|Home | About | Journals | Submit | Contact Us | Français|
Katherine A. Zukotynski, Department of Imaging, Dana-Farber Cancer Institute, 44 Binney Street, Boston, MA 02115, Phone: 617-632-3604, Fax: 617-632-4891, kzukotynski/at/partners.org
To assess 18F-labeled 2-fluoro-2-deoxy-D-glucose (18F-FDG) uptake in children with a newly diagnosed diffuse intrinsic brainstem glioma (BSG) and to investigate associations with progression-free survival (PFS), overall survival (OS) and MRI indices.
Two Pediatric Brain Tumor Consortium (PBTC) therapeutic trials in children with newly diagnosed BSG were designed to test radiation therapy combined with molecularly targeted agents (PBTC-007: Phase I/II study of gefitinib; PBTC-014: Phase I/II study of tipifarnib). Baseline brain 18F-FDG PET scans were obtained in 40 children in these trials. Images were evaluated by consensus of two PET experts for intensity and uniformity of tracer uptake. Associations of 18F-FDG uptake intensity and uniformity with both PFS and OS were evaluated as well as associations with tumor MRI indices at baseline (tumor volume on FLAIR, baseline intratumoral enhancement, diffusion and perfusion values.
In the majority of children, BSG 18F-FDG uptake was less than gray matter uptake. Survival was poor irrespective of intensity of 18F-FDG uptake, with no association between intensity of 18F-FDG uptake and PFS or OS. However, hyperintense 18F-FDG uptake in tumor compared to gray matter suggested poorer survival rates. Patients with 18F-FDG uptake in ≥ 50% of the tumor had shorter PFS and OS compared to patients with 18F-FDG uptake in < 50% of tumor. There was some evidence that tumors with higher 18F-FDG uptake were more likely to show enhancement; and when the diffusion ratio was lower the uniformity of 18F- FDG uptake appeared higher.
Children with BSG where 18F-FDG uptake involves at least half the tumor appear to have inferior survival compared to children with uptake in <50% of tumor. A larger independent study is needed to verify this hypothesis. Intense tracer uptake in the tumors compared to gray matter suggests decreased survival. Higher 18F-FDG uptake within the tumor was associated with enhancement on MRI. Increased tumor cellularity as reflected by restricted MR diffusion may be associated with increased 18F-FDG uniformity throughout the tumor.
Brainstem tumors account for 10–20% of intracranial tumors in children. (1, 2) The most common is the diffuse intrinsic brainstem glioma (BSG), usually a fibrillary (World Health Organization (WHO) Grade 2) or malignant (WHO Grade 3 or 4) astrocytoma involving the pons. (3, 4) Standard magnetic resonance imaging (MRI) has high diagnostic sensitivity when a T2 hyperintense tumor expands and diffusely infiltrates the pons (4), but is limited in assessing tumor metabolic activity. (5–7) Biopsies are not routinely performed due to potential morbidity, limiting correlations amongst tumor grade and outcome. Radiation therapy (RT) is the standard treatment, with improved neurological function, but median progression-free survival (PFS) remains <6 months and median overall survival (OS) approximates 10 months. (8) Investigations of molecular signaling agents have yet to improve the dismal prognosis. (9–15)
Positron emission tomography (PET) using 18F-labeled 2-fluoro-2-deoxy-D-glucose (18F-FDG) fused with MRI can demonstrate metabolically active disease (16) and can be helpful in diagnosis and follow-up of children with brainstem gliomas. (17) The intensity of 18F-FDG uptake in adults with brain tumors may reflect malignancy grade and predict survival. (18–21) Preliminary data indicates that 18F-FDG uptake in children may be associated with malignancy grade and add prognostic information in BSG (22–23) and related brain tumors. (24–25)
The aim of this study is to assess the prognostic value of baseline intensity and uniformity of 18F-FDG uptake in a large series of children with newly diagnosed BSG and to investigate associations amongst 18F-FDG uptake, PFS and OS, and baseline MRI indices of tumor volume, enhancement, diffusion and perfusion.
