Hypoxia and Glucose Metabolism in Malignant Tumors: Evaluation by [¹⁸F]Fluoromisonidazole and [¹⁸F]Fluorodeoxyglucose Positron Emission Tomography Imaging

Solid tumors develop regions of hypoxia when they outgrow their blood supply. Distribution of oxygen within a tumor usually shows a gradient, diminishing toward the less vascularized center of the tumor (1, 2). Hypoxia induces an aggressive phenotype, increasing metastatic potential, promoting tumor progression, and limiting the effectiveness of radiation therapy and some chemotherapeutic agents (3, 4). These characteristics are related to induction of a number of biological changes that include angiogenesis, p53 mutation, and altered regulation of cellular proliferation and glucose metabolism (5). Most solid cancer types can give rise to hypoxic regions. Significant (radiobiological) hypoxia has been detected in nearly 40% of patients with head and neck cancer (6, 7, 8). Several studies have shown smaller but significant levels of hypoxia in soft tissue sarcomas (9, 10), breast cancer (11, 12), and glioblastoma multiforme (13).

Fluorodeoxyglucose (FDG) positron emission tomography (PET) imaging has emerged as an important clinical tool for cancer detection, staging, and monitoring of response and is routinely used in the clinical management of several cancer types (14, 15). FDG uptake is largely proportional to the rate of glucose metabolism, and elevated uptake is therefore a matter of enhanced glycolysis, which can be quantified (15, 16).

More than 70 years ago, Warburg discovered that malignant cells generally have increased glycolyis, even under aerobic conditions, and suggested this was a common characteristic of tumors. Subsequent reports have indicated that aerobic glycolysis is the result of a number of genetic changes (17, 18). Hypoxic cells often use enhanced glycolysis to maintain production of energy because ATP can be produced from glucose metabolism without requiring O₂. Glycolysis can be increased as much as 2-fold in acute hypoxia and is usually associated with increased expression or modified forms of glucose transporters (GLUT 1 and GLUT 3) and increased levels of hexokinase (19) as well as intracellular mitochondrial redistribution of that enzyme. Evans et al. (20), using Ki-67 immunocytochemistry, showed a reduction in cell proliferation in hypoxic regions of a tumor. With chronic, unabated hypoxia, cell proliferation is appreciably slowed in an effort to conserve energy and adapt to reduced supplies of oxygen, micronutrients, and fuel sources such as glucose (20, 21, 22). However, these changes are not uniformly expressed and can show significant heterogeneity within and between tissue types (23).

Hypoxia is widely recognized to be an independent predictor of clinical outcome (3, 24), but hypoxia in tumors is not predicted by tumor size, grade, extent of necrosis, or blood hemoglobin status. Thus, determination of tumor hypoxia carries information independent of current clinicopathological parameters and can provide important prognostic information. Nitroimidazoles are a class of electron affinic molecules that were first shown by Chapman (25) to accumulate in hypoxic cells in cultures and in vivo. PET imaging of tissue hypoxia (pO₂ ≤ 3 mm Hg) using ¹⁸F-labeled fluoromisonidazole (FMISO), the most widely used nitroimidazole imaging agent, was developed and validated by our group. It is a noninvasive way to localize and quantify hypoxia that can be repeated when needed (26, 27, 28).

Bentzen et al. (29), in evaluating the role of FMISO and FDG in identifying hypoxia in animal studies, found that FMISO was able to identify hypoxia, but they could not find a similar role for FDG. Mankoff et al. (12), in studying blood flow and metabolism in locally advanced breast cancer, found wide variation between the two parameters; tumors with high metabolic rates responded poorly to treatment, independent of blood flow.

Although glycolysis is increased by hypoxia, there are reasons to believe that hypoxia and glucose metabolism, as measured by FMISO and FDG PET imaging, respectively, may not follow a simple relationship. Glucose metabolism is influenced by a variety of processes other than hypoxia; for example, cellular proliferation and maintaining ion gradients will result in elevated glucose metabolism in the absence of hypoxia (30). Moreover, the poor blood supply that causes hypoxia might lead to significantly reduced delivery of nutrients such as glucose in chronic situations and hence decreased glucose uptake (31). For these reasons, we investigated the relationship of regional hypoxia and glucose metabolism by comparing FMISO and FDG uptake on PET images. The study included different tumor types, ranging from those that are frequently hypoxic (head and neck cancer and glioblastoma multiforme) to those that are less frequently hypoxic (breast cancer). Our hypothesis was that high glucose metabolism could be induced by factors other than hypoxia, and, in some hypoxic tumors, limited substrate delivery might limit increases in energy metabolism induced by hypoxia. These will result in regions that show varying patterns of discordance in the uptake of these two tracers.

