Advances in Brief 2-[F]Fluoro-2-deoxyglucose and Glucose Uptake in Malignant Gliomas before and after Radiotherapy: Correlation with Outcome

Purpose: To examine whether quantitative 1-[C]glucoseor 2-[F]fluoro-2-deoxyglucose (FDG)-positron emission tomography performed before and/or after radiotherapy (RT) of malignant gliomas correlates with treatment outcome. Changes in metabolism between the start and finish of RT, and immediate post-RT studies have received little attention. Experimental Design: Adults with malignant gliomas were imaged within 2 weeks before and/or 2 weeks after RT. Four patients were imaged only before RT, 12 only after RT, and 14 both before and after RT. Each 1-[C]glucose and FDG study included arterial plasma sampling. Kinetic parameters, glucose metabolic rate (MRGlc), and FDG metabolic rate (MRFDG) were estimated by an optimization program based on a three compartment, four rate constant model. Changes in MRGlc or MRFDG from pre-RT to post-RT were calculated for the 14 patients studied at both times. Overall survival was examined, and survival was computed relative to historical controls in matched prognostic classes. Results: Low pre-RT MRGlc (P < 0.02) or MRFDG (P < 0.03), or an increase from preto post-RT in MRGlc (P < 0.004) or MRFDG (P < 0.006) are correlating with longer survival (4 patients still alive). Strikingly, the post-RT studies (n 26) showed no correlation between MRGlc or MRFDG and survival (P 0.73 and P 0.46 respectively). Conclusions: Low MRGlc or MRFDG before RT probably indicates less aggressive disease. An increase in MRGlc or MRFDG from preto post-RT in the tumors of patients with longer survival could be because of one or more of the following or other reasons: (a) apoptosis of tumor cells in response to RT requires energy; (b) decreased tumor cell density by the RT leaving normal cells with higher metabolism; or (c) inflammatory cells infiltrate and take up glucose or FDG where tumor cells are dying. Quantitative 1-[C]glucose or FDG uptake in the early weeks post-RT correlates poorly with survival. Introduction PET imaging with FDG is being used clinically in the management of malignant gliomas for grading, planning biopsies, and distinguishing recurrent disease from radionecrosis (1–12). In fact, the Therapeutics and Technology Assessment Subcommittee of the American Academy of Neurology has stated that, “cerebral glucose metabolic studies are extremely useful in the management of brain neoplasms (13).” However, from investigations undertaken to date it is not clear whether quantitative FDG-PET has a role in determining the response of malignant gliomas to therapy or how accurately FDG-PET studies after, rather than preceding, RT predict long-term treatment outcome (6, 14–19). In previous reports we presented measurements of metabolic rates in gliomas and normal brain determined with PET imaging of FDG as well as 1-[C]glucose (20, 21). Our approach involved dynamic PET imaging of the kinetic behavior of the two tracers. We used compartmental mathematical models designed for the two tracers individually to calculate the MRGlc and the MRFDG for glioma and normal brain ROIs. From these, we determined the regional lumped constant for FDG as the ratio MRFDG/MRGlc. The two tracers, FDG and 1-[C]glucose, behave differently in transport, phosphorylation, and glycolysis. Although FDG uptake is proportional to the glucose metabolic rate, an accurate MRGlc cannot be calculated from FDG data without an accurate lumped constant. Nevertheless, it is important to emphasize that MRFDG and MRGlc individually and separately are valid quantitative assessments of metabolism although they are not identical. Quantitation of Received 11/12/01; accepted 1/3/02. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 Supported by NIH Grant CA 42045. 2 To whom requests for reprints should be addressed, at Department of Neurology, Box 356465, University of Washington, Seattle, WA 98195. Phone: (206) 543-2340; Fax: (206) 685-8100; E-mail: aspence@u. washington.edu. 3 The abbreviations used are: PET, positron emission tomography; DG, deoxyglucose; FDG, 2-[F]fluoro-2-deoxyglucose; MR, metabolic rate; Glc, glucose; RT, radiotherapy; ROI, region of interest; BCNU, 1,3bis(2-chloroethyl)-1-nitrosourea; HPLC, high-performance liquid chromatography; CT, computed tomography; MRI, magnetic resonance imaging; TAC, time-activity curve; RTOG, Radiation Therapy Oncology Group. 