The institutional review boards of PBTC institution approved the studies before patient enrollment; continuing approval was maintained throughout the studies. Patients or legal guardians gave written informed consent; assent was obtained as appropriate. The two clinical trials of children with newly diagnosed BSGs were designed to investigate the efficacy of concurrent radiation therapy with molecular targeting agents. PBTC-007 was a Phase I/II study of gefitinib (ZD1839, Iressa®, AstraZeneca, Wilmington, DE); PBTC-014 was a Phase I/II study of tipifarnib (R115777, Zarnestra®, Johnson and Johnson Pharmaceutical Research and Development, Raritan, NJ). Analyses in this report are restricted to the baseline PET and MRI studies. Images were acquired at participating institutions, electronically transferred to the PBTC Operations and Biostatistics Center and, after de-identification, to the PBTC Neuroimaging Center (NIC) for analysis. (26) Investigators at the NIC were masked to patient outcome at the time of image evaluation.
Standard MRI was acquired at each institution using a 1.5 T scanner with axial FLAIR, axial T2, axial diffusion, axial T2* perfusion and axial T1 postgadolinium imaging. Baseline MRI was performed within two weeks prior to treatment. FLAIR images were obtained with 4-mm contiguous slice thickness using the sequence TR/time of inversion/TE = 10,000/2,200/162 ms. Axial T2-weighted fast spin echo (FSE) were obtained with TR/ETE = (4000–6000)/80–100, ETL = 10–16, RB = ± 16 kHz, FOV = 18–24 cm, slice thickness/gap = 4/0 interleaved, NEX = 2, matrix = 256 × 192, flow compensation option, frequency direction A/P. Diffusion images were single-shot echoplanar spin echo images TR/TE = 2000/80 ms, 128 × 128 matrix, b-factor of 5/1000 s/mm2, 3 directions (x,y,z) for trace imaging, receiver bandwidth of ± 64 kHz, frequency direction was R/L with a slice thickness/gap of 4/0. Perfusion imaging consisted of axial echoplanar imaging, gradient echo mode, single shot, matrix =128 × 128, TR/TE = 1500/45–60 ms), FOV = 18–24 cm, NEX = 1, slice thickness 4/0, frequency direction R/L, 45 to 60 phases, 10 phases prior to bolus injection of gadolinium diethylenetriamine penta-acetic acid (DTPA) 0.1 mmol/kg. Post-gadolinium axial T1-weighted spin-echo images were 4-mm contiguous slices of the whole head using repetition time (TR)/echo time (TE) = (500–700 ms)/minimum full.
18F-FDG was available across all PBTC institutions. Since the PET scans were acquired in a multi-center consortium on a variety of scanners (GE Advance NXI, GE Discovery LS, GE Discovery STE, Philips G-PET, Siemens HR+ and Siemens HiRez Bioscan) with specified spatial resolutions of 4.0–5.0 mm, considering the filtering as a consequence of the reconstruction algorithms, the spatial resolution was in the 7–10 mm range.
Consistency of PET data was maintained by adherence to a standard quality assurance program with daily blank scans and quarterly normalization, calibration, and preventive maintenance. Two phantoms were imaged at each site to ensure consistent quantitation as previously described. (25, 27) Baseline 18F-FDG PET of the brain was acquired on all subjects within 2 weeks prior to therapy. Patients fasted for 4 hours prior to the PET. The baseline brain PET was acquired in 3D mode for 10 minutes, 40–60 minutes following the intravenous administration of 5.55 MBq/kg of FDG (minimum dose 18 MBq and maximum dose 370 MBq). Attenuation correction was performed using either a 3 minute segmented transmission scan with 68Ge/68Ga rods or a CT based approach, depending on whether the scanner was a PET or PET/CT scanner, respectively. The acquired data were reconstructed using Fourier rebinning followed by a 2D ordered subset expectation maximum reconstruction algorithm.