A total of 49 patients were included in this review. These patients underwent pretherapy FMISO and FDG PET scans as part of several ongoing research studies and were recruited from the Veterans’ Administration Puget Sound Health Care System, University of Washington Medical Center and Harborview Medical Center. Signed informed consent, as approved by the University of Washington human subjects and radiation safety committees, was obtained in all cases. Human use of [¹⁸F]FMISO is covered by an Investigational New Drug approval. Patients with advanced cancer were selected for these studies to increase the chance of including tumors with significant hypoxia. FDG scans were obtained after an overnight fast (blood sugar ≤ 150 mg at the time of imaging). No special patient preparation was required for FMISO imaging. Patients were imaged in the supine position with head immobilization (for brain tumor and head and neck cancer patients) achieved using moldable foam. Both FMISO and FDG PET studies were obtained within a few days of each other and a few days before therapy. In addition to clinical examinations, all patients had routine staging procedures that included computed tomography or magnetic resonance scans, laryngoscopy, and biopsy as appropriate for the tumor type.

All scans were performed on an Advance Tomograph (GE Medical Systems, Waukesha, WI) operating in two-dimensional high sensitivity mode with 35 imaging planes covering a 15-cm axial field of view. Performance characteristics of the tomograph have been published elsewhere (32, 33). Imaging included a 25-min transmission scan of the tumor site with a Ge-68 rotating sector source. Calibration was performed using vials containing known activities of ¹⁸F imaged separately from the patient and reconstructed using the same filter as the emission images. All head and neck and torso images were reconstructed with a Hanning filter after scatter correction, resulting in a reconstructed spatial resolution of approximately 12 mm (32). Brain images were collected in three-dimensional mode and were reconstructed with a spatial resolution of 6 mm.

[¹⁸F]FDG was prepared as described in Ref. 34 and injected i.v. (3.7 MBq/kg or 0.1 mCi/kg) into patients in a fasting state. Patients returned 45 min after injection and were positioned in the tomograph, and a scan of the tumor was obtained. In patients with tumors outside of the brain, an additional four to five axial fields of view (15 cm each) were obtained that included the base of skull and extended to below the liver. Emission scans (each of 7 min in duration) followed by transmission scans (3 min each) were obtained. All patients were premedicated with lorazepam as a muscle relaxant to reduce FDG uptake in skeletal muscle.

[¹⁸F]FMISO was prepared as described previously (35). Venous access lines were established in each arm, one for FMISO injection and the other for blood sampling. Patients received i.v. injection with 3.7 MBq/kg (0.1 mCi/kg) [¹⁸F]FMISO [maximum 370 MBq (10 mCi)]. Data acquired from 120–140 min postinjection were reconstructed with scatter, randoms, and attenuation correction and used to determine the tumor hypoxic volume (HV) as described below. These images were reconstructed with the same software and filter size used for FDG. During emission tomography, four venous blood samples were obtained at intervals of 5 min. Whole blood samples of 1 ml each were counted in a Cobra multichannel gamma well counter (Packard Corp. Meriden, CT) that is calibrated each week in units of cpm/μCi. Blood activity was averaged and then expressed as μCi/ml, decay corrected to time of injection.

Images were directly processed to provide a maximum standardized uptake value [SUV_max = maximum tissue activity (μCi/ml)/injected dose (mCi/kg)] for the tumor volume. Conventional images (computed tomography, magnetic resonance) were used to delineate tumor boundaries for drawing regions of interest (ROIs). Only the primary sites of disease were included in this analysis. For each tumor, we analyzed all of the pixels in all tumor planes.

Images were transferred to Macintosh workstations for image analysis. Generous ROIs were drawn over the entire tumor volume. All imaging planes containing tumor were analyzed. Data initially available as counts/pixel were decay corrected to the time of injection and converted to μCi/ml tissue for each pixel in the selected ROI. Because the blood values were also expressed as μCi/ml, this allowed a pixel-by-pixel calculation of tissue:blood (T:B) activity ratio for all image planes. The number of pixels in the tumor volume with a T:B ratio ≥ 1.2, indicating significant hypoxia, was determined and converted to ml units to measure the HV. Because FMISO has a partition coefficient of almost unity, in normal oxygenated tissues the concentrations in tissue and blood are essentially identical. The cutoff ratio of 1.2 was based on T:B ratios measured previously with the GE Advance tomograph in normal human brain and muscle, where >90% of the values fall below 1.1 (36). This very simple model analysis is one of the strengths of the FMISO study.

For the pixel-by-pixel comparison, the FMISO and FDG scans were coregistered using software developed by Minoshima et al. (37). A ROI was drawn over the tumor regions in each slice of the FDG scan where tumor uptake was visualized. Correlative images (magnetic resonance, computed tomography) were used as necessary to aid in the delineation of tumor boundaries. The values were scaled to the maximum pixel for each anatomical image data set to provide unitless values with a maximum of 1. The ROI was mapped to the coregistered FMISO T:B image, and the T:B ratio was extracted for each pixel within the ROI. The coregistration and mapping procedure allowed direct comparison of pixel values in the FMISO and FDG images. Pixel-wise pairs of FDG SUV and FMISO T:B values were transferred to an Excel spread sheet (Microsoft Corp., Redmond, WA) for correlation analysis.

