Bevacizumab (BEV), a humanized anti–vascular endothelial growth factor antibody, is currently applied to the treatment of glioblastoma, with biological effects such as normalization of the tumor vasculature with restoration of disrupted blood-brain barrier (BBB).1 Restoration of a disrupted BBB can reduce permeability from tumor vessels, thereby making MRI with contrast media for assessment of response to BEV difficult. The Response Assessment in Neuro-Oncology criteria were developed to address such difficulties in assessment using conventional MRI.2,3 However, the assessment of biological effects within tumor during chemotherapy with BEV has been the subject of much debate.4–6
Assessment of biological response from BEV in glioblastoma has been attempted using PET with a variety of tracers.7–11 PET has been found to allow earlier and more accurate assessment of therapeutic response compared with conventional MRI.7–9,11–13 On the other hand, vigorous attempts have been made to apply MR perfusion imaging to the assessment of therapeutic response after BEV therapy.14,15 Dynamic susceptibility contrast-enhanced MR perfusion imaging and dynamic contrast-enhanced MR perfusion imaging are promising techniques for assessing therapeutic response after BEV therapy.14,15 However, these methods might be unsuitable for assessing response from BEV therapy, because they substantially depend on the permeability of the BBB, given that contrast medium is required in each procedure.15 In contrast, arterial spin labeling (ASL) perfusion imaging is theoretically unaffected by the condition of the BBB due to the use of magnetically labeled arterial blood water as a freely diffusible endogenous tracer instead of contrast medium.16,17 With this major advantage of not needing contrast media, ASL has been extensively utilized to measure blood flow within gliomas for various objectives, including malignancy grading17,18 and differential diagnosis.19,20 Furthermore, this modality may be suitable for assessing biological effects within glioma from BEV treatment. To clarify whether ASL can assess biological responses to BEV therapy as reliably as PET with amino acid tracers, the present study compared findings on ASL to those on PET with 11C-methyl-L-methionine (11C-met-PET) as a classically standard amino acid tracer, in patients with recurrent glioblastoma treated with BEV therapy.
PATIENTS AND METHODS
Patients and Treatments
This prospective study complied with the precepts established by the Declaration of Helsinki and was approved by the ethics committee at our institute (No. H22–96). Inclusion criteria for patients are as follows: aged 20 years or older with recurrent glioblastoma with wild-type isocitrate dehydrogenase based on the pathological diagnosis at initial treatment; Karnofsky performance scale of 70% or greater; more than 12 months since completion of radiotherapy in the initial treatment; recurrent glioblastoma localized in cerebral white matter; and voluntary provision of written informed consent to participate. Twenty-four patients (17 men, 7 women; median age, 54.8 years; range, 22–76 years) were enrolled in this study.
After initial treatment comprising surgical tumor resection, radiotherapy, and chemotherapy with temozolomide (TMZ), all patients had been receiving oral TMZ as maintenance therapy according to Stupp's regimen.21 At enrollment into this study, glioblastomas presented first relapse in 18 patients and second relapse in the remaining 6 patients. All patients were treated with IV BEV (10 mg/kg every 2 weeks from day 1) in addition to a dose-dense TMZ regimen (oral administration, 100 mg/m2 per day for 7 days every 2 weeks from day 2). The day of starting BEV was defined as day 1. Each biweekly BEV followed by TMZ for 7 days was defined as 1 course. This therapy (BEV/TMZ) was continued until progressive disease (PD) was identified, as determined according to the current Response Assessment in Neuro-Oncology criteria,3 which classifies therapeutic response into complete response, partial response (PR), stable disease (SD), or PD. No patients were treated with corticosteroids during BEV/TMZ therapy. Outcomes for each patient were evaluated using progression-free survival (PFS), defined as the period between day 1 and the day on which PD was first identified. Treatment decisions and management were conducted by 3 investigators (T.B., Y.S., K.O.).
Procedures for ASL and PET
We performed ASL and 11C-met-PET in 3 phases: within 7 days before starting BEV/TMZ (baseline); at around 4 weeks (immediately before the third course of BEV/TMZ); and at around 8 weeks (immediately before the fifth course).
