Bone Scan Index as an imaging biomarker to predict overall survival in the Zeus/SPCG11 study
Reza et al.—2016 ASCO Annual Meeting
Background: The Zeus/SPCG11 study is a randomized controlled clinical trial with the aim to assess the efficacy of zoledronic acid (ZA) in preventing bone metastases in high-risk prostate cancer (PCa) patients. Bone Scan Index (BSI) reflects the tumour burden in bone calculated from bone scintigraphy, and has recently been validated as an imaging biomarker in metastatic PCa patients. The purpose of this study was to investigate if change in BSI during treatment may serve as a useful imaging biomarker to predict clinical outcome in the Zeus/SPCG11 study population. Methods: We retrospectively selected as our BSI-study cohort all patients with bone scan image data of sufficient quality to allow for both baseline and 48-months follow-up BSI-assessments. BSI data was obtained by using the automated quantification software EXINI Bone BSI, in a blind fashion, without knowledge of any clinical data or treatment randomization. Clinical data on age, overall survival (OS) and prostate-specific antigen (PSA) concentration in blood at baseline and upon follow-up was collected separately. Association between clinical data and BSI change during treatment was evaluated using Cox proportional-hazards regression models, Kaplan-Meier estimates of the survival function and Log rank test. Discrimination between prognostic variables was assessed using the concordance index (C-index). Results: The 176 patients who fulfilled the inclusion criteria presented showed baseline characteristics similar to those of the final total Zeus/SPCG11 study population (N = 1,433)(p= 0.83). In our BSI-study cohort (N = 176), patients with a BSI change < 0.3 had a significantly longer median survival time compared to patients with BSI increase of > 0.3 (32.3 months and 18.2 months respectively) (p < 0.001). In the Cox analysis BSI change from baseline to follow-up was significantly associated with OS (p < 0.01 and C-index = 0.6) while age and PSA change were not significantly associated with OS. Conclusions: BSI change during treatment was associated with OS in high-risk PCa patients from the Zeus/SPCG11 study. BSI may be a useful imaging biomarker in future clinical trials involving PCa patients with bone metastases.
Automated bone scan index as an imaging biomarker in metastatic castration resistant prostate cancer (mCRPC) patients treated with radium-223
Anand et al.—2016 ASCO Annual Meeting
Background: There are no validated imaging tools to assess response to the Radium-223 (Ra 223). In this study, we explored the analytically validated automated bone scan index (auto-BSI) as an imaging biomarker in mCRPC Patients (Pts) treated with Ra-223. Methods: In a multi-center retrospective study, available bone scans of mCRPC Pts treated with Ra 223, were collected at baseline (BL) and at treatment follow-up (Tx-FU). The auto-BSI was analyzed using the EXINIboneBSI version 2. Logistic regression, cox regression and Kaplan Meier analyses evaluated the auto-BSI associations with PSA response, defined as 50% decline from BL; pain response, defined as decrease in the VAS score from BL; Alkaline phosphatase (ALP) response, defined as 30% decline from BL; and with overall survival (OS). Results: Of144 Pts treated with Ra223, 66 (46%) had a BL bone scan and 25 of them also had a Tx-FU bone scan taken 18 to 30 weeks after the first treatment. No significant differences were observed in BL characteristics of Pts with bone scans versus without. The median number of ra223 doses were 6 (1-6). Median BL auto-BSI was 2.4 (range:0.01–33.2) and median Tx-FU auto-BSI was 0.5 (range:0-20.1). The BL auto-BSI was associated with OS (HR = 1.4; 95%CI = 1.1-2.0) and with ALP response (HR = 1.6; 95%CI = 1.04-2.4), p = 0.02. BL auto-BSI was not associated with pain, p = 0.48, or with PSA response, p = 0.76. OS at 9 months with BL auto-BSI < 5 was 76% vs. 40% in Pts with BSI > 5. In Tx-FU analysis, 17 (68%) of the 25 Pts had ≥ 30% decrease in auto-BSI from BL; Only 5 (20%) Pts were PSA responders, and all 5 Pts had ≥ 30% decrease in auto-BSI. Conclusions: Baseline auto-BSI was associated with OS and with ALP response. The findings encourage prospective investigations to validate the role of BSI as an imaging biomarker in mCRPC Pts being treated with Ra-223.
