OtherClinical Investigations

Assessment of Hypoxia and Perfusion in Human Brain Tumors Using PET with 18F-Fluoromisonidazole and 15O-H2O

Matthias Bruehlmeier, Ulrich Roelcke, Pius A. Schubiger and Simon Mensah Ametamey
Journal of Nuclear Medicine November 2004, 45 (11) 1851-1859;

Abstract

Hypoxia predicts poor treatment response of malignant tumors. We used PET with 18F-fluoromisonidazole (18F-FMISO) and 15O-H2O to measure in vivo hypoxia and perfusion in patients with brain tumors. Methods: Eleven patients with various brain tumors were investigated. We performed dynamic 18F-FMISO PET, including arterial blood sampling and the determination of 18F-FMISO stability in plasma with high-performance liquid chromatography (HPLC). The 18F-FMISO kinetics in normal brain and tumor were assessed quantitatively using standard 2- and 3-compartment models. Tumor perfusion (15O-H2O) was measured immediately before 18F-FMISO PET in 10 of the 11 patients. Results: PET images acquired 150–170 min after injection revealed increased 18F-FMISO tumor uptake in all glioblastomas. This increased uptake was reflected by 18F-FMISO distribution volumes >1, compared with 18F-FMISO distribution volumes <1 in normal brain. The 18F-FMISO uptake rate K1 was also higher in all glioblastomas than in normal brain. In meningioma, which lacks the blood–brain barrier (BBB), a higher K1 was observed than in glioblastoma, whereas the 18F-FMISO distribution volume in meningioma was <1. Pixel-by-pixel image analysis generally showed a positive correlation between 18F-FMISO tumor uptake at 0–5 min after injection and perfusion (15O-H2O) with r values between 0.42 and 0.86, whereas late 18F-FMISO images (150–170 min after injection) were (with a single exception) independent of perfusion. Spatial comparison of 18F-FMISO with 15O-H2O PET images in glioblastomas showed hypoxia both in hypo- and hyperperfused tumor areas. HPLC analysis showed that most of the 18F-FMISO in plasma was still intact 90 min after injection, accounting for 92%–96% of plasma radioactivity. Conclusion: Our data suggest that late 18F-FMISO PET images provide a spatial description of hypoxia in brain tumors that is independent of BBB disruption and tumor perfusion. The distribution volume is an appropriate measure to quantify 18F-FMISO uptake. The perfusion–hypoxia patterns described in glioblastoma suggest that hypoxia in these tumors may develop irrespective of the magnitude of perfusion.

Hypoxia is defined as a reduction of intracellular oxygen pressure (pO2) as a result of decreased supply of and increased demand for oxygen. pO2 levels in normal tissue generally exceed 40 mm Hg, and severe hypoxia may be present in malignant tumors (1). Levels of pO2 < 3 mm Hg are associated with impaired response to radiotherapy (2). Two different forms of tumor hypoxia are recognized. Diffusion-limited chronic hypoxia may develop as a result of increased intercapillary distances, and acute hypoxia can result from occlusion of large tumor vessels (3). Either form of hypoxia has several implications for the further evolution of tumors (e.g., by induction of signaling cascades that promote angiogenesis, growth, and cell migration) (4). Tumor hypoxia may also lead to necrosis, which is mandatory to establish the diagnosis in glioblastoma multiforme (5). In brain tumors, the presence of hypoxia has been inferred from pathologic examination of tumor tissue, from animal models, and from MRI of malignant gliomas that indicate necrosis (5–7).

In vivo measurement of hypoxia in individual patients is of clinical interest. It could provide insight into the natural course and pathophysiology of tumors, possibly assisting in the planning of radio- and chemotherapy and in the evaluation of treatment response. Noninvasive methods to measure tumor hypoxia with PET or SPECT are available with 18F-fluoromisonidazole (18F-FMISO) or other radioligands (8–13).

18F-FMISO is a nitroimidazole derivative, which to our knowledge was the first PET agent used for hypoxia detection in a larger cohort of patients with tumors outside the central nervous system (14). 18F-FMISO PET can image tumor hypoxia by increased 18F-FMISO tumor uptake, because 18F-FMISO metabolites are trapped exclusively in hypoxic cells (15). In tumor hypoxia, the 18F-FMISO tumor concentration measured by PET typically exceeds 18F-FMISO plasma concentration as measured during PET. Because 18F-FMISO is relatively lipophilic, with an octanol water coefficient (log P) of 0.4, it diffuses through cell membranes and shows a passive distribution in normal tissue. Therefore, with the exception of the liver, the intestines, the kidney, and the bladder, the 18F-FMISO concentration in normal, nonhypoxic tissue is always lower than in plasma.

The results of 18F-FMISO PET studies of 3 patients with glioblastoma multiforme have been reported by Valk et al. (16). In the pilot study presented here, we measured hypoxia in different human brain tumors using PET and 18F-FMISO. In addition, because gliomas show a large range of tumor blood flow, we investigated the influence of tumor perfusion on 18F-FMISO kinetics and on the presence of hypoxia by means of 15O-H2O PET (17). 18F-FMISO PET scans were performed quantitatively to determine the 18F-FMISO distribution volume and transport rate constants in normal brain and in brain tumors.

