Role of fluorine-18 fluorodeoxyglucose PET/CT in head and neck oncology: the point of view of the radiation oncologist

Squamous cell carcinoma is the most common malignant tumour of the head and neck. The initial TNM staging, the evaluation of the tumour response during treatment, and the long-term surveillance are crucial moments in the approach to head and neck squamous cell carcinoma (HNSCC). Thus, at each of these moments, the choice of the best diagnostic tool providing the more precise and larger information is crucial. Positron emission tomography with fluorine-18 fludeoxyglucose integrated with CT ( 18 F-FDG-PET/CT) rapidly gained clinical acceptance, and it has become an important imaging tool in routine clinical oncology. However, controversial data are currently available, for example, on the role of 18 F-FDG-PET/CT imaging during radiotherapy planning, the prognostic value or its real clinical impact on treatment decisions. In this article, the role of 18 F-FDG-PET/CT imaging in HNSCC during pre-treatment staging, radiotherapy planning, treatment response assessment, prognosis and follow-up is reviewed focusing on current evidence and controversial issues. A proposal on how to integrate 18 F-FDG-PET/CT in daily clinical practice is also described.


INTRODUCTION
Head and neck squamous cell carcinoma (HNSCC) is the most common malignant tumour of the head and neck (HN). 1 Radiotherapy has a well-established role both in the exclusive and in the adjuvant setting. 2,3 Initial diagnosis and staging of HNSCC is based on physical examination, chest imaging, HN endoscopy, and HN CT or MRI. Clinical guidelines for HNSCC recommend different imaging approaches for each phase of disease. 2-4 Moreover, modern imaging modalities have an essential role in the tumour response after treatment and follow-up. [5][6][7] Each of the currently available imaging techniques present different levels of sensibility and specificity, and it is essential for the radiation oncologist to choose the better one, depending on the clinical scenario.
Positron emission tomography with fluorine-18 fludeoxyglucose integrated with CT ( 18 F-FDG-PET/CT) rapidly gained clinical acceptance and has become an important tool in routine clinical HN oncology. According to the National Comprehensive Cancer Network guidelines, PET or PET/CT is suggested (individualized cases) for Stage III (T3, N0, M0 or T1-3N1M0) and Stage IV (T1-T4, N0-N3, M0-M1) due to the possibility of stage migration. 2 The Ontario guidelines suggest that 18 F-FDG-PET/CT is also indicated when the primary site is unknown or in the staging of locally advanced disease. 4 It is noteworthy that not all HN guidelines agree 8  The aim of this review is to explore the most important available literature dealing with the role of 18 F-FDG-PET/CT imaging in HNSCC in pre-treatment staging, radiotherapy planning, treatment response assessment, prognosis and follow-up. Evidences and controversies have been summarized, with particular attention to the point of view of the radiation oncologist.

METHODS AND MATERIALS
We performed a comprehensive literature search of the MED-LINE database without any limits to identify relevant studies (published up to the 31 October 2015) dealing with the topic of this review. We used the keywords "HNSCC" or "head and neck cancer" AND "PET-CT" or "PET" with different combinations: (a) staging, (b) clinical impact, (c) radiotherapy planning, (d) treatment response and (e) prognosis. The titles and abstracts were examined for potentially eligible studies for full-text retrieval. Results have been presented in Tables 1-5. In addition, most significant articles are detailed in the text. Additional sources were identified from references cited in the articles identified by electronic searching. Meta-analyses have been also used as source of articles and briefly described in the text when necessary.
The inclusion criteria were: articles comparing diagnostic performance (for staging and treatment response) between PET/PET-CT and conventional imaging (CT, MRI and ultrasonography); articles evaluating the role of PET vs conventional imaging for radiotherapy planning; and articles that evaluated pre-treatment PET/CT metabolic parameters to predict the outcome of patients with HNSCC undergoing radical treatment.
The exclusion criteria were: articles regarding the role of PET limited to nasopharynx carcinoma, thyroid or salivary gland tumours (considered as specific clinical entities); studies including ,10 patients; and non-English written articles.
Thereafter, the articles have been classified depending on their main topic in order to be considered in each of the sections of this article (pre-therapeutic staging, impact on treatment decisions, monitoring treatment response, radiotherapy planning and prognosis value).

PRE-THERAPEUTIC STAGING Local tumour extension (T stage)
The correct assessment of the size and extent of a primary lesion at staging is crucial to plan surgery and radiotherapy. Indeed, infiltration of adjacent structures is an important issue in clinical routine. For example, the transgression of the midline on the tongue complicates surgery 9 or can modify the clinical target volume in radiotherapy treatment planning (RTP). The initial assessment of the local tumour extension is generally performed with clinical examination and endoscopy.
Even though 18 F-FDG-PET/CT detects primary HNSCC with high sensitivity (.95%), 10,11 contrast-enhanced (CE) CT and MRI have been considered the primary imaging modalities for evaluating T stage of HNSCC due to their superior anatomical resolution and tissue contrast.
