Molecular PET imaging for biology-guided adaptive radiotherapy of head and neck cancer

Abstract

Integration of molecular imaging PET techniques into therapy selection strategies and radiation treatment planning for head and neck squamous cell carcinoma (HNSCC) can serve several purposes. First, pre-treatment assessments can steer decisions about radiotherapy modifications or combinations with other modalities. Second, biology-based objective functions can be introduced to the radiation treatment planning process by co-registration of molecular imaging with planning computed tomography (CT) scans. Thus, customized heterogeneous dose distributions can be generated with escalated doses to tumor areas where radiotherapy resistance mechanisms are most prevalent. Third, monitoring of temporal and spatial variations in these radiotherapy resistance mechanisms early during the course of treatment can discriminate responders from non-responders. With such information available shortly after the start of treatment, modifications can be implemented or the radiation treatment plan can be adapted tailing the biological response pattern. Currently, these strategies are in various phases of clinical testing, mostly in single-center studies. Further validation in multicenter set-up is needed. Ultimately, this should result in availability for routine clinical practice requiring stable production and accessibility of tracers, reproducibility and standardization of imaging and analysis methods, as well as general availability of knowledge and expertise. Small studies employing adaptive radiotherapy based on functional dynamics and early response mechanisms demonstrate promising results. In this context, we focus this review on the widely used PET tracer ¹⁸F-FDG and PET tracers depicting hypoxia and proliferation; two well-known radiation resistance mechanisms.

Over the last decades, treatment modalities for locally advanced head and neck squamous cell carcinoma (HNSCC) have shifted from mainly surgical to radiotherapy, increasingly with the addition of systemic treatments such as chemotherapy or biologically modifying agents [1,2]. Intensity-modulated radiation therapy (IMRT) delivers high conformal doses and facilitates dose escalation to the tumor while reducing doses to normal tissues. Although treatment options have expanded, the locoregional recurrence rate is still relatively high and five-year survival rate usually below 50% [3]. New possibilities are needed to improve outcome for this patient group. Currently, treatment decisions are based on several patient – as well as tumor–parameters deduced from clinical and imaging diagnostics. Individual biological parameters are rarely taken into account, although these are dominant factors in the eventual tumor response to therapy. Also, corrections for morphological and biological tumor changes during (chemo)radiotherapy are seldom applied. In recent years, knowledge about predictive and prognostic biomarkers has increased. Using molecular and functional imaging techniques such as positron emission tomography (PET)-computed tomography (CT) scanning and magnetic resonance (MR) imaging, critical tumor characteristics such as metabolic activity, proliferation, hypoxia and vascularization can be assessed in a non-invasive manner. Definition of biological characteristics before and early during therapy may help to individually adapt and optimize treatment schedules for patients in order to procure a better prognosis and decrease treatment toxicity [4]. Information from imaging modalities can be combined to form a treatment plan, and monitoring of biological response can be used to adjust plans accordingly, instead of basing the entire schedule on a momentary situation before treatment. PET can be implemented to identify specific tumor (sub)volumes with increased radiation resistance that are to receive an escalated radiation dose, to investigate applicability of specific treatment alternatives, or possibly to ascertain cases where treatment de-escalation is an option.

In this review, we discuss developments in the field of molecular PET-CT imaging that can aid the improvement of radiotherapy delivery in HNSCC and thus the improvement of long-term outcome and reduction of toxicity. We focus on the widely used PET tracer ¹⁸F-FDG and PET tracers depicting hypoxia and proliferation, which are well-known mechanisms responsible for radiation resistance.

¹⁸F-FDG PET

PET with the glucose analogue 2-[¹⁸F] fluoro-2-deoxy-D-glucose (¹⁸F-FDG) is accepted as a powerful molecular imaging method exploiting the increased metabolic activity of cancer cells. Research is still focused on unearthing the molecular mechanisms underlying the cancer cells’ altered glucose metabolism [5]. Uptake of ¹⁸F-FDG has been assessed for correlation with several biological characteristics in tumors, such as glycolysis, glucose transporter-1 (GLUT-1) and hypoxic markers [6,7], proliferation [8], epithelial growth factor receptor (EGFR) [9], protein kinase B (AKT) [10,11] and combinations of several markers [12], with conflicting results. Overall, 18F-FDG uptake in malignancies reflects multifactorial mechanisms of increased metabolic activity and glucose utilization, performed by the glucose transporters and enzymes in the glycolytic pathway, which in turn are regulated by different signaling pathways triggered by endogenous and exogenous stimulators. Attempts to attribute 18F-FDG uptake in different malignancies to expression of one specific protein are therefore debatable.

In HNSCC, ¹⁸FDG-PET can lead to TNM stage differences and treatment strategy changes by detecting lymph nodes and distant metastases not discovered using other imaging modalities [13]. However, ¹⁸F-FDG PET requires cautious interpretation due to uptake in non-malignant tissues caused by peritumoral inflammation and physiologic changes in the head and neck region as well as limited sensitivity in evaluation of cervical node (micro)metastases [14,15]. Nonetheless, qualitative 18F-FDG PET is increasingly implemented before, during and after radiotherapy of HNSCC.