MATERIALS AND METHODS

Patients

Eleven patients were selected at the Cantonal Hospital (Aarau, Switzerland). T2-weighted and gadolinium-enhanced T1-weighted MR images were available for all patients. All patients, with a single exception, exhibited either residual or recurrent tumor after surgery, with signs of tumor progression on MR images or appearance of a new lesion. Of the 7 patients with glioblastoma, 6 completed radiotherapy before the PET study, and 5 patients had received 4–6 cycles of temozolomide chemotherapy. At the time of the PET study, 5 patients were receiving oral dexamethasone (mean, 4 mg/d). The meningioma patient was included to address the 18F-FMISO kinetics in a brain tumor that lacks a blood–brain barrier (BBB). All patients gave written consent for participation in the study. The study protocol was approved by the Ethical Committee at the Cantonal Hospital Aarau.

18F-FMISO and 15O-H2O PET Scanning

The radiosynthesis of 18F-FMISO was performed according to the 2-step procedure reported by Lim et al. (18). The total synthesis time was approximately 120 min, and radiochemical purity was >99%. Specific radioactivities were also always >30 GBq/μmol.

All PET studies were performed on a whole-body scanner (Advance; General Electric Medical Systems), where 35 reconstructed planes cover an axial field of view of 14.6 cm and the reconstructed in-plane resolution is 7 mm. Catheters were placed into a radial artery under local anesthesia and in a vein of the contralateral arm. For PET, the patient’s head was fixed in a special holder. First, a 10-min transmission scan with 2 rotating 68Ge line sources was performed for attenuation correction. Then, 500–700 MBq 15O-H2O were injected intravenously as a bolus, and a static 60-s PET scan was started when the radioactivity in the head exceeded 100,000 counts per second as measured by the scanner (a 15O-H2O scan could not be performed in 1 patient because of a malfunction of the cyclotron during the PET imaging). After waiting 10 min to allow sufficient decay of 15O, an average of 291 MBq 18F-FMISO (range, 123–421 MBq) were infused intravenously during 3 min using a constant volume infusion pump. 18F-FMISO infusion was preferred to a bolus injection, because it permitted manual drawing of blood samples for precise measurement of the arterial input curve, including the radioactivity peak after injection. A device to record the arterial input curve online in real time was not available. For this reason, arterial blood measurements were not performed during the 15O-H2O PET scan. Simultaneous with the start of 18F-FMISO infusion, a dynamic PET scan was acquired, consisting of 31 frames with a total scan duration of 90 min. The PET time frames were 9 × 20, 4 × 30, 2 × 60, 4 × 120, 9 × 300, and 3 × 600 s. During the PET scan, 26 arterial blood samples of 4 mL each were drawn, and the radioactivity of both whole blood and plasma was measured using a well counter. The blood samples were drawn at 0.5, 1, 1.5, 2, 2.5, 3, 3.5, 4, 4.5, 5, 5.5, 6, 7, 8, 9, 10, 12.5, 15, 17.5, 20, 25, 30, 40, 50, 60, and 75 min after injection. After resting approximately 45 min, the patients were repositioned on the PET scanner. A second 10-min transmission scan was acquired, followed by a static 20-min emission scan exactly 150–170 min after injection of 18F-FMISO.

Generally, image processing used a minimum of filtering. In the case of dynamic 18F-FMISO scans, filtering was not necessary, because adding the last 3 PET time frames resulted in sufficient image quality to draw the regions of interest (ROIs). The 15O-H2O images required a 3-dimensional gaussian filter with a full width at half maximum (FWHM) of 8 mm and for the static 18F-FMISO images a filter of FWHM 5 mm. Blood measurements were decay corrected to the time of 18F-FMISO injection. We performed a standard cross-calibration between PET images and blood radioactivity by measuring the radioactivity of a known amount of 18F-FMISO in a water phantom with both the PET scanner and a γ-well counter.

Image Coregistration and ROIs

For coregistration of 15O-H2O images with the first 18F-FMISO scan, no spatial corrections were necessary because of the fixed head position during both scans. The 31 frames of the first 18F-FMISO PET scan were examined for motion artifacts. A spatial correction was necessary to coregister the second 18F-FMISO scan with the first. For that purpose, the last 3 frames of the first scan (60–90 min after injection) were added, allowing delineation of the brain contours and those tumors that accumulated 18F-FMISO. These landmarks were used for a manual coregistration of both 18F-FMISO scans using the software PMOD (PMOD Technologies Ltd.) (19).

For an analysis of 18F-FMISO kinetics in normal brain, we placed ROIs of identical size in the contralateral frontal cortex, frontal white matter, and cerebellum. ROI sizes were 7.28 (312 voxels), 4.04 (173 voxels), and 18.23 mL (781 voxels) for white matter, frontal cortex, and cerebellum, respectively. Tumor ROIs in all glioblastomas (n = 7) and in the hemangioblastoma were placed on areas with increased 18F-FMISO uptake as determined by late images. These ROIs did not cover the whole tumor area but included presumably hypoxic tumor areas. In another patient with a meningioma, the tumor was delineated only in early PET images 0–30 min after injection, and these images were consequently used to define the tumor ROI. Tumor ROIs were placed on multiple transverse slices and then summed up to a 3-dimensional ROI. Sizes of these tumor ROIs were 0.86, 1.63, 2.12, 0.37, 2.12, 1.24, and 4.34 mL (37, 70, 91, 16, 91, 53, and 186 voxels, respectively) for patients 1–7, 0.86 mL (37 voxels) for patient 10, and 1.70 mL (73 voxels) for patient 11.

