VOOR MACHTELD EN ASTRID

Transcriptie

1 VOOR MACHTELD EN ASTRID

2

3 Are not our lives too short for that full utterance which through all our stammerings is of course our only and abiding intention? Joseph Conrad, Lord Jim.

4

5 Dankwoord Acknowledgments Als doctorandus wens je dat ieder proefschrift op de wereld komt zoals Pallas Athene, volmaakt geboren uit het hoofd van Zeus. In werkelijkheid gaat er telkens een lange en vaak niet geheel rimpelloze intellectuele zwangerschap aan vooraf. Ik prijs me dan ook gelukkig dat ik kon rekenen op verscheidene bekwame en toegewijde vroedvrouwen, die me hebben begeleid en het eindresultaat stelselmatig hebben bijgeschaafd. De onvolmaaktheden die desondanks overblijven zijn uiteraard enkel en alleen mijn verantwoordelijkheid. Allereerst wil ik mijn promotor, Professor Nuyts, bedanken. Sandra, vanaf de eerste dag heb je mij vertrouwd, gesteund, en standvastig de weg gewezen. Je academische en menselijke voorbeeld waren 4 jaar lang een betrouwbare leidraad en ik hoop dat je mijn vriend en mentor zal blijven. Mijn doctoraat is onlosmakelijk verbonden met herinneringen aan de vergaderingen op je bureau waarbij eerst het noodzakelijke werk werd besproken, gevolgd door updates over huizen en families. Ook mijn co-promotor, Professor Hermans, moet ik bedanken voor de steun, waardevolle inzichten en al de tijd die hij voor mij heeft vrij gemaakt. Professor Van den Bogaert wil ik graag bedanken voor het getoonde vertrouwen toen ik 4 jaar geleden schoorvoetend aanklopte met de vraag naar wetenschappelijk werk. Gedurende mijn onderzoek heeft hij mij erg geholpen met zijn uitgebreide expertise in hoofd- en halskanker, zijn ervaring in het schrijven en publiceren van wetenschappelijke artikels en met de nodige praktische steun. Ook ons huidige diensthoofd, Professor Haustermans, ben ik zeer dankbaar voor haar raad en haar voortdurende bekommernis om mijn werk. I would like to thank the members of the jury, Professors Grégoire, Levendag, Pameijer, Stroobants and Dymarkowski, for their valuable suggestions on this manuscript and for making time for the public defense. Professor Tejpar expertly guided this manuscript through its final stages, for which many thanks. The Flemish League against Cancer (VLK) supported me during the first year of my PhD, while the Research Foundation Flanders (FWO Vlaanderen) provided me with an excellent scholarship during the final years. Obviously, life would have been much harder without this financial support. Our group also received grants from the Belgian Foundation against Cancer. Mijn werk zou aanzienlijk moeilijker zijn geweest zonder de steun van mijn collega s assistenten en de staf radiotherapie-oncologie en algemene medische oncologie, de verpleging op de 5

6 Dankwoord Acknowledgments bestralingsafdeling en de dienst klinische fysica. Het zou alleszins beduidend minder aangenaam zijn geweest zonder de mede-ballingschap van Tom in het Louvre (dat van Leuven jammer genoeg) en het gezelschap van Nele in de wondere wereld der slikstructuren en brachiale plexussen. Maarten, onze onderzoeksperiodes hebben slechts kort overlapt, maar toch lang genoeg om de kortstondigheid ervan te betreuren. Je bent bovendien een beetje de erfgenaam van dit werk, dus maak er wat moois van! Vincent en Frederik, ik mis het samenwerken met jullie nu al. Jullie kameraadschap is ongetwijfeld één van mijn mooiste herinneringen aan de afgelopen 4 jaren en het levende bewijs dat volmaakte symbiose bestaat, al was ik misschien eerder een commensaal. Mijn vrienden hebben me de laatste tijd wellicht minder gezien dan vroeger, maar misschien nog steeds vaker dan hen lief was. Ik was dan ook meestal met mijn hoofd bij hoofd- en halskanker. Alleszins bedankt voor jullie geduld, geweldige afleidingsmaneuvers en ongetwijfeld buitengewoon oprechte interesse in de finesses van intensiteits-gemoduleerde radiotherapie. Ik denk daarbij in het bijzonder aan An en Wim, die samen met mij haast ervaringsdeskundigen zijn geworden in de etiologie van hoofd- en halskanker. Alles voor de wetenschap! Gelukkig zijn we dankzij onze strenge partners ondertussen weer op het rechte pad Mijn familie heeft me altijd volop gesteund en getracht te volgen waar ik mee bezig was, waarvoor veel dank. Heel bijzondere dank voor Luc, voor alle kansen en steun. Mijn ouders, jullie hebben altijd mijn flanken verdedigd en mijn achterhoede gedekt, wat zou ik zonder jullie beginnen? Dit doctoraat is er enkel dankzij jullie. Dit boek is opgedragen aan Machteld en Astrid, die het beste zijn wat me ooit is overkomen. Porque mi amor por ti es total y es para siempre. Mijn grote dank gaat uit naar alle patiënten die belangeloos hebben meegewerkt aan alle studies die in dit boek beschreven staan, vaak ten koste van grote last voor henzelf. 6

7 List of abbreviations AAO-HNS: American academy of otolaryngology/head and neck surgery ADC: apparent diffusion coefficient AF: accelerated fractionation AJCC: American joint committee on cancer AUC: area under the curve BOT: base of tongue BTV: biological target volume CHART: continuous hyperfractionated accelerated radiotherapy CI: confidence interval Cx: chemotherapy CR: complete response CRT: chemoradiotherapy CT: computed tomography CTV: clinical target volume Cu-ATSM: 60 Cu(II)-diacetyl-bis-N 4 -methylthiosemicarbazone CTCAE: common terminology criteria for adverse events DAHANCA: Danish head and neck cancer study group DC: distant control DCE-MRI: dynamic contrast-enhanced magnetic resonance imaging DFS: disease-free survival DSS: disease-specific survival DVH: dose-volume histogram DW-MRI: diffusion-weighted magnetic resonance imaging ECOG: Eastern cooperative oncology group EGFR: epidermal growth factor receptor ENT: ear-nose-throat EORTC: European organization for research and treatment of cancer ES: ethmoid sinus ESO: esophagus ESS: endoscopic sinus surgery FEES: fiberoptic endoscopic evaluation of swallowing FETNIM: [ 18 F]-fluoroerythronitroimidazole FDG: [ 18 F]-fluorodeoxyglucose FMISO: [ 18 F]-fluoromisonidazole 7

