Positron emission tomography (PET) scans of primary brain tumors were performed in pediatric patients to examine whether metabolic characteristics could be used as an index of clinical aggressiveness.
Positron emission tomography (PET) scans of primary brain tumors were performed in pediatric patients to examine whether metabolic characteristics could be used as an index of clinical aggressiveness.
Twenty-seven pediatric patients with untreated primary central nervous system neoplasms were studied with PET scans using 2-[18F] fluoro-2-deoxy-D-glucose (FDG) and/or L-[methyl-11C] methionine (MET). Metabolic characteristics as assessed with FDG and MET standardized uptake values (SUV) and SUV-to-normal brain ratios were compared with histopathology and selected histochemical features such as proliferation activity (Ki-67MIB-1) and apoptotic, vascular, and cell density indices. The median followup time was 43 months.
The accumulation of both FDG and MET was significantly higher in high-grade than in low-grade tumors, but a considerable overlap was found. The accumulation of both tracers was associated positively with age. High-grade tumors showed higher proliferative activity and vascularity than the low-grade tumors. In univariate analysis, FDG-PET, MET-PET, and apoptotic index were independent predictors of event-free survival.
We found that both FDG and MET uptake in pediatric brain tumors are associated with malignancy grade. However, no clear limits of SUVs and SUV-to-normal brain ratios can be set between low-grade and high-grade tumors, which makes the assessment of malignancy grade using metabolic imaging with PET scan difficult in individual cases. Although FDG-PET and MET-PET do not compensate for histopathologic evaluation, they may give valuable additional information especially if invasive procedures to obtain histopathologic samples are not feasible. Cancer 2002;95:1376–86. © 2002 American Cancer Society.
Brain tumors are the most common solid neoplasms in children. They constitute a heterogenous group of tumors in terms of histologic typing, prognosis, and response to therapy. The majority of childhood tumors occur along the central neural axis and infratentorially.1 The most common infratentorial childhood tumors are cerebellar astrocytomas, medulloblastomas, ependymomas, and brain stem gliomas, whereas benign and malignant astrocytomas and ependymomas manifest most commonly in a supratentorial location.1 Even though the majority of childhood brain tumors are low-grade malignant tumors that respond well to therapy, both diagnosis and treatment are complicated by tumor location near critical brain structures, which limits the possibility for diagnostic biopsy.
In adults, positron emission tomography (PET) scan is widely used for grading and characterizing brain tumors2 and for evaluating the treatment response.3 2-[18F] fluoro-2-deoxy-D-glucose (FDG) and L-[methyl-11C] methionine (MET) are the two most frequently used tracers for investigating glucose utilization and amino acid transport of brain tumors, respectively. Specifically, FDG-PET is used to grade malignancy,4, 5 predict survival,6, 7 and to assess tumor recurrence and response to treatment.2, 8, 9 Although MET-PET may also be helpful in tumor grading,10 it is even more suitable for delineation of tumor extent.11–13 Several authors have suggested that concurrent use of FDG-PET and MET-PET may be more informative for diagnostic and prognostic purposes in adults than either method alone.9, 14–17
Little is known about the metabolism of brain tumors in children. Hoffman et al.18 analyzed visually the accumulation of FDG in 17 children with various brain tumors located in the posterior fossa. The degree of FDG accumulation did not predict tumor histology, although a tendency to a higher accumulation of FDG characterized more aggressive tumor types such as medulloblastoma.18 In another study of 15 pediatric patients with primary, residual, recurrent, and metastatic brain tumors, glucose metabolism in the tumors was assessed with FDG-PET.19 The authors found that FDG accumulation was higher in children with medulloblastoma than with other types of primitive neuroectodermal tumors or infratentorial gliomas. Studies on MET metabolism in pediatric brain tumors are even more scarce. In a study of 13 pediatric patients with histologically heterogenous treated and untreated brain tumors, O'Tuama et al.20 found that methionine delineated well tumor extent. These authors suggested that the intensity of MET uptake would correlate with tumor grade.
