Subjects
From a pool of 436 patients with prostate cancer who underwent bone scintigraphy in our hospital from 1 July 2015 to 30 May 2016, we retrospectively selected 170 male patients who also underwent skeletal quantitative SPECT/CT. Clinical indications to refer patients for bone scintigraphy in our hospital were at initial staging and restaging of patients with prostate cancer, and monitoring therapeutic effects in bone metastases in castration-resistant prostate cancer. For precise diagnosis, whole-body SPECT/CT scans were performed on cervical-to-thoracic and/or lumbar-to-pelvic segments, when any localised uptake in cranial, vertebral and pelvic bones was observed during the evaluation of the initial staging or restaging of the patients with prostate cancer (Helyar et al., 2010)
We selected patients from our database on the basis of data from the picture archiving and communication system. Inclusion criteria were: (1) first-time patients who underwent bone quantitative SPECT/CT because of suspected bone metastasis for completion of initial staging or restaging, usually showing more than intermediate risk (according to the National Comprehensive Cancer Network risk categorisation), indicated by either high serum prostate-specific antigen (PSA > 10 ng/mL) or high Gleason’s score (GS > 7) or clinical stage >T2b on prostate biopsy; and (2) follow-up patients who had already been diagnosed with bone metastases from prostate cancer and received therapies for re-evaluation of bone metastases.
All patients were positively diagnosed with prostate cancer by prostate biopsy. Subjects included 112 patients who underwent bone scintigraphy for the first time, as well as 58 follow-up patients who had confirmed bone metastases. From our medical records, we obtained basic profile data, including age, body-weight (BW), and height, as well as blood biochemistry data, including serum alkaline phosphatase (ALP), PSA, and serum creatinine (Cr), acquired at the same time or within 2 months (14.4 ± 14.0 days) of the SPEC/CT test. This retrospective cohort study was reviewed by our hospital institutional review board. Documentation of informed consent from each patient was waived as the study was observational and data were obtained as part of daily clinical practice.
Bone SPECT/CT
The anterior and posterior whole-body scan and quantitative bone SPECT/CT co-registered/hybrid imaging was performed using a SPECT/CT scanner (Symbia Intevo, Siemens Healthcare K. K., Tokyo, Japan), 143 to 307 (average 204 ± 27) minutes after administration of 548 to 1170 (average 850.6 ± 168.1) MBq of 99mTc-methylene diphosphonate (MDP; FUJIFILM RI Pharma, Co. Ltd., Tokyo, Japan). For acquiring whole-body images, SPECT/CT scanning was performed using the low-energy high-resolution collimator with a SPECT condition of continuous rotation at 6°, 10 s/view, 10 min/rotation, and a low-dose CT condition of 130 kV and 50 mAs with CARE-DOSE application (Siemens Healthcare K. K., Tokyo, Japan). The acquisition range of SPECT/CT scanning was thoracic and/or abdominal to the pelvic regions, collected by one or two scans. CT images were generated with a 5-mm slice thickness using a smooth reconstruction kernel. SPECT images were reconstructed using the CGZAS algorithm (xSPECT bone mode) based on CT zonal mapping with the conditions of iteration 48, subset 1, and a Gaussian filter of smoothness 5 mm. The reconstruction was performed with 2.54-mm cubic 256 × 256 matrices, integrating a scatter correction using dual-energy scatter window subtraction and an attenuation correction based on attenuation maps derived from CT data. To measure precise bone uptake, a CGZAS reconstruction with CT zonal mapping consisting of 5 zones, air, fat tissue, soft tissue, medullary bone, and bone cortex was used, and reconstruction was performed based on X-ray attenuation mapping (Gnesin et al., 2016). The spatial resolution of the skeletal SPECT images was improved by zonal mapping, especially at the margins of the bone. Regular calibration of the SPECT/CT system was performed with an internal 57Co point-source phantom. The SPECT reconstructed values were decay-corrected to the time of injection and final values of quantitative radioactivity concentration were obtained.
Image interpretation
Three experienced physicians, who were certified by both the board of nuclear medicine and the board of diagnostic radiology, diagnosed skeletal lesions using a combination of SPECT and CT images. The findings of SPECT/CT were compared with the results of clinical follow-up every 3–6 months for 1 year after imaging, based on the consensus of the physicians as the gold standard. Clinical follow-up was performed by clinical examination, medical reports, imaging results, and PSA. Hot foci of degenerative changes were considered as skeletal degenerative diseases, such as osteophytes, degenerative intervertebral joint disease, degenerative tear of the marginal area of the vertebral body, or vertebral compression without sclerotic metastasis on CT images. Then, hot foci of bone metastases were identified in demarcated sclerotic bone metastatic areas, or in areas that could not be explained by the presence of degenerative changes. Skeletal disease extent was evaluated based on whole-body images according to literature, and scored on a 5-point scale, ranging from 0 to 4, to quantify the extent of disease (EOD) (Soloway et al., 1988).