Fused PET/MR images were obtained using a HERMES™ workstation (Hermes Medical Solutions, Stockholm Sweden) and a mutual information approach with the PET data re-sampled along the planes of the MRI. This method provides excellent results with median errors on fused images of less than 2 mm. (28) For each case, the quality of the image registration was assessed subjectively based on alignment of the cortical surface and gray matter.
Anatomic tumor extent was evaluated by a pediatric neuroradiologist using axial T1 post-contrast, FLAIR and T2 weighted MR images. Diffusion image/ROI analyses were performed using ImageJ (US National Institutes of Health, Bethesda, Maryland, USA). From the apparent diffusion coefficient map (ADC), a region of interest (3–5 mm in diameter) within the solid part of the tumor was determined using the T1, FLAIR, T2 and post-gadolinium T1 sequences as reference. In turn, the mean ADC of the region of interest (ROI) was divided by the mean ADC value from a ROI in the normal frontal white matter to obtain the normalized ADC value within tumor. The perfusion images were transferred to a Sun UltraSPARC II™ workstation (Sun Microsystems, Santa Clara, California); relative cerebral blood volume (rCBV) maps were generated from the dynamic susceptibility-weighted perfusion MRI data. ROIs (3–5 mm in diameter) were placed in the highest regions of perfusion in the tumor from the generated relative cerebral blood volume maps; the mean rCBV of the tumor ROI was divided by the mean rCBV of a region of interest obtained from the frontal white matter to get a normalized tumor rCBV value.
All PET and fused PET/MR images were evaluated by a pediatric neuroradiologist and nuclear medicine physicist for intensity and uniformity of tracer uptake in the tumor. Intensity was graded on a 5-point scale (grade 1: no uptake, grade 2: uptake similar to normal white matter, grade 3: between normal white and gray matter, grade 4: similar to normal gray matter and grade 5: greater than normal gray matter). Uniformity was defined as the percentage of the tumor (as delineated on the FLAIR MRI) demonstrating FDG uptake and was graded on a 4-point scale (grade 0: <25%, grade 1: 25–50%, grade 2: 51–75%, and grade 3: >75%). Uniformity was determined by visual estimation from the corresponding multiplanar MR and PET images through the tumor. For cases with negligible 18F-FDG uptake in tumor, uniformity was considered to be Grade-0. Figure 1 illustrates the grading scheme for intensity and uniformity of tracer uptake. Two-dimensional (2D) image analysis was also performed. The axial image through the tumor containing the maximum activity per pixel (highest FDG uptake) was identified, and a 2D ROI was manually drawn based on the FDG definition of the tumor. Due to the wide range of uptake in these tumors, a set threshold could not be used; ROIs were defined subjectively. Because the PET data analyzed in this study were from a multicenter trial, they were transmitted by the different centers to the NIC in a variety of formats, with pixel values represented as either raw counts, activity concentration (in Bq/ml), or SUV. To standardize the ROI values, they were normalized by values obtained in a comparison region as previously reported. (25) A slice at the level of the thalamus and basal ganglia was chosen, and a ROI in the temporoparietal region was used for normal gray matter values. The mean and maximum pixel values within the tumor ROI were normalized by those for normal gray matter and normal white matter to provide ratios of tumor/gray matter and tumor/white matter.
Cox Proportional Hazard (PH) Models were utilized to investigate possible associations of PET variables with PFS and OS distributions. Subjects were followed for up to 3 years from the time of enrollment. PFS was measured from treatment start date to the earliest of date of progression or death; similarly, OS was measured from treatment start date until date of death. Patients who did not experience an event for PFS or OS were censored at their last follow-up date. Since Cox PH models are known to produce spurious results if fewer than 10 events per covariate are available, associations with PFS and OS were only explored when at least nine events were available for a given neuroimaging variable for the univariable Cox models.
Log-rank and chi-square tests were used when statistically appropriate. An ordinal logistic regression model was used to evaluate the association of subjective BSG 18F-FDG uptake intensity and uniformity grading with 2D continuous PET variables. Linear-by-linear association tests were used to explore the associations of baseline 18F-FDG uptake and uniformity with baseline tumor volume on FLAIR/T2, baseline diffusion values, and baseline tumor perfusion values. The Exact Cochran-Armitage trend test was used to examine 18F-FDG uptake versus volume of tumor enhancement.