To the best of our knowledge, there are no previous studies in which a systematic and thorough semiquantitative analysis of FDG and MET accumulation in brain tumors was performed in pediatric patients. Furthermore, because the prevalent location, histopathology, and treatment preferences of brain tumors in children differ from that of adults, the findings in PET scan studies obtained in adults cannot be applied directly to children. Therefore, the current study was designed to address whether tumor metabolic characteristics as depicted by FDG-PET and MET-PET could be used as an index of clinical aggressiveness in pediatric patients with primary brain tumors. We compared the sensitivity of FDG and MET for disclosing tumor malignancy grade and prognosis by comparing PET scan findings with histopathologic and immunohistochemical analyses and followup data.
We studied 27 pediatric patients with newly diagnosed brain tumors. The mean age of the children was 9.0 ± 1.0 years (range, 0.5–16.0 years). There were 19 low-grade and 8 high-grade tumors. Radiologic evaluation was performed presurgically with contrast-enhanced magnetic resonance imaging (MRI) scans in all patients. Twenty-six of the patients underwent surgical procedures. Diagnoses were verified histologically either in surgical (n = 26) or autopsy specimens (n = 1). The tumors were graded according to the World Health Organization (WHO) classification.21 Seventeen tumors were located supratentorially and 10 were located infratentorially. All PET scan studies were performed before any oncologic treatments were administered. Nine patients were on antiepileptic treatment, which was either oxcarbazepin only (n = 5), combined oxcarbazepin and vigabatrin (n = 2), combined oxcarbazepin, vigabatrin, and natriumvalproate (n = 1), or phenytoin only (n = 1). Ten patients received corticosteroids for treatment of tumor-induced brain edema and 11 patients underwentventriculostomy before the PET scan studies to alleviate intracranial pressure. Clinical and histopathologic data of the patients are shown in Table 1. The median followup for survivors was 43 months (range, 34–112 months).
|ID no.||Age at study (yrs)||Gender||Diagnosis||Grade||Tumor Location||Type of surgery||CRT (Gy)||Chemotherapy||Follow-up (mos)||Clinical status|
|1||5.4||m||Astrocytoma||IV||Left temporal lobe||B||50||+||13||DOD|
|11||15.0||m||Pleomorphic xanthoastrocytoma||II||Right temporal lobe||TR||—||−||46||NED|
|12||13.1||m||Dysembryoplastic neuroepithelial tumor||I||Left temporal lobe||TR||—||−||40||NED|
|13||11.2||m||Ganglionglioma||I||Left temporal lobe||TR||—||−||112||NED|
|14||13.8||f||Ganglionglioma||I||Left temporal lobe||SR||—||−||68||NED|
|15||9.8||f||Ganglionglioma||I||Left temporal lobe||TR||—||−||35||NED|
|16||2.0||f||Pilocytic astrocytoma||I||Chiasma - III ventricle||SR||50||+||62||AWD|
|17||4.5||m||Pilocytic astrocytoma||I||Chiasma - n. opticus||B||—||−||44||NED|
|18||4.7||f||Pilocytic astrocytoma||I||Chiasma - n. opticus||SR||—||−||47||NED|
|19||0.5||f||Pilocytic astrocytoma||I||Chiasma - hypothalamus||SR||45||−||34||NED|
|20||3.4||m||Pilocytic astrocytoma||I||Chiasma - hypothalamus||B||45||−||43||AWD|
|21||13.6||f||Pilocytic astrocytoma||I||Left temporal lobe||TR||—||−||53||NED|
|22||4.3||m||Pilocytic astrocytoma||I||Left temporal lobe||TR||—||−||72||NED|
|23||6.9||m||Pilocytic astrocytoma||I||Right temporal lobe||TR||—||−||38||NED|
|24||3.7||f||Pilocytic astrocytoma||I||Cerebellum, right||TR||—||−||36||NED|
|27||2.2||m||Astrocytoma seu gliosis||I||Cerebellum, left||TR||—||−||59||NED|
Of 27 patients, 23 and 21 patients underwent MET-PET and FDG-PET imaging, respectively. Nineteen patients underwent both MET-PET and FDG-PET studies. The studies were performed before the start of definitive treatment, which was never delayed because of enrollment in the study. FDG and MET were used as tracers in random order in the two PET scan studies that were performed under postabsorptive conditions. Plasma glucose level was always determined before the start of the imaging scans, the average value being 5.2 ± 0.2 mmol/L. Of 44 PET scan studies, 8 were carried out in children who were sedated with either diazepam or midazolam or who received general anesthesia using either tiopental or propofol. The sedatives were used in five patients younger than 5 years of age because of poor cooperation. All sedatives were administered before the start of the transmission scan, i.e., 12–15 minutes before the tracer was injected. Only oral diazepam (n = 3) and intravenous midazolam (n = 1) were used in FDG studies.
