123I metaiodobenzylguanidine (MIBG) uptake predicts early relapse of neuroblastoma using semi-quantitative SPECT/CT analysis

Objective 123I metaiodobenzylguanidine (MIBG) scintigraphy is a useful tool for the diagnosis of neuroblastoma (NB). MIBG uptake is correlated with norepinephrine transporter expression; hence, it is expected that high-MIBG tumors would be more highly differentiated and have a better prognosis than those with lower expression. We have introduced a method of assessing MIBG accumulation semi-quantitatively using SPECT/CT fusion images. The purpose of this study was to evaluate the relationship of 123I MIBG uptake measured by semi-quantitative values of SPECT/CT and early relapse of NB. Methods We studied the cases of 11 patients (5 males and 6 females, age 5–65 months, median age 20 months) with histopathologically proven NB between April 2010 and March 2015. The early-relapse group was defined as patients who had relapsed within 3 years after the first 123I MIBG SPECT/CT exam. Other patients were classified as the delay-relapse group. Uptake of MIBG was evaluated using the count ratio of tumor and muscles. T/Mmax and T/Mmean were defined as follows: T/Mmax = max count of tumor/max count of muscle, T/Mmean = mean count of tumor/mean count of muscle. Results The average T/Mmean values of the early-relapse group and delay-relapse group were 2.65 ± 0.58 and 7.66 ± 2.68, respectively. The T/Mmean values of the early-relapse group were significantly lower than those of delay-relapse group (p < 0.05). The average T/Mmax of the early-relapse group and delay-relapse group were 8.86 ± 3.22 and 16.20 ± 1.97, respectively. There was no significant difference in T/Mmax values between the two groups. Conclusions Low 123I MIBG uptake using semi-quantitative SPECT/CT analysis was correlated with early relapse of NB.


Introduction
Neuroblastoma (NB) is one of the most common but very rare solid malignant tumors of early childhood and occurs in approximately 150 children each year in Japan [1]. NB is derived from primitive neural crest cells found in several areas of the body and is the most common extracranial solid tumor in children age 5 or younger [2]. 123 I metaiodobenzylguanidine (MIBG) scintigraphy is a useful tool for the diagnosis of NB [3]. Hybrid single-photon emission computed tomography/computed tomography (SPECT/CT) is expected to provide additional anatomical information, and more efficient diagnostic capability [4].
The scoring method due to the distribution of bone metastases has been proposed and used as a prognostic method for predicting prognosis using 123 I MIBG planar image of NB [5,6]. On the other hand, it is considered that it is difficult to predict prognosis from visual evaluation of planar images or SPECT images of 123 I MIBG examination [7,8].
Interpretation of 123 I MIBG planar images and SPECT images is usually qualitative. Several semi-quantitative scoring systems for 123 I MIBG examination of NB have been proposed and explored; however, systems based on planar imaging findings and SPECT findings are not typically included [5,[9][10][11][12][13][14][15]. Fred et al. [8] reported that the liver-to-tumor count ratio of 123 I MIBG SPECT images was useful for differentiating NB from other neuroblastic tumors. Although the liver is commonly used as reference organ, MIBG tends to accumulate in the liver because it is metabolized there. In addition, liver uptake of MIBG is not homogeneous; large differences between the right and left lobes have been reported [16], and it is difficult to designate an appropriate region of interest (ROI) or volume of interest (VOI) as the reference. For these reasons, the liver is not an appropriate reference site for analysis when using MIBG.
In the present study, we introduced a new index that uses muscle as a reference. 123 I MIBG accumulation of muscle is higher than the background, but relatively small because it is not metabolized in muscle. Because of this relatively low accumulation in the muscle, it is difficult to analyze using conventional planar images; not until the more recent development of SPECT/CT fusion images was this technique feasible.
The purpose of this study is to evaluate the ability of the new semi-quantitative values of 123 I MIBG SPECT/ CT to predict early-relapse lesions of NB.