Since reported p-values are not adjusted for multiplicity, the usual 0.05 level cannot be used to determine ‘statistical significance’; therefore, each p-value reported must be considered in light of the multiplicity adjusted significance level of 0.00076.
A total of 106 children enrolled in PBTC 007 Phase I/II and PBTC 014 Phase I/II. Of these, 40 children (38%), aged 3.4 to 18.7 years, 26 girls and 14 boys, had both baseline brain 18F-FDG PET and MRI. Twenty-three children were enrolled in PBTC-007 phase I/II (phase I: 9 & phase II: 14), and seventeen, in PBTC-014 phase I/II (phase I: 6 & phase II: 11).
The 12-month PFS rates (± one standard error (SE)) were 15.0±5.2% and 12.1±3.8%, respectively, for patients with and without baseline PET scans; 12-month OS rates (± one SE) were 49.0±7.8% and 42.5±6.0%,, respectively for patients with and without baseline PET scans. There was no evidence that the subset of children with baseline PET was biased with respect to PFS distribution (p=0.52) or OS distribution (p=0.52) (Figure 2).
Table 1 shows the 1-year PFS and OS estimates corresponding to the intensity of 18F-FDG uptake on the baseline PET. In 33 of the 40 children evaluated (83%), FDG uptake was less than normal gray matter uptake. The current data show no association between intensity of FDG uptake, PFS (p=0.36) or OS (p=0.48) (Figure 3).
Table 2 shows the 1-year PFS and OS rates (±SE) corresponding to the uniformity of BSG 18F-FDG uptake. 18F-FDG uptake was seen in more than half the tumor in 17 of the 40 children (43%). When more than 50% of the tumor was 18F-FDG-avid, PFS and OS appeared to be decreased (Figure 4). One-year PFS (±SE) estimates were 21.7±7.9% and 5.9±4.0%, respectively, for patients with <50% 18F-FDG-avid tumor versus those with ≥ 50% (exact Log-Rank test p=0.031); one-year OS (±SE) estimates were 54.9±10.2% and 41.2±11.2%, respectively (exact Log-Rank test p=0.086). There was some evidence of a linear positive association between intensity of 18F-FDG uptake and uniformity of 18F-FDG uptake (Table 3), with a Linear-Linear Association test p=0.0003, suggesting that as 18F-FDG uptake increases compared to normal tissue, the proportion of 18F-FDG-avid tumor also increases.
Continuous PET variables were not found to be associated with PFS. Although not significant based on the multiplicity adjusted significance level, the available data showed some evidence in this exploratory analysis that all continuous PET suggest associations with OS (Table 4). Patients with higher 18F-FDG uptake values than comparative tissues (gray matter, white matter, or entire brain) seemed to have earlier death on average.
There is some evidence of an association between 2D continuous PET variables and intensity of 18F-FDG uptake. Specifically, for higher values of continuous PET variables the odds of being in a higher 18F-FDG uptake intensity grade were increased (Figure 5A). In particular, a higher ratio of tumor 18F-FDG uptake to gray matter uptake was associated with a higher 18F-FDG uptake intensity grade (p ≤ 0.001). There was some evidence of association between 2D continuous PET variables and uniformity of 18F-FDG uptake, although this was not as strong (Figure 5B). For example, a higher ratio of tumor 18F-FDG uptake to gray matter uptake was associated with a higher BSG 18F-FDG uptake uniformity grade (p = 0.03). There was no trend of association between intensity of 18F-FDG uptake on the baseline PET with tumor size on MRI, perfusion ratio or diffusion ratio. There was no trend of association between uniformity of 18F-FDG uptake on the baseline PET with tumor size on MRI or perfusion ratio. However, there was some evidence that uniformity of 18F-FDG uptake was associated with tumor diffusion ratio (Figure 6). Specifically, the diffusion ratio was lower on average for higher levels of uniformity (p=0.025). There was also some evidence that for higher 18F-FDG uptake, enhancement is more likely to be observed in the tumor (Table 5; exact Cochran-Armitage trend test p = 0.032). When only those tumors that had enhancement at baseline were considered, there was no association with 18F-FDG uptake (Linear-to Linear association test p-value = 0.40)
Clinical trials of children with newly diagnosed BSGs have been designed in the PBTC to investigate combinations of radiation therapy with investigational chemotherapeutic agents. In particular, two trials (PBTC-007 and PBTC-014) were designed to study the efficacy of concurrent radiation therapy with the molecularly targeted agents gefitinib and tipifarnib, respectively. Gefitinib is a selective inhibitor of epidermal growth factor receptor (EGFR), a protein that may be overexpressed in neoplastic disease leading to the activation of the Ras signal transduction cascade and uncontrolled cell proliferation. Tipifarnib is a farnesyltransferase inhibitor that may also affect the Ras signal transduction cascade and therefore cell proliferation.