8 List of abbreviations GL: glottic larynx GORTEC: groupe oncologie radiothérapie tête et cou GTV: gross tumor volume HF: hyperfractionation HNC: head and neck cancer HPV: human papilloma virus HV: hypoxic volume ICRU: international commission on radiation units and measurements IMRT: intensity-modulated radiotherapy IPC: inferior pharyngeal constrictor muscle IS: initial slope IV: intravenous LC: local control LN: lymph node LRC: loco-regional control MACH-NC: meta-analysis of chemotherapy in head and neck cancer MDADI: MD Anderson dysphagia inventory MLC: multileaf collimator MPC: middle pharyngeal constrictor muscle MRI: magnetic resonance imaging MS: maxillary sinus MV: megavoltage NPC: nasopharynx cancer NPV: negative predictive value OAR: organs at risk OS: overall survival OTT: overall treatment time PEG: percutaneous endoscopic gastrostomy PET: positron emission tomography PPS: performance status scale PPV: positive predictive value PTV: planning target volume QoL: quality of life QLQ: quality of life questionnaire 8

9 List of abbreviations RC: regional control ROC: receiver-operating characteristics ROI: region of interest RP: retropharyngeal RPS: retropharyngeal space RR: relative risk RT: radiotherapy RTOG: radiation therapy oncology group SCC: squamous cell carcinoma SGL: supraglottic larynx SGS: salivary gland scintigraphy SEF: salivary excretion fraction SIB: simultaneous integrated boost SIC: signal-intensity curve SNC: sinonasal cancer SPC: superior pharyngeal constrictor muscle SPECT: single photon emission computed tomography SUV: standardized uptake value SWOG: Southwest oncology group T/B: tissue to blood ratio TGF-α: tumor growth factor α TROG: trans-tasman radiation oncology group TSE: turbo spin-echo UES: upper esophageal sphincter VF-MBS: videofluoroscopy with modified barium swallow 9

10

11 Table of contents Dankwoord Acknowledgments... 5 List of abbreviations... 7 Table of contents I. General Introduction 1. Head and neck cancer Epidemiology Treatment Radiotherapy Concomitant chemotherapy Radiotherapy for head and neck cancer: from 2D to 3D to IMRT Conformal radiotherapy Promises & pitfalls of IMRT Target definition Functional imaging in head and neck cancer FDG-PET DW-MRI Towards biological conformality? Reducing toxicity Xerostomia Dysphagia II. Objectives 1. General objective Specific aims Clinical application of IMRT Target definition Reducing toxicity III. Results A. Clinical application of IMRT 1. IMRT for pharyngo-laryngeal cancer Material & methods Patient selection

12 Table of contents Pre-treatment evaluation Simulation and treatment delivery Target volumes Dose prescription Organs at risk Chemotherapy Evaluation of toxicity and response Statistical analysis Results Experience with the hybrid fractionation schedule Impact of adding concomitant chemotherapy Intensity-modulated chemoradiotherapy Discussion IMRT for sinonasal cancer Material & methods Patient selection Pre-treatment evaluation Treatment Conventional and conformal radiotherapy IMRT Evaluation of toxicity and response Statistical analysis Results Conventional and conformal radiotherapy Initial results with post-operative IMRT Update of IMRT and comparison to conformal radiotherapy Discussion B. Target definition: selection of the nodal target volume 1. Retropharyngeal lymph nodes in oropharynx cancer Retropharyngeal lymph nodes Material & methods Patient population CT studies

13 Table of contents Treatment Statistical analysis Results Incidence of RP nodal involvement Anatomic localization of pathologic RP nodes Prognostic influence of pathologic RP node Discussion DW-MRI for lymph node staging & impact on radiotherapy planning Material & methods Patient population Imaging Surgery, topographic correlation, and histopathologic analysis Nodal GTV and CTV delineation Statistical analysis Results Neck dissection results Nodal staging Nodal GTV delineation Nodal CTV delineation Discussion C. Target definition: delineation of the primary tumor 1. Value of repeated functional imaging with FDG-PET, FMISO-PET, DW-MRI, and DCE-MRI 1.1. Material & methods Study design Image acquisition Image analysis Statistical analysis Results Imaging results Patterns of failure Discussion Predictive value of DW-MRI for dose-painting Material & methods

14 Table of contents Study design Image analysis Correlation to treatment outcome Statistical analysis Results Treatment outcome DW-MRI versus volumetric assessment: lesion-based analysis Correlation of DW-MRI and volumetric assessment with LRC Discussion D. Reducing toxicity 1. Radiation-induced xerostomia in HNC patients a literature review Pathophysiology Measurement Impact on quality of life Prevention Cytoprotectants Salivary gland-sparing radiotherapy Salivary gland transfer Treatment Conclusion DW-MRI to evaluate major salivary gland function before and after radiotherapy Material & methods Study design DW-MRI examinations Salivary gland scintigraphy Statsitical analysis Results Radiotherapy DW-MRI Discussion Dysphagia after CRT for HNC: dose-effect relationships for the swallowing structures Material & methods Patient selection and characteristics

15 Table of contents Evaluation of late dysphagia Clinical and dosimetric factors Statistical analysis Results DVH analysis Acute toxicity Late toxicity Quality of life results Discussion Evidence-based organ-sparing IMRT in head and neck cancer Material & methods Preventing xerostomia Sparing the parotid glands Sparing the submandibular glands Sparing the oral cavity and minor salivary glands Preventing dysphagia Summary IV. General conclusions and future perspective 1. IMRT for sinonasal cancer IMRT for pharyngo-laryngeal cancer Clinical application of IMRT Target definition Selection of the nodal target volume Delineation of a biological target volume Reducing toxicity Preventing xerostomia Preventing dysphagia Future perspective Summary Samenvatting Reference list Bibliography