For a pixel-per-pixel correlation between 15O-H2O and 18F-FMISO PET images, a different set of tumor ROIs was defined to minimize the possibility of including surrounding normal brain tissue. These ROIs were placed only on a single transverse slice showing the maximum tumor diameter, so that it covered the whole tumor area with variably increased 18F-FMISO uptake.

PET Data Analysis

Decay-corrected time–activity curves were derived from the first 18F-FMISO PET scan from 0 to 90 min after injection. The radioactivity of normal brain and tumor was plotted against time after injection and the time–activity curves were used for kinetic modeling to calculate the 18F-FMISO transport rates and distribution volume in 2- and 3-compartment models (Appendix). In addition, we applied Logan plots in normal brain and in tumor, with DVLogan as a measure of the distribution volume of 18F-FMISO independent of assumptions about the number of underlying tissue compartments (20). The second, static 18F-FMISO PET scan was also evaluated by ROI analysis (i.e., ratios of tumor and cerebellum radioactivity were calculated).

15O-H2O PET images were analyzed semiquantitatively, and perfusion in tumor hypoxia was graded as: 1 = lower than white matter; 2 = equal to white matter; 3 = between white matter and cortex; 4 = equal to cortex; 5 = higher than cortex. For the purpose of comparing hypoxia with perfusion, hypoxic tumor areas were delineated on late 18F-FMISO PET images, and the resulting ROIs were transferred to the coregistered 15O-H2O PET images. In addition, a pixel-per-pixel correlation between 15O-H2O and 18F-FMISO images was performed. Pixel values in a tumor ROI were normalized to their own maximum, and relative 18F-FMISO uptake at 0–5, 60–90, and 150–170 min after injection was plotted against relative 15O-H2O uptake.

Measurement of 18F-FMISO Stability in Plasma and Urine

For measurement of 18F-FMISO plasma stability, 0.6 mL of a solution containing 3% perchloric acid (60%; Merck), 1% Na2S2O5 (Merck), and 0.1% ethylenediaminetetraacetic acid (Fluka) were added to 0.5 mL plasma for protein precipitation. The solution was centrifuged at 4,800 rpm during 5 min, and 1 mL of the supernatant was submitted to high-performance liquid chromatography (HPLC) using a Merck Hitachi L-6200A Intelligent Pump and a Bondclone 10 C18 column. We used a solution containing 80% water and 20% acetonitrile for elution; flow was set to 0.5 mL/min. The eluate was collected by a 2112 Redirac fraction collector in 100 samples during a total elution time of 30 min. The radioactivity of each sample was measured in a γ-well counter (Cobra II Auto Gamma; Camberra Packard) and was plotted against elution time. 18F-FMISO metabolites in urine were measured similarly, with the exception that no protein precipitation was performed. In addition, the retention time of intact 18F-FMISO was determined in several HPLC experiments using an aliquot of the 18F-FMISO injection solution. Using knowledge of the normal retention time of intact 18F-FMISO, radioactive metabolites in plasma and urine were visually discriminated from the 18F-FMISO peak on the HPLC plots, and the amount of metabolites and intact 18F-FMISO were expressed as a percentage of total radioactivity in the chromatogram.

RESULTS

Clinical data are presented in Table 1 . Examples of MR and 18F-FMISO PET images are shown in Figure 1, where the glioblastoma multiforme exhibited increased 18F-FMISO uptake at late scan time (150–170 min after injection). In contrast, the meningioma showed higher 18F-FMISO uptake than surrounding brain only during the first 30 min after injection. Time–activity curves of the same 2 patients are depicted in Figure 2, showing that the radioactivity in the glioblastoma begins to exceed plasma radioactivity approximately 30 min after injection, whereas the radioactivity in the meningioma did not exceed plasma radioactivity at any time. In normal brain, the radioactivity of cortex and white matter slowly approached the radioactivity in plasma, a finding that is representative for all patients.

FIGURE 1.
FIGURE 1.

(A and B) Gadolinium contrast-enhanced T1- and T2-weighted MR images in patient with glioblastoma multiforme in left temporal lobe. (C) Late PET images show 18F-FMISO tumor uptake 150–170 min after injection. (D) Gadolinium contrast-enhanced T1-weighted MR image of patient with meningioma in left temporal lobe. (E) 18F-FMISO uptake in meningioma is visible in early PET image 0–5 min after injection but not in late PET images 150–170 min after injection (F). PET images are normalized to their own maximum.

FIGURE 2.
FIGURE 2.

Time–activity curves of 18F-FMISO PET scans in patient with glioblastoma (A) and patient with meningioma (B), corresponding to PET images shown in Figure 1. Plasma = radioactivity concentration in arterial plasma during PET.

View this table:
TABLE 1

Summary of Clinical Data

In normal brain, the kinetics of 18F-FMISO generally satisfied a 2-compartment model with the distribution volume (DV′) < 1 in all regions, indicating a passive distribution and reflecting a homogeneous 18F-FMISO uptake in the cerebrum and cerebellum at late scan time (Table 2). In tumor, increased 18F-FMISO uptake was observed at late scan time in all glioblastoma multiforme and in the hemangioblastoma but not in the anaplastic astrocytoma (World Health Organization tumor classification III [WHO III]), a fibrillary astrocytoma (WHO II), or a meningioma. Tumor areas with increased uptake showed an 18F-FMISO distribution volume (DVTOT) > 1 in the 3-compartment model (Table 3). Determination of the 18F-FMISO distribution volume using a Logan plot resulted in DVLogan values that were slightly smaller than DVTOT in 5 of 7 tumors. Notably, DVLogan was <1 in 2 patients despite increased 18F-FMISO uptake and DVTOT > 1. Ratios of tumor and cerebellum radioactivity in the static late PET image were comparable with DVTOT values (Table 3).