Since it is not possible to exactly define the size and extent of a primary lesion based on 18 F-FDG uptake, the PET (alone) images are not suitable to define the T stage of a patient. In addition, the main limitation of hybrid PET/CT, if performed with low-dose unenhanced CT, is its inability to accurately assess the extent of tumour spread and its relationship with adjacent structures.  69 88 Moreover, although PET/CT performed with contrast-enhanced CT provides both anatomical and metabolic details at the same time, there is no clear recommendation for routine use of PET/ CT in initial T staging.
Some authors showed the high potential value of PET/CT to identify the local tumour extension. 11 was no statistical difference in the detection of the primary lesion between WB PET/CT and HN PET/CE-CT protocols. 12 A major limit of this study is the lack of comparison with the MRI performance. Moreover, 66% of the patients participating in the study presented an oropharyngeal carcinoma, making difficult to infer these data in HN neoplasms of different origin.

Mandibular invasion
The presence or absence of mandibular invasion is a major determinant in both therapeutic approach and prognosis of HNSCC. 13 CT and MRI are commonly used to evaluate the status of the mandible. CT has been reported to be the most accurate method in evaluating discrete cortical bone involvement. 14 However, MRI is superior to CT for evaluating tumour invasion into medullary cavity of the mandible. 15 Gu et al 16  PET/CT was found to be more specific than MRI (83% vs 61%, respectively, p 5 0.0015) but less sensitive (78% vs 97%, respectively, p 5 0.0391). Given the low positive-predictive value (PPV) of MRI, a positive MRI scan should incite to confirm data Cancer from unknown primary The incidence of cervical metastases from unknown primary cancer (UPC) has been estimated to be around 2-9%. The absence of information about the primary tumour strongly influences the therapeutic approach (i.e. bilateral tonsillectomies, additional pharyngeal mucosa field irradiation). 17 PET/ CT can identify approximately 30% [18][19][20][21][22][23][24] of tumours in patients presenting cervical lymph node metastases from UPC, in whom the primary was not detected by the comprehensive diagnostic work-up including endoscopy and conventional imaging methods (CT or MRI). Table 1 summarizes the diagnostic performance of PET/CT in these studies. [18][19][20][21][22][23][24][25][26][27][28][29][30] It is noteworthy that it should be performed before examination under anaesthesia for targeted panendoscopy and biopsy, avoiding potential false positives due to the inflammation that usually follows these kinds of procedures. 26  The largest prospective study evaluating the diagnostic performance of PET/CT in UPC has been published by Johansen et al. 26 The authors report data about 60 patients presenting a primary tumour detection rate of 30%, and the sensitivity and specificity rates were 86% and 69%, respectively. However, this study had several limits, namely, it used three different PET engines, and among them one was PET/CT, whereas in two cases, it was PET alone. Furthermore, in some cases, PET was acquired for WB, in other cases for half body. Furthermore, the authors did not perform extensive comparison with other imaging modalities.
In summary, physical examination remains essential 5 for primary tumour assessment (especially regarding superficial tumour extension in the mucosa) while CE-CT and MRI continue to be the reference imaging modalities, mostly due to the lack of shown superiority of PET over morphological examination; however, PET seems to be a promising staging tool, in particular, when morphological examinations suffer from artefacts due to dental implants. Finally, when compared with conventional imaging, PET/CT is recommended to identify primary tumours in patients presenting with cervical lymph node metastases with unknown primary. Nevertheless, it needs opportune integration with other diagnostic procedure to exclude the relatively high risk of falsepositive findings.

Lymph node involvement (nodal staging)
The information about nodal involvement is crucial in HNSCC, as it strongly influences the treatment and prognosis of the patients. 2,3 Current non-invasive staging techniques include clinical examination, ultrasonography, CE-CT and MRI. The criteria adopted in the evaluation of the nodal status are the size, CE and radiological aspect of the nodes (presence of necrosis, analysis of the capsule to identify any sign of extracapsular extension). 33 These techniques could define positive nodes with high specificity, but present limitations in the evaluation of small lymph nodes. [34][35][36][37][38] The overall diagnostic accuracy (using pathology as the reference standard) of CT and MRI for detecting metastases in the clinical node negative neck (cN0) is relatively low. Sensitivities range from 14% to 80% for CT and from 29% to 85% for MRI, and specificities range from 80% to 100% for both CT and MRI. 33, [35][36][37] The sensitivity and specificity of the imaging techniques influence the clinical practice of the radiation oncologist; patients presenting an expected risk of nodal involvement exceeding 20% 39 undergo a prophylactic treatment of the neck, including a neck dissection or unilateral and/or bilateral neck irradiation. 39,40 Considering these rates of microscopic involvement, it means that there are at least two thirds of the patients who are treated on the nodal areas without presenting a nodal involvement, only because it is not possible to predict with a better accuracy the "real nodal status of the patients".  c Medium and high impact: the impact on patient management in these studies was classified as follows: low impact (treatment modality and delivery unchanged); medium impact (change in the treatment within the same therapeutic modality: changes in the type of surgery on primary cancer and/or neck dissection, and changes in the dose or radiotherapy fields); or high impact (change in treatment intent and/or treatment modality: curative to palliation, surgery to chemoradiation or vice versa). d Not available. e In these two studies, treatment decisions were made by a tumour board (pre-PET staging and treatment management plan, and post-PET staging and treatment plan). Therefore, the pre-PET treatment decision and post-PET treatment decision were both made by the tumour board. f This accuracy was calculated for the cases presenting discordant TNMs (n 5 100) between conventional work-up and from adding a PET (not for the whole population of the study). g The overall accuracy regarding the whole population of the study. h Meaning that the clinical original clinical decision (pre-PET) adopted by the tumour board was changed in approximately one out of five (9.5% 1 16.7% 5 26.2%) patients participating in the study, due to the PET/CT.