971 Vol. 8, 971–979, April 2002 Clinical Cancer Research Research. on April 14, 2017. © 2002 American Association for Cancer clincancerres.aacrjournals.org Downloaded from MRFDG and MRGlc each can be studied to correlate with prognosis and response to therapy. Simply stated, one would expect that successful RT of a metabolically active tumor would kill tumor cells and be associated with a reduction of metabolic rate measured by any one of several PET methods such as with FDG, 1-[C]glucose, or [O]O2 (22–24). Specifically, the tumors of patients that responded to treatment would show reduced metabolism and, conversely, the tumors of patients who did not respond would show unchanged or increased metabolic rates. Starting with this hypothesis we initiated PET studies of patients with malignant gliomas before and after they received their initial RT. All of the patients had both 1-[C]glucose and FDG studies. The PET measurements were correlated with overall survival or survival relative to appropriate prognostic class (25). The PET studies with either FDG or 1-[C]glucose preceding RT correlated inversely with survival, in agreement with a previous study of patients imaged before any RT or chemotherapy (16). Higher metabolic rate before treatment was associated with shorter survival. Unexpectedly, patients whose tumors showed increased metabolism after RT compared with before RT survived longer than those whose tumors showed decreasing metabolism. Lastly, PET studies with either FDG or 1-[C]glucose after RT showed no correlation whatsoever with survival. Materials and Methods Patients. Thirty patients were studied, all with supratentorial malignant gliomas and all within 2 weeks before and/or 1–3 weeks after RT, except for 1 post-RT patient studied 6 weeks after RT. Fourteen patients were imaged both before and after RT, 4 only before RT and 12 only after RT. Two cases were multicentric, and 5 had bilateral tumors with involvement of the corpus callosum. Table 1 shows additional details. Twenty-four patients received 59.4 Gy conformal external beam RT in 33 fractions at 1.8 Gy/fraction. Five patients received different fractionation and total dose prescriptions between 50 and 67.8 Gy, and 1 patient in the pre-RT group received only 3.6 Gy in 2 fractions before he died. The majority of patients received one or more chemotherapy regimes after RT. The majority of these treatments were nitrosourea-based with procarbazine/1-(2chloroethyl)-3-cyclohexyl-1-nitrosourea/vincristine, BCNU/cisplatin, or BCNU alone. All of the patients signed informed consent. Radiopharmaceutical Synthesis. 2-[F]Fluoro-2deoxyD-glucose was synthesized by the method of Hamacher et al. (26). The radiochemical and chemical purity of the product was measured by analytical HPLC using an aminopropyl normal phase column (Alltech and Associates, Inc.) with a CH3CN:H2O (93:7 v/v) mobile phase, and refractive index and radioactivity detection of the effluent. Silica gel TLC with a CH3CN:H2O (95:5 v/v) mobile phase was also used to assess radiochemical purity, which was consistently 99% for both methods. The synthesis of 1-[C]glucose (D isomer) followed the method of Shiue and Wolf (27, 28) as modified recently by Dence et al. (29). Typically, 1.7 Ci of [C]cyanide at end of bombardment yielded 35–40 mCi of 1-[C]glucose at the end of synthesis. The glucose was separated from mannose and any other impurities using an Aminex HPX-87P, 30 cm 7.8 mm, HPLC column (Bio-Rad Laboratories) at 70°C and eluted with sterile USP water. The radiochemical and chemical purity of the product was measured by analytical HPLC using another Aminex HPX-87P column at 70°C eluted with deionized water, and with refractive index and radioactivity detection of the effluent. PET Devices and Procedures. Two different PET systems were used over the course of this study. In all of the cases the preand post-RT scans for a given patient used the same tomograph. The first PET scans were obtained on a time-offlight PETT Electronics SP-3000 device containing four rings of BaF2 detectors with 320 crystals in each ring (30–32). Axial collimation of photons in the tomograph allowed direct and cross-plane data to be collected, yielding seven image planes. This PET device acquired emission data in