The 11C-met-PET scan was performed on a different day from MRI. All 11C-met was synthesized using the solid-phase 11C-methylation method according to a report by Pascali et al.22 The radiochemical purity of the produced 11C-met was greater than 99%. All procedures for the preparation of 11C-met were performed by an investigator (K.T.). At 30 minutes after IV injection of 11C-met with a dose of 325 to 398 MBq (mean, 6.8 MBq/kg body weight), PET was performed using a PET/CT system (SET3000 GCT/M; Shimadzu, Kyoto, Japan). PET scans were reconstructed under the following conditions: field of view, 256 mm2; matrix, 128 × 128; pixel size, 2.0 × 2.0 mm2; and slice sickness, 2.6 mm. Estimations for 11C-met-PET scans complied with a previous study.11 Briefly, average SUV was determined automatically for each region of interest (ROI) of 6 mm in diameter with ROIs placed on the area of highest accumulation of 11C-met within the tumor, and on 3 regions in apparently normal cerebral white matter in the contralateral hemisphere. The ratio of SUV in tumor to normal tissue (mean value of 3 regions) was calculated for each met-PET scan as SUVT/N. We observed trends in changes in SUVT/N over the course of the 8 weeks. The rate of change was calculated using the following formula: (SUVT/N from later PET − SUVT/N from earlier PET)/SUVT/N from the earlier PET (%). To verify the reliabilities of T/N ratios, we conducted volumetric analysis for each scan. We extracted pixels showing SUVT/N ≥ 1.6 and counted the total number of extracted pixels on all slices in each scan. This threshold was based on a previous report.11 Tumor volumes on PET were calculated by multiplying the number of pixels and pixel size using an in-house program. Estimations on PET images were performed by one investigator (T.S.), who was blinded to the results of conventional MRI and ASL.
Arterial spin labeling was performed using a 3.0-T MRI system (Discovery MR750; GE Healthcare Japan, Tokyo, Japan). Radiological technicians supervised by one investigator (M.S.) performed ASL scans with the following sequences: 3-dimensional fast spin echo, pseudo-continuous ASL; repetition time, 4347; echo time, 10.5 milliseconds; field of view, 240 × 240 mm2; matrix, 128 × 128; pixel size, 1.875 × 1.875 mm2; slice sickness, 3 or 4 mm. In the first performance of ASL as the baseline, an appropriate inversion time was selected from 4 different times (1025, 1525, 2025, and 2525 milliseconds) set in the MRI machine for each patient. Technicians performed scanning under each inversion time in ascending order (from 1025 to 2525 milliseconds) and determined the inversion time providing the greatest contrast in signal intensity between tumor and apparently normal brain on the contralateral side as the appropriate inversion time, while watching on a monitor. Appropriate inversion time was also used at scanning ASL at 4 weeks and 8 weeks for each patient. On color images converted from original images, ROIs of 6 mm in diameter were manually placed on the region showing the highest blood flow within the tumor and 3 regions in apparently normal white matter in the hemisphere contralateral to the side of the tumor. Regions of interest were transferred to the cerebral blood flow (CBF) map, and average relative CBF (rCBF) was automatically determined using software included in the clinical recording system. The normalized ratio was then calculated for each patient, as the average rCBF on tumor divided by the mean value of 3 average rCBFs on apparently normal brain (rCBFT/N). We observed trends in changes also in rCBFT/N over the course of the 8 weeks. Trends for the rate of change in rCBFT/N during the 8 weeks were estimated using the same procedure as PET: (rCBFT/N from later ASL − rCBFT/N from earlier ASL)/rCBFT/N from the earlier ASL (%). Volumetric analysis was conducted also for each ASL scan. In this analysis, pixels showing rCBFT/N ≥ 1.3 were extracted, and the total number of extracted pixels was counted. This threshold was based on a previous report of differentiation between recurrent high-grade gliomas and non–high-grade gliomas using rCBFT/N on ASL.23 Tumor volumes were calculated by multiplying the number of pixels and pixel size using an in-house program. Calculations of tumor volume were conducted by one investigator (F.Y.), who was blinded to PET findings. After ASL scanning, we performed conventional MRI and assessed therapeutic responses at 4 weeks and at 8 weeks in comparison with baseline MRI.