Review article: Role of PET/CT in head and neck cancer patients BJR   Although PET/CT examination protocols without the use of contrast medium have been utilized, increasing evidence supports the use of CE-CT as a part of routine PET/CT protocols. 46,47 Recently, there have been several reports of the possible superiority of PET/CE-CT over standard PET/CT in different clinical settings, including better local 48 and nodal analysis. 49 Other studies, 50,51 confirmed the high accuracy of nodal staging by PET/CT, in particular, if CE-CT is used during PET protocol.
In summary, 18 F-FDG-PET has high diagnostic performance in the overall nodal staging of patients with HNSCC. When compared with conventional imaging, PET/CT is similar or superior for detecting cervical nodal metastases (Table 2). However, this modality is not yet accurate enough to replace the accuracy of neck dissection in the identification of occult cervical metastasis in patients with cN0. [36][37][38][41][42][43] Detection of distant metastasis The presence of distant metastases is the most important predictor of patient survival in several cancers. Overall incidence of distant metastasis in HNSCC is relatively low (2-18%). 52 Distant metastases frequently occur in the lungs and are routinely detected by chest CT (73% sensitivity and 80% specificity). 53 It is noteworthy that early detection of metastasis has a major impact on patient management avoiding unnecessary radical treatments. 54 Regarding the detection of bone metastases, Yi et al 59 showed in a recent meta-analysis on .3000 patients, a sensitivity and a specificity of 81% and 99% for PET, and of 89% and 99% for PET/CT, respectively, which are better than the results obtained by bone scintigraphy. Bone scintigraphy relies on the osteoblastic response to bone destruction by cancer cells and the accompanying increase in blood flow. Therefore, 18 F-FDG-PET is more efficient than bone scintigraphy for bone lesion detection considering their frequently lytic character. 59 Table 3 summarizes available studies comparing the performances of different imaging approaches to detect distant metastasis of HNSCC. 52,53,[58][59][60][61][62][63] Globally, all these studies indicate that PET 6 CT is superior to conventional imaging. 18 F-FDG-PET shows higher accuracy (90-95%) than CT for the detection of distant metastasis. 60,61,63 Given the very high negative-predictive value (NPV), it suggests that in case of negative PET scan, other imaging techniques are not necessary. Nevertheless, the PPV for detecting SPT or distant metastasis is around 60%, suggesting that additional diagnostic methods are still necessary to exclude false-positive results. 62 Senft et al 61 assessed the added value of 18 F-FDG-PET (to chest CT) in the screening of distant metastases in patients with HNSCC and high-risk factors (more than or equal to three lymph node metastases, bilateral lymph node metastases, lymph node metastases of $6 cm, low jugular lymph node metastases, regional tumour recurrence and SPT). 145 consecutive patients with HNSCC underwent chest CT and 18 F-FDG-PET. 18 F-FDG-PET improved pre-treatment screening of distant metastasis compared with chest CT, showing higher sensitivity (53% vs 37%) and PPV (80% vs 75%). Moreover, the authors showed that the sensitivity of the combination of CT and 18 F-FDG-PET was higher (63%) than the sensitivity of each of these techniques alone.
Ng et al 60  In summary, when compared with conventional imaging, PET/CT is a valuable tool to rule out the presence of distant metastases in HNSCC, especially in locally advanced tumours. 52,53,[60][61][62]64 Second primaries SPTs are detected in almost 10% of patients with HNSCC, 64,65 particularly in patients who smoke and/or in patients who are negative for human papillomavirus. 66 The identification of synchronic or metacronic SPT 67 could occur both at the HN region (more frequently) and/or elsewhere (lungs, oesophagus, colon etc.), and it can influence the therapeutic approach 61 and the prognosis of patients (especially those presenting with HNSCC). 68 Strobel et al 64  According to these results 64,65 with a negative 18 F-FDG-PET/CT, the extent of endoscopy can be reduced to the area of the primary tumour.