list mode format with timing markers that allowed selection of time binning of data after acquisition. The limiting resolution at the center of the field of view was 5 mm in the transaxial plane and 7.5 or 11 mm in the axial plane. The interplane distance was 15 mm. The second scanner was a General Electric Advance whole body positron emission tomograph providing 35 image planes of data over a 15 cm axial field of view (33–35). The tomograph includes 18 rings of bismuth germanate oxide detectors with 672 crystals/ring. The system sensitivity in two-dimensional mode (axial septa in place) is 135 kcps/mCi/ml. The limiting transaxial resolution is 4.1 mm with a slice thickness of 4 mm. Imaging Procedure. Patients fasted for at least 9 h before the scans. Before the PET scans all of the patients had either X-ray CT or MRI scans without and with contrast injections. From the scout images of these studies axial image planes were selected for the PET scans to correspond with the planes containing the greatest tumor areas. After head immobilization was secured, patients were positioned in the tomograph. Alignment of the axial PET scan planes with those selected from the CT or MRI images was accomplished in the SP3000 by taking a lateral skull radiograph overlain with grid lines that corresponded to the planes of the tomograph. Patient and tomograph position and angulation were adjusted so that the PET tomograph planes corresponded to the desired CT or MRI axial planes. A system of laser beams then allowed for advancing the head of the patient into the tomograph to maintain the exact positioning in relation to the rings. Head positioning in the General Electric Advance tomograph was similar except no scout films were required. An attenuation scan was obtained with a rotating sector source of Ge around the brain and tumor-containing region. While this was underway, an i.v. line was introduced for isotope injection, and a wrist radial artery line was inserted for plasma sampling for the isotope TAC. The arterial line was connected Table 1 Clinical data and timing of PET studies Male Female GM AA Age range Post-RT chemotherapy Preand post-RT 6 8 9 5 30–65 13/14 Pre-RT 3 1 4 0 37–58 1/4 Post-RT 7 5 11 1 27–68 10/12 a Glioblastoma multiforme. b Anaplastic astrocytoma. 972 FDGand Glucose-PET in Gliomas before and after RT Research. on April 14, 2017. © 2002 American Association for Cancer clincancerres.aacrjournals.org Downloaded from to an automated blood sampler, which could be preprogrammed for the desired sampling sequence (36). Before scanning and isotope injection a blood glucose level was drawn and analyzed by a Glucose Analyzer II (Beckman Instruments, Fullerton, CA). This was repeated several times after isotope injection. After completion of the transmission/attenuation scans and placement of the vascular accesses, tomograph emission scan acquisition was started 1 min before injection of radioactive tracers. Calibration of the tomograph for Ci/ml was accomplished by imaging a 10-cm diameter cylinder of known activity, as determined by a dose calibrator (Capintec, Ramsey, NJ), under similar conditions as the patient imaging protocol (30 cm field of view; 6 mm Hanning filter). For some studies, calibration was performed with the patient in the tomograph during the patient imaging procedure. After a 1-min emission scan, 1-[C]glucose (typically 20 mCi) in 10 or 20 ml of sterile normal saline was injected i.v. over 1 or 2 min. 1-[C]Glucose imaging data were collected and reconstructed in time bins as follows: 4 of 20 s; 4 of 40 s; 4 of 60 s; 4 of 180 s; and 14 of 5-min duration. Arterial blood was sampled at a frequency similar to the dynamic image acquisition. The 1-ml blood samples were centrifuged, and then 0.5 ml of plasma was pipetted and counted for total plasma radioactivity using a Cobra multichannel gamma counter (Packard Corp., Chicago, IL). Three 1-ml samples of the calibration cylinder were also obtained for well counting, which allowed us to convert sample cpm/ml to Ci/ml. A FDG study followed each 1-[C]glucose study. Typically we injected 7–10 mCi of FDG in 10 ml of normal saline over 1 or 2 min. Emission data and blood samples were collected as described above for 1-[C]glucose. Image Analysis. ROIs in the gliomas were selected to include the contrast-enhancing volume and adjacent nonenhancing