In Vivo Quantification of Hypoxic and Metabolic Status of NSCLC Tumors Using [¹⁸F]HX4 and [¹⁸F]FDG-PET/CT Imaging

Lung cancer has the highest death rate among leading cancer types (1). Standard treatment for advanced-stage non–small cell lung cancer (NSCLC) is the combination of radiotherapy and chemotherapy, administered either sequentially or concurrently (2). Tumor cell hypoxia has a negative impact on cancer treatment effectiveness, it promotes resistance to radiotherapy and chemotherapy, increases the metastatic potential, and is therefore related to a poor prognosis (3–5). Tumor hypoxia is present in the majority of NSCLCs, which can be visualized and quantified using functional PET imaging with radiolabeled 2-nitroimidazoles (6, 7).

3-[¹⁸F]Fluoro-2-(4-((2-nitro-1H-imidazol-1-yl)methyl)-1H-1,2,3-triazol-1-yl)propan-1-ol ([¹⁸F]HX4) is a relatively new nitroimidazole with attractive pharmacokinetic properties that has successfully completed preclinical and clinical testing (8–10).

In standard clinical practice, a combination of anatomic computed tomography (CT) and functional 2[¹⁸F]fluoro-2-deoxy-d-glucose (FDG)-PET imaging is frequently used to visualize, detect, and stage malignancies. In addition, FDG-PET can be used to identify subvolumes with a high metabolism, which are more susceptible to local recurrence after (chemo)radiotherapy (11, 12). Aerts and colleagues (11, 13) showed that the residual tumor volume after radiotherapy is mainly located within the pre-radiotherapy high FDG-uptake volume. However, 30% of the residual volume did not correspond to the high FDG volume. This may be caused by tumor regrowth in pre-radiotherapy hypoxic tumor subvolumes located outside the high FDG volume. Therefore, it is of great interest to investigate the correlation between both unfavorable biologic features (high tumor metabolism, hypoxia) because they can be used to predict treatment outcome. In addition, imaging-derived tumor features have the potential to guide treatment with hypoxic modifiers or radiotherapy dose painting (14–16). The uptake of FDG in the cell is dependent on the overexpression of glucose transporters (GLUT), which can be upregulated in the absence of oxygen, through the HIF1α-mediated pathway (17). This may suggest a possible overlap between volumes of high FDG uptake and tumor hypoxia, even though they represent different biologic properties of tumors.

The aim of this study was to perform an extensive comparison of tumor metabolism, using FDG, and hypoxia, using HX4, to fully characterize the relationship between both PET tracers on a global tumor and voxel level for primary NSCLC and the regional lymph node metastases.

FDG- and HX4-PET/CT images of 25 patients with NSCLC (17 male and 8 female) were acquired before the start of external beam radiotherapy. The average age of the patients was 63 years (range, 40–82 years). Tumor stage ranged from IIB-IV, pathology being adenocarcinoma (N = 13), squamous cell carcinoma (SCC; N = 5), and large cell carcinoma (N = 7). Patients were treated with radical radiotherapy (N = 3) or chemoradiation (N = 22), with the majority of patients receiving at least one cycle of chemotherapy before PET imaging and before the start of radiotherapy (Supplementary Table S1). PET data were acquired in the translational research part of two phase II trials [PET-Boost, NCT01024829 (18); Nitroglycerin (NCT01210378)], both having identical PET imaging procedures. The clinical trials were approved by the appropriate medical ethical review committee, and all patients provided written informed consent before study entry.

HX4 was produced as described in previous publications (8–10, 19). After intravenous administration of 429 ± 57 MBq HX4, PET/CT imaging was performed at 4 hours postinjection (p.i.) for 20 to 30 minutes in a single bed position centered around the primary tumor. HX4-PET/CT images were acquired on a Gemini TF 64 scanner (Philips Healthcare), with a spatial resolution of approximately 5-mm FWHM. We performed CT-based attenuation correction and scatter correction (SS-SIMUL), and reconstructed PET images using -ordered subset iterative time-of-flight reconstruction technique (BLOB-OS-TF) with three iterations and 33 subsets in a 144 × 144 matrix and voxel sizes of 4 × 4 mm.