In summary, PET/CT is an accurate method detecting second primaries, with a high NPV. 64,65,[67][68][69] Nonetheless, it should be stressed that due to a low PPV (approximately 60% 65 ) in this setting, additional diagnostic methods are necessary to exclude false-positive results (inflammation and hyperplasia in the HN region or intestinal polyps can result in false positives). Moreover, whenever possible PET/CT should be performed before endoscopy and biopsy to avoid false-positive results. 26

POSITRON EMISSION TOMOGRAPHY/CT AND CLINICAL IMPACT ON TREATMENT DECISIONS
Although PET/CT imaging is effective for the staging of HNSCC, its impact on patient management is somehow controversial. Indeed, to date, the overall impact of PET/CT on treatment decisions in HNSCC has been rarely explored compared with the number of studies assessing the impact of PET on staging. However, there are four prospective trials 54-57 that have specifically analyzed the impact of PET/PET-CT in the treatment approach of HNSCC. These studies followed the same methodology and are detailed in Table 4. These studies addressed at the same time the issue of the impact of PET on the initial staging and management of patients with HNSCC: globally, the PET changed the original treatment plan in approximately 30% of patients. [54][55][56][57] The largest trial, published by Lonneux et al 55 included 233 patients (Stages I-IV) and reported a modification in the original treatment plan in 32 patients (13.7%). In 12 patients (5.2%), the modification was classified as medium (the therapeutic modality remained the same, but PET altered the treatment planning). In 20 patients (8.6%), the impact of PET on patient management was classified as high (change in treatment intent and/or treatment modality, e.g. curative to palliation, surgery to chemoradiation and so on). Interestingly, one of the studies 57 assessed together the usefulness of PET/CT for staging and its overall impact on management plans specifically in patients with Stages III and IV HNSCC where the treatment plan was altered in 22/84 (26%) patients (Table 4). These results are in line with the current guidelines. 2,3 In summary, PET/CT should be included [54][55][56][57] in the routine diagnosis of patients with Stages III-IV 57 HNSCC, as it significantly improves staging accuracy and also has a marked impact on management plans.

RADIOTHERAPY PLANNING
CT is the primary imaging modality in RTP. The CT images are acquired with the patient in the supine position, immobilized with an individual head support and a rigid customized mask to increase positioning accuracy and to prevent movement during image acquisition. All other imaging modalities (such as PET or MRI) are considered as secondary images. 70 The secondary images in RTP will have to be registered (fused) to the primary planning CT scan. When the PET and RTP CT images are acquired on separate scanners (often in a position that does not correspond to real treatment position), a registration module in the RTP computer system can be used to fuse images. Regardless of the imaging data used (i.e. separate PET or PET/CT) correct co-registration of the PET data with the CT data used for RTP must be verified, since the difference in spatial localization of tumour may lead to false estimation of the gross tumour volume (GTV). Ideally, the fusion process can be executed automatically by a hybrid PET-CT-dedicated RTP scanner performed with the corresponding immobilization devices and reproducing radiation delivery conditions. 70 Recognizing the potential of PET/CT-guided treatment planning, some authors recently investigated the role of PET in RTP, specifically for the correct delineation of lymph nodes. Schinagl et al 71 compared the volume of metastatic lymph nodes between 18 F-FDG-PET/CT segmentation (by ten methods) and CT with the volume as determined by pathological examination. They concluded that beyond the detection of lymph node metastasis (staging), PET has no additional value over CT for the delineation of lymph nodes.
Despite the limited role of PET-CT in improving the contouring of nodes, 71 it seems to have a main role in improving the definition of the primary tumour GTV. Indeed, PET/CT information is frequently integrated in RTP. Nevertheless, the use of 18 F-FDG-PET for target volume delineation in RTP for HNSCC has been mainly evaluated in single institution studies. 5,31,[72][73][74][75] Different segmentation methods have been proposed. Visual interpretation of the PET signal, considered the most intuitive method for segmentation, has been commonly applied in many studies. The main limit of this approach is that it is a highly operator-dependent process, and it is influenced by window-level settings. 5,[74][75][76] This is one of the major weakness in the use of PET-CT in the target volume delineation of HNSCC. This variability could be reduced by using a more objective methodology: isocontouring based on a fixed standardized uptake value (SUV) such as a SUV of 2.5-3 g l 21 or relative thresholds such as a percentage of the maximum tumour intensity (40% SUV max , 50% SUV max ). 75,77 Nevertheless, according to this method, several structures containing a high physiological 18 F-FDG uptake, such as the tonsillar area or the vocal cords, can be incorrectly included in the segmented area. Therefore, models using a fixed threshold relying on SUV are somehow debatable. 73 To overcome this issue, several authors successfully developed advanced adaptive relative threshold segmentation methods based on maximal tumour uptake, background uptake, tumour dimensions and tumour grade. 31,75,78,79 Thereafter, other methods including gradient-based 73 detections have been introduced. In brief, this method relies on the watershed transform and hierarchical cluster analysis, to allow a better estimation of the gradient intensity. Interestingly, this method allows automatic delineation and therefore is an operatorindependent process. Most studies comparing GTV definitions using 18 F-FDG-PET against CT or MRI reported a decrease in the GTV, especially when using more sophisticated segmentation methods. 5,31,[72][73][74]76,80 However, few groups have validated delineation process using different imaging modalities against surgical resection specimens. 31,[81][82][83][84][85] In general, all imaging modalities overestimated the tumour extension compared with surgical specimen. Nevertheless, none of the image modalities (CT, MRI or PET) completely encompassed the surgical specimen volume because of an underestimation of superficial tumour extension in the mucosa, 31 as also reported by Ng et al. 81 According to Daisne et al, 31 the GTV delineated from 18 F-FDG-PET applying an adaptive signal-to-background method was significantly smaller than GTV delineated by CT or MRI. In addition, GTV-PET was the closest volume to the pathological GTV obtained from surgical specimen. On average, the PET delineated smaller volumes than CT or MRI. Nevertheless, GTV contours at PET were not totally encompassed by those delineated with CT or MRI.