The study was reviewed and approved by the Joint Committee on Ethics of the Turku University and the Turku University Central Hospital. The nature, purpose, and potential risks of the study were explained to all subjects and their legal guardians before they gave their informed consent to participate.
The PET scan studies were performed by using either an eight-ring ECAT 931/08 tomograph (Siemens/CTI, Knoxville, TN; n = 33) or an 18-ring GE Advance whole-body tomograph (General Electric Medical Systems, Milwaukee, WI; n = 11). The ECAT scanner has an axial resolution of 6.7 mm and a spatial resolution of 6.5 mm.22 The GE Advance has an axial resolution of 4.3 mm and a spatial resolution of 4.3 mm.23
FDG synthesis was a modification of the method reported by Hamacher et al.24 The radiochemical purity of FDG always exceeded 99%. MET was synthesized as described by Långström et al.25 and Någren and Halldin.26 The radiochemical purity of MET was measured as described by Någren and exceeded 96%.27
Before the emission scan was performed, a 10-minute transmission scan for correction of photon attenuation was performed with a removable ring source containing 68Ge. Following a bolus injection of FDG (median dose, 2.1 MBq/kg), an emission scan was acquired for 55 minutes (4 × 30, 3 × 60, and 0 × 300-second frames; n = 14). In seven patients, a 20-minute late accumulation scan (4 × 300-second frames) was performed after 30 minutes of FDG injection. An intravenous bolus injection of 5.5 MBq/kg MET was followed by either a dynamic acquisition for 40 minutes (4 × 30, 3 × 60, 5 × 180, and 4 × 300-second frames; n = 14) or a 20-minute steady-state scan beginning 20 minutes after MET injection (4 × 300-second frames; n = 9). The entire brain could be scanned with a 10.8-cm (ECAT) or 15-cm (GE Advance) axial field of view of the two tomographs. A light foam rubber head holder was used for fixation of the head.
All data were corrected for decay and the measured photon attenuation and reconstructed into a 256 × 256 (ECAT) or a 128 × 128 (GE Advance) matrix. The final in-plane resolution in reconstructed and Hann-filtered images was 9.5 mm for the ECAT scanner and 7.4 mm for the GE Advance scanner.
Visual analysis was performed by two independent observers (M.U., L.M.). Anatomic localization of the tumor was confirmed by visually comparing the PET scan images with the corresponding MRI scans. The tumor was graded as hypermetabolic when the accumulation of tracer in the tumor was higher than that of the surrounding brain tissue, eumetabolic when there was no distinguishable accumulation of tracer within the tumor area or when there was disagreement between the two observers, and hypometabolic when there was less radioactivity in the tumor compared with the surrounding brain tissue.