MRI is often diagnostic in BSG, anatomically defining tumor extent, and serially used to assess response and status during and after therapy. However, precise tumor metabolic evaluation on standard MRI is limited. Several papers in adults have suggested that functional imaging with 18F-FDG PET complements anatomic MR imaging in the evaluation of brain tumors by identifying metabolically active disease. DiChiro suggested that the intensity of 18F-FDG uptake was associated with the malignant tumor grade (18) and that intensely 18F-FDG-avid disease reflected high-grade disease and decreased patient survival. (19) Results have varied regarding the association between 18F-FDG uptake intensity and prognosis. DeWitte et al. concluded that in adults with a low-grade glioma, increased tracer uptake suggested a poor survival. (29) This group later showed that in adults with a high-grade glioma, 18F-FDG uptake was not an independent predictor for prognosis. (30) The results of adult studies may not be applicable to the pediatric population. Ultimately, PET appears to provide a helpful non-invasive tool for the evaluation of brain tumor metabolic activity. (20, 31–33)
Recently, studies have suggested that PET may be useful for the evaluation of children with brainstem gliomas. Kwon et al. reported 18F-FDG PET uptake in 12 children with the suggestion that hypermetabolic tumors were more likely to reflect a glioblastoma compared to little or no 18F-FDG uptake in anaplastic astrocytomas or low-grade astrocytomas. (22) Pirotte et al. reported 20 children with newly diagnosed BSG, all of whom had PET guided stereotactic biopsy, indicating that PET guidance improved the diagnostic yield of stereotactic biopsy sampling and that PET data might carry prognostic value. All tumors with high FDG uptake were malignant and associated with a shorter survival time than tumors with absent or moderate FDG uptake.(23) Williams et al. suggested that 3D maximum and mean tumor 18F-FDG uptake were associated with progression-free survival in pediatric supratentorial anaplastic astrocytomas when using 3D PET analysis techniques. (25)
In our evaluation, there was no evidence of association between the intensity of 18F-FDG uptake and PFS or OS, when evaluated objectively (Figure 3). Less than 20% of children survived progression-free 12 months into the trial, and less than 20% were alive at 24 months (Figure 2). The overall survival was slightly lower for children with intense FDG uptake (Figure 3). The techniques used showed evidence of an association between 2D continuous PET variables and subjective measures of intensity or uniformity of FDG uptake.
There was the suggestion that in tumors with metabolic activity in >50% of the tumor volume, there was apparently inferior PFS and OS (Figure 4). In addition, both higher maximum tumor to gray matter ratios and higher mean tumor to gray matter ratios appeared to be associated with decreased survival (Table 4). This may mean that more intense tracer uptake in tumor compared to normal gray matter suggests decreased survival. A larger study is needed to prospectively verify this hypothesis.