FMISO and FDG uptake was investigated in a global manner and by analyzing pixel-wise correlations. The following image parameters were used in the analyses: FDG – SUV and SUV_max; FMISO – HV and T:B and T:B_max. Correlation between FDG and FMISO uptake was analyzed in two ways, globally (using SUV_max and HV) and on a pixel basis (with SUV and T:B ratio). Spearman rank correlation was also used to test for correlation. The statistical significance of differences between these measures was analyzed using the Kruskal-Wallis test. The relationship between FMISO and FDG uptake was investigated using the separate median values as the cutoff for HV and T:B ratio.

We hypothesized that the correlation between FMISO and FDG uptake would result in four possible scenarios (Fig. 1): group I, low degree of hypoxia and low glucose metabolism; group II, low degree of hypoxia but high glucose metabolism; group III, high degree of hypoxia and high glucose metabolism; and group IV, high degree of hypoxia but low level of glucose metabolism. We analyzed our data based on this fundamental hypothesis, looking for trends in the frequency of distribution of correlations in each group.

The 49 previously untreated patients analyzed for this study all had biopsy proven malignancies at the time of imaging. Head and neck cancer patients had a median age of 57 years (range, 47–79 years), and a T₁ (3 patients), T₂ (5 patients), T₃ (7 patients), or T₄ (11 patients) primary tumor. Soft tissue sarcoma patients had a median age of 41 years (range, 25–72 years) with a mean tumor volume of 253 ml (range, 23–1742 ml); breast cancer patients had a median age of 45 years (range, 32–66 years), and all had stage IV disease; and glioblastoma multiforme patients had a median age of 50 years (range, 31–65 years). Only pretherapy scans were used for these analyses. Examples of patient studies, showing corresponding planes of FDG and FMISO images in axial projection, are given in Figs. 2, A−D.

Details of results for the FMISO and FDG data comparisons are given in Table 1. Statistical associations between FMISO (HV and T:B_max) and FDG SUV_max uptake are tabulated. When the whole tumor images were compared, correlation coefficients between SUV_max and HV were 0.63 for head and neck cancer, 0.47 for breast cancer, 0.38 for glioblastoma multiforme, and 0.32 for soft tissue sarcoma. Table 1 also reports Spearman correlations between the overall tumor FDG quantitation (SUV_max) and its HV. Overall, the correlation was small (r = 0.24) but nonetheless significant, with a P of <0.005. When the four tumor types are analyzed separately, there is little evidence to support an association between FDG and FMISO in any tumor type, apart from the head and neck tumors, where the nominal P was 0.038. Figs. 3 and 4 show the relationship between FDG SUV_max and FMISO distribution relative to median values for both HV and T:B_max. Tables 2 and 3 show the distribution of correlation based on tracer uptake.

Table 1 also reports the average pixel-by-pixel correlations observed within each tumor type. All of these average correlations are positive and are highly significant. There is clear evidence of linear association between FMISO and FDG on a pixel-by-pixel basis. However, the strength of this association varies across the tumor types with the ordering sarcoma < brain < breast < head and neck. In fact, the differences between average correlations within the tumor types are highly significant (P < 0.005) using the rank-based Kruskal-Wallis test.

Our results (Tables 2 and 3) indicate that all of the possible scenarios (see Figs. 3 and 4) of the relationship between FMISO and FDG in our hypothesis (Fig. 1) can occur in tumors, although groups I (low degree of hypoxia and low glucose metabolism) and III (high degree of hypoxia and high glucose metabolism) show the expected preponderance. The presence of distribution in the other groups [II (low degree of hypoxia but high glucose metabolism) and IV (high degree of hypoxia but low level of glucose metabolism)] is explained by enhanced glycolysis due to factors other than hypoxia and the lack of enhanced glycolysis even in the presence of significant hypoxia, respectively. Hypoxia is heterogeneously distributed within a tumor (38).

Hypoxia can be a significant problem in the management of many solid tumors (3). Hypoxia-inducible factor 1α, which is overexpressed in response to hypoxia, is the primary transcription factor mediating a number of physiological and biological changes that include aerobic glycolysis and slowing of proliferation (39). This study was prompted by the greater degree of variation in FMISO and FDG uptake in several tumor types seen on PET scanning. Whereas it is reasonable to anticipate that hypoxia results in increased glycolysis, the results in our study indicate that there can be a wide variation in these two phenomena.