16

17 I. GENERAL INTRODUCTION

18

19 I. General Introduction 1. Head and neck cancer 1.1. Epidemiology Head and neck cancer (HNC) is a broad term that encompasses all epithelial malignancies arising from the upper gastro-intestinal and airway tracts, thus including tumors originating from a wide variety of sub-sites such as the nasal cavity and paranasal sinuses, nasopharynx, oral cavity, oropharynx, larynx, hypopharynx, or salivary glands. It constitutes the fifth most common malignancy worldwide, representing about 6% of all cancers and yearly accounting for an estimated 47,560 new cases in the United States alone and at least 500,000 worldwide [1]. The overall majority of these epithelial malignancies are squamous cell carcinoma (SCC), for which the most important risk factors are tobacco and alcohol consumption [2 4]. Chronic exposure of the upper aero-digestive tract to these carcinogens is thought to produce field cancerization, a process in which patients are at risk for developing cancer at different mucosal sites [5, 6]. This occurs through the promotion of chromosomal deletion and point mutations of certain genes such as the p53 tumor suppressor gene. Transformation of clonally independent pre-malignant lesions progresses in a multi-step fashion through hyperplasia, dysplasia, and in situ carcinoma to invasive malignant lesions [7]. Other risk factors include viral infection, occupational exposure, radiation, dietary factors, and genetic susceptibility [2, 4]. Interestingly, there is increasing evidence documenting human papillomavirus (HPV), mainly HPV type 16 and to a lesser extent type 18, as an important cause of specific subsets of head and neck cancer, perhaps establishing a particular entity [8, 9]. Malignancies of the nasal cavity and paranasal sinuses (sinonasal cancer, SNC) represent a particular sub-site of HNC, with different epidemiology, treatment, and prognosis. They are relatively rare, representing 3 to 5% of all head and neck cancers and less than 1% of all malignancies [10, 11]. Worldwide, SCC is also the most frequent pathology in SNC [10 12]. Limited anatomical access makes early diagnosis difficult, while the presence of air-filled spaces permits asymptomatic growth until the invasion of adjacent structures produces symptoms. Therefore, most patients present with advanced disease, often extending into the base of skull, by the time of diagnosis [13, 14]. 19

20 I. General Introduction 1.2. Treatment Treatment for HNC is highly complex, not only because of the variety of disease subsites, but also because of the intricate anatomy, with normal and tumoral structures often in close proximity, and the importance of preserving organ function [15]. Patients require careful evaluation by a multidisciplinary team to determine optimal management. While radiotherapy (RT) and surgery remain the two main treatment options, systemic therapy has recently become an integral part of multidisciplinary treatment [16, 17]. A new class of targeted agents, the epidermal growth factor receptor (EGFR) inhibitors, has also shown clinical benefit in this disease [18, 19]. The choice of modality depends upon primary site, clinical stage, medical condition of the patient, resectability of the tumor, patient preference, and the experience of the treatment centre. Table 1. American joint committee on cancer stage groupings for HNC. Clinical Stage T-stage N-stage M-stage Stage 0 Tis N0 M0 Early stage: Stage I T1 N0 M % of cases - Single modality Stage II T2 N0 M0-5-yr survival 60 80% Stage III T3 N0 M0 Loco-regionally advanced stage: T1-3 N1 M0 - > 50% of cases Stage IVa T4a N0-1 M0 - Multimodality approach T1-4a N2 M0-5-yr survival 30 40% Stage IVb T4b any N M0 any T N3 M0 Stage IVc any T any N M1 < 10% of cases, palliation Approximately 30% to 40% of patients present with early (stage I and II) disease (Table 1). These patients can be effectively treated with either surgery or definitive radiotherapy. Both modalities result in similar rates of loco-regional control and survival, so the choice is usually based on the morbidity and functional outcome that can be expected [20]. However, more than half of patients present with loco-regionally advanced (stage III or IV) disease at diagnosis. Management of these patients requires aggressive and concerted 20

21 I. General Introduction measures, and remains a clinical challenge. Patients can be treated with complete surgical excision followed by post-operative (chemo-)radiotherapy or with primary (chemo-)radiotherapy. Until recently, 5-year survival rates were reported to be below 30% for patients with stage IV disease and 40% for all loco-regionally advanced tumors [21]. However, intensification of radiotherapy treatment using altered fractionation schedules and/or concomitant chemotherapy has resulted in significantly improved loco-regional control and survival rates. Malignancies of the nasal cavity and paranasal sinuses pose a major therapeutic challenge, as these tumors are usually diagnosed at an advanced stage and often lie very close to, or even invade, critical normal tissue structures such as eyes, optic nerves, optic chiasm, and brainstem. At the Leuven University Hospitals, patients with sinonasal cancer are treated in a multidisciplinary way, consisting of a combination of radiotherapy and surgery, since many decades. Primary RT as a definitive management is generally reserved for those patients who are medically or surgically inoperable or who refuse an operation. Traditionally, external resection (lateral rhinotomy with medial maxillectomy or cranio-facial resection) has been the surgical procedure of choice [22, 23]. Since 1992, endoscopic sinus surgery (ESS) is used whenever possible, since it results in comparable local control rates, but with lower morbidity and shorter hospital stay [24] Radiotherapy Radiation therapy for definitive treatment of head and neck cancer was conventionally given in daily fractions of Gy, to total doses of Gy over six or seven weeks. This schedule evolved through empirical experience, as a balancing act between tumor control and toxicity. However, there is emerging clinical and biologic evidence that loco-regional control can be improved by altered fractionation schedules, exploiting the differential sensitivity of tumor cells and normal tissue to radiation to increase therapeutic gain [25]. The European organization for research and treatment of cancer (EORTC) trial was the first to show improved loco-regional control with altered fractionation compared to conventional RT and thereby provided the starting platform for further research into fractionation [26]. Altered fractionation schemes can be loosely divided into three types: hyperfractionation (HF), accelerated fractionation (AF), and hybrid fractionation schedules. Hyperfractionation is based on the principle that, because late-responding normal tissues are more sensitive to fraction size than tumors, decreasing the size of each fraction (to Gy) should permit using a higher total radiation dose (typically Gy) without increasing 21

22 I. General Introduction late morbidity. In practice, multiple treatments with small fractions per day are given over approximately the same treatment duration. Accelerated fractionation attempts to reduce tumor repopulation as a major cause of RT failure. Clinical support for tumor cell proliferation during therapy is found in studies which have shown a loss in tumor control with prolongation of overall treatment time (OTT) in HNC [27]. In AF, the overall treatment time is shortened without reducing total dose or fraction size by giving conventional sized fractions more than once daily. EORTC 22811, comparing a fractionation schedule with three fractions per day to conventional RT, showed no differences in outcome between the treatment arms [28]. The Danish head and neck cancer study group (DAHANCA) trials 6 & 7 demonstrated that by using six instead of five fractions per week, it is possible to significantly improve 5-year loco-regional control from 60% to 70% [29]. In hybrid fractionation the treatment time is also reduced, but with changes in other variables such as fraction size, total dose, and time distribution. Continuous hyperfractionated accelerated radiotherapy (CHART) consists of an intensive course in which the OTT is greatly shortened together with a substantial decrease in total dose. In concomitant-boost RT, irradiation of the boost volumes (i.e. macroscopic tumor) is delivered on the same day as the elective volumes, in two different fractions. One important randomized trial by the radiation therapy oncology group (RTOG), comparing 3 types of altered fractionation, showed a significant loco-regional control and survival benefit with both hyperfractionation and concomitant-boost RT, but not with accelerated fractionation [30]. This observation is consistent with the results of a large meta-analysis by Bourhis et al., showing an absolute survival benefit of 8% at 5 years for schedules with increased total dose (HF), compared to only 2% for AF. The use of any type of altered fractionation schedule leads to an absolute survival benefit of 3.4% at 5 years [31]. The effect of altered fractionation is more pronounced on the primary tumor than on nodal disease. In the meta-analysis, there was a 23% reduction in the risk of local failure, with an absolute benefit of 8.5% at 5 years. For regional control, there was a 13% risk reduction and an absolute benefit of only 1.9% at 5 years. No effect on distant metastases could be detected [31]. It is clear that altered fractionation not only increases loco-regional control, but also survival, and that conventional RT can no longer be considered standard of care for advanced HNC [25]. Most altered fractionation schedules significantly increase acute toxicity, predominantly mucositis and dysphagia [25, 26, 28 30]. Late toxicity is generally not increased after modestly accelerated or hyperfractionated RT, although very accelerated trials did show sharply elevated late toxicity [26, 28]. These considerations formed the basis for a hybrid 22