The injected activity of FDG was based on the patient's body weight according to the national guidelines (20). PET/CT imaging was performed 1 hour after intravenous administration of FDG. FDG-PET/CT scans were acquired using a Biograph 40 PET/CT scanner (Siemens Healthcare). Scatter and attenuation corrections were applied. PET images were reconstructed using OSEM 2D (Ordered Subset Expectation Maximization, four iterations, eight subsets) and a Gaussian filter of 5 mm. A respiratory correlated CT was performed, with the mid-ventilation scan selected for the attenuation correction and fusion with the FDG-PET.

HX4- and FDG-PET/CT scans were acquired in the same week for all except 1 patient. The median interval between both PET scans was 3 days (range, 1–14 days). No interventions (e.g., radiotherapy or chemotherapy) were performed between the FDG- and HX4-PET scans. Both scans were acquired with the patient positioned in radiotherapy position, on a flat tabletop using a laser alignment system with arms in an arm-support positioned above the head.

For all patients, gross tumor volumes (GTV), including the primary lesion (GTV_prim) and involved lymph nodes (GTV_ln), were defined on the FDG-PET/CT scan by two experienced radiation oncologists in consensus. GTV_prim and GTV_ln were analyzed separately. Lesions with a size <5 cm³ were excluded because of potential partial volume effects.

The FDG-PET/CT was rigidly registered to the HX4-PET/CT using registration software developed in-house. The rigid transformation was determined by the registration of the FDG-CT to the HX4-CT; the same transformation was subsequently applied to the FDG-PET scan and the GTVs. A volume of interest in the aorta was defined as background region.

The maximum and mean standardized uptake values (SUV_max and SUV_mean), corrected for body weight, were determined within the GTV for both FDG- and HX4-PET. For the HX4-PET, calculations were made of the maximum tumor-to-blood ratio (TBR_max), defined as the SUV_max in the tumor divided by the SUV_mean in the aorta, the HX4 high-fraction (HX4-HF) and HX4 high-volume (HX4-HV), both defined as the fraction/volume of the GTV with a TBR > 1.4. For the FDG-PET, calculations were made of the FDG high-fraction (FDG-HF) and FDG high-volume (FDG-HV) based on the PET-Boost trial strategy, using the GTV volume with an SUV above 50% of the SUV_max (18).

This classification for defining HX4-HV and FDG-HV, as a fraction of the total GTV, was used to subdivide regions of a tumor into four classes: (i) FDG-low and HX4-low, (ii) FDG-high and HX4-low, (iii) FDG-low and HX4-high, and (iv) FDG-high and HX4-high. To evaluate the effect of the threshold definition on tumor subdivision, a calculation was made of the average distribution using alternative thresholds. The HX4 threshold varied from TBR > 1.3 to TBR > 1.5; the FDG threshold ranged from SUV > 30% to SUV > 70% of SUV_max.

A visual and voxel-wise comparison of the FDG and HX4 uptake within the GTV was performed to compare spatial uptake patterns in the primary lesions. On the basis of the voxel-wise analysis, we separated lesions into three groups. First, in lesions showing a high correlation between the FDG and HX4 uptake, the hypoxic volume was entirely within the high metabolic volume. Second, in lesions showing a moderate correlation between the FDG and HX4 uptake, there was only a partial overlap between the HX4-HV and the FDG-HV. Third, in lesions showing a different uptake pattern between the two tracers, there were two distinct regions of FDG-HV and HX4-HV.