To obtain comparable data, a static scan was generated for all patients 30 minutes after FDG injection (a scan summed more than four 300-second–frames, corresponding time 30–50 minutes) and this scan for used for all analyses. In the MET-PET studies, the time frame of 20–40 minutes was used in all cases. The regions of interest (ROIs) were drawn on the entire tumor area and on the seemingly normal cortical and subcortical regions that served as reference areas. The size of the whole tumor ROI was determined by visual comparison with the corresponding contrast-enhanced T1 MRI scans. Reference ROIs were drawn on two or three planes over the frontal cortex and white matter (FDG-PET) or over the whole brain area (MET-PET). Whether reference ROIs were drawn on two or three planes was determined by individual anatomy and head circumference. A contralateral mirror ROI was placed in a unilateral tumor without demonstrable mass effect on the contralateral side. In addition to the ROIs encompassing the whole tumor, semiautomatic software was applied for defining the highest tracer accumulation within the whole tumor ROI (ROImax).28 The ROImax size was adjusted to the resolution of the two scanners. Tracer accumulation in the ROIs was analyzed as the standardized uptake value (SUV), which is the activity concentration in an ROI at a fixed time point divided by the injected dose normalized to the patient's measured weight.29 In addition to SUVs in tumors, tumor-to-frontal cortex SUV ratios were established for both tracers. A tumor-to-white matter SUV ratio was established for FDG only and a tumor-to-whole brain area SUV ratio was established for MET only.
Ki-67 labeling was performed by using monoclonal anti–Ki-67 (MIB-1, IgG1 subclass, Immunotech S.A., Marceille, France) and an avidin-biotin immunoperoxidase technique as described previously,30, 31 with slight modifications. Briefly, the 5-μm–thick paraffin-embedded sections of tumors were incubated in citrate buffer during heating in a microwave oven. The bound antibody was visualized by using diaminobenzidine (DAB) as the chromogen.
The Ki-67MIB-1 indices were quantified visually by point-counting on tumor areas expressing subjectively the highest number of immunopositive nuclei. The assessment proceeded in adjacent microscopic fields of a ×40 objective lens along horizontal and vertical axes perpendicular to each other until 1000 cells were counted (0.3 ± 0.0 mm2). Only neoplastic cells were included in the quantification of Ki-67–positive cells. Necrotic and hemorrhagic areas and the section borders were also omitted from quantification. The results are expressed as the percentage of Ki-67–positive cells per 1000 tumor cells.
Monoclonal rabbit anti-human von Willebrand factor (vWF; 1:1000; Dakopatts A/S, Glostrup, Denmark) was used as the marker for endothelial cells.32 The bound primary antibodies were visualized using avidin-biotin-peroxidase with DAB as the chromogen.
Tumor vascularization was quantified by counting the number of immunohistochemically identifiable capillaries per square millimeter in the adjacent microscopic fields using a ×10 objective lens along the horizontal and vertical axes perpendicular to each other in up to 30 fields of view (0.29 ± 0.0 mm2).
In situ detection of apoptotic cells in paraffin-embedded 5-μm–thick tissue sections was performed as described earlier.33 Briefly, the deparaffinized tissue sections were digested with proteinase K (10 μg/mL). Terminal deoxynucleotidyl transferase end-labeling was performed using digoxigenin-labeled dideoxy-UTP. Phosphatase-conjugated anti-digoxigenin antibodies bound to dideoxynucleotides were visualized by 5-bromo-4-chloro-3-inodyl-phosphate as the substrate and nitroblue tetrazolium as the chromogen.
A distinct color reaction within tumor cells represented apoptotic DNA fragmentation. The quantification was started on tumor areas expressing subjectively the highest number of immunopositively stained cells. The assessment proceeded in adjacent microscopic fields using a ×25 objective lens along the horizontal and vertical axes perpendicular to each other in up to 30 fields of view (3.2 ± 0.3 mm2). The results are expressed as the number of positive cells per square millimeter of tissue section area.
