Bone scan index on bone scintigraphy and radiation therapy for bone metastases from cancers other than prostate and breast cancers: a retrospective observational study

A low bone scan index that is associated with a better prognosis in patients with bone metastases from prostate or breast cancer, the former often being osteolytic, has been established. In this study we aimed to use new automatic analysis software (VSBONE BSI; Nihon Medi-Physics, Tokyo, Japan) to investigate whether the pre-radiation therapy bone scan index, derived from bone scintigraphy images, is a prognostic indicator in patients undergoing radiation therapy for bone metastases from cancers other than breast or prostate cancer.

Bone is a common site for metastasis of malignant tumors, 50-75% of malignant tumors metastasizing to bone (1,2). It has been reported that 30-40% of lung cancers metastasize to bone during their clinical course (3). The modality of choice for evaluation of the presence and extent of bone metastases is bone scintigraphy (BS) using radiopharmaceuticals such as 99m Tc-labeled phosphonates or phosphates because the whole body can easily be scanned in one session. However, because radiopharmaceuticals accumulate at sites of hydroxylapatite deposition, non-metastatic bone lesions caused by trauma and degeneration can appear as hot spots (1,4). In recent years, hot spots have been characterized and quanti ed using an arti cial neural network (ANN) and an index for the total amount of bone metastases, called the bone scan index (BSI), has been developed (5,6). For example, bone metastases from prostate cancer are often osteoblastic and readily appear as hot spots. Many reports on the relationship between BSI and prognosis indicate that the prognosis is better in patients with lower BSI (7,8,9). However, there have been few reports on BSI and malignant tumors other than prostate cancer. When bone metastases are detected in patients with prostate or breast cancers, the median survival is expected to be years with recent new endocrine therapy and molecular-targeted drugs. In contrast, the median survival is as short as approximately 7 months in patients with lung cancer and hepatocellular carcinoma (10,11,12,13). Therefore, radiation therapy (RT) is often administered to relieve symptoms of bone metastasis from lung cancer and other cancers. In this study, we determined the pre-irradiation BSI and prognosis in patients who underwent RT for bone metastases from malignant tumors other than prostate and breast cancers, such as lung cancer and hepatocellular carcinoma. We used new analysis software VSBONE BSI (Nihon Medi-Physics, Tokyo, Japan) to automatically calculate BSI. To the best of our knowledge, this is the rst time this software has been used in the clinical study.

Methods
We retrospectively reviewed relevant data of patients treated with local RT for bone metastases between 2013 and 2019, and we selected 51 patients whose bone metastases had been con rmed by computed tomography (CT) or magnetic resonance imaging (MRI) and who had also undergone BS. This study was approved by our institutional Review Board; informed consent was obtained from all patients. We have read the Declaration of Helsinki and followed its guidelines in this investigation. The clinical data of all 51 patients are summarized in Table 1. Age at the time of BS ranged from 32 to 85 years (median 66 years) and there were 33 men (64.7%) and 18 women (35.3%). The most common primary tumor was lung cancer (21 patients, 41.2%) and the second most common liver (nine patients, 17.6%). Seven patients had solitary bone metastasis (13.7%) and 44 multiple bone metastases (86.3%). According to the CT ndings, bone metastases were classi ed into the following three types (14,15): osteoblastic (n = 6, 11.8%), osteolytic (n = 15, 29.4%), and mixed (n = 30, 58.8%). Concurrent visceral metastases were detected in 36 patients (70.6%). Abbreviations: RT, radiation therapy.

Bone scintigraphy
Each patient was injected with 740 MBq of 99m Tc-methylene diphosphonate (Fuji lm RI Pharma, Tokyo, Japan) or 99m Tc-hydroxymethylene diphosphonate (Nihon Medi-Physics, Tokyo, Japan). Whole-body scintigraphy was performed ≥ 3.5 hours after the administration, using a gamma camera at a speed of 12 or 13.3 cm/min with a low-energy high-resolution collimator, 512 × 1024 or 256 × 1024 matrix size, zoom factor 1.00, and 139 keV photopeak with 20% window or 140.5 keV photopeak with 10% window. One of the following two gamma camera units was used; Forte (Philips Japan, Tokyo, Japan) or Discovery NM/CT 670 (GE Healthcare Japan, Tokyo, Japan). Raw image data were transferred to a PC with VSBONE BSI software and analyzed.

VSBONE BSI
The database of VSBONE BSI comprises data of pairs of anterior and posterior BS images of Japanese patients with prostate cancer (16,17). Additionally, skeletal segmentation from 246 patients with prostate cancer and hot spot extraction from 896 patients with prostate cancer were performed, using a butter ytype network of fully convolutional networks. Fully convolutional networks have no threshold for hot spots unlike ANN. Hot spots suspected of being metastases are depicted in red, whereas lesions suspected of being false positives are depicted in blue. From these, the BSI is automatically calculated. The BSI of the 51 patients in this study ranged from 0.00-11.86% (mean 2.50%).

