Whole-Tumor Perfusion CT Parameters and Glucose Metabolism Measurements in Head and Neck Squamous Cell Carcinomas: A Pilot Study Using Combined Positron-Emission Tomography/CT Imaging

Materials and Methods

Results

All PET/CT and PCT studies were sequentially acquired without adverse effects and were suitable for further evaluation. The summary statistics for the PET and PCT measurements in the healthy-appearing muscle tissue and in the tumors are demonstrated in Table 1. A representative case of PET/CT and PCT imaging in a patient with primary skin SCCA is shown in Fig 1.

• Download figure
• Open in new tab
• Download powerpoint

Fig 1.

Tissue perfusion CT map (A) and PET/CT SUV_mean map (B) of a 54-year-old woman with a SCCA in the skin on the left temporal side. Blood flow (A) map shows a hypervascularized area in the subcutaneous region with infiltration of the cutis (arrows). The increased perfusion of the neoplastic tissue is coupled with increased SUV uptake (B) in the FDG PET/CT imaging (arrows).

All functional PET/CT (SUV_mean and SUV_max) and PCT parameters (mean BF, BV, MTT, and PS) in tumors were very different from the corresponding parameters in healthy-appearing muscle tissue at a very high level of statistical significance (.0001 < P < .001). In the pooled ROIs (tumor and muscle tissue ROIs) group, significant correlations were observed among SUV_max and SUV_mean and the tissue perfusion parameters (Table 2).

Unlike the pooled ROIs in which high correlation coefficients were caused by the linear relationship between healthy and pathologic values,¹² further correlation coefficient analysis only in the tumor ROIs revealed limited correlations as follows: between BF and SUV_mean (r = 0.63; P = .011), between BF and SUV_max (r = 0.57; P = .02), and between PS and SUV_mean (r = 0.53; P = .04). Regression analysis between SUV parameters and BF showed SUV_max = 0.09 × BF + 7.2 (R² = 0.33; P = .02) and SUV_mean = 0.05 × BF + 2.22 (R² = 0.45; P = .011) (Fig 2). In a similar fashion, regression analysis between PS and SUV_mean revealed SUV_mean = 0.05 × PS + 5.36 (R² = 0.35; P = .04).

• Download figure
• Open in new tab
• Download powerpoint

Fig 2.

Regression analysis of the mean BF values in the tumor (triangles) and healthy tissue (circles) with the SUV_mean. There is a positive significant correlation (r = 0.63; P = .011) in the tumor ROIs. Linear regression analysis showed SUV_mean = 0.05 × BF + 2.22 (R² = 0.45; P = .011) (regression line). The curved dashed lines show the 95% confidence intervals for the tumor ROIs means. The dashed line depicts the line of perfect agreement (equation line).

The BV values in the tumor ROIs did not demonstrate any significant correlation with the SUV_mean and SUV_max (r = 0.06–0.14; P ≥ .62). The MTT values in the tumors showed a negative, but not significant, correlation with the SUV_mean(r = −0.3; P = .27) and a weak nonsignificant correlation with SUV_max (r = 0.16, P = .4).

Regarding the tumor-to-nontumor (muscle) SUV_mean and SUV_max ratios, descriptive statistics showed that the tumor-to-muscle SUV_mean ratio was 9.45 ± 3.55 (8.43–10.78), whereas the SUV_max ratio was 17.58 ± 4.32 (15.48–19.04). Correlation coefficient analysis between BF in tumor tissue and SUV ratios of tumor to muscle demonstrated significant correlations: BF–Ratio SUV_mean: r = 0.64; P = .01; BF–Ratio SUV_max: r = 0.53; P = .04. Regression analysis of the above significantly correlated parameter showed Ratio SUV_mean = 0.14 × BF–3.48 (R² = 0.42; P = .01) and Ratio SUV_max = 0.14 × BF + 4.51 (R² = 0.29; P = .04) (Fig 3). The correlation coefficients between the other perfusion CT parameters and SUV ratios of tumor to muscle were not significant and ranged from −0.08 (MTT-Ratio SUV_mean) to 0.35 (BV–Ratio SUV_max; P ≥ .29).

• Download figure
• Open in new tab
• Download powerpoint

Fig 3.

Regression analysis of the mean BF values in the tumor with the SUV_mean ratios of tumor-to-nontumor tissue. There is a positive significant correlation (r = 0.64; P = .01) in the tumor ROIs. Linear regression analysis showed Ratio SUV_mean = 0.14 × BF–3.48 (R² = 0.42; P = .01) (regression line). The curved dashed lines show the 95% confidence intervals for means. The dashed line depicts the line of perfect agreement (equation line).