Values of T/N ratio, rates of change, and tumor volumes were compared between ASL and PET at each time point or each phase using Mann-Whitney U test. We examined correlations in T/N ratio, rates of change, and tumor volumes between ASL and PET at each time point or at each phase (early phase from baseline to 4 weeks, late phase from 4 weeks to 8 weeks, and entire course from baseline to 8 weeks), using Spearman correlation coefficient by rank test. Differences in rates of change for the time course between SUVT/N and rCBFT/N were analyzed by repeated-measures 2-factor analysis of variance. Progression-free survival was compared between patients recruited at first and second relapse using log-rank testing. All patients were divided into 2 groups: those with PFS equal to or above the median value for all patients, and those with PFS below the median. Accuracies of predicting patients with longer PFS were assessed for 11C-met-PET and ASL, using receiver operating characteristic (ROC) analysis. Receiver operating characteristic analysis was conducted for SUVT/N and rCBFT/N at each phase, and also for fluctuations in SUVT/N and rCBFT/N at each phase. Furthermore, patients were divided into 2 groups using cutoff values showing maximum sensitivity as well as maximum specificity for predicting patients with long PFS. We compared PFS between 2 groups using log-rank testing. All data were analyzed using PASW Statistics version 18 software (SPSS Japan, Tokyo, Japan). Values of P < 0.05 were considered significant in all analyses.
All patients received at least 4 courses of therapy over the 8 weeks, and completed the full schedules of conventional MRI, 11C-met-PET, and ASL. Highest accumulation of 11C-met and highest blood flow with ASL were seen at nearly the same regions on color images at each time point in all patients (Fig. 1). Values of SUVT/N and rCBFT/N were 2.37 ± 0.35 and 2.06 ± 0.30 at baseline, 1.87 ± 0.50 and 1.51 ± 0.44 at 4 weeks, and 2.07 ± 0.66 and 1.63 ± 0.66 at 8 weeks, respectively. Values of SUVT/N were significantly larger than those of rCBFT/N at all time points. Significant correlations were found between SUVT/N and rCBFT/N at all time points (Figs. 2A–C). Rates of change for SUVT/N and rCBFT/N were −21.8% ± 13.9% and −27.0% ± 16.7% in the early phase, +12.1% ± 19.0% and +7.6% ± 25.3% in the late phase, and −12.9% ± 23.3% and −16.6% ± 28.0% over the 8 weeks, respectively. No significant differences were found in rates of change between SUVT/N and rCBFT/N at all phases. Rates of change for SUVT/N and rCBFT/N showed significant correlations in all phases (Figs. 2D–F). Tumor volumes extracted from PET and ASL were 6.1 ± 5.5 cm3 and 3.1 ± 3.3 cm3 at baseline, 2.9 ± 3.8 cm3 and 1.8 ± 3.0 cm3 at 4 weeks, and 7.5 ± 9.6 and 5.9 ± 9.2 cm3 at 8 weeks, respectively. Tumor volumes from PET were significantly larger than those from ASL at baseline (P = 0.03), whereas no significant differences were identified at 4 weeks and 8 weeks. At all time points, significant correlations were found between tumor volumes from PET and ASL (Figs. 2D–F). Changes in SUVT/N and rCBFT/N during the 8 weeks after initiating BEV showed similar trends, with transient reductions at the early phase, followed by rapid rebounds at the late phase (Fig. 3). Analysis using repeated-measures 2-factor analysis of variance showed that SUVT/N and rCBFT/N fluctuated independently in parallel throughout 8 weeks (P = 0.26).