Mean ± 1 SD were reported for all parameters. Linear and multiple linear regressions were performed to correlate the GTV-based parameters (SUV_max, SUV_mean, TBR, HF, and HV) and to quantify the voxel-wise comparison of the FDG and HX4 uptake. Pearson correlation coefficients were calculated. A P value of <0.05 was assumed to be statistically significant.

This study analyzed the overall FDG and HX4 uptake in the primary tumor and lymph nodes of 25 patients with NSCLC. All GTV_prim (N = 25) and 19 GTV_ln were larger than 5 cm³ and all were used for the analysis. The average values of the GTV, FDG, and HX4 parameters are shown in Table 1. The subclassification, based on tumor pathology, showed no significant differences for any of the FDG or HX4 parameters (examples are shown in Supplementary Fig. S1). The FDG-HV was larger than the HX4-HV in 24/25 GTV_prim and in all GTV_ln. Potential correlations between FDG- and HX4-PET-based parameters were investigated.

The correlation coefficients for the primary tumors are shown in Table 2. The majority of the FDG- and HX4-PET-based parameters showed a significant correlation with the primary tumor volume. Note that the HX4-HV was significantly correlated with the tumor volume, while the HX4-HF was not. The FDG-SUV_mean correlated positively with all HX4-PET parameters. The FDG-SUV_max only showed a significant correlation with HX4-SUV_max (R = 0.54; P < 0.01), HX4-TBR (R = 0.55; P < 0.01), and HX4-HV (R = 0.66; P < 0.001). The highest correlations were observed when comparing the FDG-HV with the HX4-based parameters: HX4-SUV_max (R = 0.63; P < 0.01), HX4-TBR (R = 0.62; P < 0.01), and HX4-HV (R = 0.76; P < 0.0001). Two examples are shown in Fig. 1. From Fig. 1B one can appreciate that, although there is a correlation between FDG-SUV_max and HX4-TBR_max, it is not possible to distinguish the nonhypoxic lesions by using only the FDG-SUV_max parameter.

A multiple linear regression was performed to test the interaction between primary tumor volume and FDG parameters to predict the hypoxic volume. Using the parameters primary tumor volume and FDG-SUV_max to predict HX4-HV, we observed a correlation coefficient of 0.74 (R² = 0.55) with a significant contribution of both FDG-SUV_max (P < 0.01) and primary tumor volume (P = 0.03). Adding the interaction term (FDG-SUV_max* primary tumor volume) to the model increases the correlation coefficient to 0.82 (R² = 0.67).

For the involved lymph nodes, GTV_ln volume has a large effect on the correlation coefficients between the HX4 and FDG parameters (Supplementary Table S2). The multiple linear regression using GTV_ln volume and FDG-SUV_max to predict HX4-HV (R = 0.96) therefore showed a significant contribution only for the GTV_ln volume (P < 0.001) and not for FDG-SUV_max (P = 0.26).

The average distribution within the primary tumor based on the four previously predefined categories is shown in Table 3 and visualized in Fig. 2A. On average, the FDG-HV is 42% ± 21% of the GTV_prim, of which 10% ± 12% is hypoxic. On average, 3% (range, 0%–31%) of the GTV_prim is hypoxic but outside the FDG-HV, representing 24% (3.2%/13.6%) of the total hypoxic volume.

The effect of alternative thresholds on the average distribution of FDG and HX4 within the primary tumor is shown in Supplementary Table S3 and visualized for two examples in Fig. 2B and C. This figure shows that the hypoxic percentage of the GTV (HX4-HF) outside the high FDG area (FDG-HF) is relatively stable.

Tracer uptake above the background level in both PET scans is essential for comparing the overlap of FDG-HV and HX4-HV. All primary lesions showed FDG uptake with an SUV_max > 3.5; however, only 19 out of 25 primary lesions expressed an HX4 uptake (TBR > 1.4). These 19 lesions were selected for further analysis.