Geets et al 73 validated a gradient-based method in seven patients with laryngeal carcinoma. The calculated volumes for laryngeal tumours according to this methodology 73 were compared with the macroscopic specimens and, additionally, with the volumes obtained applying the source-to-background ratio developed by Daisne et al. 31 Interestingly, the gradient-based method proved to be more accurate than the source-to-background ratio but neither the threshold-based nor the gradient-based volumes encompassed completely the laryngeal specimens.
Interestingly, in a recent multicentric prospective study by Leclerc et al, 86 the primary tumour was automatically delineated on the 18 F-FDG-PET images using a gradient-based method previously described by this group. 73 They confirmed that the use of 18  The goal of the study was to compare three semiautomatic PET segmentation methods for derivation of PV in primary HNSCC on sequential 18 F-FLT PET images before and during chemoradiation. The following semiautomatic segmentation methods were applied to sequential PET scans: background-subtracted relative-threshold level, a gradient-based method using the watershed transform algorithm and hierarchical clustering analysis and a fuzzy locally adaptive Bayesian algorithm. The authors 88 concluded that fuzzy locally adaptive Bayesian algorithm (FLAB) was the best performing method for segmentation of the PV on repeat 18 F-FLT PET/CT scans during chemoradiation. FLAB is less sensitive to image noise than the other segmentation approaches tested in the study. This finding may have other potential implications for radiotherapy indicating that FLAB is a promising candidate for radiation target volume adaptation based on sequential 18 F-FLT PET scanning.
Currently, there is still no consensus (national/international) between institutions regarding the best method to use for delineation. Therefore, data from 18 F-FDG-PET can complement other diagnostic imaging modalities for management decisions and guidance of RTP, but it cannot replace physical examination or MRI/CT scans to achieve significant details such as assessing invasion of tumour-surrounding tissues. 85 Moreover, defining the primary tumour boundaries with PET is a difficult task.
In summary, current evidence is based on numerous heterogeneous small studies with changing methodology for different research questions. PET-based RTP is a promising modality to improve contouring accuracy. PET is the imaging modality that defines the closest volume to the pathological specimen. The main drawback is the lack of standardized method for functional volume segmentation, which highly influences the size and shape of the resulting GTV. Currently, the most accurate segmentation method seems to be the gradient-based method validated by Geets et al. 73 However, it may not completely encompass the tumour specimen volume. 31,73 This issue is more relevant when considering superficial mucosal spread (evaluable by physical examination). 5,81 Therefore, even in some contemporary HNSCC study protocols (European Organisation for Research and Treatment of Cancer-1219; NCT01880359), PETbased target volume delineation is not allowed. Before PET can reliably be incorporated into routine high-precision RTP, operator-independent segmentation tools have to be developed and validated (international consensus), and also, clinical effect on outcomes should be reflected in clinical studies.

MONITORING RESPONSE TO CHEMORADIOTHERAPY: RESIDUAL DISEASE AND RECURRENCE (FOLLOW-UP)
Early detection of residual or recurrent disease following radiotherapy is a diagnostic challenge owing to post-treatment anatomical distortions, mostly related to oedema and fibrosis. 89 The key role of a diagnostic tool evaluating treatment efficacy is to correctly identify patients requiring salvage-tailored treatments. Moreover, an early detection of the relapse could help in the selection of patients who could be successfully retreated. 90 In this setting, 18 F-FDG-PET/CT is an interesting modality to evaluate response to treatment, as it can assess metabolic activity-rendering malignant process. 89 Isles et al 91 preformed a meta-analysis reporting that 18 F-FDG-PET (without CT) is a highly accurate tool for monitoring response and detecting relapse after chemoradiotherapy (for both the primary site and lymph nodes). Moreover, several studies have demonstrated that 18 F-FDG-PET/CT also has a higher accuracy in the detection of recurrent lesions compared with CT/ MRI. 6,[92][93][94][95][96] These results obtained with PET/CT are not significantly different from those obtained with PET alone.
The timing of PET/CT after the treatment is crucial. 6,[97][98][99][100][101] It is widely accepted that PET has a high NPV (around 90%) if it is performed at least 8 weeks after chemoradiotherapy. Therefore, a negative PET scan after treatment appears to be a consistent predictor of the absence of residual tumour. 102 According to other reports, more accurate evaluation is possible when PET/CT is performed 8-12 weeks after treatment. 6,97 The meta-analysis of Gupta et al showed a weighted mean (95% CI)-pooled sensitivity, specificity, PPV and NPV of post-treatment 18 100,101 Intuitively, delaying a response evaluation tool would surely increase its accuracy. However, there is no homogeneous data for optimal window for salvage treatment; probably, it would be wise not to postpone salvage surgery beyond a clinically reasonable point.
There is debate regarding the need for elective neck dissection after radical chemoradiotherapy. There are two prospective studies 98,99 addressing the status of neck adenopathy of nodepositive HNSCC that had 18  Imaging data were compared with clinicopathological outcomes. For cervical lymph nodes, sensitivity was 100%, specificity was 98%, PPV was 92% and NPV was 100%. Therefore, both these prospective studies concluded that PET-guided management of the neck after chemoradiotherapy appropriately spares neck dissections in patients with complete response or presenting with PETnegative residual CT lesions. 98,99 Recently, Mehanna et al 103 published a prospective, randomized controlled trial assessing the non-inferiority of PET/CT-guided surveillance (evaluation was performed 12 weeks after definitive chemoradiation. Neck lymph node dissection was only indicated when PET/CT presented an incomplete or equivocal response) to planned neck dissection in a total of 564 patients with locally advanced HNSCC (Stage N2 or N3 disease), who underwent chemoradiation for primary treatment.