Target volume delineation for radiotherapy

The use of ¹⁸F-FDG PET for target volume delineation in radiotherapy planning for HNSCC has been investigated in single institution studies [13,16–25]. The simplest method for segmentation, visual interpretation of the PET tumor signal, has been commonly applied in many studies, but is highly operator dependent and influenced by window level settings [13,19,22,24]. Other investigators used fixed thresholds based on standardized uptake value (SUV) to segment PET tumor volumes, such as a SUV of 2.5 or 40%/50% of the maximum tumor intensity [23,26]. Models using a fixed (relative) threshold relying on SUV are questionable, notwithstanding tumoral uptake heterogeneity; the head and neck region contains several structures that tend to have a high physiological ¹⁸F-FDG uptake, such as the vocal cords and the tonsillar area, which can be erroneously included in the segmented area. Furthermore, SUV can show variation depending on the scanning protocol, blood glucose levels and other patient factors.

Various (semi-)automated PET segmentation algorithms have been proposed to reduce inter- observer variation. Several investigators used advanced adaptive relative threshold segmentation methods based on maximal tumor uptake and/or background uptake [17,18,25,27]. Other techniques have been introduced to refine HNSCC (semi-) automated segmentation, such as methods using image gradients to define tumor areas [28,29], halo-based contouring [21], graph-based segmentation using information from co-registered PET and CT together [30], or an adaptive region growing/dual front active contour model [31]. Several studies addressed impact to the gross tumor volume (GTV) when using ¹⁸F-FDG PET segmentation alongside CT images; reported changes mostly concerned a decrease in GTV, especially for the more sophisticated segmentation methods [13,17–21,23,24].

A comparative study by Schinagl et al. [32] of five different segmentation methods reported that PET-segmented volumes frequently extended outside CT delineations, while being smaller on average. Few groups have validated HNSCC delineation using different imaging modalities against surgical resection specimens. In these studies, although describing various PET segmentation methods, ¹⁸F-FDG PET-defined GTVs were better related to surgical resection specimens than CT- or MRI- defined GTVs [17,33], but still did not encompass the entire pathological GTV. A gradient-based method using the watershed transform and hierarchical cluster analysis, developed by Geets et al. [20], outperformed the adaptive signal-to-background ratio thresholding method by overestimating macroscopic pathological tumor volume by only 20% instead of 68%. Zaidi et al. [34] tested nine different segmentation methods against surgical specimens and found the best performance (i.e. underestimation of the actual volume by an average of 6%) for a spatial wavelet-based algorithm, which incorporates spatial information during the segmentation process. As more specialized segmentation tools that exploit the differences in metabolism between tumors and surrounding tissues are being developed, the use of ¹⁸F-FDG PET for correct primary tumor delineation alongside CT/MRI diagnostics can become a substantial asset. This not only holds true for primary tumors, but also for the correct identification and delineation of lymph node metastases [35]. Furthermore, ¹⁸F-FDG PET directed dose distribution could lead to better sparing of organs at risk, such as the parotid glands [18]. However, it is imperative that validated independent and robust methods, which of yet seem to function in highly specialized study settings, become readily available for the clinical practice.

IMRT planning and adaptive radiotherapy based on ¹⁸F-FDG PET

Integration of ¹⁸F-FDG PET in IMRT planning has been described as beneficial for treatment individualization and dose escalation [36–38]. Groups have reported two-year overall survival of 80–90% and locoregional control of 70–80% after implementation of ⁸F-FDG PET-CT-based IMRT [39,40]. In a study of 10 HNSCC patients who underwent five ¹⁸F-FDG PET-CT scans before and during therapy, Castadot et al. [29] found not only changes in the volume, but also in the position of target volumes and organs at risk during concomitant chemoradiotherapy. This may give rise to adaptive strategies, where patients are re-imaged and re-planned during therapy. However, implementation of adaptive schemes may not be straightforward. Automated delineation methods based on signal-to-background ratios may erroneously expand PET tumor volumes far beyond the actual tumor area during (chemo)therapy, because of decreasing SUV values in tumors and increasing background signals due to induced inflammation during therapy [41,42]. The use of an adequate delineation method during treatment is mandatory for planning adaptation. Geets et al. [28] found that, using a gradient-based algorithm on five pre- and per-treatment ¹⁸F-FDG PET scans in 10 HNSCC patients treated with chemoradiotherapy, PET-segmented target volumes reduced significantly during treatment. Adaptive IMRT lead to a decrease in the high-dose volumes compared to pre-treatment CT planning, with little further impact on selected organs at risk, proving this approach useful for dose escalation schemes.