View this table:
TABLE 2

18F-FMISO PET Results in Normal Brain

View this table:
TABLE 3

PET Results in Tumor

Visual comparison of 15O-H2O with late 18F-FMISO PET images in 6 patients with glioblastoma showed a large range of tumor perfusion within areas with increased 18F-FMISO uptake (i.e., increased 18F-FMISO uptake was found both in hypo- and hyperperfused tumor areas (Fig. 3; Table 3). Generally, increased 18F-FMISO uptake was found in the tumor margin but not in the tumor center. Tumor centers of all glioblastomas showed decreased radioactivity in both 15O-H2O and 18F-FMISO PET images.

FIGURE 3.
FIGURE 3.

(A–C) 15O-H2O PET perfusion images in 3 patients with glioblastoma multiforme. (D–F) Corresponding late 18F-FMISO PET images show tumor hypoxia in low perfusion, in intermediate perfusion with an inverse pattern compared with hypoxia, and in high perfusion. PET images are normalized to their own maximum.

Pixel-per-pixel correlations between 15O-H2O and 18F-FMISO images were performed in 5 of the glioblastoma patients (example in Fig. 4). At 0–5 min after injection, 18F-FMISO uptake was positively correlated with perfusion in 4 of 5 cases. The correlation coefficients (r values) in these patients were 0.69, 0,02, 0.42, 0.74, and 0.86. This correlation was no longer evident at 60–90 min after injection in 3 of 5 patients but persisted in 2 patients. With a single exception, no correlation was found at 150–170 min after injection, when the r values were 0.02, 0.10, 0.61, 0.06, and 0.01. No inverse correlation between 18F-FMISO uptake and perfusion was observed.

FIGURE 4.
FIGURE 4.

Pixel-by-pixel correlation of relative tumor perfusion with 18F-FMISO accumulation in hypoxic area of glioblastoma multiforme, where 18F-FMISO uptake was measured at 0–5 min (A), 60–90 min (B), and 150–170 min (C) after injection.

HPLC analysis of 7 patients showed that 97%–99% of plasma radioactivity represented intact 18F-FMISO 3 min after injection. Most of the 18F-FMISO in plasma (92%–96%) was still intact 90 min after injection (in 1 patient after 75 min), and 1 hydrophilic metabolite, accounting for 4%–6% of plasma radioactivity, was detected. In contrast, higher 18F-FMISO catabolism was found in urine, where radioactive metabolites accounted for up to 17% at 95 min after injection.

DISCUSSION

In this pilot study, we used 18F-FMISO PET to investigate hypoxia in brain tumors. 15O-H2O PET was performed to compare the pattern of tumor hypoxia with perfusion. We found increased 18F-FMISO uptake, indicating the presence of tumor hypoxia, in all 7 glioblastoma patients. Tissue radioactivity in glioblastoma exceeded plasma radioactivity by a factor of up to 2 within 90 min after injection. The 18F-FMISO tumor distribution volumes varied between approximately 1 and 2. In contrast, the distribution volume in normal brain was always <1. The increased 18F-FMISO distribution volume in tumor resulted in observable 18F-FMISO retention in late PET images (150–170 min after injection). Increased 18F-FMISO uptake was also found in the hemangioblastoma patient, but 18F-FMISO did not accumulate in an anaplastic astrocytoma (WHO III), in a fibrillary astrocytoma (WHO II), or in a meningioma.

The presence of hypoxia appears to be a common feature of glioblastoma multiforme (21), and intracerebral high-grade gliomas have been found to accumulate 18F-FMISO in an animal model (22). In humans, hypoxia in glioblastoma multiforme has also been directly measured with Eppendorf needle electrodes. Not only do the glioblastomas show severe hypoxia, but surprisingly low-grade astrocytomas and peritumoral brain tissue do so as well (23). Moreover, various endogenous markers of hypoxia are found to be overexpressed in brain tumors. For example, in glioblastoma multiforme and, to a lesser degree, in grade II and III astrocytomas, a heterogeneous pattern of the hypoxia-inducible factor 1α (HIF-1α) overexpression was found (24). Increased HIF-1α levels were observed in tumor areas adjacent to necrosis (e.g., in pseudopalisading cells). Consistent with that, we found increased 18F-FMISO tumor uptake generally in the periphery but not in the center of a glioblastoma multiforme. The latter is expected, because only viable cells are able to accumulate 18F-FMISO, and delivery to necrotic tissue is low.