Patients were considered to have incomplete nodal responses when PET/CT performed 12 weeks after treatment showed high 18 F-FDG uptake (with or without enlarged lymph nodes in the neck). In addition, results of PET/CT presenting mild or no 18 F-FDG uptake in enlarged lymph nodes or mild 18 F-FDG uptake in normal-sized nodes were classified as equivocal responses. The rest of the PET/CT scans were classified as complete responses. Patients showing an incomplete or equivocal response in the neck but presenting a complete response in the primary location underwent neck lymph node dissection within 4 weeks after PET/CT.
The survival rate was similar (2-year overall survival rate of 84.9% and 81.5% in the surveillance group and in the plannedneck dissection group, respectively) between patients who underwent PET/CT-guided surveillance policy and patients undergoing a planned surgery. Indeed, the hazard ratio for death (upper boundary of the 95% CI for the hazard ratio, ,1.50; p 5 0.004) favoured PET/CT-guided surveillance policy.
Moreover, surveillance resulted in considerably fewer operations (approximately 80% of patients were spared neck dissection compared with planned dissection surgery; 54 vs 221), and it was more cost effective. The per-person cost saving was £1492 (approximately $2190 in US dollars), with an additional 0.08 quality-adjusted life years per person.
However, these authors recommended that patients with an equivocal 18 F-FDG uptake should continue to undergo neck dissection. In addition, when extrapolating these results to daily clinical practice, it should be noted that in this study, only a small number of patients [17/564 (3%)] presented N3 disease. Therefore, a direct extrapolation of a PET/CT-guided surveillance policy to patients presenting N3 (Stage IVb) disease should not be indicated owing to the small number of such patients recruited in the study. 18 F-FDG-PET/CT could have a potential interesting role in the follow-up of patients with HNSCC. Despite that, the clinical advantages and economic costs of this issue have not yet been largely addressed. One of the largest studies has been published by a group from Pittsburg. 90 They evaluated 388 patients retrospectively to assess the recurrence rate after radical chemoradiotherapy among patients who underwent PET/CT surveillance. Tumour recurrence was detected in 110 patients (73 asymptomatic and 37 symptomatic). Indeed, 95% (95% CI, 87-98%) of asymptomatic recurrences were observed within 2 years of follow-up. The authors proposed to evaluate patients for recurrence with PET/CT at 2, 5, 8 and 14 months post-treatment. The reason for this protocol is because their study demonstrated that PET/CT detected almost all HNSCC recurrences within 2 years.

IN SUMMARY
(1) The overall diagnostic performance of 18 F-FDG-PET/CT for response assessment is good, but its PPV is not optimal. By contrast, the NPV is particularly high and negative post-treatment PET/CT is very suggestive of absence of viable disease that can guide daily clinical management decisions. In this context, the timing of PET/CT after the end of the treatments is a crucial issue. Available evidences suggest waiting a minimum of 8 weeks before restaging with PET/CT, and preferably 12 weeks to increase the NPV.
(2) Available evidences suggest that this strategy is safe to avoid neck dissection in patients presenting negative PET/CT after CRT. 98,99,102 The safety of this attitude is also confirmed by the results of the PET-NECK study, 103 where PET/CT-guided active surveillance showed similar survival outcomes compared with planned neck dissection, and considerably fewer neck dissections, and it was more cost effective. However, extrapolation of a PET-CTguided surveillance policy to patients with N3 (Stage IVb) disease cannot currently be justified. PROGNOSTIC VALUE Treatment outcome of HNSCC cancer remains heterogeneous. Identification of novel pre-treatment factors (other than tumour stage, lymph node involvement, anatomical subsite or human papillomavirus status) that potentially predict long-term outcome is of great interest.
Quantifying the prognostic value of PET is challenging. In general, the results of prognostic value of SUV remain undetermined because of the small sample of most of these studies. Moreover, it should be considered that HNSCC prognosis also depends on the initial tumour site; data regarding the prognostic value of PET/CT depending on different tumour locations is scarce in the literature.
Two meta-analyses have been conducted to estimate the effect of SUV on the prognosis of HNSCC. First, Zhang et al 104 analyzed the potential of SUV [SUV max and mean SUV (SUV mean )] as a prognostic marker. These authors concluded that increased SUV max/ mean of the primary tumour is a poor prognosis factor and has a potential value in predicting local control, disease-free survival and overall survival. Thereafter, Xie et al 105 performed another meta-analysis to evaluate the prognostic value of SUV, confirming that low primary tumour SUV was associated with better survival prognosis. It should be stressed that SUV estimates suffer from poor reproducibility between centres because of the lack of standardization of the acquisition and processing protocols.