The assessment of cell density was performed in the sections immunostained to visualize Ki-67MIB-1. Neoplastic and nonneoplastic cells were quantified usinf a ×40 objective lens along the horizontal and vertical axes perpendicular to each other until 1000 cells were counted (0.3 ± 0.0 mm2). Endothelial cells were not analyzed. The results are expressed as the number of cell per square millimeter.
Spearman's correlation coefficients were calculated for the SUVs in tumors and for tumor-to-reference area SUV ratios in relation to each other and to immunohistologic parameters. For group comparisons between malignancy grades, patients were divided into low-grade (WHO Grade I–II) and high-grade tumor (WHO Grade III–IV) groups. Group comparisons were performed with the Mann–Whitney test because the distributions were skewed. A univariate Cox model was used for survival analysis. Multiregression analyses were not performed due to the small number of patients experiencing some event during the followup period. All statistical analyses were performed using the SPSS software (SPSS, Chicago, IL). The results are expressed as mean ± the standard error of the mean. Two-sided P values less than 0.05 were statistically significant.
The median followup time was 43 months (range, 34–112 months). All patients completed the followup. All patients with low-grade tumors are still alive (100%, n = 19) whereas two of eight patients (25%) with high-grade tumors are still alive after followup times of 24 and 50 months, respectively (Table 1). Currently, Patient 6 is the only survivor in the high-grade group (patient had anaplastic germinoma) who is disease free. In contrast, only two patients (11%) in the low-grade group have experienced disease recurrence.
In the visual analysis of MET-PET studies, 22 of 23 tumors (96%) were graded as hypermetabolic. In FDG-PET studies, 11 of 21 tumors (52%) were graded as hypermetabolic, 8 (38%) as eumetabolic, and 2 (10%) as hypometabolic. FDG accumulation was higher in five of six high-grade tumors than in the surrounding brain tissue. In addition, four of five optic tract and two cerebellar pilocytic astrocytomas showed high FDG accumulation. There was no disagreement between the interpretations of the two observers in any case.
Tumor-to-frontal cortex (both tracers), tumor-to-white matter (FDG), and tumor-to-whole brain area (MET) SUV ratios were significantly higher in the high-grade than in the low-grade tumors (P < 0.05). In contrast, only the FDG (P < 0.05) SUV ratio in tumors was associated positively with malignancy grade (Table 2).
|Mean (± SEM)||Maximum (± SEM)|
|Low-grade tumors||High-grade tumors||Low-grade tumors||High-grade tumors|
|MET tumor||2.4 (0.3)||3.7 (0.7)||2.9 (0.4)||4.5 (0.9)|
|MET tumor/frontal cortex||1.5 (0.1)||2.3 (0.3)a||1.5 (0.1)||2.2 (0.2)a|
|MET tumor/whole brain||1.7 (0.1)||2.6 (0.3)a||1.3 (0.1)||1.8 (0.2)a|
|FDG tumor||3.8 (0.4)||7.8 (1.6)a||4.5 (0.5)||9.6 (2.0)a|
|FDG tumor/frontal cortex||0.6 (0.1)||1.3 (0.3)a||0.6 (0.1)||1.3 (0.3)a|
|FDG tumor/white matter||1.1 (0.1)||2.4 (0.5)a||1.0 (0.1)||2.2 (0.5)a|
The SUVs for both tracers correlated significantly with age in all normal reference areas (Fig. 1) and even in tumors (r = 0.51–0.59, P < 0.02–0.005 for FDG and r = 0.48, P < 0.03 for MET). In contrast, none of the FDG or MET tumor-to-normal reference area SUV ratios correlated with age in any area.
Mean Ki-67MIB-1 indices depicting proliferative activity of cells were 2.9 ± 0.6% for low-grade and 23.6 ± 7.7% for high-grade tumors (P = 0.002). The respective apoptotic indices were 0.8 ± 0.2 and 151.9 ± 47.2 apoptoses/mm2 (P < 0.0001). vWF indices were 37.5 ± 9.2 and 121.8 ± 58.4 capillaries/mm2 (ns) and cell density indices were 4298 ± 426 and 7395 ± 2215 cells/mm2 (ns) in low-grade and high-grade tumors, respectively.