To quantify the pharmacokinetics of 18F-FMISO in normal brain and tumor, we used dynamic 18F-FMISO PET data and the measured arterial 18F-FMISO concentration for kinetic modeling with standard 2- and 3-compartment models. The calculated distribution volumes and transport rate constants (Appendix) characterize 18F-FMISO as a PET agent. These parameters are especially important in the brain, because the state of the BBB can affect a radioligand’s tissue delivery, confounding PET images. Our results confirm this effect for 18F-FMISO. Increased K1 values, ranging between 0.067 and 0.143 mL/min/g, were found in all glioblastomas, compared with values for K1 of 0.010 mL/min/g in white matter and 0.023 mL/min/g in cortex, indicating a relatively slow entrance of 18F-FMISO into normal brain. The increased 18F-FMISO tumor uptake in glioblastoma cannot be an exclusive effect of BBB disruption resulting in increased K1 values. Increased K1 can account only for higher 18F-FMISO uptake rates as long as tissue radioactivity does not exceed plasma radioactivity and cannot explain the continuing tracer accumulation at later scan times. In the meningioma, for example, which lacks the BBB, we found the highest value for K1 of all brain tumors, as expected. The increased K1 permitted delineation of the meningioma in an early 18F-FMISO PET image (Fig. 1), but the tumor did not exhibit subsequent 18F-FMISO accumulation and was not visualized in the late PET image. Accordingly, the 18F-FMISO distribution volume in the meningioma was not increased despite a high K1 value. Therefore, we assume that 18F-FMISO accumulation in glioblastoma at late scan time does not result from an increased K1 but from the presence of a second kinetic tissue compartment with an additional uptake rate, k3. Thus, it is likely that late 18F-FMISO PET images provide more than an image of the disrupted BBB. However, increased K1 values in glioblastoma may confound 18F-FMISO PET images at early times after injection.

We propose the total distribution volume (DVTOT) of a 3-compartment model for absolute quantification of 18F-FMISO tumor uptake, where values >1 indicate tumor uptake due to k3 > 0 and, therefore, suggest the presence of hypoxia. Alternatively, the 18F-FMISO distribution volume can be determined graphically using a Logan plot, avoiding assumptions about the number of underlying tissue compartments. Finally, because the distribution of 18F-FMISO in the brain is very uniform, simple radioactivity ratios using ROI analysis can be used for relative quantification of static 18F-FMISO at late times after injection. Ratios of tumor and cerebellum radioactivity likely contain the same information as DVTOT for the detection of tumor hypoxia (Table 3).

We assume that no other active mechanism than the binding of bioreductive metabolites, dependent on the presence of hypoxia, leads to increased 18F-FMISO tumor uptake. Additional factors (e.g., tumor viability, the presence of necrosis, or cellular density) may modulate the degree of 18F-FMISO tumor uptake. Moreover, cellular 18F-FMISO uptake and retention depend nonlinearly on the pO2 (25). Therefore, although DVTOT or DVLogan allow quantitative estimates of 18F-FMISO uptake, they cannot be directly related to tumor pO2.

An often debated issue in the use of 18F-FMISO concerns stability in vivo, because this is a prerequisite for specific tumor uptake in hypoxic areas. Although a relatively high metabolism of 18F-FMISO in plasma and urine in mice hinders tumor imaging (26), we find most of the plasma radioactivity to be intact 18F-FMISO until 90 min after injection in humans and >80% of 18F-FMISO to appear as intact parent substance in the urine. We conclude that nonspecific radioactivity uptake in human tumors as a result of 18F-FMISO plasma metabolites is not relevant, and no correction for metabolites is needed if 18F-FMISO plasma concentration is measured to provide an arterial input curve. In this respect, the use of 18F-FMISO represents no disadvantage compared with other nitroimidazole derivatives.

An important aspect of this study was investigating the influence of perfusion on 18F-FMISO uptake in brain tumors (i.e., we wanted to exclude the possibility that 18F-FMISO PET images simply reflect tumor perfusion). Pixel-per-pixel comparisons of 18F-FMISO with 15O-H2O PET images yielded a positive correlation of early 18F-FMISO uptake with perfusion (Fig. 4A), as for almost any radioligand used in nuclear medicine. This correlation was only minimal or already lost at 60–90 min after injection (Fig. 4B), and 18F-FMISO images at 150–170 min after injection were independent of perfusion with a single exception (Fig. 4C). These results are a precondition if 18F-FMISO retention at late time in brain tumors is to provide information other than perfusion. Static PET images should therefore not be obtained earlier than 2–2.5 h after injection.

We found different patterns of hypoxia and perfusion in patients with glioblastomas (Fig. 3), showing apparently that hypoxia occurs independently of perfusion. All glioblastomas investigated showed hypoxia. Only 2, however, revealed hypoperfusion, whereas hypoxia occurred in tumor areas with intermediate or high perfusion in the other 4 tumors. As illustrated in Figure 3, these patterns were observed in the tumor periphery, where tissue is considered most active with regard to proliferation and infiltration of surrounding brain tissue. This hypoxia–hyperperfusion pattern is unexpected, because in general one would expect hypoxia to occur in low perfusion. The latter has been demonstrated in the myocardium of dogs, where Martin et al. (27) demonstrated an inverse correlation between perfusion and 18F-FMISO uptake after complete or partial occlusion of the left anterior descending coronary artery. Animal studies using human glioma xenografts also demonstrated an inverse relationship between the density of perfused vessels and the fraction of hypoxic tumor cells (7). That study showed that the intensity of hypoxia staining corresponds to increasing distances of tumor cells from vessels, making an inverse correlation between perfusion and hypoxia conceivable. Groshar et al. (28) used 99mTc-hexamethylpropyleneamine oxime and 123I-iodoazomycin arabinoside SPECT in patients to demonstrate an inverse relation between perfusion and hypoxia in nonglial tumors. However, all 4 glioblastoma patients in that study showed decreased perfusion and absence of hypoxia.