Another concept in the range of PET-based IMRT planning is theragnostic “dose painting by numbers”, where voxel-wise dose prescription and escalation is related to PET tracer uptake to procure a non-uniform radiation dose distribution [43]. A phase I trial at the University of Ghent used adaptive IMRT planning based on dose painting by numbers according to ¹⁸F-FDG PET voxel intensities [44]. Median total doses of 80.9 Gy or 85.9 Gy, in a total of 32 fractions, were planned to the high-dose target volumes remaining after 20 fractions. The researchers concluded that although treatment to the 85.9 Gy dose level is feasible, development of late onset mucosal ulcers designated the 80.9 Gy dose level as maximum tolerated dose. Per-treatment re-planning can be beneficial, as described in a prospective trial involving adaptive CT-based IMRT planning mid-treatment [45]. A dosimetric benefit and no negative effects on outcome events were found in 22 patients after a median follow-up of 31 months. The aforementioned studies show that treatment plans can be adapted during therapy, following the metabolic response pattern.

Prognostic value of ¹⁸F-FDG PET before, during and after radiotherapy in HNSCC

Increasing numbers of treatment centers use ¹⁸F-FDG PET scans for routine diagnostic staging in HNSCC patients. Investigators have looked at the prognostic value of such imaging. Studies in often non-uniformly treated cohorts of HNSCC patients reported that a high pre-treatment ¹⁸F-FDG uptake was associated with poor outcome [46–49]. A recent prospective study, however, did not report an overall consistent prognostic PET-parameter (SUV- or metabolic volume-based) for tumors within different regions of the head and neck area in 77 patients treated with (chemo)radiotherapy [50].

Others have focused on the ability of ¹⁸F-FDG PET-CT to provide prognostic information and to serve as a tool for treatment response assessment after completion of (chemo)radiotherapy [51]. A meta-analysis concluded that best accuracy in detecting residual or recurrent disease is achieved three months after completion of therapy [52]. ¹⁸F-FDG PET could be used for decisions on salvage neck-nodal dissections after (chemo)radiotherapy [53]. However, post-therapy assessment is useless when aiming for early treatment modification to improve outcome or reduce overtreatment. Brun et al. [54] reported more complete remissions and better five-year overall survival in HNSCC patients with tumors showing a low metabolic rate on ¹⁸F-FDG PET scans performed 5–10 days after one cycle of neoadjuvant chemotherapy (n = 10) or after a median of 24 Gy radiotherapy (n = 37). One recent study in 37 HNSCC patients treated with chemoradiotherapy reported superior two-year overall survival and locoregional control for patients with a decrease in ¹⁸F-FDG PET SUV_max of 50% or more after 10–20 Gy compared to patients who had a lesser decrease [55]. Conversely, another group did not observe a correlation between reduction of ¹⁸F-FDG uptake after two weeks of chemoradiotherapy and outcome in 26 patients [56]. They found a prognostic value for ¹⁸F-FDG PET-CT performed 8–12 weeks after therapy with regard to disease specific survival and relapse free survival. The prognostic value of ¹⁸F-FDG PET before or during treatment therefore remains debated.

In conclusion, information from widely available ¹⁸F-FDG PET can complement other diagnostic modalities for treatment decisions and guidance of radiotherapy planning, but it cannot replace clinical examination or CT/MRI scans to obtain important details. Defining the primary tumor boundaries with PET is difficult. For example, it seems impossible to define superficial tumor spread as is mostly found by clinical examination [17,24]. To assess invasion of tumor surrounding tissues, it seems best to use the combined qualities of fused PET-MRI scans [57]. Tumor limits can be misconstrued due to ¹⁸F-FDG uptake in surrounding non-malignant tissue or due to a decrease in tumor-to-background ratio during therapy. Highly specialized segmentation methods seem to comply best with histological resection specimens as compared to CT or MRI in a few small studies, and can result in an accurate reduction of GTV and in sparing of normal tissues for radiotherapy planning. Larger multi-institutional studies should generate the most robust imaging quality and segmentation methods [58–60]. Adaptive IMRT planning per-treatment is possible with ¹⁸F-FDG PET determination of metabolic HNSCC activity, but more knowledge is needed on the potential volume and spatial shift in ¹⁸F-FDG uptake during therapy and its correlation with actual tumor activity as opposed to an inflammatory reaction. Also, more should be known about the relation of ¹⁸F-FDG uptake with the site of HNSCC tumor recurrence, which is situated predominantly but not exclusively inside the pretreatment PET-derived target volume in small retrospective series [38,61]. The prognostic value of ¹⁸F-FDG PET before and early during treatment may be a valuable asset in assigning and redirecting therapy, but the first outcomes are derived from small heterogeneous non-randomized studies that show conflicting results and have not produced validated and applicable cut-off values for the clinic.

Hypoxia

Hypoxia is an important mechanism of radioresistance in HNSCC [62]. It also impacts on the tumor micro-environment by stimulating angiogenesis and metastatic potential [63]. A high hypoxic fraction in HNSCC is associated with a decreased clinical outcome after radiotherapy [64,65]. Several options to modify hypoxia have been proven successful in improving therapy outcomes, such as accelerated radiotherapy with carbogen and nicotinamide (ARCON) [66] and the hypoxic radiosensitizer nimorazole [67]. Also, dose escalation can be applied to the hypoxic tumor regions. A recent meta-analysis showed an overall beneficial effect of the addition of hypoxic modification to radiotherapy of HNSCC [68].