Patients
Inclusion criteria were histopathologically proven neuroblastoma, a maximum interval of 50 days between 123 I MIBG SPECT/CT examination and surgical intervention and no tumor-specific treatment before the time of scintigraphy. 123 I MIBG SPECT/CT examinations were performed between April 2010 and March 2015 in 11 consecutive pediatric patients meeting our criteria (Table1). The group consisted of 5 males and 6 females (age 5-65 months, median age 20 months). At the time of diagnosis patients were staged according to the International Neuroblastoma Staging System (INSS) [17] and International Neuroblastoma Risk Group (INRG) risk classification [5]. According to the International Neuroblastoma Pathological Classification (INPC) of NBs, almost all NBs in our study were subtypes of poor differentiation. MYCN amplification was proven by histopathological study. Vanillyl mandelic acid (VMA), homovanillic acid (HVA) and neuron-specific enolase (NSE) before treatment were also measured. Patients were followed up after MIBG examination (270-3016 days, median 1749 days). Treatments for patients were selected based on the clinical staging (surgical resection: 6, chemotherapy: 10, auto peripheral blood stem cell transplantation: 5, radiotherapy: 3). We defined the early-relapse group as patients who had relapsed within 3 years after the first 123 I MIBG SPECT/CT exam. We defined the date of relapse as when the pediatrician comprehensively diagnosed recurrence based on clinical findings, histological examination, imaging studies and other information. Other patients were classified as the The whole-body imaging protocol consisted of spot images of the whole-body dorsal/ventral view in a matrix of 256 × 256 pixels. The whole-body scan was obtained at a table feed rate of 7 cm/min. SPECT images of the abdomen were available in all patients, with acquisitions in 50 views (40 s/view) and iterative reconstruction by OSEM in 8 iterations, with 10 subsets for a final matrix of 128 × 128 pixels. Scatter correction with using triple energy window (TEW) method and CT-based attenuation correction was also used for SPECT image reconstruction. To avoid motion artifacts, light sedation was necessary in all patients. CT scans were also performed with a 5-mm section thickness (130 keV, 30 mAs).

Image analysis
Individual data analyses were performed by two experienced observers (Y.K. and S.B.). 123 I MIBG SPECT/CT images were reviewed on a SYNAPSE workstation (FUJIFILM Medical, Tokyo, Japan). To assess the anatomical site and the exact extent of disease, all patients underwent CT or MRI before or after 123 I MIBG SPECT/CT with a mean interval of 5 days. The VOI of the primary tumor lesion was determined by consensus of two observers (Figs. 1, 2). VOI of muscle was designated individually in each patient using a 1-cm-diameter sphere of the latissimus dorsi muscle (Fig. 3). These VOIs were designated using an IntelliSpace Portal 6.0 workstation (Koninklijke Philips N.V., Amsterdam, The Netherland). The maximum and mean counts of both tumor and muscle VOIs were noted. The mean CT Hounsfield unit value of each tumor was also measured. The tumor-to-muscle ratios using the max counts and mean counts of VOIs (T/ Mmax and T/Mmean) were calculated as semi-quantitative parameters using the following formulas:

Statistical analysis
Data are presented as single values or as means ± standard errors of the mean unless otherwise specified. Semi-quantitative parameters were compared between the early-relapse and delay-relapse group using Mann-Whitney's U test. Relapse-free survival analysis was performed using cut-off values of the semi-quantitative parameters as determined by ROC analysis. If value of two groups was clearly separated by ROC analysis, cut-off value was settled as minimal integer between two groups. Correlation between semiquantitative parameter and each other conditions as MYCN amplification, existence of metastases, VMA and HMA was evaluated using Chi-square test or correlation coefficient. p values < 0.05 were considered to indicate statistical significance.
T∕Mmax = max count of tumor ∕ max count of muscle, T∕Mmean = mean count of tumor ∕ mean count of muscle.

Results
The results of measurements of all 11 patients are given in Table 2. All patients but one survived the entire follow-up period.
T/Mmax values for the early-relapse and delay-relapse groups were 8.86 ± 3.22 and 16.20 ± 1.97, respectively. There was no significant difference in the T/Mmax between the two groups. T/Mmean values for the early-relapse group and delay-relapse group were 2.65 ± 0.58 and 7.66 ± 2.68, respectively. The T/Mmean value of the early-relapse group  was significantly lower than that of the delay-relapse group (p < 0.05) (Fig. 4). Fig. 5. The survivals of the early-relapse and delay-relapse groups were clearly separated using the cut-off value of 4.0. The log-rank test showed a statistically significant difference between the high-and low-T/Mmean groups (p < 0.01).

Relapse-free survival analysis of T/Mmean is shown in
There was a statistically significant relationship between T/Mmean and MYCN amplification using Chi-square test (Table 3; p < 0.01). T/Mmean values of all patients with positive MYCN amplification were lower than 4.0.
There was no significant relationship between T/Mmean and the presence of distant metastases (Table 4). All cases of the early-relapse group appeared to be free of distant metastases at the time of the MIBG examination.