Improved therapy response assessment in patients with bone metastases of prostate cancer using an expert system for bone scan interpretation
Päsler et al.—2015 EANM Congress
Aim: For the evaluation of treatment response in bone metastases of prostate cancer, biomarkers like prostate specific antigen (PSA) and visually interpreted bone scans have limitations in early and accurate assessment of the further course of disease. Hence we investigated if a semi-quantitative parameter of systemic bone involvement - the bone scan index (BSI) - enables an improved response assessment. Methods: In 43 patients with prostate cancer (73?6yrs; median Gleason score: 8, range 5-10) 157 bone scan pairs with a median interval of 3 months were performed for the assessment of response to chemotherapy. Whole body bone scanning was acquired 2-3 hours after injection of 7 MBq Tc-99m-MDP per kg body weight with a large field dual-head gamma-camera. Scans were evaluated (i) visually by an experienced nuclear medicine specialist and (ii) by the artificial neural network based determination of the BSI using the expert system EXINI bone (EXINI, Lund, Sweden). Progression of metastatic bone involvement was defined by (i) occurrence of two new lesions in the visual interpretation (VI) by a nuclear medicine specialist and (ii) different thresholds for the relative increase of the BSI by 5, 10 or 25% (BSI-5%, BSI-10%; BSI-25%) between two scans. Finally, assessments according to bone scanning were compared to changes of prostate specific antigen (PSA) in serum. Results: Based on VI and BSI-25% progression was detected in a similar magnitude (28 vs. 27%, p=0.7440). Though, PSA increases suggested progression significantly more frequently (50%, p<0.0001). Using BSI-5% and BSI-10% significantly higher rates of progression compared to VI (49 and 43 vs. 27%, p<0.0001) were detected - not significantly different to the rate of PSA increase. However, assessment of progression diverged between PSA and BSI changes in 38% of the cases (for BSI-5%) and 39% (for BSI10%), respectively. Conclusion: A relative increase of 5 - 10% in the BSI - in a frequency comparable to PSA - is superior to the visual interpretation of scintigraphy in the assessment of bone metastatic progression. Thus bone scintigraphy in conjunction with expert system evaluation holds potential for more accurate response monitoring in these patients. Nevertheless, results have to be correlated to longer follow up, in order to explain differences in progression assessment between BSI and PSA in a considerable number of cases.
Validation of the imaging biomarker Bone Scan Index - influence of image quality on reproducibility
Kaboteh et al.—2015 EANM Congress
Objective: Bone Scan Index (BSI) represents the tumor burden as percentage of total skeletal mass. In several studies, BSI has shown to be associated with survival in prostate cancer patients. To clinically qualify BSI as an imaging biomarker a comprehensive analytical validation is required. In this study, we assessed the reproducibility of BSI in assessing bone scans with varying image quality, measured as total number of Counts. Method: Two repeat whole-body bone scans were obtained from each of 48 patients with prostate cancer. Both bone scans were done within four hours after a single intravenous injection of 600 MBq 99mTc MDP. In 21 patients the same scan speed was used in both scans, adjusted so that routine bone scans were in accordance with the EANM procedure guidelines (anterior and posterior images each contain >1.5 million counts). In 27 patients, the second scan was acquired with double scan speed, resulting in a reduction of counts in the images by 50%. The BSI values were generated automatically using the software EXINI Bone BSI. A reproducibility threshold of automated BSI for consistent measurement of change in bone scan was defined as 95th percentile of the absolute difference. Result: The mean BSI was 2.29 in the ?same scan speed? group and the mean absolute difference between the two BSI values was 0.16 (SD 0.28), the reproducibility threshold was observed at 0.37. In the group with one double speed scan the mean BSI was 1.98. The mean differences between the low counts scan and standard counts scan were 0.08 (SD 0.29) (p=0.17). Five of the 27 patients (19%) showed differences greater than the reproducibility threshold of 0.37. Conclusion: Bone scans with low counts are less reproducible and it is important that EANM procedure guidelines are followed for this type of quantitative analysis. Within reproducibility threshold of 0.37, BSI can be used to produce a consistent quantitative analysis of bone scans in patients with prostate cancer.