On the basis of the voxel-wise analysis, we observed that in 10 lesions, less than 1 cm³ of the HX4-HV was outside the FDG-HV (group 1; Fig. 3A). In seven lesions, 2 to 12 cm³ of the HX4-HV was outside the FDG-HV (group 2; Fig. 3B). Finally, in 2 patients, a clearly distinct uptake pattern was observed between the two tracers and hypoxic volumes of 46 and 102 cm³ were observed outside the FDG high-uptake region, which were 73% and 78% of the total HX4-HF, respectively (group 3; Fig. 3C). The primary tumor volume was significantly correlated to the group the lesion was assigned to (R = 0.75; P < 0.01).

This study was initiated to assess the correlation of (spatial) uptake patterns of hypoxia (using HX4-PET) and tumor metabolism (using FDG-PET) in primary NSCLC and associated lymph node metastases. Both biologic features are known to have an adverse impact on treatment outcome in NSCLC. FDG-PET is routinely used in clinical practice for staging, radiotherapy planning, and treatment response monitoring, while the use of hypoxia (HX4) PET imaging is still limited to clinical trials. We show in 25 patients with NSCLC, with different histopathologic subtypes, that HX4-PET imaging provides additional information to FDG-PET, which can be used to individualize patient treatment.

The relationship between HX4- and FDG-PET was investigated at a tumor level by comparing the overall uptake within the GTVs of the primary tumor and lymph node metastases. Significant correlations were observed between GTV-, HX4-, and FDG-PET image parameters. Previous studies comparing the overall uptake of hypoxia PET and FDG-PET showed varying results. No correlations were observed by Bollineni and colleagues (7) or Cherk and colleagues (21), while Vera and colleagues (22) reported a significant correlation. Gagel and colleagues (23) compared FDG and FMISO uptake to the gold standard of hypoxia measurements (pO₂ polarography) and observed a moderate correlation for FMISO but no correlation for FDG. However, because both FDG-SUV_max and GTV are predictors of survival in NSCLC (24–26) and the amount of tumor hypoxia is related to outcome after radiotherapy (27), the reported correlation between hypoxia and FDG-PET is plausible.

Information about hypoxia on a tumor level can be used in clinical practice to select patients who may benefit from hypoxia modification before or during anticancer treatment. Previous studies have shown that the addition of hypoxic modification during radiotherapy results in an increased therapeutic benefit (28). Recently, Arrieta and colleagues (29) investigated in patients with NSCLC the use of nitroglycerin (an organic nitrate that causes vasodilatation, increased blow flow, and reduces the expression of HIF1-α) in combination with chemoradiation. In this study, promising response rates were observed; however, there was also (mild) increased toxicity (e.g., headache, hypotension) due to nitroglycerin administration. Another promising compound is the hypoxia-activated prodrug TH-302 that releases bromo isophosphoramide mustard, a potent DNA-alkylating agent, in hypoxic regions. Saggar and Tannock (30) recently demonstrated that TH-302 administered together with chemotherapy enhances the antitumor effect but also increases toxicity. From these recent studies, we acknowledge the therapeutic effect of additional antihypoxia treatment, but also the importance to limit unnecessary toxicity by selecting patients who will benefit from these modifications. We show that we can noninvasively visualize and quantify tumor hypoxia, using HX4-PET, in patients with NSLSC. In addition, our results show that patients with a larger tumor size and higher FDG uptake are more likely to have a larger hypoxic volume. This combination (GTV size and FDG uptake) could be used as a surrogate for hypoxia PET imaging; however, despite the correlation between hypoxia and FDG parameters, the distinction between hypoxic and nonhypoxic tumors based on FDG-PET can be misleading, because nonhypoxic tumors are present in a broad range of FDG uptake (FDG-SUV_max, 3.5–17.5 also shown in Fig. 1B).