Globally, studies evaluating the prognostic utility of PET/CT in HN cancer are quite heterogeneous. Most of them have focused mainly on the SUV max . Some studies have demonstrated worse clinical outcomes with higher pre-treatment SUV max . [106][107][108] Other studies observed the correlation of survival with several PET data such as SUV mean or metabolic tumour volume (MTV). 109,110 Kitajima et al reported 111 that the pre-treatment SUV max of nodal disease (rather than the primary tumour) in patients with laryngeal cancer was prognostic of recurrence. However, a prospective trial 7 conducted at the MD Anderson Cancer Center, Houston, TX, evaluated this particular question and failed to demonstrate any significant clinical correlation between pre-radiotherapy PET-CT SUV parameters and treatment outcomes.
According to other authors, SUV mean 109,112 may be a better prognosis marker than SUV max , with inferior disease-free survival in patients presenting higher pre-treatment SUVmean . 109 These results could be explained considering that SUV max reflects the highest intensity of 18 F-FDG uptake as measured in the highest pixels within a concrete region of interest, whereas the SUV mean represents the average of the intensity of the uptake providing a more global picture of tumour metabolism than SUV max . 109 Nevertheless, a potential pitfall of SUV mean is the lesser degree of reproducibility relative to SUV max . 109 More recently, there has been an increasing interest in the use of volumetric parameters of metabolism such as the MTV and total lesion glycolysis (TLG), which weights the volumetric burden and volumetric activity of tumours. Pak et al 113 conducted a metaanalysis (13 studies including 1180 patients) on volumetric parameters addressing the prognostic value of MTV and TLG in patients with HN cancer. Despite the various methods adopted between studies, these authors concluded that MTV and TLG are accurate prognostic indicators of outcome in patients with HN cancer. Indeed, high MTV and TLG increased the risk of disease progression and death. The studies evaluated in the meta-analysis and more recent contemporary studies 110,112,[114][115][116][117][118][119][120][121][122][123][124][125][126][127][128][129] are described in Table 5.  In summary, meta-analysis and several studies showed that PETquantified data such as SUV max , SUV mean , MTV, TLG are strongly and negatively correlated with survival. However, given that some uncertainty still exists on which of these parameters are the best predictors, prospective trials are needed to definitively settle this issue.
POSITRON EMISSION TOMOGRAPHY/MRI: ADDING NEW POSSIBILITIES Since information provided by PET/CT and MRI is complementary in many clinical situations, it seems to make sense to combine the two modalities. The high soft-tissue contrast and the different functional imaging techniques of MRI might help to ameliorate the informative value of a hybrid imaging study. Consequently, the discussion about potential applications of this new hybrid technology in oncological imaging and especially in the HN has been generated. 130 The feasibility 131 and diagnostic performance (sensitivity, specificity and accuracy) of clinical PET/MRI have been demonstrated in a significant number of studies, 51,85 but technological and logistic challenges, such as errors in attenuation correction and long scan duration continue to be a major focus of the PET/ MRI literature. 132 PET/MRI has many potential advantages over PET/CT (including perineural spread of tumours and the infiltration of important anatomical landmarks, such as the pre-vertebral fascia and great vessel walls; lower radiation exposure, higher soft-tissue contrast and several functional techniques). 133 Realizing this potential in clinics will likely require new radiopharmaceuticals and applications other than WB cancer staging. 132 Although PET/MRI is still in the early stages of clinical development, it is clear that the clinical adoption of PET/MRI is slower than that of PET/CT. This slower evolution is not only due partly to ongoing technological challenges (e.g. accurate attenuation correction of PET images) but also to the complex logistics of combining a WB PET scan with WB or organ-specific MRI. Key applications of PET/MRI that provide information that is clinically relevant and different from that provided by PET/CT still need to be defined. 134 Further studies that exploit the functional MRI and molecular PET capabilities may report substantial contributions of PET/ MRI for treatment response predictions, radiotherapy planning, tumour phenotyping and treatment monitoring. 131,135,136 Future research involving larger patient series is needed to assess the true impact of this technique in HNSCC and will show whether PET/MRI outperforms PET/CT, MRI, diffusionweighted MRI or their combination. Table 6 summarizes some of the ongoing clinical trials (www. ClinicalTrials.gov) dealing with the issue of PET/CT in HNSCC treated with radiotherapy. Unknown primary: 18 F-FDG-PET/CT is recommended to identify primary tumours in patients presenting with cervical lymph node metastases with unknown primary 2.

FUTURE APPLICATIONS OF POSITRON EMISSION TOMOGRAPHY/CT IN RADIATION ONCOLOGY
Pre-treatment staging: Multiple studies have demonstrated that adding PET (or PET/CT) to a conventional work-up resulted in a higher staging accuracy; nodal classification is improved, especially in terms of specificity; and 18 F-FDG-PET or PET/CT detects more distant metastases and second malignancy than conventional staging 3.
Clinical impact: A more refined staging has demonstrated to have clinical impact on treatment decisions. PET/CT should be included in the routine diagnosis of patients with Stages III-IV HNSCC, as it significantly improves staging accuracy and also has a marked impact on management plans. However, there is no evidence with regard to a possible benefit in outcomes due to PET/CT

4.
Radiotherapy: The use of 18 F-FDG-PET translates into smaller GTV for the primary tumour volumes than with the use of CT or MRI. PET can complement other diagnostic imaging modalities for management decisions and guidance of radiotherapy planning, but it cannot replace physical examination or MRI/CT. There is no current standardized method for functional volume segmentation recommended for daily practice

5.