All FDG and MET SUVs in tumors and tumor-to-reference area SUV ratios were positively associated (data not shown). The highest correlation was observed between whole tumor ROI for FDG and MET (r = 0.76, P = 0.001). Apoptotic index correlated significantly with both tumor-to-frontal cortex (r = 0.44, P = 0.038) and tumor-to-whole brain MET SUV ratios (r = 0.48, P = 0.021). In contrast, no significant associations were observed between Ki-67MIB-1 index or vWF index and any of the SUVs or SUV ratios.
FDG tumor-to-frontal cortex SUV ratio (P = 0.037) and apoptotic (P = 0.047) and vWF (P = 0.027) indices were also higher in patients who experienced recurrent or progressive disease during the followup period than in those who had an event-free followup (Table 3). In Cox univariate regression analysis, FDG (R = 0.29, P = 0.012) and MET (R = 0.28, P = 0.021) tumor-to-frontal cortex SUV ratios and apoptotic index (R = 0.39, P = 0.002) were associated with disease progression (Table 4).
|Patients with event-free survival Mean ± SEM (25th and 75th quartiles)||Patients with recurrent disease Mean ± SEM (25th and 75th quartiles)||P|
|FDG-PET 0.6 ± 0.1||1.1 ± 0.3||< 0.05|
|(0.3, 0.9)||(0.6, 1.4)|
|1.6 ± 0.2||2.2 ± 0.3||ns|
|(1.2, 1.9)||(1.5, 2.9)|
|Ki-67MIB index (%)|
|5.2 ± 2.2||16.7 ± 7.1||ns|
|(1.9, 4.1)||(0.8, 28.3)|
|Apoptotic index (apoptoses/mm2)|
|7.1 ± 6.2||118.8 ± 48.1||< 0.05|
|(0.0, 1.7)||(0.5, 224.5)|
|von Willebrand factor index (capillaries/mm2)|
|34.3 ± 8.9||145.8 ± 65.2||< 0.05|
|(22.0, 44.0)||(45.5, 111.5)|
|Cell density index (cells/mm2)|
|4434.0 ± 441.6||6780.3 ± 2034.9||ns|
|(3090.1, 4905.1)||(3991.9, 5249.0)|
|Ki-67MIB index (%)||0.17||0.052|
|Apoptotic index (apoptoses/mm2)||0.39||0.002|
|von Willebrand factor index (capillaries/mm2)||0.25||0.055|
|Cell density index (cells/mm2)||0.13||0.086|
The patients who died during the followup period (n = 6) had significantly higher FDG (P = 0.019) and MET (P = 0.046) tumor-to-frontal cortex SUV ratios, as well as higher Ki-67MIB-1 (P = 0.049) and apoptotic (P = 0.003) indices compared with the patients who survived the followup period.
This study was undertaken to determine whether tumor metabolism as examined using PET scan could be used as an index of clinical aggressiveness in pediatric patients with primary brain tumors. We imaged 27 pediatric patients with untreated brain tumors using both FDG and MET and compared the findings using PET scan with cell density, proliferative activity, apoptotic rate, and vascularization of the tumors. The accumulation of FDG and MET was significantly higher in high-grade tumors than in low-grade tumors in accordance with previous studies on adult patients. Of special interest was the finding that accumulation of both tracers in tumors and in normal brain tissue was associated positively with age. PET scan findings also correlated significantly with the apoptotic index, but not with the proliferation index, cell density, or vascularization. Metabolic activity determined by PET scan correlated significantly with event-free survival, indicating that noninvasive assessment of clinical behavior of pediatric brain tumors may be feasible with PET scan.