23 I. General Introduction fractionation schedule that was implemented in 2001 at the Leuven radiotherapy department. A schedule was developed that is feasible in daily practice and results in adequate loco-regional control, with acceptable toxicity. Patients receive 20 daily fractions of 2 Gy (40 Gy), followed by 20 fractions of 1.6 Gy twice daily (32 Gy) up to a total dose of 72 Gy (Figure 1). Figure 1. The hybrid fractionation schedule used at the University Hospitals Leuven Concomitant chemotherapy Almost in parallel with the investigations into altered fractionation, clinical trials with the use of systemic chemotherapy (Cx) in HNC demonstrated successful anti-tumor activity [16]. Therefore, chemotherapy was increasingly added to primary radiotherapy, with most clinicians opting for a concurrent delivery of the two modalities. The rationale for this was that tumor cell clonogens could be sensitized to radiation by the concurrent delivery of Cx, thereby possibly enhancing loco-regional control. A large investigation from the meta-analysis of chemotherapy in head and neck cancer (MACH-NC) collaborative group, as well as its updates in 2004 and 2009, showed an absolute survival benefit of 6.5% at 5 years for concomitant chemoradiotherapy (CRT), making this approach the new standard of care [32 34]. The largest benefit was observed with platinum-based Cx, and no significant difference was seen between mono- or poly-chemotherapy [32 34]. Although it was initially hypothesized that by exposing both tumor and circulating micro-metastases to the drugs cytotoxic effect the incidence of metastases could also be diminished, no such effect was actually observed [32 34]. Obviously, the sensitizing effect of Cx is not selective for tumor cells, and adjacent normal tissues within the radiation field are also subjected to more effective and thus more toxic RT. Consistently, CRT trials report an increased incidence of toxicities, with mucositis, dysphagia, 23

24 I. General Introduction and dermatitis being the most prominent. In the majority of the randomized trials incorporated in these meta-analyses a conventional fractionation schedule was used. Supported by the apparent success of both altered fractionation and concomitant chemotherapy, more recent trials investigated the combination of both approaches. Some large studies showed a significant benefit when platinum-based Cx was combined with either hyperfractionated or concomitant-boost RT [35 37]. Based on these data, it became evident that the combination of altered fractionation and concomitant chemotherapy has the potential to further improve both loco-regional control and survival for HNC [38]. This approach was also adopted in the Leuven University Hospitals, where cisplatin (100 mg/m 2 ) is added in week 1 and 4 of the radiation treatment since Radiotherapy for head and neck cancer: from 2D to 3D to IMRT 2.1. Conformal radiotherapy The head and neck region is a complex site for RT, because normal (e.g. salivary glands or swallowing muscles) and tumoral structures generally lie in close proximity of each other. The introduction of conformal radiotherapy with three-dimensional treatment planning (3D-RT) on computed tomography (CT) scans signified a first major improvement over conventional twodimensional radiotherapy (2D-RT), where the treatment portals are based on a radiographic simulation film. Forward treatment planning is employed for both 2D-RT and 3D-RT, which essentially consists of the radiation oncologist designing the RT fields and the radiation physicist generating an optimized dose distribution. In contrast, inverse treatment planning is employed with intensity-modulated radiotherapy (IMRT) and consists of identifying the target volumes and organs at risk (OAR) on the planning CT. The dose to the tumor is specified, as well as the maximum acceptable doses to adjacent normal structures, and the physicist uses that information to produce the optimal plan, i.e. one that ensures target coverage by the prescribed radiation dose while reducing doses to the organs at risk. The basic principle behind IMRT is the use of intensity-modulated beams, i.e. beams with several dose intensity levels, essentially adding an additional dimension to the treatment planning. Usually, 5 to 7 non-opposing, co-planar beams are combined to sculpt the high-dose areas around the target volumes, with steep dose fall-off immediately outside these regions, thus allowing highly conformal radiation dose delivery. The most popular way to perform spatial modulation of beam intensity is through the use of a multileaf collimator (MLC), although other 24

25 I. General Introduction solutions exist (e.g. tomotherapy). A MLC consists of a high number (up to 120) of individual leaves that can move independently of each other. Step-and-shoot IMRT occurs when then leaves move when the beam is off, but do not move when the beam is on. Another solution is dynamic IMRT, whereby the leaves move continuously at various speed during irradiation Promises & pitfalls of IMRT The major advantage of IMRT is a more conformal dose distribution with steep dose gradients between the target volumes and the critical OAR, which should result in decreased toxicity and could possibly allow for the delivery of a higher radiarion dose to the tumor (i.e. dose-escalation), in an effort to improve loco-regional control [39]. The disadvantages of IMRT are: an increased risk of a marginal miss, a less homogenous dose distribution, a higher cost, and a higher total body dose because of leakage through the collimator and internal scatter as a result of increased beam-on time [40]. Consequently, IMRT should only be used when there appears to be a clear advantage over 3D-RT. Implementation of IMRT in clinical routine requires adequate compensation for setup uncertainties, appropriate selection and accurate delineation of target volumes, proper dose prescription with application of rigorous dose-volume constraints for organs at risk, and quality control of both the clinical and physical aspects of the whole procedure [41]. Indeed, there is increasing concern within the radiation oncology community that inappropriate and inept use of IMRT could very well result in worse results than before its wide-spread introduction [42 44]. At the Leuven department, IMRT has been implemented in selected HNC cases since Target definition The success of highly conformal radiotherapy techniques, such as IMRT, in the sparing of normal tissues and/or in dose-escalation relies heavily on accurate disease localization in individual patients. This is certainly true for HNC, where normal and tumoral structures are often in close proximity. According to the international commission on radiation units and measurements (ICRU) 50 guidelines, radiation oncologists use 3 different target volumes for radiotherapy planning: 1) gross tumor volume (GTV): gross extent of the malignancy as determined by all available means; 2) clinical target volume (CTV): GTV together with areas of possible subclinical microscopic disease; 3) planning target volume (PTV): CTV plus a margin to ensure that the CTV receives the prescribed dose [45]. 25