Comparing the imaging modalities of MRI and PET, there was evidence that uniformity of BSG FDG uptake was associated with tumor diffusion ratio values (Figure 6). Specifically, when the diffusion value was lower, the uniformity of 18F-FDG uptake was higher. This suggests that increased tumor cellularity likely represents more tumor viable cells and higher 18F-FDG uptake throughout the tumor. Indeed, data reported by Palumbo et al. on 15 adults with metastatic brain lesions suggested that hypercellular tumors may have increased impedance to water diffusion resulting in low ADC and high 18F-FDG uptake. (34) Holodny et al. reported 21 adults with pathologically proven glial tumors of the brain and found that ADC maps appear to provide unique information that may be analogous to 18F-FDG PET with increased 18F- FDG uptake corresponding to lower ADC values. (35)
Only a minority of diffuse intrinsic brain stem gliomas had intense 18F-FDG uptake. Without a biopsy, the histologic milieu in children with a diffuse intrinsic brainstem glioma is unknown. However, these tumors at baseline often have increased diffusion likely reflecting a combination of tumor cellularity and vasogenic edema. (36) In this study, the association between uniformity of 18F-FDG uptake and diffusion suggests that those tumors with lower 18F-FDG uptake have lower tumor cellularity at baseline.
Tumor size and perfusion values were not associated with baseline 18F-FDG uptake or uniformity. However, there was a suggestion that with higher 18F-FDG uptake, tumor enhancement is more likely (Table 5), which may reflect more aggressive disease since the majority of brainstem gliomas do not enhance. (37)
The principal limitation of this study is the sample size. Only 40 subjects within the two protocol studies received baseline 18F-FDG PET, making it difficult to draw significant inferences from this evaluation; the reported cohort is, however, larger than other papers published in the literature. (22, 23) All children in this study had poor survival, meaning small differences in survival reflected by changes in intensity or uniformity of 18F-FDG uptake may have been difficult to appreciate statistically. Without biopsy, it is unknown whether these tumors represent a molecularly and pathologically heterogeneous group of tumors which could also affect results. Future studies correlating PET with post-mortem tissue sampling may be helpful although may be limited by possible changes in the tumor over time during treatment. Evaluation of a larger patient population is needed to establish statistical significance of the parameters we have studied. Subsequent studies should also evaluate the relative difference between baseline and follow-up PET scans in the clinical evaluation of children with these brain tumors. Future studies may also investigate the use of PET radiopharmaceuticals besides 18F-FDG, including 11C labeled methionine, 18F F-DOPA, 18F labeled choline and 18F-fluorothymidine (FLT), which may theoretically provide better sensitivity for cellular proliferation.
The majority of children with BSG demonstrate 18F-FDG uptake by tumor that is less than uptake by normal gray matter. Survival is poor irrespective of the intensity of BSG 18F-FDG uptake; there is no association between intensity of 18F-FDG uptake, PFS or OS. Our data suggest that intense tracer uptake may be seen in association with more uniform tracer uptake spread throughout the tumor and that intense tracer uptake in tumor compared to the normal gray matter uptake suggests decreased survival. When 18F-FDG uptake is seen in at least half the tumor, PFS and OS are subjectively shorter relative to children with 18F-FDG uptake in only a small portion of tumor.
Comparing MRI with PET, increased uniformity of 18F-FDG avidity may be associated with increased tumor cellularity. Further evaluation with more patients is needed to establish statistically significant associations that can provide prognostic information for clinical management.
We acknowledge the other site PET physicians including Dr. Randall Hawkins (University of California, San Francisco), Dr. James Mountz (Children’s Hospital of Pittsburgh), Dr. Satoshi Minoshima (Seattle Children’s Hospital), Dr. David Earl-Graef (Children’s National), Dr. Stewart Spies (Children’s Memorial), and Dr. John Butman (National Institutes of Health). We acknowledge Cynthia Dubé for manuscript preparation.
Research Support: This work was supported in part by NIH grant U01 CA81457 for the Pediatric Brain Tumor Consortium (PBTC), The Pediatric Brain Tumor Consortium Foundation (PBTCF), The Pediatric Brain Tumor Foundation of the United States (PBTFUS) and American Lebanese Syrian Associated Charities.