A mismatch between hypoxia and perfusion may be characteristic of malignant brain tumors, which exhibit a discrepancy between high vascularization and a comparably low, possibly inefficient perfusion (29). Differentiation between diffusion-limited (chronic) and perfusion-limited hypoxia has been proposed (3). Whether these 2 types are reflected in the patterns observed in our present study and that of Groshar (28) cannot be answered, because the findings of Rijken et al. (7) and Tochon-Danguy et al. (22) were obtained in animals. It is not known whether patterns of hypoxia and perfusion change with the evolution of glioblastoma. Their impact on prognosis and response to therapy remains a subject for future study.

CONCLUSION

We performed 18F-FMISO and 15O-H2O PET in brain tumors to measure tumor hypoxia and perfusion. Increased 18F-FMISO tumor retention at late scan time was found predominantly in glioblastoma multiforme, and this was associated with an increased 18F-FMISO tumor distribution volume. This finding could not be explained by nonspecific perfusion effects or by BBB disruption alone. We used the 18F-FMISO distribution volume as a quantitative criterion for hypoxia, where values >1 indicated active binding and suggested the presence of tumor hypoxia. 18F-FMISO accumulated in both hypo- and hyperperfused tumor regions, suggesting that hypoxia in glioblastoma may develop irrespective of the magnitude of perfusion. Simple tumor-to-cerebellum ratios at late scan time provide a good estimate of the 18F-FMISO distribution volume in the clinical setting.

APPENDIX

Description of Compartment Models

Kinetic modeling was performed with dynamic 18F-FMISO PET data during the first scan from 0–90 min after injection using the commercially available software PMOD (19). The tissue time–activity curves measured by PET and the arterial plasma concentration were used to quantify transport rates and distribution volume of 18F-FMISO with standard 2- and 3-compartment models. The 3-compartment model consisted of 1 intravascular and 2 tissue compartments, where the transport rate constants K1 and k2 describe the 18F-FMISO uptake and washout rates from plasma to the first tissue compartment, respectively, and the transport rate constants k3 and k4 represent the 18F-FMISO exchange between the first and the second tissue compartment. The intravascular compartment was set to a fixed volume of 0.05 (i.e., spillover by blood radioactivity was assumed to account for 5% of tissue radioactivity). Curve fitting determined the transport rates K1, k2, k3, and k4. Briefly, a model tissue curve was calculated using the arterial plasma radioactivity curve and the model transport rates K1–k4, and the fitting algorithm tried to minimize the difference between the measured tissue radioactivity and the model curve. We used an iterative fitting algorithm with Marquardt–Levenberg optimization, a standard option in the software PMOD (19,30). The total 18F-FMISO distribution in both tissue compartments was calculated as DVTOT = K1/k2 × (1 + k3/k4). Alternatively, we used a 2-compartment model consisting of an intravascular compartment with a constant volume of 0.05 and a single tissue compartment. Parameters of the 2-compartment model were K1′ and k2′, and the 18F-FMISO distribution volume was calculated as DV′ = K1′/k2′. This model has only 2 free parameters, K1′ and k2′, and, therefore, promises fits of higher robustness.

Model Evaluation and Results

The goodness of fit was evaluated using the Akaike criterion (AIC) as described by Hawkins et al. (31). Generally, low AIC values indicate the appropriateness of the model to explain PET data with a minimum of free parameters. Introduction of an excess parameter does not impair the model fit but leads to an increased AIC value. This indicates that the parameter is not necessary to explain PET data and, therefore, not reliable. On the other hand, too few free parameters may prevent the model curve from being fitted well to PET data, resulting in increased AIC values.

To determine the optimal model to explain the 18F-FMISO kinetics in brain tissue, we successively increased the degrees of freedom of the 3-compartment model by stepwise introduction of K1, k2, k3, and k4, until the minimum AIC value was reached. For example, if fitting with a free k3 resulted in a lower AIC value than a fit with k3 kept at zero, this was interpreted as evidence for the presence of a k3 and, therefore, of a second kinetic tissue compartment. Consequently, we used the 3-compartment model only if the AIC criterion indicated the presence of 2 kinetic tissue compartments. This was true in all tumors and also in normal brain in some instances. In all other cases, the 2-compartment model was used instead, particularly in normal brain. In addition, we used the AIC to determine reversibility or irreversibility of 18F-FMISO tissue uptake. Thus, we checked whether the AIC increased if k2 or k4 was held constant at zero in the 2- or 3-compartment model, suggesting reversibility of binding, because k2 and k4 are >0.

In normal brain, 18F-FMISO generally distributed passively in a single reversible tissue compartment with DV′ < 1 (Table 2). Reversibility of 18F-FMISO uptake could always be established (i.e., keeping k2 at zero resulted in significantly worse fits with higher values for the AIC). However, kinetic analysis showed the presence of 2 tissue compartments in the frontal cortex in 4 patients and in the cerebellum in 2 patients. We have no explanation for the sporadic presence of 2 18F-FMISO tissue compartments in normal brain. In no cases did analysis of the AIC indicate the presence of a second tissue compartment in white matter. There was good agreement between DVTOT and DV′ (Table 2). Thus, the choice of the model did not play an important role in the resulting value of the 18F-FMISO distribution volume in normal brain. Determination of the 18F-FMISO distribution volume by a Logan plot yielded values that were comparable with the 2- and 3-compartment models in the frontal cortex and cerebellum, but values in white matter were somewhat lower. In tumor, increased 18F-FMISO uptake at late scan times was associated with DVTOT > 1 (Table 3). In these instances, analysis of the AIC indicated the presence of 2 kinetic tissue compartments (i.e., the goodness of fit always improved with the 3-compartment model).