Automatic bone scan index for therapy response assessment of radium-223-dichloride (Ra-223) therapy in advanced prostate cancer
Sakretz et al.—2015 EANM Congress
Aim: Evaluation of progression and therapy response is needed to improve and optimize individualized therapy management of patients suffering from metastatic castration resistant prostate cancer (mCRPC). However, there is a lack of consistent data for therapy response assessment to optimize treatment. Treatment decisions based on PSA-value alone are inadequate, therefore other clinical and radiologic criteria indicating disease progression should be taken into account. Therefore we evaluated the utility of an automatic bone scan index (aBSI) to assess response to Ra-223-therapy in patients with mCRPS with bone metastases, in comparison to established response markers, e.g. number of lesions or PSA. Materials and Methods: Nine patients with clinical progress of bone metastases from mCRPC, who underwent 6 cycles of Ra-223-therapy were retrospectively included in this pilot study. Whole-body bone scans with Tc-99m-DPD were performed and blood samples were taken approximately four weeks before and after 6 cylces of Ra-223-therapy. The aBSI was calculated using EXINIboneBSI (EXINI Diagnostics AB, Sweden). Changes of aBSI, number of lesions, levels of PSA and alkaline phosphatase (AP) were compared pre- and post-therapy. Correlation between aBSI and PSA as well as AP was evaluated by calculation of Pearson?s correlation coefficient. Results: Therapy response / progress was defined as decreasing / increasing aBSI, less / more than two new lesions in whole-body bone scan and lower / higher PSA level compared to baseline. Response was found in 2/9 (22 %), stable disease in 1/9 (11%) and progress in 6/9 (67%) of the patients. The median aBSI was 2,67?2,62 at baseline scan and 2,71?2,73 at control scan after 6 cycles of therapy. The median number of lesions was 18?40 and 22?42, respectively. The median PSA level was 38,5?339,19 and 46,7?364,63, respectively. Median alkaline phosphatase level was 1,28?1,87 and 1,23?2,18, respectively. The ∆aBSI (mean 0,04?0,86) showed a statistically significant correlation with ∆PSA (mean 60,79?117,2; r = 0,76, p = 0,014) and weaker one with ∆AP (mean -0,28?1,86; r=0,59, p = 0,093). Limitations are the small patient sample-size, the retrospective analysis and high inter-individual variability. Conclusions: The aBSI might be a useful and promising data analysis tool for the assessment of therapy response to Ra-223. Studies in larger cohorts are needed.
Routine usefulness analysis of computer assisted diagnosis (CAD) software for bone scan index
Ferrer et al.—2014 EANM Congress
AIM: We purchased a CAD software, commercially available, in order to help to detect metastases in bone scans. This retrospective study investigates lesion detection performance of such CAD tool as compared to nuclear physician diagnosis. MATERIAL & METHODS: In 2013, 245 patients (112 F/133 M) with different types of cancer underwent one to several whole-body (WB) bone scans (388). 14832 lesions were detected by the CAD system on anterior and posterior views using an intermediate sensitivity among 3 levels (low, intermediate, high). Trained physicians were able to alter CAD decision (confirmed by SPECT/CT) for each lesion from high to low (HL) and low to high (LH) metastatic probability. XML files stored by CAD software were analyzed to extract relevant informations such as skeletal localization, number of lesions, maximum counts per lesion. RESULTS: The number of examinations was 179 (151 resp.) for prostate (breast resp.). The 58 remaining examinations were performed for other types of cancer. Number of metastases was 3903 for prostate cancer and 2082 for breast cancer. Prostate cancer metastases were mostly located in ribs, thoracic spine (TS) and pelvis (21.7, 10.9 and 9.7 % resp.) whereas in TS, pelvis, skull and lumbar spine (LS) for breast cancer (21.0, 18.1, 12.9 and 11.3% resp.). Whatever the pathology, HL and LH changes were 11.8% of detected lesions and the most frequent modifications appeared in TS, pelvis, skull and LS (17, 15, 13 and 9% resp.). For the most frequent examinations i.e. breast (prostate resp.) cancer, overall changes were 5.5% (3.7% resp.) of detected lesions. In breast cancer, the most frequent lesion alterations were respectively 8.7, 7.8, 7.2 and 6.6% for skull, pelvis, TS and ribs. The same analysis performed for prostate cancer lead to 8.2, 7.2, 3.8 and 3.1% for ribs, TS, skull and pelvis respectively. The mean of maximum counts in lesions manually changed by physicians was 25% lower than the respective value in lesions automatically labeled by CAD system as highly metastatic (Wilcoxon-test p < 10e-6). CONCLUSION: This study showed that CAD system is an effective tool for physicians in their daily work as less than 12% of lesions were falsely identified. Some improvements would be helpful e.g. labeling unlabeled foci, as some ribs lesions, in breast cancer especially, were not identified at all, making any changes impossible. Maybe, a learning phase based on breast cancer WB scans would have helped the CAD system to perform better.