Image analysis and quantification
Quantitative measurement was performed on a Syngo Via workstation (Siemens Healthcare K. K., Tokyo, Japan) by an experienced physician specializing in nuclear medicine. SUV based on BW was calculated according to the equation given below; relative weight in voxels of interest (VOI) was assumed to be 1 g per 1 cm3.
$$ SUV=\frac{\left( concentration of radioactivity in\ VOI\right)/\left( volume of\ VOI\right)}{\left( injected radioactivity\right)/\left( body weight\right)} $$
We inspected and selected up-to 6 lesions from the visually hottest spots on SPECT images. The actual hottest 3 lesions were defined by comparison of SUVmax for bone metastases. SPECT/CT images were analysed and each spherical VOI area on the lesion for SUV measurement was set manually on each hot spot, ensuring that the whole area of each hot location was included. Other hot activity areas, such as renal pelvis, ureteral tracts, and bladder were carefully excluded. Diagnosis of each skeletal lesion detected as a hot spot was based on CT scans of the bone window. Skeletal hot lesions were divided into two groups; those diagnosed as signifying bone metastases and those that were caused by degenerative changes. The three hottest spots of bone metastasis or the hottest spot of degenerative changes for each patient were included in further analysis. One visually normal thoracic (T) and/or lumbar (L) vertebral body in each patient was defined as a control. Each spherical VOI was set on the T and/or L vertebral body without hot spots when possible, and were size-matched in each case.
The maximum, peak, and average SUVs (SUVmax, SUVpeak, and SUVave, respectively) of 99mTc-MDP were obtained in visually normal vertebral bodies and skeletal lesions on SPECT/CT images based on the BW method. SUVpeak was defined as the average of the greatest value in a 1-cm sphere in the VOI. In addition, each metabolic volume (MV) and SUVave was obtained as the volume and average SUV at 40% threshold of the SUVmax in the bone metastases and degenerative changes groups. The three of the hottest lesions in the bone metastases group and the hottest lesion in the degenerative changes group were selected for further analysis according to the SUV measurement. If the patient showed no obvious abnormal uptake, the SUV in a normal vertebral body was allocated to the degenerative changes group.
Comparison of SUVs between the bone metastases and degenerative changes groups
In a patient-based analysis, the highest SUV in the bone metastases and degenerative changes groups for each patient was included for comparison. Each SUV in the bone metastasis group was compared with that in the degenerative changes group. In a lesion-based analysis, analysed lesions included up-to three of the highest SUVs for each patient in the bone metastasis group, the highest SUVs for each patient in the degenerative changes group, and SUVs in all T and L.
Evaluation of the effect of lesion size on SUVs
For evaluation of the effect of size on SUVs, bone metastasis and degenerative changes lesions were divided into 4 categories according to MV, as follows: <10, 10 to <20, 20 to <30, and ≥30 cm3. Each SUV from the bone metastases group was compared to an SUV from the degenerative changes group in the corresponding size category.
Discrimination accuracy of SUVs for bone metastasis compared to degenerative changes in hot spots
Discrimination accuracy of each SUV in the bone metastases group in hot spots was compared with that in the degenerative changes group by a receiver-operator characteristic curve (ROC) analysis. In a patient-based analysis, ROC analyses were performed with and without the ALP and PSA parameters. In a lesion-based analysis, discrimination accuracy of each SUV for bone metastases versus degenerative changes groups was compared by ROC analysis.
Sample number and statistical power planning
We assumed that the average SUVmax of bone metastatic lesions would be 10-fold greater than that of other lesions based on our preliminary small sample data (not published). The sample size was fixed by the statistical power analysis. Assuming that the size d, alpha, power, and allocation ratios of patients with bone metastasis to those without were 0.6, 0.05, 0.95, and 0.5, respectively, the required total sample size was estimated to be 166. Thus, we chose to include data from 170 subjects, which was slightly higher than the required number.
Statistical analysis
Statistical analysis and graphing of data were performed using SPSS ver. 24.0 (IBM Corp.) and Prism 7.0b (GraphPad software, Inc.) software. Data distribution was analysed using the Shapiro-Wilk normality test. Data was expressed as average ± standard deviation. Differences between the SUVs of multiple groups were analysed using the Kruskal-Wallis test with Dunn’s multiple comparisons. The differences in basic profile data, blood biochemistry data, and SUVs, between the bone metastases and degenerative changes groups, were analysed by the Mann-Whitney U test. A p value of less than 0.05 was considered statistically significant for all the analyses. A box-and-whisker plot was constructed, with the bottom and top of the boxes representing the first and third quartiles, and the data minimum and maximum values as the ends of whiskers. The median was marked as a line in the box.