26 I. General Introduction For target volume delineation, anatomical imaging, i.e. computed tomography and magnetic resonance imaging (MRI), remains the most widely used modality. CT is broadly available, has a high spatial resolution, does not suffer from geometric distortion, and provides intrinsic information on the electronic densities of various tissues information that is used in dose calculation algorithms. CT allows clear delineation of tumors that border to air-filled cavities, fat tissue, or bone. However, CT lacks contrast resolution for differentiation between normal soft-tissue structures and tumor extent. This limitation can lead to significant inter- and intra-observer variations in tumor volume delineation of HNC [46]. Furthermore, CT images are degraded by the presence of dental fillings, which can complicate interpretation in oropharyngeal or oral cavity tumors. MRI with T1- or T2-weighted sequences is more accurate than CT for evaluating soft-tissue extent of HNC and is considered the standard imaging technique for nasopharyngeal and sinonasal cancer [47, 48]. However, no clear benefit of MRI was observed in pharyngo-laryngeal tumors [49]. It is to be expected that biochemical and molecular imaging will play an increasing role in radiation treatment planning of HNC, transforming both tumor delineation and target volume selection. 3. Functional imaging in head and neck cancer 3.1. FDG-PET The most commonly used functional imaging modality is undoubtedly [ 18 F]- fluorodeoxyglucose (FDG) positron emission tomography (PET). The increased glucose metabolism of cancer cells, as compared to normal tissues, is responsible for the enhanced uptake of FDG in malignant growth. The use of FDG-PET, and in particular integrated PET/CT, for radiotherapy planning in head and neck cancer has considerably increased in recent years [50]. The pivotal study in that respect was performed by Daisne et al. [51]. The authors compared pre-operative primary tumor (GTV) delineation based on CT, MRI, and FDG-PET with the surgical specimen ( gold standard ) in 9 laryngectomy patients. Although all imaging techniques overestimated the true GTV, FDG-PET appeared to be the most accurate modality. It should be noted however, that substantial parts of the surgical specimen (on average 10% on CT, 9% on MRI, and 13% on FDG-PET) were missed on each modality. Consequently, in-field failures can occur outside the FDG-avid regions of the anatomical GTV, and it appears prudent not to use FDG-PET as the sole instrument for GTV delineation [52]. However, CT, MRI, and FDG-PET can each add complimentary data to improve primary tumor delineation (Figure 2). 26

27 I. General Introduction Furthermore, the determination of the volume and shape of the tumor from PET images remains the subject of much controversy. Obviously, the easiest method consists of visual interpretation of the PET images [53, 54]. However, since the threshold level of the PET image, which depends on the display windowing, strongly influences the visual assessment of tumor boundaries, this is a rather subjective approach. The most commonly used objective method relies on the choice of a fixed threshold, i.e. a given percentage (usually 40 or 50%) of the maximal activity within the tumor, for distinguishing between malignant and normal tissues [55]. However, it was observed that the adequate threshold for fitting a surgical specimen, used as the gold standard, in a PET image varied between 36% and 73% of the maximal activity [51, 56]. A more dependable option, developed and validated by the Brussels group, is the use of an adaptive threshold, on the basis of the signal-to-background ratio [51, 57]. Figure 2. Contouring of the primary tumor in a patient with a ct4an1 squamous cell carcinoma of the piriform sinus on (a) contrast-enhanced CT, (b) gadolinium-enhanced T1-weighted MRI, and (c) FDG-PET, using the signal-to-background algorithm [51, 57]. Lymph node staging is at least as important as primary tumor delineation for radiotherapy planning of HNC, since it has implications on both therapeutic and elective target volumes. Currently, prophylactic coverage of large portions of potentially normal neck tissue is necessary to avoid undesirable marginal failure, negating some of the possible organ-sparing advantages of IMRT. Although anatomical imaging has adequate specificity, it lacks sufficient sensitivity and negative predictive value to allow for the individualized exclusion of clinically negative, at-risk nodal regions [58]. FDG-PET shows better sensitivity, although at the cost of some falsepositive results due to inflammation [59 65]. Moreover, FDG-PET has recognized spatial resolution limitations and cannot reliably identify disease < 0.5 cm in diameter [51, 59 65]. Consequently, results with FDG-PET do not differ significantly from those with CT or MRI [50]. 27

28 I. General Introduction 3.2. DW-MRI Diffusion-weighted magnetic resonance imaging (DW-MRI) is an imaging technique able to detect molecular diffusion, i.e. the Brownian motion of water molecules in biologic tissues. Thus, DW-MRI can characterize tissue and generate image contrast based on differences in water mobility. Diffusion-weighted images are obtained by applying pairs of opposing magnetic field gradients around the refocusing pulse of a T2-weighted sequence. Water molecules will be dephased by the first gradient and rephased by the second gradient. If the water molecules are stationary, no net dephasing is expected. Movement of the tissue water molecules between the two opposing gradients will result in dephasing, depicted as signal loss on the diffusion-weighted images [66]. This signal loss will be proportional to the amount of water molecule movement and the strength of the gradients (b-value). By repeating the sequence with different b-values, the observed signal loss can be quantified using the apparent diffusion coefficient (ADC). DW-MRI has already shown its value in tumor detection and treatment response evaluation of HNC [67, 68]. Some reports have also suggested its ability to accurately discriminate between different causes of cervical lymphadenopathy [69 71]. A pilot study by Vandecaveye et al. on the use of DW-MRI for cervical nodal staging in HNC found a sensitivity of 84% and a specificity of 94% per lymph node (LN) in surgically treated patients, using an ADC threshold of 0.94 x 10-3 mm 2 /sec. The technique appears especially promising in the detection of sub-centimetric lymph nodes (still 76% sensitivity and 94% specificity per LN), by definition rarely identified on conventional imaging and also challenging for FDG-PET because of its limited spatial resolution [72] Towards biological conformality? Recently, the notion of an additional biological target volume (BTV) was introduced [73]. As IMRT permits lowering the dose to the organs at risk outside the target volume, the maximum dose becomes restricted by the presence of dose-limiting structures within the target volume, such as cartilage, connective tissue, nerves, and bone. It was therefore suggested that, rather than requiring dose uniformity within the PTV, traditionally the mainstay of conformal RT planning, dose-escalation on areas of increased radioresistance within the tumor should be attempted in an effort to improve loco-regional control (Figure 3). Functional imaging, i.e. all modalities that offer information on factors that influence treatment outcome (e.g. proliferation, hypoxia), will provide the biological target for this dose-painting [74]. 28