Hypoxia can be measured by invasive methods such as polarography electrodes or immunohistochemical staining for hypoxia-related markers in tumor biopsies [69]. These methods are prone to sampling errors and have a limited use because of their invasive and technically demanding nature. Biopsies only represent a fraction of the total tumor volume. The extent of hypoxia can vary widely between but also within head and neck tumors. Also, results of hypoxia-modification within individual tumors cannot be monitored rapidly using repeated invasive procedures. Non-invasive hypoxia imaging can provide an attractive substitute. In the field of PET imaging, multiple hypoxia-related markers have been tested during the years. Initially ¹⁸F-FDG was thought to be a hypoxic marker, since hypoxic (tumor) cells display an elevated glycolytic activity. However, this depends on the type of glucose metabolism preferred by a tumor; since most HNSCC display an aerobic glycolysis with high glucose utilization even under well-oxygenated circumstances, ¹⁸F-FDG uptake is not specifically related to the level of hypoxia [70,71].

Nitroimidazole-based PET tracers

Nitroimidazole-based compounds such as misonidazole and pimonidazole are exogenous markers that selectively bind to hypoxic cells after administration and are frequently used for immunohistochemical staining of tumor sections. Labeled nitroimidazoles can be used as PET tracers of hypoxia. The first and best-known is [¹⁸F] Fluoromisonidazole (¹⁸F-FMISO) [72]. In 10 HNSCC xenograft models, autoradiography of ¹⁸F-FMISO compared with the fraction of pimonidazole-stained tumor on immunohistochemistry revealed that ¹⁸F-FMISO accumulation depended on the type of hypoxia distribution pattern; ribbon-like, patchy or homogeneous, with the highest correlation found in ribbon-like hypoxia [73]. Studies evaluating ¹⁸F-FMISO uptake in cervical lymph node metastases displayed conflicting correlations with pO₂ polarography measurements [74,75]. Several authors stress the methodological limitations of ¹⁸F-FMISO and other nitroimidazole-based PET tracers. ¹⁸F-FMISO is reduced under hypoxic conditions by intracellular nitroreductase enzymes and bound to cellular macromolecules. ¹⁸F-FMISO displays a slow washout of unbound tracer from background tissues, creating a relatively low contrast in images. The spatial separation of hypoxic cells from perfused vessels results in long diffusion times for the tracer to reach hypoxic areas. Spatial information in small tumors is hampered by the inherent PET resolution and by the fact that tracer uptake only occurs in viable hypoxic cells often constituting a small subpopulation of the tumor. There has been debate regarding the ideal imaging time of ¹⁸F-FMISO, but overall reliable imaging can take place from two hours after injection onwards, when normal tissues have equilibrated with plasma and hypoxic tissues still show retention of ¹⁸F-FMISO [76] (Figure 1). SUV-defined contrast between HNSCC and background tissues is likely optimal four hours after injection [77], but kinetic modeling can make early quantification more reliable, more appropriate for heterogeneous tumors and therefore more adjusted for individually adapted treatment planning [78]. In order to procure a better contrast between tumor and background, efforts have been made to find nitroimidazole compound tracers with a faster diffusion in tumors as well as a faster whole body clearance than FMISO. Studies focused on more lipophilic nitroimidazoles, such as EF3 and EF5 based on the radiosensitizer etanidazole, resulted in tracers with a good penetration and diffusion in tumors but a simultaneous limited clearance and therefore a varying potential in tumor detection [79,80].

Figure 1. ¹⁸F-FMISO PET-CT scan of a cT4N0M0 oral cavity tumor 30 min after injection with diffuse uptake in tumor and normal tissues (A–C) and 3 hours after injection with specific tumor retention (D–F). Transversal (A + D), coronal (B + E) and sagittal (C + F) planes.

A more lipophilic and consequentially more rapidly clearing nitroimidazole compound, ¹⁸F- fluoroazomycin arabinoside (¹⁸F-FAZA), displayed a good correlation between PET/autoradiography uptake and pimonidazole and Hoechst (perfusion marker) immunohistochemical staining in xenograft models [81]. In two clinical pilot studies, seven of 11 and six of nine HNSCC patients, respectively, showed adequate tumor PET-imaged uptake of ¹⁸F-FAZA two hours after injection [82,83]. In the DAHANCA 24 study, ¹⁸F-FAZA PET-CT scans prior to and during (chemo)radiotherapy combined with nimorazole, resulted in an identifiable hypoxic tumor volume in 25 of 40 and six of 13 HNSCC patients, respectively [84]. If a hypoxic tumor volume was discernible on the scan procured during therapy, the location was similar but the total volume smaller compared to the pre-therapy scan. During a median follow-up of 19 months there was a significant better disease free survival for patients with non-hypoxic tumors on ¹⁸F-FAZA PET-CT than for patients with hypoxic tumors; 93% versus 60%, respectively (p = 0.04).