Discussion

The biologic basis of FDG-PET is the increased glucose metabolism of malignant cells compared with noncancerous tissue. After administration of the radiopharmaceutical, FDG is transported into the tumor cell, phosphorylated by hexokinase, and then trapped inside the tumor cell, as it is neither processed in the glycolytic pathway nor able to leave the cell. Thus, it can be visualized on PET scan. The rate of FDG uptake in primary lung cancer has correlated with the degree of differentiation of the tumor; doubling time of the tumor; proliferation rates; and, thus, aggressiveness,^13,14 whereas the overexpression of glucose transporters has been associated with poor prognosis in patients with oral SCCA.¹⁵ For glucose to be taken up and used by SCCA, an adequate vascular supply, including neoangiogenesis-induced vasculature, is necessary. PCT provides information about tissue perfusion in the intravascular and extravascular extracellular space, which has been shown to correlate also with histopathologic grade and angiogenic stimuli, such as vascular endothelial growth factor.^16,17 The relationship between these 2 imaging techniques and whether their roles may be complementary in tumor imaging of different organs have remained controversial topics, partly because direct comparison of these modalities by means of combined imaging has so far not been performed in humans.

The regulation of glucose use in cells is complex and involves various levels and different metabolic pathways. For example, increased glycolysis can be a response to an increase in cellular energy, increase in cell proliferation and synthesis rates, and activation of specific oncogenic pathways even in the presence of adequate oxygen (Warburg effect).¹⁸ Tumor hypoxia occurs when a rapidly growing tumor eventually surpasses the host–tissue-derived vascularization that can only provide microcirculatory functions in selected parts of the tumor. Where the tumor must depend on tumor-derived neovascularization, the delivery of oxygen and nutrients is severely compromised, giving rise to chronic or diffusion-limited hypoxia, which is exacerbated by deterioration of blood rheology, platelet aggregation, and increased interstitial pressure within the tumor. Hypoxia is also believed to induce a shift from aerobic to anaerobic glycolysis, which is associated with production of lactate and lower pH values. However, evidence from a direct comparison of SUV with hypoxia markers contradicts this view,¹⁹ and discrimination between chronic and acute hypoxic environment may play an important role in understanding the decrease or increase of metabolic activity in the neoplastic tissue, respectively.¹⁸ Moreover, tumor hypoxia seems to be a spatially and temporally heterogeneous phenomenon, resulting from the combined effect of many factors, including tumor type and volume, disease site (specific organ or tissue), regional microvessel attenuation, BF, oxygen diffusion and consumption rates, etc. Thus, a consistent pattern has not always been observed between local glucose use and hypoxia as monitored by FDG-PET, fluoromisonidazole-PET imaging, and pO₂ measurements.^20,21

In our study, we correlated values and ratios of FDG SUV and mean tissue perfusion (BF) as well as other PCT-associated parameters (BV, MTT, and PS) obtained, thereby eliminating any measurement inaccuracies during the coregistration of values obtained from separate PET and CT scans. The mean SUV_mean of the SCCA in our study is slightly lower than those reported by Hirasawa et al⁸ in 16 patients with head and neck tumors and by Miles et al²² in 2 cases with SCCA of the lung. On the other hand, the mean SUV_mean of the SCCA in our study is higher compared with the values reported by Tateishi et al¹⁰ in 12 patients with SCCA of the lung. The PCT values in tumor tissue reflected also the neovascularity, were in close agreement with those reported in the literature, and were significantly different from those in the muscle tissue.^1-3,23 Relatively large SDs in the tumor PS values and muscle BF and PS parameters, as also shown in the initial work of Gandhi et al,³ were evident in our population without affecting the differentiation between the 2 types of tissue.

In our study, there was a statistically significant correlation between SUV_mean and SUV_max and mean BF in the tumor tissue. This correlation indicates a coupling between 2 different functional parameters regarding SCCA physiology. This coupling may be anchored in the paradoxical situation in which tumor cells growing in conditions of normal oxygen tension (which may be provided by adequate blood supply) can also show an elevated glycolytic rate (Warburg effect).²⁴ Moreover, a positive relationship between increased perfusion and tissue oxygenation, which is subsequently expressed in glucose metabolism, has been observed in breast and lung tumors.^10,25 It is notable that Mankoff et al²⁵ showed a positive correlation between SUV and BF measured with ¹⁵H₂O-PET (considered the criterion standard in determination of BF) in patients with advanced breast cancer. Other PCT studies^8,9 in liver and head and neck cancers have shown that a negative correlation between SUV and tissue perfusion may also exist. In these cases, it has been postulated that the increasing neoangiogenesis of the tumor results in inadequate oxygen supply, which subsequently leads to an uncoupling of glucose uptake and BF.