29 I. General Introduction Figure 3. Whereas at present the RT target is defined as the anatomical GTV, biological imaging techniques may provide information for defining a BTV to improve dose targeting to certain regions of the tumor, as long as accurate registration is available. Some arguments exist to assume that FDG-avid regions within the tumor are of increased radioresistance. First of all, FDG-PET could hypothetically be considered as a crude marker for tumor proliferation, even if experimental data are currently lacking. Secondly, both in vitro and in vivo studies have shown increased FDG uptake in hypoxic circumstances, although a clinical study did not demonstrate a clear correlation between glucose metabolism and hypoxia [75]. However, results from several trials suggest that pre-treatment FDG uptake is correlated with outcome in HNC patients [76 78]. Madani et al. have taken the next step, and performed a phase I FDG-PET-guided dose-escalation study in HNC, which appeared well-tolerated [79]. Tumor hypoxia has been shown to be one of the major factors affecting treatment resistance in head and neck cancer [80]. As surrounding oxygen levels fall below 5 mmhg, cells become progressively more resistant to radiotherapy. The difference in radiosensitivity between 29

30 I. General Introduction aerobic and hypoxic cells is typically in the range of 2.5 to 3 (= oxygen enhancement ratio). In the absence of oxygen, radiation-induced radicals in DNA may be reversed by donation of hydrogen from non-protein sulfhydryls, leading to less net DNA damage, and thus less cell kill, for the same dose. Even a small fraction of hypoxic cells can dominate the radiotherapy response of the tumor, since the radiosensitive, aerobic cells will be rapidly eliminated, leaving the radioresistant, hypoxic cells. Non-invasive PET imaging evaluating the gross disease can provide serial quantitative measurement of hypoxia, providing a promising target for dose-painting. A number of potential exogenous hypoxic cell markers, labeled with positron-emitting radionuclides, have been studied, including [ 18 F]-fluoromisonidazole (FMISO), 60 Cu(II)-diacetyl-bis-N 4 - methylthiosemicarbazone (Cu-ATSM), [ 18 F]-fluoroerythronitroimidazole (FETNIM), and several others [81 84]. Of these tracers, FMISO is certainly the most developed. FMISO is a derivative of the nitroimidazole group of compounds, which have been investigated as hypoxic cell sensitizers. FMISO enters cells by passive diffusion, where it is reduced by nitroreductase enzymes to become trapped in cells with reduced tissue oxygen partial pressure. When oxygen is abundant in normally oxygenated cells, the parent compound is quickly regenerated by reoxidation and metabolites do not accumulate. However, in hypoxic celles, the low oxygen partial pressure prevents reoxidation of FMISO metabolites, resulting in tracer accumulation. Because FMISO accumulates in hypoxic cells with functional nitroreductase enzymes, this can only happen in viable hypoxic cells but not dead necrotic cells. The group from University of Washington was the first to determine the hypoxic fraction of tumors with FMISO-PET [81, 82]. Researchers from the same institution performed pre-treatment FMISO-PET scans in 73 HNC patients. Significant hypoxia was present in 58 (79%) patients, and both the degree of hypoxia and the size of the hypoxic volume were independent predictive factors for survival [85]. These data imply that FMISO-PET can be used to estimate the burden of hypoxia in HNC, and could provide a promising target for dose-painting [86, 87]. DW-MRI is a functional imaging technique that displays the extent and direction of random water motion in tissues, providing information on tissue cellularity and the integrity of cellular membranes. Preclinical and clinical data indicate a number of potential roles of DW-MRI in the characterization of malignancy, including determination of lesion aggressiveness and monitoring response to therapy [88]. Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) assesses changes in signal intensity after the intravenous injection of a paramagnetic contrast agent that reduces the T1 (longitudinal relaxation time) value of blood and thus enhances signal intensity on T1-30

31 I. General Introduction weighted imaging. Serial T1-weighted imaging samples the blood signal at intervals of a few seconds, for a period of up to several minutes after injection [89]. The temporal changes in signal intensity obtained by DCE-MRI are related to the underlying permeability and perfusion of the tumor microenvironment and the interstitial pressures within the tumor, all of which are known to influence response [90]. 4. Reducing toxicity The organs at risk in the head and neck region include spinal cord, brainstem, salivary glands, swallowing structures, and mandible, amongst others. For nasopharyngeal and sinonasal cancer, the eyes, optic nerves, optic chiasm, and temporal lobes of the brain are also at risk. Exceeding the tolerances of these structures can lead to cord or brainstem dysfunction, xerostomia, dysphagia, osteoradionecrosis, blindness, or brain necrosis. Thus, organ-sparing IMRT requires the appropriate selection and accurate delineation of multiple avoidance structures. However, defining dose-response curves, allowing the definition of reliable radiation dose constraints below which a complication is unlikely to occur, is a complicated process. Most available data are based on retrospective analyses, which often insufficiently correct for confounding clinical factors, and which use different endpoints to define a complication in disparate patient populations. Therefore, these analyses do not allow definitive conclusions. Moreover, there is increasing concern that inappropriate sparing of normal tissue, perhaps due to an overemphasis in recent literature on toxicity prevention rather than on tumor eradication, could lead to avoidable marginal recurrences [42 44]. Obviously, shielding clinically negative, at-risk regions from elective radiation to prevent damage to healthy tissue should be approached with extreme care: adequate selection of patients is of crucial importance and loco-regional recurrences need to be carefully evaluated and reported. Since it is clear that both xerostomia and dysphagia are the main causes of decreased quality of life (QoL) after radiotherapy for HNC, most studies focused on the prevention of these two complications [91] Xerostomia Xerostomia is considered to be the most prominent complication after radiotherapy for head and neck cancer. Radiation-induced damage to the salivary glands alters the volume, consistency, and ph of secreted saliva. Saliva changes from thin secretions with a neutral ph to thick and tenacious secretions with increased acidity. Patients suffer from oral discomfort or 31

32 I. General Introduction pain, find it difficult to speak, chew or swallow, and run an increased risk of dental caries or oral infection. Ultimately, this can lead to decreased nutritional intake and weight loss. Radiationinduced xerostomia not only significantly reduces the quality of life of potentially cured patients, but also poses a major new health problem for them [92] Dysphagia Post-radiotherapy dysphagia is due to neuromuscular fibrosis and radiation-induced edema, leading to abnormal motility of deglutition muscles such as impaired pharyngeal contraction and laryngeal elevation. Probably, sensory changes in the oral cavity and the pharynx also play a role by changing the patient s perception of swallowing. Swallowing dysfunction after radiotherapy is correlated with compromised quality of life, and can lead to lifethreatening complications such as aspiration pneumonia [93 95]. Because the risk of radiation-induced dysphagia is associated with the use of concomitant chemotherapy and accelerated fractionation schedules, its incidence has considerably increased in recent years [96 98]. Currently, the impact of late dysphagia on health-related QoL after radiotherapy is at least as important as that of permanent xerostomia [91]. These observations have led some to suggest that late dysphagia is the dose-limiting toxicity of chemoradiotherapy and constitutes the main obstacle towards further treatment intensification [99, 100]. 32