Results
Univariate analysis revealed that the BSI was signi cantly higher in patients with lung cancer than in those with other tumors (mean BSI 3.26% vs. 1.97%; P = 0.009) ( Fig. 1 and Table 2). The BSI was signi cantly lower in patients with solitary and osteolytic bone metastases than in those with other types of metastases (P = 0.035 and P = <0.001, respectively) ( Fig. 2 and Table 2). Abbreviations: BSI = bone scan index; CI = con dence interval; RT = radiation therapy. *Signi cant difference between two groups (P < 0.05) Overall survival time after BS ranged from 25 days to 2113 days with a median of 221 days. A high BSI (≥ mean) was not correlated with prognosis ( Fig. 3 and Table 3). Both univariate and multivariate analyses showed that the younger the patients were when BS was performed or bone metastases diagnosed, the better the prognosis (P = 0.016 and P = 0.036, respectively) ( Fig. 4 and Table 3). Abbreviations: CI = con dence interval; RT = radiation therapy. * Signi cantly different (P < 0.05)

Discussion
Bone metastases often occur in patients with prostate or breast cancer. Postmortem examination reportedly reveals bone metastases in 68% of patients with prostate cancer and 73% of those with breast cancer (19). In particular, bone metastases in patients with prostate cancer tend to be osteoblastic and are highly detectable as hot spots on BS (1,20). BSI was rst developed for characterization and quantitation of these hot spots, and a method using ANN, which was trained on a patients' database, was established (5,21). In Japan, BONENAVI software (Fuji lm RI Pharma) was developed using ANN trained on a database of Japanese patients with prostate cancer (22). BONENAVI BSI automatically detects hot spots and makes calculations by combining abnormal areas with a high probability of representing metastases. Recently, databases of Japanese patients with breast cancer and lung cancer have been added and reported to show good accuracy (23). In this study, we used VSBONE BSI, new analysis software, to automatically calculate of BSI for the rst time, to the best of our knowledge. VSBONE BSI software also uses ANN trained on databases of Japanese patients and automatically calculates BSI on the basis of the same basic principles as does BONENAVI BSI (16,17). In this study of patients with cancers other than prostate and breast cancers, BSI was signi cantly higher in those with bone metastases from lung cancer than in those with bone metastases from other malignant tumors. In general, metastases from other malignant tumors including hepatocellular carcinoma, are predominantly osteolytic, whereas metastases from lung cancer are more likely to have increased hydroxylapatite deposition. In this study, there were patients with BSI 0.00% despite multiple bone metastases from hepatocellular carcinoma having been identi ed by CT. Possible reported explanations for bone metastases not being detected by BSI are that 2.4% of bone metastases from hepatocellular carcinoma are osteolytic and there are soft tissue components in 38% of patients (13,24). Our nding of signi cantly lower BSI with osteolytic lesions in this study is consistent with past reports that BONENAVI BSI of mild osteoblastic lesions from prostate, breast, and lung cancers are signi cantly lower than osteoblastic or mixed types metastases (14).
Many studies have reported that high BSI is a poor prognostic factor in patients with castration-resistant prostate cancer (4,7,9). Some studies have also reported that multiple bone metastases from lung and other cancers have a poorer prognosis than solitary bone metastases (12,25). We found no correlation between BSI and prognosis in patients in this study, including those with lung cancer; however, osteolytic lesions may have been underestimated.
Few studies have been conducted on BS and RT to date. As for imaging modalities used to evaluate metastatic lesions before treatment with RT for bone metastases, no difference in therapeutic effects between BS, MRI, and 18 uorodeoxyglucose positron emission tomography have been reported (26). We have reported, we believe for the rst time, that age at the time of BS is a signi cant favorable prognostic factor for RT for bone metastases, suggesting that long-term survival may be expected even in young patients with high BSI. For example, radium-223 (Xo go; Bayer Pharmaceuticals, Whippany, NJ, USA), which is incorporated into newly formed osteoblastic bone matrix, is associated with longer survival in patients with bone metastases from prostate cancer (27). If radium-223 is approved for treating bone metastases from malignant tumors other than prostate cancer and covered by health insurance in Japan, we can expect prolongation of survival in patients with lung cancer and high BSI.

Conclusions
The only signi cant association with survival identi ed in this study was for age at the time of bone scintigraphy and at time of diagnosis of bone metastases. In particular, we found no association between bone scan index and survival in the whole study cohort.

Declarations
Ethics approval and consent to participate Not applicable.

Consent for publication
Not applicable.

Availability of data and material
The dataset used during this study are available from the corresponding author on reasonable request.

Competing interests
Drs. Ishibashi, Maebayashi, Kimura, and Okada declare that they have no competing interests.

Funding
No funding was received speci cally for this work.
Authors' contributions NI collected the patients' data and treated the patients and was a major contributor to writing the manuscript. YK and MO interpreted the BS images. TM treated the patients. All authors read and approved the nal manuscript.
castration-resistant prostate cancer and bone metastases: results from a phase 3, double-blind, randomised trial. Lancet Oncol. 2014;15 (7):738-46. Figure 1 Representative example of a 62-year-old man with lung cancer and multiple bone metastases. Computed tomography showed multiple osteoblastic lesions (a,b). Bone scintigraphy showed multiple hot spots (c) and VSBONE BSI showed these lesions as red and blue. The BSI was extreme high at 6.45% (d).

Figure 2
Representative example of a 76-year-old man with hepatocellular carcinoma and multiple bone metastases. Computed tomography showed multiple osteolytic soft tissue masses (a,b). Bone scintigraphy failed to detect these lesions (c) and VSBONE BSI showed hot spots as blue only. The BSI was undetectable at 0.00% (d).

Figure 3
Kaplan-Meier survival curves of patients with BSI < mean 2.50% or ≥ mean. The difference between the two groups is not statistically signi cant (P = 0.322).