Previous PCT studies have not examined any correlation between PS values (as a marker of neoangiogenesis) and SUV. It has also been hypothesized that the negative correlation may be because of the hypoxia induced by the massive tumor cell proliferation that distances cells from the vasculature. Such an influence of the size of the tumor on the SUV-BF correlation has been recently observed in lung tumors.²² In contrast to the aforementioned results of an inverse correlation between perfusion and SUV, the results of our study are distinct because the positive correlation between blood glucose metabolism and perfusion was observed in mostly moderately large (T2–3) SCCAs and was accompanied with a weak, but statistically significant, correlation with PS values, which indicated a coupling of the neoangiogenesis with the glucose metabolism. A previous study of separate PET and MR imaging studies in liver metastases of colorectal cancer showed a nonsignificant, negative correlation between rate constant κ_ep (a parameter akin to PS) and the tumor-to-nontumor SUV_mean ratio.¹⁹ We believe that the different modalities and histologic characteristics of the tumor may be the reasons for this discrepancy. Another study in patients with lung cancer showed a positive correlation between SUV and PS.²⁶ Future studies would need to examine the relationship between SUV and PS (ie, a potential mismatch between PS and SUV would indicate a peritumoral inflammation since it can theoretically have increased SUV but not increased neo-angiogenesis).²⁷ This relationship has to be elucidated in future studies. Our PS results should be interpreted cautiously because the applied postprocessing software may provide a wide range of PS relative to BF in the tumor capillaries, which, in turn, may lead to an inaccurate determination of both parameters.²⁸ Finally, our results may not apply in other head and neck tumor types nor in neoplastic tissue in other organs because of differences in histologic features, tumor size, degrees of cellularity, and necrosis.²⁹

The SUV values in our study did not correlate with BV and MTT in the tumors, though a negative correlation between BF (as well as SUV) and MTT may exist because of the newly formed tumor vasculature, characterized by extensive arteriovenous shunts, that may lead to a rapid washin and washout of the contrast agent and FDG-labeled blood.¹ We attribute the lack of correlation between BV and SUV to the different nature of the measured values. The “lumped” BV value in our postprocessing protocol expresses the whole intravascular and extravascular extracellular BV, and thus correlation with cell metabolism may not be achievable.

Initial work on the predictive value of PCT and PET has been already separately performed. Zima et al⁶ investigated the response prediction of advanced SCCA treated with induction chemotherapy, whereas Hermans et al²³ showed a predictive value of BF in the outcome of SCCA treated with radiation therapy. In a similar fashion, SUV measurements performed by Schwartz et al³⁰ showed an association between baseline primary tumor FDG SUV and SCCA outcomes, but no predictive value of the nodal SUV, whereas a recent study of nasopharyngeal carcinoma showed that patients having tumors with high ¹⁸F-FDG uptake had a significantly lower 3-year disease-free survival rate than patients with lower tumor uptake.³¹ The results of our study shed light on the biology of head and neck SCCA and may potentially find application in the dual-technique assessment (which can be rapidly performed in the same setting) of untreated head and neck SCCAs that could determine the contribution (or added value) in predicting the behavior and prognosis of a tumor. For instance, the noninvasive dual-technique pretreatment assessment of SCCAs and the degree of correlation (match or mismatch) between SUV and SCCA perfusion may provide additional information about tissue oxygenation status, which is of significant importance for both radiation and chemotherapy (including induction chemotherapy) outcomes. In such cases, a pixel-to-pixel evaluation of the tumor tissue may be advantageous. Thus, the therapeutic options may potentially be tailored on an individual basis. Beyond this point, the dual functional assessment of SCCA may provide an improved monitoring tool because the evident role of the FDG-PET/CT technique may be enhanced by PCT, for which some preliminary experience exists.¹

Our findings were limited by the small patient population and the subsequent lack of a sufficient number of small versus advanced SCCAs. Furthermore, partial volume averaging during the manual delineation of the tumor tissue in FDG PET/CT and PCT was an inevitable limitation. The mean value of perfusion and SUV parameters indicate the value of these variables over the ROI, but it does not reflect the heterogeneity of a tumor. Although tumor heterogeneity was not taken into account in our study, the data were corrected for gross necrosis. Statistical analyses, such as histogram analysis or pixel-to-pixel evaluation, may be used to characterize spatial heterogeneity in future studies. Correlation of our findings with tumor oxygenation measurements by oxygen-sensitive needle electrodes could be performed to determine the relationship of functional imaging and tumor hypoxia.

Conclusions

Our study contributes to the understanding of the relationship between 2 noninvasive promising functional imaging modalities (PET and PCT). The increased and positively correlated SUV and tissue perfusion indices may be based on the increased tissue oxygenation and the lack of any acute or chronic hypoxic conditions in the assessed head and neck SCCAs. Our results may lead to better understanding of tumor pathophysiology, pretherapeutic planning, and follow-up imaging of the patient.