Another nitroimidazole-based tracer with better water solubility than ¹⁸F-FMISO, ¹⁸F-HX4 (3-[¹⁸F]fluoro-2-(4-((2-nitro-1H-imidazol-1-yl)methyl)-1H-1,2,3,-triazol-1-yl)-propan-1-ol), based on the 2-nitroimidazole pharmacophore, is currently undergoing clinical evaluation. In eight of 12 HNSCC patients, Cheng et al. [85] found similar tumor-to-muscle ratios on ¹⁸F-HX4 PET scan performed 1.5 hours after injection compared to ¹⁸F-FMISO scans performed two hours after injection on the following day. Two lesions showed no tracer uptake, while two tumors only showed uptake of either ¹⁸F-HX4 or ¹⁸F-FMISO.

One study describes dynamic PET imaging with the fluorinated nitroimidazole ¹⁸F-labeled fluoroerythronitroimidazole (FETNIM) [86].

Other hypoxia PET tracers

An alternative lipophilic PET tracer of hypoxia is based on a metal complex of radioactive copper with ATSM, diacetyl-bis(N4-methylthiosemicarbazone) [87]. The precise mechanisms of hypoxic cell retention of Cu-ATSM and other processes affecting Cu-ATSM stability are unclear [88]. While a clinical study supported the potential of ⁽⁶²⁾Cu-ATSM as prognostic marker of radiotherapy outcome [89], additional investigations need to establish Cu- ATSM as a specific marker of hypoxia.

Quantifiable PET imaging of endogenous hypoxia-associated markers, such as the membrane protein carbonic anhydrase IX that is upregulated by HIF-1α under hypoxic conditions, is feasible using specific monoclonal antibody (fragments) such as G250 [90,91]. Even though endogenous markers are attractive for use in the clinic, no clinical studies have been undertaken as of yet.

Hypoxia instability

One issue influencing hypoxia imaging with any PET tracer is the instability of hypoxia in HNSCC. A shift in hypoxic regions of head and neck tumors develops over time, with life-times of hypoxic cells spanning from several hours to multiple days [92]. It is argued that the dynamic cellular and micro-environmental processes influencing hypoxia in tumor areas dictate the uptake and retention kinetics of different PET hypoxia tracers, with a highly variable non-linear curve after administration followed by an approximately linearly sloped signal [93]. Repeated ¹⁸F-FMISO PET imaging performed three days apart and a mean 162 minutes (range 117–195) after injection in 13 HNSCC patients showed variability in spatial tracer uptake – only six patients showed well-correlated tracer uptake distribution [94]. However, a recent study showed highly stable hypoxic tumor areas in 10 of 11 HNSCC patients, on ¹⁸F-FMISO PET scans performed four hours after injection with a two-day interval [95]. Multiple effects, such as a more stabilized tumor- to-blood ratio four hours after injection as opposed to 2–3 hours, a longer time-span between the repeated scans of the former study and differences in imaging- and reconstruction-protocols, might partly explain these outcome differences. Nevertheless, hypoxia regions may show fluctuation during a radiotherapy course and this should be taken into account if hypoxic tumor volumes are utilized for radiotherapy planning.

It is hard to argue a preference for any specific hypoxia (-related) PET tracer for HNSCC based on the currently available data. Most knowledge has been gathered concerning the characteristics of ¹⁸F-FMISO. However, more extensive clinical testing might designate another nitroimidazole-based compound, such as ¹⁸F-FAZA or ¹⁸F-HX4, as the most convenient and reliable tracer in terms of uptake and clearance, earliest post-injection time for imaging, simplicity and cost of production and general applicability in multi-institutional settings. This is unclear as of yet and efforts should focus on defining the best tracer at hand instead of developing yet another hypoxic marker in small (pre-)clinical settings.

Hypoxia PET for radiotherapy planning

Hypoxia PET imaging can be incorporated into the radiotherapy planning process to apply a radiotherapy boost for focal hypoxia. In in silico studies, Chao et al. [96] and Dalah et al. [97] demonstrated that IMRT dose to hypoxic regions defined by high ⁶⁰Cu-ATSM or ⁶⁴Cu-ATSM uptake in head and neck tumors could be escalated while normal tissues were spared. The concept of ¹⁸F-FMISO or ¹⁸F-FAZA PET-guided dose escalation in HNSCC using IMRT was applied by multiple groups [98–104]. Boost doses up to 84 Gy could be achieved without exceeding normal tissue tolerance in silico (however, as mentioned before, 80.9 Gy seems to be the maximum tolerated dose to avoid late-onset mucosal ulcers inside the GTV area [44]). The hypoxic volume demarcation was determined using set tumor uptake/background tissue ratio thresholds or by visual contouring. Grosu et al. [100] also escalated doses to ¹⁸F-FAZA positive cervical node subvolumes, but found that specific demarcation was not feasible in 39% of patients when hypoxic areas were diffusely distributed. Most planning studies relied on the assumption of hypoxic tumor region stability, which may not reflect the actual situation. Lin et al. [105] observed that IMRT planned boost doses of 84 Gy on ¹⁸F-FMISO positive tumor areas did not cover dissimilar hypoxic areas on PET images obtained three days later (before start of treatment) in four of seven HNSCC patients; the average equivalent uniform dose to the hypoxic tumor volume decreased from 87 Gy to 80 Gy.