Conventional MRI showed complete response in 1 patient, PR in 9 patients, and SD in 14 patients at 4 weeks, and PR in 2 patients, SD in 19 patients, and PD in 3 patients at 8 weeks. Tumors in all patients deteriorated after 8 weeks, with a median PFS of 127 days (range, 51–335 days). Progression-free survival showed no significant difference between patients registered to this study at first relapse (median PFS, 127 days) and second relapse (median PFS, 103; P = 0.47). In ROC analyses, the cutoff value leading to the optimal balance of sensitivity and specificity for predicting patients with long PFS (≥127 days) in SUVT/N was 2.45 at baseline, 1.67 at 4 weeks, and 1.87 at 8 weeks (Figs. 4A–C). The areas under the curve (AUCs) for SUVT/N at baseline, 4 weeks, and 8 weeks were 0.44, 0.66, and 0.73, respectively. In ASL, the cutoff leading to the optimal balance of sensitivity and specificity was 2.10 at baseline, 1.41 at 4 weeks, and 1.89 at 8 weeks (Figs. 4D–F). The AUCs for rCBFT/N at baseline, 4 weeks, and 8 weeks were 0.41, 0.65, and 0.66, respectively. In predictions using T/N ratio, 11C-met-PET at 8 weeks showed the best accuracy, with 76.9% sensitivity and 81.8% specificity. In comparisons of PFS between groups of patients assigned by cutoffs determined from ROC analysis, only patients with SUVT/N less than 1.87 at 8 weeks showed a significantly longer PFS (Fig. 5). In ROC analysis for the rate of change in SUVT/N, the cutoff value leading to the best match of sensitivity and specificity was −16.3% in the early phase, +17.2% in the late phase, and −13.2% over 8 weeks. The AUCs in the early phase, late phase, and entire phase were 0.80, 0.66, and 0.76, respectively (Figs. 6A–C). For ASL, the best cutoff was −19.2% in the early phase, +10.3% in the late phase, and −12.7% in the entire phase (Figs. 6D–F). The AUC was 0.71 in the early phase, 0.68 in the late phase, and 0.62 in the entire phase. Between the 2 groups of patients divided by these cutoff values, all comparisons showed significant differences in PFS (Fig. 7). Given the results of AUCs, however, rate of change in SUVT/N in the early phase offered the most accurate predictor of patients with long PFS.
In the present study, the areas showing the highest blood flow on ASL maps resembled those regions showing the highest accumulation on 11C-met-PET. Furthermore, significant correlations were identified in T/N ratios, fluctuations in ratios, and tumor volumes between 11C-met-PET and ASL at all time points and all phases. These findings might be explained by assuming that high cell turnover consisting of amino acid metabolism facilitates neovascularization within glioma.24–26 However, the types of information provided from 11C-met-PET and ASL are inherently different. Uptake of amino acid tracers for PET does not only depend on cell proliferation but also expression of amino acid transporters including L-type amino acid transporter-1/2.27 In addition, ASL has drawbacks, including low signal-to-noise ratio and motion artifacts. Previous reports have suggested that PET with amino acid tracers cannot be replaced by MR perfusion imaging, because accumulation of amino acid tracer on PET is not reflected by blood volumes measured with MR perfusion imaging.28,29 In those reports, T/N ratios were larger on PET than on dynamic susceptibility contrast-enhanced MR perfusion imaging, and tumor volumes were also larger on PET than on perfusion imaging, when a cutoff of 1.6 was used to extract pixels for both modalities. Likewise in this study, values of SUVT/N were significantly larger than those of rCBFT/N at all time points, and tumor volumes on PET were larger than those on ASL at baseline. Results of no differences in tumor volumes at 4 weeks and 8 weeks between PET and ASL might result from different thresholds (1.6 for PET and 1.3 for ASL) for extraction of pixels and/or the influences of BEV treatment.
A trend for change, a transition from reductions in the early phase to rebound in the late phase during BEV therapy, was demonstrated in some previous reports using PET with amino acid tracers.7,11,30 Uptake of 11C-met is reported to be associated not only with cell density but also with microvessel density within glioma.31,32 A possible reason for the rapid reduction in the early phase may be not a cytocidal effect from BEV, but rather a reduction in permeability caused by BEV.7,11 One study documented that BEV had already caused reductions in vessel density at 24 hours after initiating BEV, leading to reduced rCBF on ASL.33 A rapid reduction in 11C-met uptake at 4 weeks might be attributable to reductions in intratumoral vessel density, because changes in SUVT/N and rCBFT/N fluctuated in parallel over the 8 weeks. On the other hand, a possible reason for the upward rebound in rates of change during the later phase might have been increased microvessel density. Another study documented that blood flow as measured by ASL and vessel density dropped at 14 days, then rebounded at 77 days and 45 to 84 days after administering sorafenib, an antiangiogenic inhibitor.34 Increases of SUVT/N in the late phase might have been caused by increased microvessel density, which was suggested by increase of rCBFT/N. Histological investigations of microvascular structures during the 8 weeks should be undertaken in the future.