Monitoring response to treatment: PET/CT is recommended to check for residual disease at least 8 and, preferably, 12-16 weeks after definitive chemoradiotherapy in node-positive HNSCC (NPV .90%), avoiding unnecessary neck dissections in patients presenting complete response. Mehanna et al 103 indicated that patients with incomplete (presenting high 18 F-FDG uptake at 12 weeks after chemoradiotherapy, with or without enlarged lymph nodes in the neck), or equivocal response (mild or no 18 F-FDG uptake in enlarged nodes or mild 18 F-FDG uptake in normal-sized nodes) should undergo neck dissection. Few patients in this trial had N3 (Stage IVb) disease [17/564 (3%)]. Therefore, a PET-CT-guided surveillance policy to patients presenting N3 disease is not currently justified due to the small number of patients presenting N3 disease in this study

6.
Surveillance: PET/CT is not indicated in routine follow-up 7.
Prognosis: Patients presenting increased SUV (SUV max , SUV mean ) or higher MTV or TLG seem to have worse prognosis (higher risk of treatment failure). There is no currently evidence to support that this type of patients should receive different treatment approach. Nevertheless, these parameters could be used to stratify patients in future clinical trials 8.
To further validate incorporating PET/CT into daily practice also, clinical effect on outcomes and cost effectiveness should be reflected in clinical studies 18 F-FDG, fluorine-18 fludeoxyglucose; GTV, gross tumour volume; HNSCC, head and neck squamous cell carcinoma; MTV, metabolic tumour volume; NPV, negative-predictive value; PET, positron emission tomography; SUV, standardized uptake value; SUV max , maximal SUV; SUV mean , mean SUV; TLG, total lesion glycolysis.
Dose escalation to 18 F-FDG-avid subvolumes of the tumour as well as adapting the RTP during treatment regarding the functional tumour changes induced during radiation are two important issues that are currently under investigation. [137][138][139] Madani et al 140 performed a Phase I trial to establish the maximum tolerated dose, when the dose was escalated in 18 F-FDG-PET GTV within the anatomically (CT/MRI) based GTV. They demonstrated the feasibility of heterogeneous dose delivery with dose escalation up to 77.5 Gy for 18 F-FDG-avid tumour areas (the so called "dose painting" approach). Another approach would be adaptation of the biological target volume during the course of radiotherapy in order to reduce Figure 1. Proposal to incorporate positron emission tomography (PET)/CT (contrast-enhanced CT) into routine clinical practice for head and neck squamous cell carcinoma. † Cervical metastases from unknown primary cancer (UPC). *Evaluate to include MRI to assess softtissue invasion or perineural spread if needed depending on primary tumour location. 16 V Prognosis: uptake parameters and volumetric parameters can potentially help in identifying patients with worse outcomes (still an area of active investigation). ¥ Radiotherapy (RT) plan: PET/CT may be helpful, improving contouring accuracy. However, standardized method is lacking. p Follow-up with conventional imaging (CT/MRI) is recommended. When equivocal findings additional PET/CT can be performed. m Once recurrence is detected, additional PET/CT may be of interest to improve patient counselling; restaging the tumour and planning additional therapy, especially when "aggressive" interventions (extended surgery or reirradiation) may be needed. S However, Mehanna et al 103 indicated that patients with incomplete [patients presenting high fluorine-18 fludeoxyglucose ( 18 F-FDG) uptake at 12 weeks after chemoradiotherapy, with or without enlarged lymph nodes in the neck] or equivocal response (mild or no 18 F-FDG uptake in enlarged nodes or mild 18 F-FDG uptake in normal-sized nodes) should undergo neck dissection. In addition, it should be noted that few patients in this trial had N3 (Stage IVb) disease [17/564 (3%)]. Therefore, a PET-CT-guided surveillance policy to patients presenting N3 disease is not justified due to the small number of patients presenting N3 disease in this study. CR, complete response; QT, chemotherapy. the treated volume as radiotherapy progresses. Duprez et al 138 used adaptive intensity-modulated radiotherapy planning based on dose painting by numbers according to 18 F-FDG-PET voxel intensities concluding that replanning was possible reaching a total dose of 80.9 Gy. It is noteworthy that these kinds of approaches require extreme attention, as a slight shift of the anatomy will not only cause a mismatch of dose and intratumour anatomy but also a higher dose into nearby healthy tissue. Moreover, another unsolved issue is the monitoring of the shift of high SUV regions during the course of treatment.
Last but not least, the possibility of targeting radiation resistance within the tumour on the basis of biological information (intratumoural hypoxic and proliferation states) obtained from functional imaging with other tracers than 18 F-FDG is an emergent strategy. [141][142][143] However, this issue is still under investigation and should still be considered as experimental. 144,145 CONCLUSION PET/CT is an important diagnostic tool in HN oncology, especially in the initial staging and in monitoring response to definitive chemoradiotherapy. Moreover, approaches of neck dissection sparing based on the results of restaging PET/CT are safe and should be implemented in daily clinical practice. The summary of indications and controversies are detailed in Table 7. Based on the results of our review, we summarized in Figure 1 a proposal for integrating PET/CT in daily clinical practice.