response as significant prognosticators with a c-index of 0.621 (Table 2). By adding the aBSI response to this basic model, aBSI response was the strongest and independent predictor for OS (P= 0.015) followed by PS (P= 0.02), with the c-index of the model improving from 0.621 to 0.660. Interestingly, a decrease in aBSI was found in 18 (56.3%) of 32 patients with stable disease in BS findings, indicating a potential discrepancy in detecting changes involving bone regions between subjective BS-dependent and objective aBSI-based analysis.

Figure 4.

Relationships of aBSI and aBSI response with OS probability. A, The low aBSI value group (aBSI <3.0%) were found to have a longer OS than the high aBSI value group (aBSI ≥3.0), although there was no significant difference between the groups (P=0.061). B, By contrast, aBSI response was significantly associated with OS (P=0.016).

Table 2. Difference in the c-index between the models with and without aBSI response determined by a Cox proportional hazards model after age adjustment
VariableCo-efficientChi-squared test P Hazard ratio95% CIc-index
Performance status1.6615.5770.01821.6611.090–2.5290.621
 0 or 1 vs ≥2      
PSA response2.5954.3270.03752.5951.057–6.372 
 <90% vs ≥90%      
Performance status2.3305.4290.01981.6731.085–2.5800.660
 0 or 1 vs ≥2      
PSA response0.8793.3950.06542.4080.945–6.133 
 <90% vs ≥90%      
aBSI response0.9825.9280.01492.6701.211–5.886 
 Decrease vs increase      


Bone scans, the most frequently used imaging technique to detect bone metastasis, are able to detect such metastasis up to 18 months earlier than plain film analysis [6,21]; however, one of the major drawbacks of conventional BS systems is the high potential for subjectivity when evaluating target bone regions. RECIST define osteoblastic bone metastasis as non-measurable [22], thus a therapeutic response in a bone region is likely to be analysed in a subjective manner and the analyst might overlook a significant therapeutic effect. In addition, BS images reflect the secondary effects of a tumour on the skeleton and return false-positive results in patients with degenerative changes, inflammation, Paget's disease and trauma [23]; therefore, a definite diagnosis may often require radiography, MRI, or CT examinations of the area of interest.

The BSI system was developed to overcome the drawbacks of BS by using retrospective data analysis to focus on the differences in accumulation of radioactive materials among tumour metastasis, inflammation and normal physiologically associated potentials. Imbriaco et al. [8] developed and established this novel approach to provide reliable and quantitative assessments of bone involvement using BS, with minimum intra-observer and inter-observer variations [9]. The CAD system for BS applied in the present study, BONENAVI®, quickly calculates BSI, and does not require manual intervention.

We found a stepwise increase in aBSI values in parallel with EOD progression, which was in line with the morphological nature of EOD. The EOD grade system proposed by Soloway et al. [24] and Matzkin et al. [25] is well known to correlate with survival in metastatic PCa. In the present study, multivariate analysis showed that chemotherapeutic response as evaluated by a change in aBSI value was significantly and independently correlated with OS. By contrast, bone metastasis using EOD grade was not identified as a significant factor contributing to OS in univariate analysis. Thus, aBSI monitoring may reflect and predict the response to chemotherapy, which in turn could provide the opportunity of longer survival to patients with metastatic CRPC affecting the bone.

The most widely used tumour marker for diagnosing PCa and also for choosing the appropriate treatment for the management of PCa is PSA. A recent publication has shown that (i) PSA response was significantly correlated with OS and (ii) PSA was a potential predictor of survival after first-line chemotherapy [26]. However, PSA levels might be influenced by bone metastasis, lymph node metastasis and visceral involvement, which does not always correlate with tumour burden involving the bone. In the present study, the PSA level reduction rate showed a weak but significantly positive correlation with aBSI alteration 16 weeks after starting DEC therapy. All patients in the present series had metastatic CRPC affecting the bone, thus an indicator focusing on bone alterations conferred by DEC therapy might be superior to PSA reduction rate. As shown in Fig. 3, it is possible that an alteration in aBSI value might precede PSA level reduction, while an increase in aBSI value may be a sign of disease progression that appears earlier than rising PSA level.

A limitation of the present study is the possibility that we did not completely exclude the effect of flare-up phenomenon on BS when analysing aBSI alterations at 16 weeks after starting DEC therapy. In addition, 16 weeks after starting treatment might be too soon to exclude this possibility and also to evaluate the true chemotherapeutic effects of DEC therapy on bone regions. Another concern is that the Japanese-oriented database has not been fully established, even though the present CAD system has shown high sensitivity and specificity to detect bone metastasis [14–17]. Nevertheless, 28 of 42 patients with CRPC (66.7%) showed a reduction in aBSI value, although only 10 patients (23.8%) were shown to have a chemotherapeutic effect when evaluated by a conventional approach with EOD grade. Thus, more than half of the patients with stable disease (56.3%) had an aBSI reduction at 16 weeks after starting DEC therapy, which might potentially affect the decision-making process regarding the next step in these cases: keeping with DEC therapy, moving to a different treatment, or ceasing DEC therapy with no further treatment.

Despite the limitations, this is the first report to show a promising role for aBSI to evaluate patients with CRPC with bone metastasis receiving taxane-based chemotherapy. We believe urologists should attempt to quantify BS findings to provide important prognostic information for patients with PCa with bone metastasis.

In conclusion, chemotherapy response of osseous metastasis was effectively evaluated by aBSI in patients with CRPC as compared with BS despite other variables indicating a response. Furthermore, aBSI response was shown to be a strong prognostic indicator in patients with CRPC treated with DEC therapy. Thus, we consider that aBSI can provide useful information for determining an effective therapeutic strategy for PCa patients with bone metastatic disease. Larger studies could confirm our hypothesis.


None declared.