Discussion
Our study introduces the tumor-to-muscle count ratios of T/Mmax and T/Mmean as new semi-quantitative values for 123 I MIBG SPECT/CT; they represent the ratios of the maximum and mean values of MIBG count, respectively. As a result of our analysis, the T/Mmean ratio of the delay-relapse group was found to be significantly higher than that of earlyrelapse group. This result may relate with the expression of the MIBG transporter of the tumor. It is thought that the expression of the norepinephrine transporter (NET) and vesicular monoamine transporters 1 and 2 (VMAT1 and VMAT2) are correlated with MIBG avidity in neuroblastoma [18,19]. Temple et al. [19] reported that the expression of VMAT2 is significantly lower in MYCN-amplified neuroblastomas. All patients of the early-relapsed group had MYCN amplification in this study, which contributed to their low-T/Mmean values.
Tumor necrosis also seemed to be correlated with low T/ Mmean, but we did not investigate the exact volume of the necrotic lesion in the tumor. Though CT Hounsfield units may be related with the degree of necrosis in the tumor, there was no statistical difference between the CT Hounsfield units of the tumor of the early-relapse group and those of delay-relapse group (result not shown).
The VMA and HVA of early-relapse group were lower than those of the delay-relapse group and there was weak positive correlation between T/Mmean and VMA or between T/Mmean and HVA. It is known that both VMA and HVA reflect tumor volume and catecholamine metabolism in NB [20]. Especially at VMA, Strenger et al. [21] showed that VMA was lower in the MIBG-negative group than in the   [22] reported that VMA was lower in MYCN amplification group than nonamplification group. These reports suggest that weak MIBG accumulation results in low VMA and MYCN amplification, which we believe is consistent with the results of our study.
In this study, we calculated T/Mmax or T/Mmean only for the primary tumor lesion and not for the distant metastatic lesions. We did not measure uptake in distant metastatic lesions because we believe that we may not be able to accurately detect all metastases and that the MIBG avidity and distribution may be different for primary tumors and metastatic lesions. Although the results of this study did not include cases with metastases in the early-relapse group, T/Mmean was lower in the early-relapse group than in the delay-relapse group, regardless of the presence or absence of distant metastases. This suggests that T/Mmean may be a prognostic tool for patients with unknown distant metastasis, even if the presence or absence of distant metastases is unknown. The usefulness of T/Mmax and T/Mmean in distant metastatic lesions remains uncertain.
T/Mmax values of the delay-relapse group tended to be higher than those of early-relapse group, but the difference was not significant. Although T/Mmax may also reflect the catecholamine metabolic activity of the tumor, it overestimates the uptake of MIBG when tumors have a heterogeneous MIBG distribution. This may be the reason why no significant difference was found in the results using T/ Mmax. On the other hand, the measurement of T/Mmax has the advantage that the margin of VOI does not have to be carefully considered and is more easily measured. Although not as accurate a predictor as T/Mmean, T/Mmax may have some use in prognostic evaluation.
We decide to use 123 I MIBG count of muscle for reference lesion. While the accumulation of 123 I MIBG in the muscle lesions is low, it is higher than the background count and more homogenous than liver lesions. Determining the VOI of a tumor or muscle is difficult with SPECT images alone because the boundaries of the organs are unknown or unclear. It is also difficult to apply the positional information from CT or MRI performed at different examinations. On the other hand, with SPECT/CT, SPECT and CT images are taken almost simultaneously, and the location information of the tumor and muscles are identical. This makes it possible to determine the VOI relatively accurately with SPECT/CT. This study did not include patients treated with 131 I MIBG therapy and the usefulness of T/Mmean for predicting the effect of 131 I MIBG therapy is unclear. It has been reported that the accumulation of 131 I MIBG is stronger than that of 123 I MIBG [23], 131 I MIBG may show good accumulation and therapeutic effect when the T/Mmean is high, but it needs further study.
There were several limitations in this study. First, this study was retrospective. A prospective study is needed to confirm the effectiveness of 123 I MIBG T/Mmean values at predicting relapse. Second, the small number of patients limited our analysis. This is due to the fact that NB is a very rare tumor. However, our results show a clear difference in T/Mmean between the early-relapse group and the delayrelapse group. Therefore, we can strongly expect T/Mmean to be a good prognostic indicator for treatment prognosis and we think if the 123I MIBG T/Mmean of pre-treatment NB is low, more careful post-treatment follow-up of the disease may be required, whether or not metastatic lesions are present.

Conclusions
In conclusion, we found that a low level of 123 I MIBG uptake using semi-quantitative analysis of SPECT/CT was correlated with early relapse of neuroblastoma. This parameter shows promise as a predictor of treatment response in patients with neuroblastoma.