In the visual analysis, five of six (83%) high-grade tumors presented as hypermetabolic in FDG-PET. This is in agreement with two previous studies showing that medulloblastomas in children have a high FDG accumulation.18, 19 Our study also confirms that high-grade tumors other than medulloblastomas may show high FDG accumulation in pediatric patients (Fig. 2). These data are also consistent with previous data in adults demonstrating that visual analysis of FDG accumulation provides a reasonable estimate of tumor grade.34, 35
We also found that 22 of 23 (96%) patients examined using MET-PET exhibited a higher MET accumulation in tumor than in the normal adjacent brain tissue. However, MET accumulation did not differentiate between low-grade and high-grade tumors. Therefore, in children, visual analysis of MET accumulation appears to have high sensitivity for brain tumor detection (Fig. 2), whereas MET-PET seems to have less value for assessing the malignant potential of the tumor. This is in keeping with previous data in adults.10, 16, 36 The delineation of tumor extent with MET-PET rather than with FDG-PET may, however, be advantageous for radiotherapy planning.37
One patient with glioblastoma (Patient 1) exhibited marked FDG hypometabolism in the tumor area. Large areas of necrosis within the tumor subsequently found at autopsy may explain the minimal accumulation of FDG in this case, even though the necrotizing effect of chemotherapy and radiotherapy cannot be assessed separately. In addition to high-grade tumors, four of five optic tract and two of four cerebellar pilocytic astrocytomas showed higher FDG accumulation in the tumor than in the surrounding white matter although tumor FDG uptake was lower than that in cortex. The discrepancy between high tracer accumulation and low-grade histology might reflect the high vascularization and/or high expression of glucose transporters38 in pilocytic astrocytomas. The only patient who did not show any visible accumulation of MET (Patient 27) had a small cerebellar pilocytic astrocytoma of 1 cm in diameter. Although this might indicate low metabolic activity, MET accumulation may also have been underestimated because of the partial volume effect.
Until now, no previous study has assessed quantitatively FDG and MET accumulation in parallel in pediatric patients with brain tumors. In addition, the current study is the first in which all patients were examined at the time of primary diagnosis, before any surgical or oncologic procedures. In fact, previous studies18–20 included patients with primary, residual, and recurrent tumors and PET scan imaging was performed either before, during, or after oncologic treatment. This approach may not be ideal for studies on tumor grading because both radiation and chemotherapy decrease cerebral glucose utilization in children,39, 40 adults,41 and animals.42, 43
There is no consensus on the optimal method to quantify tracer accumulation in childhood brain tumors. Therefore, we adopted methods of calculating tumor SUVs and tumor-to-reference area SUV ratios using both mean and maximum SUVs as reported in adult studies.16, 36 A study by Hustinx et al.44 questioned the value of SUVs in brain tumor characterization. They did not find a correlation between SUVs and the metabolic rate of regional cerebral glucose uptake in a group of 27 patients with treated and untreated histologically heterogenous brain tumors. Whether this was due to heterogeneity of the study population or to the effect of treatment (preoperatively vs. postoperatively) remains unclear. As suggested in the study, we used SUV ratios instead of absolute SUVs because of the high, unexplained intersubject variability of SUVs. In addition, both peak and average values of ROIs drawn in two to three planes were used to minimize the effect of tumor heterogeneity and statistical fluctuation. We used the frontal, temporal, and parietal cortices, the cerebellum, and either white matter (FDG-PET studies) or whole brain area (MET-PET studies) as the reference areas. Mean and maximum SUVs were equally useful for evaluating tumor grade (Table 2). As most tumors in the current study were localized either along the neural axis, in the temporal lobe, or in the cerebellum, tumor-to-contralateral normal brain, tumor-to-temporal cortex, and tumor-to-cerebellum SUV ratios could not be used in the majority of patients. Therefore, these data were not included in the final analyses. Since PET and MRI image correlation was not available in the current study, we chose to use the frontal cortex as the cortical reference region. Without an overlaying MRI scan, delineation of the frontal cortex is easier than delineation of the parietal cortex due to the higher rate of the regional metabolic rate of glucose in the frontal cortex.45
Previous studies have demonstrated that dexamethasone,46, 47 antiepileptic agents,48–50 and anesthetic drugs51 may reduce cerebral glucose metabolism especially in the grey matter. By calculating tumor-to-white matter SUV ratios for FDG accumulation, we excluded the effects of these confounding factors. Because MET is distributed more evenly to cortical and subcortical areas than FDG, tumor-to-whole brain SUV ratio was used instead of tumor-to-white matter SUV ratio for MET-PET studies.