Dose painting by numbers according to PET tracer uptake level in tumors in different head and neck areas was technically feasible using ¹⁸F-FMISO or ⁶¹Cu-ATSM PET images [106–108]. A kinetic model based on repeated ¹⁸F-FMISO imaging during radiotherapy indicated that individual re- oxygenation times are linked to tumor control probability (TCP) [109]. Two studies incorporated models into the radiotherapy planning system that could also adjust for tumor re-oxygenation during therapy. Thorwarth et al. [106] estimated that TCP would increase from 56% to 70% using dose painting by numbers based on hypoxia PET. Toma-Dasu et al. [107] proposed a dose-painting model that calculated prescription doses in the events of static or dynamic oxygenation status in tumors during therapy. Tumors with low and homogeneous tracer uptake could theoretically be controlled by prescription doses between 64 and 76 Gy. However, for hypoxic tumors with heterogeneous uptake, doses up to 121 Gy would be required to gain 95% TCP in the event of static oxygenation, which is not an attainable goal. When oxygenation dynamics were incorporated into the model, the 95% TCP dose for the same tumor would be 77 Gy, which is closer to doses used in current practice.

The common limitation in these studies is that they were planning exercises that were not actually delivered to patients.

Outcome prediction using hypoxia PET

The relevance of hypoxia for outcome prediction of advanced HNSCC has been demonstrated in several imaging studies. Elevated pre-therapy ¹⁸F-FMISO uptake is related to a worse prognosis for patients treated with (chemo)radiotherapy [110–113]. More hypoxic tumors could therefore be candidates for intensified treatment protocols based on baseline values. But, as mentioned before, hypoxia is a dynamic process and therapy induces re-oxygenation of HNSCC. If partial or total re-oxygenation takes place during therapy, a single pre-therapy hypoxia measurement will be insufficient to allocate or continue hypoxia targeting modalities. Lee et al. [114] found ¹⁸F-FMISO uptake in 18 of 20 stage III-IV HNSCC patients before chemoradiotherapy. In the fourth week of treatment, only two patients showed residual hypoxia, but regional/distant recurrence occurred in another patient during follow-up. In 29 patients with repeated ¹⁸F-FMISO scans during chemoradiotherapy with or without tirapazamine, all six patients with residual ¹⁸F-FMISO uptake after four weeks displayed locoregional or distant failure during follow-up [112]. Zips et al. [115] reported that, in a prospective cohort of 25 patients, ¹⁸F-FMISO imaging parameters after 1–2 weeks of chemoradiotherapy provided stronger prognostic potential for local recurrence than pre-treatment parameters. In the ¹⁸F-FMISO imaging substudy of a phase II randomized trial that randomized between concurrent chemoradiotherapy alone or combined with tirapazamine as hypoxic cytotoxin, patients with hypoxic tumors showed less locoregional failure in the tirapazamine group than in the chemoradiotherapy-only group [112]. Absence of hypoxia was associated with low risk of locoregional failure in the group treated with chemoradiotherapy alone, suggesting that this group would not benefit from more intensive therapy. This indicates that ¹⁸F-FMISO could identify patients who are most likely to benefit from a tirapazamine containing chemoradiotherapy regimen. This notion is underlined by the outcome of a phase III trial analyzing chemoradiotherapy versus chemoradiotherapy with tirapazamine in patients unselected for hypoxia, where addition of tirapazamine to chemoradiotherapy did not improve outcome [116].

In conclusion, non-invasive imaging of the therapy resistance factor hypoxia is feasible and achieved with multiple PET tracers displaying specific accumulation in hypoxic HNSCC areas. ¹⁸F-FMISO has been evaluated most extensively and seems to represent hypoxic subvolumes within HNSCC tumors adequately, but no widely applicable quantification and evaluation methods are available for the clinical practice as of yet. Several hypoxia-related PET tracers have been applied for in silico radiotherapy dose escalation planning. The hypoxic tumor subvolume provides a basis for radiotherapy boosting, but small though relevant hypoxic volumes or patterns can remain unnoticed due to the limited PET spatial resolution. Furthermore, reliable dose escalation to resistant tumor regions requires repetitive hypoxia PET imaging during therapy due to fluctuation in oxygenation status, and it is still uncertain which dose levels are necessary to eradicate hypoxic subpopulations. Hypoxia PET results have been shown to have a prognostic as well as a predictive value in small prospective series. This provides a basis for further studies allocating hypoxia modifying treatment according to hypoxia status.