tumor as defined by MRI or CT images. They were drawn from the integrated FDG images while referencing MRI or CT images. Generally ROIs were placed on the FDG uptake scans (30–60 min), and cysts and resection cavities were avoided. For nonenhancing tumors the region of biopsy and T2 signalenhancing areas were selected conservatively. The model described below used the TACs from these regions and the plasma input functions both expressed as Ci/ml. Glioma ROI TACs were analyzed by the three compartment models shown in Fig. 1. The glucose model is similar to that described by Blomqvist et al. (37); the FDG model is that described by Phelps et al. (38) as a modification of Sokoloff’s model (39) for 2-DG. The program incorporating these models used the plasma TACs as input functions. The tissue data were not decay corrected because the models were formulated to account for this. Plasma data were corrected to time of sampling, not time of injection. The 1-[C]glucose plasma data were corrected for metabolites (CO2 and lactate), assuming linear accumulation of metabolites such that 18% of plasma activity at 60 min could be ascribed to metabolites in the same fashion as reported by Spence et al. and Blomqvist et al. (21, 37). It was assumed that the metabolites remained in the vascular space and were not taken up into tissue. The predicted tissue TACs were calculated using numeric integration of the differential equations for each model. The tail of the 1-[C]glucose activity was appropriately added to the FDG activity, because the FDG was injected second. There were a total of 11 variable parameters: K1, k2, k3, and k4 for 1-[C]glucose; K1, k2, k3, and k4 for FDG; a delay term to shift the tissue activity relative to the plasma curve for both tracers; and a blood volume term to account for activity in large blood vessels. The best fit of the models to a given data set was achieved using an established nonlinear weighted least squares algorithm (40). The algorithm minimized the sums of the squares of the differences between the model output and the tissue data, weighted proportionally to the inverse square root of the count of each data point. With this optimization program, the kinetic rate constants were estimated for 1-[C]glucose and FDG for ROI TACs. The metabolic rate for each hexose was calculated with their kinetic rate constants and the plasma glucose concentration (Cp) as: MR CP (K1 k3)/k2 k3). Statistics. Survival was determined from the date of diagnosis and calculated by two approaches. The first was simply months from diagnosis. The second was based on the historical data reported by the RTOG (25). Patients were individually placed in the appropriate RTOG classes by age, neurological status, and histology. The increase (or decrease) in survival ( S) relative to the appropriate RTOG class was calculated as: S patient actual survival/median survival for their RTOG class. Data were compared by regression and univariate analysis with the Wilcoxon test. For both hexoses, Glc and FDG, the patients were ranked by MR pre-RT, the MR post-RT/MR pre-RT ratio, or MR post-RT; then survival or S in the higher 50% was compared with the lower 50%. The relation between the response (time to death) and the measured prognostic factors was evaluated using a standard statistical analysis based on the Cox proportional hazards model (41) as implemented in S-plus (Math Soft Inc., Cambridge, MA). This analysis permits the examination of the influence on survival of the MRGlc and MRFDG while controlling for the impact of tumor histology, RTOG prognostic group, treatment procedures, and other potentially relevant patient information such as age and sex. Both MRGlc and MRFDG were considered as prognostic indicators of survival. The Cox model allows us to assess the percentage change in the risk of a death associated with increasing the MR value by one unit, while keeping all of the other variables (e.g., histology and RTOG prognostic group) Fig. 1 The models for glucose and FDG metabolism showing the compartments and parameters. 973 Clinical Cancer Research Research. on April 14, 2017. © 2002 American Association for Cancer clincancerres.aacrjournals.org Downloaded from fixed. We should emphasize that this is a retrospective analysis, and the findings presented would need to be confirmed with a proper prospective study.