Cerebral glucose utilization may be age dependent although this has never been demonstrated in the healthy developing brain.52 More specifically, glucose uptake may increase to a peak value at the age of 9 and decrease thereafter to adult levels.52 In our series, the reference areas and tumor SUVs correlated significantly with age. These findings might suggest that the age dependency of glucose utilization in childhood brain tumors parallels that of the normal brain. The association between malignancy grade and the age of the patients did not solely explain this finding. While the four youngest patients were studied under anesthesia, the potential effect of anesthetic drugs might have influenced these results. As regards to quantitative analysis, the age dependency of glucose utilization can be circumvented by using tumor-to-reference area SUV ratios, which was done in the current study.
In accordance with the FDG-PET studies, we found that MET accumulation was age dependent both in normal brain tissue and in tumors. In the cortex, MET accumulation (i.e., nonnormalized SUVs) seems to increase almost linearly from early childhood to adolescence. These findings disagree with the results reported in one previous study, which found a sevenfold decrease in frontal cortex MET uptake in children between the ages of 1.8 and 71 years (n = 17).53 In the latter study, MET accumulation was measured as the percent dose accumulated in cubic centimeters in the cortical region in images obtained about 1.5 minutes after tracer injection, i.e., during the rapid uptake phase. In the current study, MET accumulation was measured 20–40 minutes after tracer injection when the steady state of MET uptake is reached. The discrepant results may be explained by the different acquisition protocols.
The histochemical studies showed that Ki-67MIB and apoptotic indices were significantly higher in high-grade tumors than in low-grade tumors. In contrast to previous studies, however, all immunohistochemical indices showed considerable overlap between low-grade and high-grade tumors. The histologic heterogeneity of tumors in our study may contribute to the discrepant findings in histochemical indices because previous studies have been done within one histologic tumor entity. In adults, cell density may be the main determinant of glucose metabolism in astrocytomas.54 Therefore, the histologic heterogeneity of tumors might also explain the overlap in FDG-PET SUVs and SUV ratios because the cell densities among different tumor types in childhood (e.g., astrocytoma, oligodendroglioma, ependymoma, and medulloblastoma) differ.21
In the Cox univariate regression analysis, FDG-PET and MET-PET and apoptotic index showed prognostic significance for event-free survival. Our findings need to be confirmed in a more homogenous study population where tumor histology and treatment are less variable.
In conclusion, the utilization of FDG and MET in pediatric brain tumors is associated positively with tumor malignancy grade. Many of the problems related to the effects of age and anesthetic and/or antiepileptic drugs on tracer accumulation can be circumvented or at least decreased by using tumor-to-frontal cortex and/or tumor-to-white matter (FDG) or tumor-to-whole brain (MET) SUV ratios for quantitative analyses. However, because of the considerable overlap of individual SUVs and SUV-to-normal brain ratios between the low-grade and high-grade tumors, determination of malignancy grade with PET scan should be made with caution. FDG-PET may, however, provide information on the malignancy grade of the tumor if it is not feasible to obtain a histologic sample. Conversely, delineation of tumor extent by MET-PET may provide additional information for local management of pediatric brain tumors, whether management includes surgery or radiotherapy.