As shown previously, ASL may allow demonstration of changes in microvasculature within recurrent glioblastoma under BEV therapy. Assessments of fluctuations in rCBFT/N for all phases showed credible prediction of patient outcomes. Nevertheless, ROC analysis suggested fluctuations of SUVT/N in the early phase as the most credible. With regard to predicting PFS from specific time points, only SUVT/N at 8 weeks offered significant prediction of PFS. Arterial spin labeling offered an influential examination for biological effects from BEV only in cases of assessing changes in measured values. Taken together, 11C-met-PET had an advantage over ASL for determining biological response from BEV in the present study.
Some limitations regarding the study results must be considered. First, the sample size in this study was small, because few patients met all the inclusion criteria for the study. Although 11C-met-PET seemed to be better in prediction of PFS than ASL in this study, further study of a large sample size is needed. In addition, this study recruited patients at both first and second relapse. However, this limitation might be permissible because no significant difference in PFS was evident between patients at first and second relapse. Second, we did not measure absolute values of ASL and 11C-met-PET in ROIs rigorously placed at identical regions using procedures such as superimposition, because values for each image were measured by investigators blinded to data from other modalities. In addition, we were interested in whether ASL alone could predict outcomes without using information provided by PET.
The rCBFT/N from ASL is probably associated with SUVT/N from 11C-met-PET in recurrent glioblastoma under BEV therapy, because several values of both modalities showed significant correlations. Changes in rCBFT/N fluctuated in parallel with SUVT/N over the course of 8 weeks, and could offer a predictor of PFS. However, 11C-met-PET may have advantages over ASL for biological response from BEV in patients with recurrent glioblastoma.
1. Jain RK. Normalizing tumor vasculature with anti-angiogenic therapy: a new paradigm for combination therapy. Nat Med
2. Wen PY, Macdonald DR, Reardon DA, et al. Updated response assessment criteria for high-grade gliomas: response assessment in neuro-oncology working group. J Clin Oncol
3. Ellingson BM, Wen PY, Cloughesy TF. Modified Criteria for Radiographic Response Assessment in Glioblastoma
Clinical Trials. Neurotherapeutics
4. Ellingson BM, Bendszus M, Boxerman J, et al. Consensus recommendations for a standardized Brain Tumor Imaging Protocol in clinical trials. Neuro Oncol
5. Galldiks N, Langen KJ, Pope WB. From the clinician's point of view—what is the status quo of positron emission tomography in patients with brain tumors? Neuro Oncol
6. Hutterer M, Hattingen E, Palm C, et al. Current standards and new concepts in MRI and PET response assessment of antiangiogenic therapies in high-grade glioma patients. Neuro Oncol
7. Schwarzenberg J, Czernin J, Cloughesy TF, et al. 3′-deoxy-3′-18
F-fluorothymidine PET and MRI for early survival predictions in patients with recurrent malignant glioma treated with bevacizumab
. J Nucl Med
8. Galldiks N, Rapp M, Stoffels G, et al. Response assessment of bevacizumab
in patients with recurrent malignant glioma using [18
F]Fluoroethyl-L-tyrosine PET in comparison to MRI. Eur J Nucl Med Mol Imaging
9. Schwarzenberg J, Czernin J, Cloughesy TF, et al. Treatment response evaluation using 18
F-FDOPA PET in patients with recurrent malignant glioma on bevacizumab
therapy. Clin Cancer Res
10. Colavolpe C, Chinot O, Metellus P, et al. FDG-PET predicts survival in recurrent high-grade gliomas treated with bevacizumab
and irinotecan. Neuro Oncol
11. Beppu T, Terasaki K, Sasaki T, et al. MRI and 11
C-methyl-L-methionine PET differentiate bevacizumab
true responders after initiating therapy for recurrent glioblastoma
. Clin Nucl Med