Relationship of bone scan index and progression-free survival after docetaxel treatment for CRPC patients with bone metastases
Mizokami et al.—2014 GU Cancers Symposium
Background: A computer-aided diagnosis system for bone scintigraphy using semiquantitative index [Bone Scan Index (BSI)] has been used to measure the tumor burden of bone metastases. We examined relationships of BSI, bone turnover marker, and prostate-specific antigen (PSA)-progression free survival (PFS) after docetaxel-treatment for castration resistant prostate cancer (CRPC) with bone metastasis.Methods: Sixteen CRPC patients with bone metastases (median age 72, range 52 to 82) were treated with docetaxel. We evaluated bone metastasis by bone scintigraphy before or around six months after docetaxel-treatment retrospectively. BSI was automatically calculated by BONENAVI software version 1 (FUJIFILM RI Pharma, Co. Ltd., Tokyo, Japan; Exini Bone, Exini Diagnostics, Sweden). Serum PSA, bone alkaline phosphatase (BAP), carboxyterminal telopeptide of type I collagen (I-CTP) were examined every months. PSA-PFS was evaluated after docetaxel-treatment and compared with baseline of BSI, BAP, I-CTP, and change of these value after treatment. Overall survival (OS) was also evaluated by these markers. The rate of patients with PFS and OS was estimated by the Kaplan-Meier method.Results: Baseline of BSI, the serum BAP, and I-CTP before docetaxel-treatment did not affect PFS. The change of BAP and I-CTP by the docetaxel-treatment also did not affect PFS. Only the change of BSI affected PFS and the median PFS of CRPC patients with increased BSI and decreased BSI was 5.5 months and 10 months, respectively (p=0.026). Although OS showed a longer tendency in CRPC patients with decreased BSI than with increased BSI, there was not the significant difference (p=0.12).Conclusions: The change of BSI affected PFS in CRPC patients with bone metastases. Bone scan and its evaluation with BONENAVI was effective to monitor the clinical course during chemotherapy.
Bone scan index as a biomarker of systemic metastatic bone involvement in prostate cancer can be easily obtained using an expert system
Namazian et al.—2013 EANM Congress
Aim: Bone scintigraphy is sensitive for the detection of metastases but its summary into one parameter of systemic bone involvement is a challenge. The bone scan index (BSI) has been suggested for this purpose, but it is complex to determine it. The present study evaluates the helpfulness of an expert system in this context. Methods: In 44 patients (69±6yrs.) scanned for bone metastases of prostate cancer (median Gleason score: 8, range 5-10) the EXINI bone software (EXINI, Lund, Sweden) has been employed to facilitate the determination of BSI. The software is based on artificial neural networks that have been trained using a large database of bone scans interpreted by fully-fledged experts. BSI as a measure of the extent of skeletal involvement and the number of metastases (NUM) were determined twice in 193 scans: (i) fully automated as provided by the software (available within seconds after data loading) and (ii) including individual selection and adaptation of lesions by a nuclear medicine physician (necessitating a few minutes). Furthermore, correlation of (i) physician modified parameters (BSI, NUM) to Gleason score and (ii) therapy related change in these parameters to change in tumor-marker PSA (prostate-specific antigen, baseline range 7-1046ng/ml) after 11 months of chemotherapy (Docetaxel) were evaluated. Results: In a small number of cases (4%) the automated scan analysis was cheated and included the bladder as a "bone metastasis". In all other cases a high correspondence between the automated and physician-modified result of analysis was observed for both parameters (BSI: r2=0.99, p<0.0001, NUM: r2=0,98, p<0.0001). There was no significant difference between automated and modified analysis for BSI (2.92 vs. 2.96, p greater than 0.05) but a small, significant difference in NUM (29 vs. 31, p<0.0001). In patients with a Gleason score ≥8, a tendency of more hotspots with increasing score was seen (r2=0.27, p=0.0197). With respect to chemotherapy monitoring the correlation to change in PSA was significant for change in BSI (r2=0.81, p<0.0001) and at the borderline of significance for NUM (r2=0.32, p=0.0569). Conclusion: In the vast majority of cases the expert system suggests lesions for BSI calculations, which require only minor modifications, and the few cases of program failure (due to bladder-activity) are easy to notice. Considering the minimal time requirement the program is most suitable for a clinical context. Moreover, preliminary analysis shows correlation of BSI with other biological parameters indicating BSI to be a valuable marker of systemic bone involvement.