33 II. OBJECTIVES

34

35 II. Objectives 1. General objective The main goal of this thesis was to evaluate the clinical application of IMRT in HNC to gain a better understanding of its strengths and weaknesses, to improve target volume definition and delineation in order to obtain better tumor control, and to increase its organ-sparing potential. 2. Specific aims 2.1. Clinical application of IMRT - To compare clinical outcome (tumor control and toxicity) of primary radiotherapy for advanced pharyngo-laryngeal cancer, with or without concomitant chemotherapy, delivered through 3D-RT versus IMRT. - To compare clinical outcome (tumor control and toxicity) of radiotherapy for sinonasal cancer delivered through 2D-RT or 3D-RT versus IMRT Target definition - To evaluate the importance of retropharyngeal lymph node involvement in oropharyngeal cancer. - To investigate the impact of nodal staging with DW-MRI on treatment planning. - To compare several functional imaging modalities (FDG-PET, FMISO-PET, DWI-MRI, DCE-MRI) regarding their suitability for GTV and BTV delineation Reducing toxicity - To review the current state of the art regarding salivary gland-sparing radiotherapy, and the value of DW-MRI as a non-invasive tool to evaluate major salivary gland function. - To assess the incidence of late dysphagia after chemoradiotherapy and to examine its correlation with clinical and dosimetric parameters. - To provide an evidence-based manual on organ-sparing IMRT. 35

36

37 III. RESULTS

38

39 A. CLINICAL APPLICATION OF IMRT

40

41 CHAPTER 1: IMRT FOR PHARYNGO-LARYNGEAL CANCER Published as: Nuyts S., Dirix P., Hermans R., Vander Poorten V., Delaere P., Weltens C., Van den Bogaert W. Early experience with a hybrid accelerated radiotherapy schedule for locally advanced head and neck cancer. Head Neck 2007; 29(8): Nuyts S., Dirix P., Clement P.M.J., Vander Poorten V., Delaere P., Schoenaers J., Hermans R., Van den Bogaert W. Impact of adding concomitant chemotherapy to hyperfractionated accelerated radiotherapy for advanced head and neck squamous cell carcinoma. Int J Radiat Oncol Biol Phys 2009; 73(4): Dirix P., Nuyts S. Value of intensity-modulated radiotherapy in stage IV head and neck squamous cell carcinoma. Int J Radiat Oncol Biol Phys; In press.

42

43 A. Clinical application of IMRT 1. IMRT for pharyngo-laryngeal cancer To evaluate a new technique or treatment, it is necessary to compare it to the previous standard. Consequently, we performed a retrospective analysis of all patients with advanced HNC treated with 3D-RT at our department. First, patients treated between 2001 and 2004 according to the hybrid fractionation schedule (72 Gy in 6 weeks) were evaluated with regard to clinical outcome and toxicity. Secondly, these results were compared to patients treated with the same fractionation schedule and concomitant Cx (cisplatinum 100 mg/m 2 every 3 weeks), from 2004 to At the same time, all patients treated with IMRT were prospectively followed Material & methods Patient selection All patients presenting with loco-regionally advanced (clinical stage III, IVa, or IVb) head and neck cancer at the Leuven University Hospitals between 2001 and 2008 were evaluated by a multidisciplinary team consisting of ear-nose-throat (ENT) head & neck surgeons, maxillofacial surgeons, head and neck radiologists, medical oncologists, and radiation oncologists. Treatment was decided between (1) primary surgery followed by post-operative radiotherapy versus (2) primary radiotherapy, according to institutional guidelines. Primary radiotherapy was delivered according to the hybrid fractionation schedule; from 2004 onwards patients also received concomitant chemotherapy. Between February 2001 and December 2008, 191 consecutive HNC patients were treated with primary radiotherapy Pre-treatment evaluation Pre-treatment evaluation consisted of complete history and physical examination, routine blood counts, liver function tests, ultrasound scan of the abdomen, chest X-ray, oesophagogastroscopy, and tumor biopsy. All patients were imaged with computed tomography and/or magnetic resonance imaging of the head and neck region. Bone scans, PET scans, and CT scans of the abdomen or chest were only obtained when clinically indicated. Staging was performed according to the 2002 TNM classification system of the American joint committee on cancer (AJCC) [101]. Dental evaluation, with the necessary interventions, was performed in each patient before the start of RT. 43

44 A. Clinical application of IMRT Simulation and treatment delivery At the time of simulation, patients were immobilized with a thermoplastic head, neck, and shoulders mask with five-point fixation (Posicast, Sinmed, Reeuwijk, The Netherlands), and treatment planning CT scans were taken. Serial CT scan slices, 3 mm thick, from the head down to the clavicles were obtained after intravenous (IV) injection of a contrast agent. The CT images were transferred to the treatment planning system, and the treatment target volumes and critical organs at risk were outlined in each CT slice as described below. Treatment was delivered with 6 or 10 megavoltage (MV) photon of a linear accelerator. A three-dimensional conformal radiotherapy technique was used in 149 (78.0%) patients. In 40 of those 149 patients, a 4-field technique was used to spare the contralateral parotid gland [102]. The other 109 patients were treated with 2 opposing lateral beams and 1 lower neck field for the supraclavicular regions. Forty-two (22.0%) patients were treated with intensity-modulated radiotherapy. The Helios software for inverse IMRT planning (sliding window), integrated in the commercial Eclipse (Varian, Palo Alto, CA) treatment planning system, was used, with the single pencil beam algorithm and modified batho heterogeneity correction. Usually, 5 to 7 nonopposing, co-planar beams were chosen to ensure that 95% of the dose encompassed the target volume Target volumes The GTV included any visible disease (primary tumor as well as lymph nodes) on imaging studies and/or physical examination. The CTV encompassed a 10 mm margin, with appropriate anatomical correction, around the GTV (CTV-boost). The elective nodal CTV was defined according to our institutional guidelines (Table 2), based on the validated proposals by Eisbruch and Chao [ ]. Nodal levels were delineated according to the consensus guidelines [106]. For the node-negative neck, the retrostyloid space was not included in the CTV and the upper limit of level II was placed where the posterior belly of the digastric muscle crosses the jugular vein [104]. For the node-positive neck, the CTV was extended to the base of skull (jugular foramen) and included the retrostyloid space [107]. The planning target volumes (PTV-boost and PTV-elective) encompassed the CTV-boost and CTV-elective plus a 5 mm margin, respectively. Modification of the PTV was done if it extended outside of the skin or overlapped with the spinal cord. 44