We analyzed the uptake patterns of these two tracers (FMISO and FDG) in two different ways: (a) global measures (HV and SUV_max); and (b) pixel analysis (SUV and T:B_max) using all of the pixels in the delineated tumor. Visual interpretation of these graphs showed different correlation between the global uptake of the two tracers for the various tumor types. However, because the ability to respond to hypoxia is tissue type dependent and genetically determined, the other traits might be exhibited in some situations (17).

We analyzed the global correlation between FMISO and FDG for all tumor types as summarized in Table 1. The maximum T:B ratio of FMISO uptake is nearly identical in different tumor types, whereas HV shows variation according to the tumor type. Our previous analysis of soft tissue sarcomas did not show an association between tumor volume and HV (28).

Using identical ROIs on coregistered PET images, our pixel-by-pixel analyses of these images showed only a marginally significant spatial correlation in the uptake of the two tracers. There is wide variation among tumor types (Table 1). This wide variation in uptake can be attributed to the functional and biochemical heterogeneity seen in hypoxic tissues (20). Within a tumor, oxygenation varies, ranging from well-perfused and oxygenated regions to areas that are underperfused and severely hypoxic. Existence of a mixture of acute and chronic hypoxia within a tumor is likely to be a significant factor in causing this heterogeneity (40). This is clearly seen in our patients with sarcoma, who show the least correlation between FMISO and FDG uptake. The discordance between FMISO and FDG might also be a reflection of vascularization and blood supply; tumors with poor blood supply have increased hypoxia but also have reduced glycolysis as a result of decreased supply of the nutrients (41).

Our study demonstrates heterogeneity in hypoxia and in energy metabolism and that the discordance between the uptake of FMISO and FDG does not appear to be a function of the tumor size, as seen in our analysis of a cohort of patients with soft tissue sarcomas (28). If that (tumor size) were the case, soft tissue sarcomas would be expected to show a greater degree of heterogeneity than head and neck cancers, which are relatively smaller. On the contrary, we noted heterogeneity and discordance in different cancer types, indicating that this phenomenon is ubiquitous in solid tumors and likely reflects the genetic response to hypoxic stress.

Similarly, if there were a simple mechanistic relationship between FMISO and FDG uptake, our pixel-by-pixel analyses would have indicated a higher correlation. However, this was not the case. FDG uptake is a function of many factors such as microvasculature, GLUT, hexokinase expression, and proliferation rate (22). FDG uptake has been correlated with aggressiveness of tumors and shown to be an indicator of poor prognosis in many different tumor types (12, 42, 43, 44, 45). Tolerance to chronic hypoxia is a hallmark of higher organisms, in an effort to adapt for times of insufficient supply of oxygen and glucose (46). Bioenergetic adaptations can vary in tumor cells exposed to chronic depletion of oxygen. Our study clearly indicates that there can be a wide variation in the relationship between hypoxia and energy metabolism and all four possible scenarios can exist in tumors, a fact that needs to be considered in the clinical and imaging characterization of tumor biology and in treatment.

The presence of widespread heterogeneity in a tumor would indicate that hypoxia is diffuse throughout the tumor and that response to hypoxia will be equally widespread and varying. A diffuse hypoxic process indicates a generalized response to hypoxic stress with associated biological changes and aggressive behavior and hence with poor prognosis. On the other hand, a lesser degree of spatial heterogeneity in hypoxia might mean a more localized and well-demarcated process. This information on the biology of tumors will also have implications on treatment selection. Whereas the former requires a more general or systemic approach such as a hypoxic cell cytotoxin (e.g., tirapazamine) to solve the problem of hypoxia (47), a more focal hypoxia might benefit from a local/regional approach [e.g., radiation dose escalation (48)].

Our study is limited by the small number of patients. However, we have included in our analysis several tumor types differing in degrees of hypoxia and several corresponding planes on both FMISO and FDG images. This provides us enough statistical advantage. Although significant heterogeneity in the uptake of these two tracers exists, it is impossible to look at any degree of variation in the uptake between individual tumor cells using current PET technology. However, it is apparent that hypoxia-induced changes in a group of cells, rather than in an individual cell, are clinically and radiobiologically more significant; an analysis of the variation between individual cells is not likely to result in any advantage for treatment using current technologies. PET FMISO imaging will likely provide a “road map” for tissue sampling in characterizing biomarker expression. Whereas FDG uptake might indicate the presence of hypoxia (49, 50), it should not be considered as a surrogate marker for hypoxia (51). Thus, using both FMISO and FDG appears to be beneficial in the evaluation of solid tumors. Additional patient studies that include other tumor types are needed before more definitive conclusions can be reached.

We thank the following for their help: Jeff Scharnhorst (RN), patient care coordinator and database manager; and all nuclear medicine technologists, radiochemists, and physicists in the Division of Nuclear Medicine.