Bone Scan Index validated using 18F-PET/CT in prostate cancer patients
Ly et al.—2011 EANM Annual Congress
Objectives Bone scan index (BSI) is a method to quantify the percentage of the skeleton affected by tumor mass on a whole-body bone scan. BSI can be used to predict survival and to monitor patients with prostate cancer. The analysis uses as input the two-dimensional bone scan images for the estimation of skeletal involvement, which is of three-dimensional nature. The purpose of this study was to validate the two dimensional BSI method using a corresponding analysis of 18F-PET/CT images in prostate cancer patients.
Method Eleven prostate cancer patients who had undergone both a whole-body bone scintigraphy (750 MBq Tc-99m MDP) and an 18F-fluoride-PET/CT were studied. BSI was calculated using the automated software EXINI bone (EXINI Diagnostics AB, Lund Sweden). A custom program was developed for the quantification of skeletal involvement in the 18F-PET/CT images. The program uses the CT images to identify, segment and quantify the skeletal volume. Abnormal 18F-fluoride uptake is automatically detected, segmented and its involvement of the total skeleton is calculated. The metastases automatically detected by the EXINI bone program were used to calculate BSI, and the corresponding abnormal uptake in the 18F-fluoride-PET/CT images were manually selected for the calculation of a corresponding “PET/CT-BSI”.
Results BSI ranged from 0.03-1.88 (mean 0.51) and PET/CT-BSI from 0.04 to 1.49 (mean 0.43). Linear regression analysis showed a correlation of r=0.98. A total of 55 metastases were included from the eleven patients. For the individual metastases, linear regression analysis showed a correlation of r=0.56 between the two BSI methods.
Conclusions The two dimensional BSI method based on whole-body bone scans correlate relatively well with the three dimensional PET/CT-BSI method on a patient level. A discrepancy in skeletal involvement between the two methods could be expected because of the added dimension in PET/CT. These findings show the feasibility of using the two dimensional bone scans to estimate BSI even though the tumor burden is three-dimensional by nature.
Prognostic value of metastasis pattern for morbidity in prostate cancer measured by quantitative bone scintigraphy
Højgaard et al.—2010 ASCO Annual Meeting
Background: Quantification of bone metastases in the skeleton expressed as Bone Scan Index (BSI) and location of these are prognostic factors for mortality and morbidity in prostate cancer (PC). The purpose of this study was to identify prognostic factors for morbidity in PC based on location and extent of metastases derived from quantified bone scintigraphies (BS). Methods: Data from 107 consecutive patients with noncurable hormone-naive PC with available pretreatment 99mTc-MBP BS from the period 2003-2007 were retrospectively analyzed. BSI and location of metastases were measured from the BS using Exini Bone 1.0 (Exini AB, Sweden). Observed BSI for 8 regions was compared to the expected BSI calculated from the red bone marrow distribution. A logistic regression model for having skeletal events (SE), defined as external beam radiation for bone pain, spinal cord compression or pathological fracture at any location in the skeleton based on pre-treatment BSI, metastases distribution and PSA, was made. Results: The average BSI for the 72 patients with metastases at diagnosis was 1.063. Metastatic distribution expressed as the average BSI in these patients for each region was compared to the expected BSI of the same regions (Table). Patients with metastases to the skull had increased risk of SEs with OR: 4.2 (95% CL: 1.1-16.4, p<0.036) compared to the baseline patient with no metastases. No other regions, nor total BSI or PSA reached significance. Conclusions: PC does not metastasize randomly to the skeleton but clusters in the regions closest to the primary tumor. Patients with skull metastases have a higher risk of SEs throughout the skeleton. Very few SEs actually occurred in the skull region but skull metastases may indicate an atypical pattern of dissemination with a more aggressive phenotype of PC with more SEs.