18F-FMISO tumor binding was found to be reversible in 4 patients, where holding k4 to zero resulted in worse data fits with increased AIC values. In 3 patients, k4 > 0 could also be determined, but the goodness of fit was not impaired if k4 was kept at zero. Therefore, reversibility of the second tissue compartment could not be proven in these patients. On one hand, reversibility of 18F-FMISO tumor binding is surprising, because the covalent binding of bioreductive 18F-FMISO metabolites in hypoxic cells is expected to be irreversible (32). On the other hand, isolated cell studies have indicated that some metabolites of radiolabeled misonidazole may “back diffuse” from the cells at a low rate, which can possibly explain our finding (33).

Acknowledgments

We thank Professor Gustav K. von Schulthess and Professor Alfred Buck for placing the PET infrastructure of the University Hospital Zurich at our disposal and for technical support. Also, we thank Thomas Berthold and Claudia Keller for assistance in PET and Erika Sinnig for laboratory work. We also thank Dr. John Missimer for assistance in editing the manuscript.

Footnotes

  • Received Jan. 8, 2004; revision accepted May 27, 2004.

    For correspondence or reprints contact: Matthias Bruehlmeier, MD, Department of Nuclear Medicine, Cantonal Hospital Aarau, CH-5001 Aarau, Switzerland.