Proliferation

In HNSCC, enhanced proliferative activity and compensatory tumor cell proliferation during treatment adversely affect outcome [117]. Various treatment regimens have been developed to counteract this effect, such as accelerated radiotherapy [118,119], chemoradiotherapy [1], or radiotherapy combined with cetuximab [120], but these approaches also increase side effects [121]. PET monitoring of proliferative activity of tumors before and during treatment may potentially assist in better patient selection and in treatment strategy modification based on early response assessment. PET of proliferation has focused on imaging of thymidine analogue tracers. Thymidine is a native nucleoside, which is incorporated into deoxyribonucleic acid (DNA). Shields et al. [122] introduced the thymidine analogue 3’-deoxy-3’-[¹⁸F] fluorothymidine (¹⁸F-FLT) as PET tracer, exploiting the activity of the enzyme thymidine kinase 1 (TK1) as measure of proliferative activity. ¹⁸F-FLT is phosphorylated by TK1 and trapped intracellularly [122]. During DNA synthesis, TK1 activity increases almost tenfold. ¹⁸F-FLT trapping is related to TK1 activity and closely associated with proliferative activity [123]. ¹⁸F-FLT has been validated against histopathology in a variety of tumor types [124–126]. In a study of 17 HNSCC patients, ¹⁸F-FLT PET SUV_max and SUV_mean could not or only weakly be correlated with immunohistochemical staining for proliferation-related markers and TK1 in resected tumor sections [127]. The discrepancy might have been due to differences in biomarker characteristics, discrepancies in resolution of the imaging modalities, and differences in quantification methods. Troost et al. [128] found that ¹⁸F-FLT PET in 10 HNSCC patients did not discriminate between metastatic and reactive lymph nodes, since the latter also displayed reactive B-lymphocyte proliferation.

A clinical study of De Langen et al. [129] exhibited the reproducibility of quantitative ¹⁸F-FLT PET measurements. Pre-clinical and clinical studies confirmed that radiotherapy reduced ¹⁸F-FLT uptake in head and neck tumors early, while no apparent changes in tumor size or morphology could be noted [130,131]. Menda et al. [132] reported kinetic ¹⁸F-FLT PET analysis of eight HNSCC patients at baseline and after five days of chemoradiotherapy. Uptake in tumor and metastatic lymph nodes showed a significant decrease after five days of treatment relative to baseline. ¹⁸F-FLT PET response was also noted after induction cetuximab in a pilot study with six patients [133]. An example of reduction of ¹⁸F-FLT uptake following cetuximab and radiotherapy is shown in Figure 2.

Figure 2. ¹⁸F-FLT PET-CT scans of a patient with a cT4N2M0 hypopharyngeal tumor before therapy (A), after induction cetuximab (B), after 1 week of radiotherapy (two doses of cetuximab, 10 Gy) (C) and after 3 weeks of radiotherapy (four doses of cetuximab, 32 Gy) (D). Note the concurrent reduction in cervical vertebra marrow proliferation after start of radiotherapy.

Troost et al. [131] demonstrated that high proliferative tumor subvolumes, as defined by ¹⁸F-FLT PET, can provide the basis for an IMRT plan with dose escalation within these regions (Figure 3). In this study concerning 10 oropharyngeal carcinoma patients, repeated ¹⁸F-FLT PET indicated that the highly proliferative volumes (delineated by 80% of SUV_max) displayed moderately stable spatial similarity between baseline PET and PET after one week of (chemo)radiotherapy, but large inter-individual differences occurred. Although SUV_max decreased significantly between scans as well, which might have influenced the segmentation method, this indicates that therapy induces spatial instability in proliferative subvolumes, similar to hypoxic subvolumes.

Figure 3. Dose-escalation to GTV_80%1 (as delineated using a cut-off of 80% of maximum tumor uptake signal in a ¹⁸F-FLT PET-CT scan performed before start of radiotherapy) and GTV_80%2 (as delineated in a ¹⁸F-FLT PET-CT scan performed after 1 week of radiotherapy) for a cT3N0M0 oropharyngeal tumor. Using IMRT with integrated simultaneous boost technique, total dose was 50.3 Gy to bilateral cervical lymph node regions (large planning target volume, red) and 68 Gy to primary tumor (small planning target volume, blue); in 34 fractions. GTV_80%1 (black) and GTV_80%2 (green) were consecutively irradiated with 2.3 Gy for 10 fractions, resulting in a dose of 71 Gy in total and a dose of 74 Gy in the overlapping region. (A and B) Dose distributions for first 2 weeks of treatment (A) and weeks 3 and 4 (B). (C) Dose distribution for remaining 14 fractions without dose-escalation. (D and E) Dose distributions of total treatment plan in transverse (D) and sagittal (E) planes. Parotid glands are delineated in sky blue and spinal cord in green. This figure was originally published in JNM. Troost EG, Bussink J, Hoffmann AL, Boerman OC, Oyen WJ, Kaanders JH. 18F-FLT PET/CT for early response monitoring and dose-escalation in oropharyngeal tumors. J Nucl Med 2010;51:866–74. © by the Society of Nuclear Medicine and Molecular Imaging, Inc.