It is important to note that a correlation at a global tumor level provides no information about the intratumoral heterogeneity. At the moment, limited data are available concerning the correlation of hypoxia PET and FDG-PET at a subvolume (e.g., voxel) level in NSCLC (7, 31). The spatial concordance and discordance of both PET modalities is of interest for radiotherapy boosting strategies. FDG-PET is already used in the context of clinical trials to boost highly metabolic tumor subvolumes (18, 32). We hypothesize that hypoxia PET imaging may be more selective in defining radioresistant voxels within the GTV, and can provide complementary information regarding the definition of radiotherapy boost volumes. A voxel-wise comparison was performed to evaluate the spatial distribution of the HX4 and FDG uptake. A reasonable correlation between both tracers was observed in the majority of patients. This is in contradiction to the previous published results of Bollineni and colleagues (7), who observed no correlation between FDG and the hypoxia PET tracer FAZA. This disagreement can probably be explained by the definition of the target lesion. Bollineni and colleagues used an FDG-based threshold to define the target lesion, thereby excluding voxels with a low FDG uptake. Conversely, Lohith and colleagues (31) reported a similar spatial distribution of [⁶²Cu]ATSM and FDG in 5 patients with an adenocarcinoma of the lung, which was not present in patients with SCC. Also, they observed a difference in intratumoral distribution between adenocarcinoma and SCC, which was not observed in our cohort of patients with NSCLC. It is well described that hypoxia leads to an increased uptake of glucose through various molecular mechanisms (33). Nevertheless, an increased glycolysis is also observed without hypoxia, e.g., by c-myc aberrations (34). From a molecular point of view, it is therefore logical that FDG uptake and hypoxia is partially overlapping and is highly dependent on the genetics of the tumor.

Thresholds were defined arbitrarily to define regions with a high or low uptake on both FDG- and HX4-PET. The high FDG-PET volume was defined on the basis of the ongoing NSCLC boost trial (18), whereas the high HX4 region was based on previous publications, indicating that a threshold of TBR>1.4 is rational to define hypoxia (6, 8, 35, 36). These thresholds showed the HX4-HV to be three times smaller on average than the FDG-HV.

This work can be used in clinical setting to divide patients with a hypoxic lesion into different groups, stratifying lesions with an agreement or disagreement between the HX4- and FDG-PET-uptake pattern. In the patients with a concordance, the use of HX4-PET has limited additional value for the selection of the radiotherapy boost volume; however, this volume could be limited to HX4-high areas only, facilitating further dose escalation without comprising the surrounding healthy tissue. In other patients, a (partial) discordance between the HX4- and FDG-PET-uptake pattern was observed. In these patients, the boost region could be adjusted to either HX4-PET or a combination of HX4- and FDG-PET with the aim to improve loco-regional control. On the basis of the current analysis, a radiation boost to the FDG-high area (SUV > 50% SUV_max) would on average miss 24% of the hypoxic volume, which seems in agreement with the residual activity after radiotherapy outside the high-FDG area as reported by Aerts and colleagues (11). Previous studies have already shown that radiotherapy dose distribution based on tumor hypoxia is possible and promising (37, 38). Currently, there are strategies available to investigate the original location of local recurrences inside the tumor volume (39). These studies will characterize the subvolumes inside the heteregenous tumor that are difficult to control. Ultimately, the effect of tumor subvolume characterization and targeting, by radiotherapy or other therapeutic interventions, needs to be assessed in a randomized trial.