12. Hutterer M, Nowosielski M, Putzer D, et al. O-(2-18F-fluoroethyl)-L-tyrosine PET predicts failure of antiangiogenic treatment in patients with recurrent high-grade glioma. J Nucl Med
13. Wardak M, Schiepers C, Dahlbom M, et al. Discriminant analysis of 18
F-fluorothymidine kinetic parameters to predict survival in patients with recurrent high-grade glioma. Clin Cancer Res
14. Schmainda KM, Prah M, Connelly J, et al. Dynamic-susceptibility contrast agent MRI measures of relative cerebral blood volume predict response to bevacizumab
in recurrent high-grade glioma. Neuro Oncol
15. Essig M, Shiroishi MS, Nguyen TB, et al. Perfusion MRI: the five most frequently asked technical questions. AJR Am J Roentgenol
16. Golay X, Hendrikse J, Lim TC. Perfusion imaging using arterial spin labeling. Top Magn Reson Imaging
17. Wolf RL, Wang J, Wang S, et al. Grading of CNS neoplasms using continuous arterial spin labeled perfusion MR imaging at 3 Tesla. J Magn Reson Imaging
18. Yeom KW, Mitchell LA, Lober RM, et al. Arterial spin-labeled perfusion of pediatric brain tumors. AJNR Am J Neuroradiol
19. Lai G, Mahadevan A, Hackney D, et al. Diagnostic accuracy of PET, SPECT, and arterial spin-labeling in differentiating tumor recurrence from necrosis in cerebral metastasis after stereotactic radiosurgery. AJNR Am J Neuroradiol
20. Ye J, Bhagat SK, Li H, et al. Differentiation between recurrent gliomas and radiation necrosis using arterial spin labeling perfusion imaging
. Exp Ther Med
21. Stupp R, Mason WP, van den Bent MJ, et al. Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma
. N Engl J Med
22. Pascali CBA, Iwata R, Decise D, et al. High efficiency preparation of [11
C]methionine by on-column [11
C]methylation on C18 Sep-Pak. J Labelled Comp Radiopharm
23. Ozsunar Y, Mullins ME, Kwong K, et al. Glioma recurrence versus radiation necrosis? A pilot comparison of arterial spin-labeled, dynamic susceptibility contrast enhanced MRI, and FDG-PET imaging. Acad Radiol
24. Fischer I, Gagner JP, Law M, et al. Angiogenesis in gliomas: biology and molecular pathophysiology. Brain Pathol
25. Ningning D, Haopeng P, Xuefei D, et al. Perfusion imaging of brain gliomas using arterial spin labeling: correlation with histopathological vascular density in MRI-guided biopsies. Neuroradiology
26. Noguchi T, Yoshiura T, Hiwatashi A, et al. Perfusion imaging of brain tumors using arterial spin-labeling: correlation with histopathologic vascular density. AJNR Am J Neuroradiol
27. Langen KJ, Galldiks N, Hattingen E, et al. Advances in neuro-oncology imaging. Nat Rev Neurol
28. Cicone F, Filss CP, Minniti G, et al. Volumetric assessment of recurrent or progressive gliomas: comparison between F-DOPA PET and perfusion-weighted MRI. Eur J Nucl Med Mol Imaging
29. Filss CP, Galldiks N, Stoffels G, et al. Comparison of 18
F-FET PET and perfusion-weighted MR imaging: a PET/MR imaging hybrid study in patients with brain tumors. J Nucl Med
30. Harris RJ, Cloughesy TF, Pope WB, et al. 18
F-FDOPA and 18
F-FLT positron emission tomography parametric response maps predict response in recurrent malignant gliomas treated with bevacizumab
. Neuro Oncol
31. Di Ieva A, Grizzi F, Tschabitscher M, et al. Correlation of microvascular fractal dimension with positron emission tomography [(11)C]-methionine uptake in glioblastoma
multiforme: preliminary findings. Microvasc Res
32. Okita Y, Kinoshita M, Goto T, et al. (11)C-methionine uptake correlates with tumor cell density rather than with microvessel density in glioma: a stereotactic image-histology comparison. Neuroimage
33. Rajendran R, Huang W, Tang AM, et al. Early detection of antiangiogenic treatment responses in a mouse xenograft tumor model using quantitative perfusion MRI. Cancer Med
34. Schor-Bardach R, Alsop DC, Pedrosa I, et al. Does arterial spin-labeling MR imaging-measured tumor perfusion correlate with renal cell cancer response to antiangiogenic therapy in a mouse model? Radiology
Keywords:Copyright © 2019 Wolters Kluwer Health, Inc. All rights reserved.
11C-methyl-L-methionine PET; arterial spin labeling perfusion imaging; glioblastoma; bevacizumab