    E-mail: matthias.bruehlmeier{at}ksa.ch

REFERENCES

  1. ↵
    Nias AHW. An Introduction to Radiobiology. 2nd ed. London, UK: Wiley & Sons; 1998.
  2. ↵
    Gray LH, Conger AD, Ebert M, et al. The concentration of oxygen dissolved in tissues at the time of irradiation as a factor in radiotherapy. Br J Radiol. 1953;26:638–648.
  3. ↵
    Airley RE, Monaghan JE, Stratford IJ. Hypoxia and disease: opportunities for novel diagnostic and therapeutic prodrug strategies. Pharm J. 2000;264:666–673.
    OpenUrl
  4. ↵
    Ikeda E, Achen MG, Breier G, et al. Hypoxia-induced transcriptional activation and increased mRNA stability of vascular endothelial growth factor in C6 glioma cells. J Biol Chem. 1995;270:19761–19766.
    OpenUrlAbstract/FREE Full Text
  5. ↵
    Kleihues P, Cavanee WK. Tumours of the nervous system. Pathology and Genetics: World Health Organization Classification of Tumors. Lyon, France: IARC Press; 2000.
  6. Bernsen HJ, Rijken PF, Peters H, et al. Hypoxia in a human intracerebral glioma model. J Neurosurg. 2000;93:449–454.
    OpenUrlCrossRefPubMed
  7. ↵
    Rijken PF, Bernsen HJ, Peters JP, et al. Spatial relationship between hypoxia and the (perfused) vascular network in a human glioma xenograft: a quantitative multi-parameter analysis. Int J Radiat Oncol Biol Phys. 2000;48:571–582.
    OpenUrlCrossRefPubMed
  8. ↵
    Chapman JD, Engelhardt EL, Stobbe CC, et al. Measuring hypoxia and predicting tumor radioresistance with nuclear medicine assays. Radiother Oncol. 1998;46:229–237.
    OpenUrlCrossRefPubMed
  9. Rasey JS, Koh WJ, Grierson JR, et al. Radiolabeled fluoromisonidazole as an imaging agent for tumor hypoxia. Int J Radiat Oncol Biol Phys. 1989;17:985–991.
    OpenUrlPubMed
  10. Lewis J, Laforest R, Buettner T, et al. Copper-64-diacetyl-bis(N4-methylthiosemicarbazone): an agent for radiotherapy. Proc Natl Acad Sci USA. 2001;98:1206–1211.
    OpenUrlAbstract/FREE Full Text
  11. Gronroos T, Eskola O, Lehtio K, et al. Pharmacokinetics of 18F-FETNIM: a potential marker for PET. J Nucl Med. 2001;42:1397–1404.
    OpenUrlAbstract/FREE Full Text
  12. Dolbier WR Jr, Li AR, Koch CJ, et al. [18F]-EF5, a marker for PET detection of hypoxia: synthesis of precursor and a new fluorination procedure. Appl Radiat Isot. 2001;54:73–80.
    OpenUrlCrossRefPubMed
  13. ↵
    Lehtio K, Oikonen V, Nyman S, et al. Quantifying tumour hypoxia with fluorine-18 fluoroerythronitroimidazole ([18F]FETNIM) and PET using the tumour-to-plasma ratio. Eur J Nucl Med Mol Imaging. 2003;30:101–108.
    OpenUrlCrossRefPubMed
  14. ↵
    Rasey JS, Koh WJ, Evans ML, et al. Quantifying regional hypoxia in human tumors with positron emission tomography of [18F]fluoromisonidazole: a pretherapy study of 37 patients. Int J Radiat Oncol Biol Phys. 1996;36:417–428.
    OpenUrlCrossRefPubMed
  15. ↵
    Whitmore GF, Varghese AJ. The biological properties of reduced nitroheterocyclics and possible underlying biochemical mechanisms. Biochem Pharmacol. 1986;35:97–103.
    OpenUrlCrossRefPubMed
  16. ↵
    Valk PE, Mathis CA, Prados MD, et al. Hypoxia in human gliomas: demonstration by PET with fluorine-18-fluoromisonidazole. J Nucl Med. 1992;33:2133–2137.
    OpenUrlAbstract/FREE Full Text
  17. ↵
    Leenders KL. PET: blood flow and oxygen consumption in brain tumors. J Neurooncol. 1994;22:269–273.
    OpenUrlCrossRefPubMed
  18. ↵
    Lim JL, Berridge MS. An efficient radiosynthesis of [18F]fluoromisonidazole. Appl Radiat Isot. 1993;44:1085–1091.
    OpenUrlCrossRefPubMed
  19. ↵
    Mikolajczyk K, Szabatin M, Rudnicki P, et al. A JAVA environment for medical image data analysis: initial application for brain PET quantitation. Med Inform (Lond). 1998;23:207–214.
  20. ↵
    Logan J, Fowler JS, Volkow ND, et al. Graphical analysis of reversible radioligand binding from time–activity measurements applied to [N-11C-methyl]-(-)-cocaine PET studies in human subjects. J Cereb Blood Flow Metab. 1990;10:740–747.
    OpenUrlCrossRefPubMed
  21. ↵
    Brat DJ, Castellano-Sanchez AA, Hunter SB, et al. Pseudopalisades in glioblastoma are hypoxic, express extracellular matrix proteases, and are formed by an actively migrating cell population. Cancer Res. 2004;64:920–927.
    OpenUrlAbstract/FREE Full Text
  22. ↵
    Tochon-Danguy HJ, Sachinidis JI, Chan F, et al. Imaging and quantitation of the hypoxic cell fraction of viable tumor in an animal model of intracerebral high-grade glioma using [18F]fluoromisonidazole (FMISO). Nucl Med Biol. 2002;29:191–197.
    OpenUrlCrossRefPubMed
  23. ↵
    Collingridge DR, Piepmeier JM, Rockwell S, et al. Polarographic measurements of oxygen tension in human glioma and surrounding peritumoural brain tissue. Radiother Oncol. 1999;53:127–131.
    OpenUrlCrossRefPubMed
  24. ↵
    Zagzag D, Zhong H, Scalzitti JM, et al. Expression of hypoxia-inducible factor 1-alpha in brain tumors: association with angiogenesis, invasion, and progression. Cancer. 2000;88:2606–2618.
    OpenUrlCrossRefPubMed
  25. ↵
    Gross MW, Karbach U, Groebe K, et al. Calibration of misonidazole labeling by simultaneous measurement of oxygen tension and labeling density in multicellular spheroids. Int J Cancer. 1995;61:567–573.
    OpenUrlPubMed
  26. ↵
    Rasey JS, Hofstrand PD, Chin LK, et al. Characterization of [18F]fluoroetanidazole, a new radiopharmaceutical for detecting tumor hypoxia. J Nucl Med. 1999;40:1072–1079.
    OpenUrlAbstract/FREE Full Text
  27. ↵
    Martin GV, Caldwell JH, Graham MM, et al. Noninvasive detection of hypoxic myocardium using fluorine-18-fluoromisonidazole and positron emission tomography. J Nucl Med. 1992;33:2202–2208.
    OpenUrlAbstract/FREE Full Text
  28. ↵
    Groshar D, McEwan AJ, Parliament MB, et al. Imaging tumor hypoxia and tumor perfusion. J Nucl Med. 1993;34:885–888.
    OpenUrlAbstract/FREE Full Text
  29. ↵
    Vajkoczy P, Menger MD. Vascular microenvironment in gliomas. J Neurooncol. 2000;50:99–108.
    OpenUrlCrossRefPubMed
  30. ↵
    Marquardt DW. An algorithm for least squares estimation of nonlinear parameters. J Soc Industr Appl Math. 1963;11:431–441.
    OpenUrlCrossRef
  31. ↵
    Hawkins RA, Phelps ME, Huang SC. Effects of temporal sampling, glucose metabolic rates, and disruptions of the blood–brain barrier on the FDG model with and without a vascular compartment: studies in human brain tumors with PET. J Cereb Blood Flow Metab. 1986;6:170–183.
    OpenUrlPubMed
  32. ↵
    Casciari JJ, Graham MM, Rasey JS. A modeling approach for quantifying tumor hypoxia with [F-18]fluoromisonidazole PET time–activity data. Med Phys. 1995;22:1127–1139.
    OpenUrlCrossRefPubMed
  33. ↵
    Rasey JS, Grunbaum Z, Magee S, et al. Characterization of radiolabeled fluoromisonidazole as a probe for hypoxic cells. Radiat Res. 1987;111:292–304.
    OpenUrlCrossRefPubMed
Back to top

In this issue

Print
Download PDF
Article Alerts
Email Article
Citation Tools
Bookmark this article

Jump to section

Related Articles

Cited By...

More in this TOC Section

Similar Articles