This study has several limitations. First, most patients received chemotherapy before the start of radiotherapy and PET imaging. Chemotherapy can reduce the amount of tumor hypoxia and downregulates metabolism, resulting in a decreased uptake of HX4 and FDG (40). However, the focus of our research is on the correlation between both imaging modalities; therefore, treatment differences between patients are less relevant. In addition, it is most important to have recent PET information before the start of (adaptive) radiotherapy. Second, we were not able to validate the current imaging observations on tumor specimens. Nevertheless, van Baardwijk and colleagues (41) showed previously that FDG-PET imaging is correlated to GLUT-1 and HIF-1α expression in patients with NSCLC and Dubois and colleagues (8) showed a high correlation between HX4-PET uptake and pimonidazole staining in a rat rhabdomyosarcoma model. Third, the study acquired PET scans in free-breathing, which might cause blurring of the PET signal. Although, both the FDG and HX4 scans were obtained in this setting, we do not expect any substantial bias for the comparison. Furthermore, advanced-stage tumors are known to show little breathing-induced motion (42, 43). Fourth, the FDG-PET/CT was rigidly registered to the HX4-PET/CT scan to compare spatial uptake patterns. Small errors in registration can have a significant effect on correlation (44). However, patients in the current study were aligned in radiotherapy treatment position providing a strong basis for accurate registration. Fifth, there was a small time interval between the FDG- and HX4-PET/CT scan. Changes in anatomy, tumor metabolism, or hypoxia may have occurred in this interval and influenced the comparison results. The time interval in our study was short (median, 3 days) and no interventions (e.g., chemotherapy or radiotherapy) were performed between the two scans, limiting the chances of anatomic or physiologic changes. Finally, the usability of a tracer for radiation dose painting is dependent on its spatial reproducibility. Aerts and colleagues (45) showed that the location of low and high FDG volumes was stable during radiotherapy. The short-time reproducibility for HX4 (2 vs. 4 hours) was confirmed, but the long-term reproducibility is still unknown (6). However, a high reproducibility has been reported by Busk and colleagues (46) and Okamoto and colleagues (47) for the alternative hypoxia tracers FMISO and FAZA.

In conclusion, there is a positive correlation between GTV-, FDG-, and HX4-uptake parameters on a tumor level. The hypoxic tumor volume is on average three times smaller than the metabolic active tumor volume. Approximately half of the lesions showed a good spatial correlation between the PET tracers. In the other cases, a (partial) mismatch was observed. Hypoxia PET imaging gives complimentary information to metabolic FDG imaging, which can potentially be used to individualize patient treatment by selecting patients for treatment with hypoxic sensitizers or hypoxia PET-based radiotherapy dose escalation.

No potential conflicts of interest were disclosed.

Conception and design: C.M.L. Zegers, W. van Elmpt, B. Reymen, M.C. Öllers, F.M. Mottaghy, D. De Ruysscher, P. Lambin

Development of methodology: C.M.L. Zegers, W. van Elmpt, E.G.C. Troost, A.D. Windhorst, P. Lambin

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.M.L. Zegers, W. van Elmpt, B. Reymen, A.J.G. Even, F.J.P. Hoebers, J. Eriksson, A.D. Windhorst, F.M. Mottaghy, D. De Ruysscher

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.M.L. Zegers, W. van Elmpt, A.J.G. Even, E.G.C. Troost, M.C. Öllers, R.M.A. Houben, F.M. Mottaghy, P. Lambin

Writing, review, and/or revision of the manuscript: C.M.L. Zegers, W. van Elmpt, B. Reymen, A.J.G. Even, E.G.C. Troost, M.C. Öllers, F.J.P. Hoebers, R.M.A. Houben, J. Eriksson, A.D. Windhorst, F.M. Mottaghy, D. De Ruysscher, P. Lambin

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.M.L. Zegers, W. van Elmpt, J. Eriksson

Study supervision: W. van Elmpt, P. Lambin

The authors thank the patients who agreed to participate, C. Overhof for handling the data management for both clinical trials, and R. Franssen for the data acquisition.

This study was financially supported by the CTMM framework (AIRFORCE project, grant 030-103), the EU 6th and 7th framework program (METOXIA, EURECA, and ARTFORCE), euroCAT (IVA Interreg; www.eurocat.info), and the Kankeronderzoekfonds Limburg of the Health Foundation Limburg and the Dutch Cancer Society (KWF UM 2011-5020, KWF UM 2009-4454, KWF MAC 2011-4970, and KWF MAC 2013-6425).

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.