Repeated ¹⁸F-FLT PET in HNSCC has been demonstrated to correlate with treatment outcome [47,134]. In a study of 48 HNSCC patients treated with (chemo)radiotherapy, Hoeben et al. [134] reported early ¹⁸F-FLT PET response between baseline and after one and three weeks of therapy. SUV_max and visually contoured proliferative volumes of the primary tumors decreased significantly between consecutive scans, while signal-to-background and 50%-of-maximum signal segmentation methods failed to delineate plausible proliferative tumor volumes as the ¹⁸F-FLT tumor uptake signal reduced during therapy. Baseline tumor SUV_max and visually delineated proliferative volume, as well as their decrease early during therapy, were prognostic for three-year locoregional control and disease free survival. Kishino et al. [135] performed ¹⁸F-FLT PET and ¹⁸F-FDG PET scans before therapy, after four weeks of radiotherapy and five weeks after completion of therapy in 28 patients. ¹⁸F-FLT decreased most significantly during radiotherapy, with a complete response in 63% of patients compared to 16% on ¹⁸F-FDG PET. Patients with residual post-treatment activity on either modality had significantly worse three-year local control than patients with a complete response. Specificity and overall accuracy of ¹⁸F-FLT PET were significantly higher than those of ¹⁸F-FDG PET (i.e. 72% and 74% vs. 19% and 30% after 40 Gy radiotherapy, respectively, p < 0.0001; and 80% and 81% vs. 48% and 57% five weeks after radiotherapy, respectively, p < 0.01). In a follow-up report, no clinical recurrences were diagnosed in patients showing a primary tumor SUV ratio < 1.5 for the mid-treatment versus post-treatment scan, while no such correlation was found for ¹⁸F-FDG PET [136].

The high specificity of ¹⁸F-FLT PET, not influenced by inflammatory processes in the primary tumor like ¹⁸F-FDG, and its ability to characterize proliferation before and early during therapy, make it an attractive tracer in the development of individualized HNSCC patient management. It could help clinical judgments concerning the addition of systemic therapy or targeted agents to radiotherapy or concerning the application of accelerated radiotherapy. However, more detailed fundamental knowledge on ¹⁸F-FLT uptake in different tumor types is needed; e.g. the extent to which tumors depend on a de novo thymidine synthesis pathway as opposed to on a thymidine salvage pathway influences the degree in which ¹⁸F-FLT uptake represents actual proliferative activity [137]. More information regarding the extent of region fluctuation of proliferation during therapy, as described by Troost et al. [131], is of essence if ¹⁸F-FLT PET were to be used for (adaptive) boost localization. Furthermore, the prognostic and predictive value in different studies has not been translated into applicable quantifiers to be used for treatment allocation or patient risk group stratification in prospective studies.

Conclusions and future perspectives

The material presented in this review exemplifies a large basis for the implementation of molecular PET imaging in the management of HNSCC patients. However, this basis is multi-faceted and reliant on numerous heterogeneous small studies reporting on a range of tracers that are diversely applied, with varying analytic methodology for divergent research questions. PET tracers imaging distinct biological tumor characteristics offer specific possibilities for individualized therapy. Dirix et al. [138] even reported the added value of combining ¹⁸F-FDG PET, ¹⁸F-FMISO PET, diffusion weighted MRI and dynamic contrast-enhanced MRI before and during chemoradiotherapy for radiotherapy planning, early response assessment and prognosis prediction in 15 patients. However, it is hardly feasible to perform standard multiple molecular/functional imaging modalities alongside the routine imaging modalities in HNSCC patients. A priori research questions should be clearly formulated and HNSCC patient care should be centralized to prevent an inexorable expansion of diagnostic and therapeutic procedures that do not adhere to evidence-based protocols.

Before implementation of molecular PET imaging in the clinical practice, several issues need to be addressed. The procedures for acquiring and processing PET have to be standardized before insertion into radiotherapy protocols [139]. If repeated imaging before and during therapy is warranted, efforts should be taken to assure patient comfort and reproducibility. Random set-up errors in patient or tumor position should be kept to a minimum [140].

Additionally, there is the issue of image resolution; a PET voxel size of approximately 4 × 4 mm cannot be optimally matched to the biological processes on a microscopic level [141]. PET spatial resolution is also diminished by the physical characteristics of the positron emitter, the inherent positron range of selected tracers and by blurring and partial volume effects. It may be difficult to distinguish heterogeneous uptake areas within tumors, certainly if signal-to-background ratios are low. However, any issues regarding spatial resolution of PET are counterbalanced by the also limited precision for dose calculation and delivery over a number of radiotherapy fractions. Such matters are certainly critical obstacles in the dependable implementation of, e.g., precision-based dose painting by numbers. Regarding the reliable delineation of target volumes, additional efforts are required. There is a definite need for a validated, reliable and robust delineation method that can be widely applied before and during therapy, applicable to multiple PET systems and tracers. A strong collaboration of radiotherapy and nuclear medicine departments in broader settings can achieve the common goal of finding an optimal method in shared large, high-quality validation sets as suggested by Lee [59].

Implementation of PET for tailoring of radiotherapy-based treatment and ultimately for improving HNSCC patients’ outcome will be feasible in the future, but significant hurdles remain to be taken.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.