45 A. Clinical application of IMRT Dose prescription Patients were treated according to the hybrid fractionation schedule to a total dose of 72 Gy. The initial target volumes, encompassing both boost and elective target volumes, were treated with daily single doses of 2 Gy, five days per week over 4 weeks to 40 Gy (days 1 to 26). On days 29 and 30, 2 sessions of 1.6 Gy per day were given with a minimal interval of 6 hours, to a total dose of 46.4 Gy. From day 31 onward, the boost target volumes were treated. Two daily fractions of 1.6 Gy were administered (again with a minimal interval of 6 hours) for another 8 days, resulting in a total tumor dose of 72 Gy. Table 2. Institutional guidelines for elective nodal CTV delineation in stage III/IV HNC. Primary tumor site Clinical stage Ipsilateral levels Contralateral levels 1. Oral cavity T3-4 and/or N+ Ib-II-III-IV Ib-II-III-IV - anterior localization + Ia + Ia - level II LN + V + V 2. Oropharynx T3-4 and/or N+ - level II LN - T4/tongue invasion 3. Glottic larynx T3-4 and/or N+ - level II LN - subglottic extension 4. Supraglottic larynx T3-4 and/or N+ - level II LN - oropharynx invasion - subglottic extension 5. Hypopharynx T3-4 and/or N+ - T4 - level II LN RP-II-III-IV-V + Ib + Ib II-III-IV + Ib & V + VI II-III-IV + Ib & V + RP + VI RP-II-III-IV-VI + V + Ib & V RP-II-III-IV-V + Ib + Ib II-III-IV + Ib & V + VI II-III-IV + Ib & V + RP + VI RP-II-III-IV-VI + V + Ib & V Abbreviations: N+ = node-positive disease; RP = retropharyngeal lymph nodes Organs at risk The spinal cord, brainstem, parotid glands, submandibular glands, and mandible were delineated in all patients. In selected patients, the inner ear, middle and external ear, temporomandibular joints, oral cavity, pharyngeal constrictor muscles, esophagus, or brachial plexus were outlined if clinically appropriate. In cases without risk of laryngeal involvement, the larynx was also outlined as an organ at risk. The cumulative dose to the spinal cord was not to exceed 50 Gy. In those patients where parotid-sparing was feasible, the mean dose to at least one parotid gland was kept below 26 Gy [102]. 45

46 A. Clinical application of IMRT Chemotherapy Between February 2004 and December 2008, 118 patients received cisplatinum (100 mg/m 2 ) intravenously, on the first day of week 1 and 4 over 1 hour. Calculated creatinine clearance was checked before each administration. Application of chemotherapy was delayed or discontinued in case of absolute neutrophil count < 1000/nL or thrombocytes < 100/nL to prevent radiotherapy breaks Evaluation of toxicity and response All patients were evaluated by their treating radiation oncologist at least once a week, or more frequently if required, during the radiotherapy course. Patients then returned for follow-up visits until regression of acute toxicity (approximately 8 weeks after the end of treatment). Acute toxicity was graded according to the common terminology criteria for adverse events (CTCAE) version 3.0. Patients were advised to use intensive mouthwashes; oral candidiasis was treated with a local nystatine suspension or systemic ketoconazole. Pain medication was started by the treating radiation oncologist when indicated. Similarly, patients were weekly seen by a specialized nurse, who advised patients in their dietary intake. In patients treated without concomitant chemotherapy, no percutaneous endoscopic gastrostomy (PEG) tube was placed before treatment. The decision to hospitalize and/or to place a PEG tube for nutritional support was left to the discretion of the treating radiation oncologist. Patients treated with CRT were consistently hospitalized during the first 2 to 3 days of weeks 1 and 4, for the administration of chemotherapy. The decision to hospitalize the patient at any other time-point was left to the discretion of the treating radiation oncologist. In the CRT patients, a PEG tube was placed at the start of treatment, to guarantee sufficient nutrition. Afterwards, patients were regularly followed at the multidisciplinary outpatient clinic: every 2 months the first 2 years after treatment, every 3 months the third year, every 4 months the fourth year, every 6 months the fifth year, and then every year. Late toxicity was graded according to the RTOG/EORTC late radiation morbidity scoring schema. Patients did not receive routine neck dissections after treatment, although surgery was considered if patients had evidence of progressive residual disease or recurrence. Response evaluation consisted of physical examination and CT scan at 2 to 3 months after treatment. Thereafter, a CT scan was done yearly or at the discretion of the treating oncologist or ENT head & neck surgeon. 46

47 A. Clinical application of IMRT Statistical analysis Patient and tumor characteristics were recorded at the start of treatment. Toxicity was prospectively scored. Follow-up data were retrospectively collected: information was sought on the date of first recurrence, distant metastasis, and/or death. The data were analyzed using the software package Statistica 8 (StatSoft Inc, Tulsa, OK, USA). Cumulative survival and tumor control rates were calculated using the Kaplan-Meier product-limited (actuarial) method. Patient groups were compared for age, gender, primary site, T-classification, N-classification, and clinical stage using the non-pared student t-test. Survival and recurrence rates were compared using the log-rank test. Incidence of toxicity was compared using the non-pared student t-test. A p value of less than 0.05 was considered statistically significant Results Experience with the hybrid fractionation schedule From February 2001 to March 2004, 73 patients with biopsy-proven HNC were treated with a hybrid fractionation schedule, without concomitant chemotherapy. Median follow-up was 21 months (range, 1 57 months) for all patients, and 43 months (range, months) for patients still alive at the close-out date (March 2006). Patient characteristics Patient and tumor characteristics are summarized in Table 3. There were 62 (84.9%) male and 11 (15.1%) female patients, with a mean age of 58.9 years (range, years); the median age was 59 ± 10,4 years. The majority of patients (54.8%) presented with a tumor of the oropharynx. In the other patients, the tumor was located in the larynx (32.9%), hypopharynx (8.2%), or oral cavity (4.1%). All patients presented with locally (> T2 tumors: 76.7%) and/or regionally (N +: 72.6%) advanced tumors. 47

48 A. Clinical application of IMRT Table 3. Patient and tumor characteristics. Characteristics No. (%) of CRT patients No. (%) of RT patients p Age 0.09 Median age (range) 56 (38 76) 59 (34 76) 60 years 30 (33.3%) 34 (46.6%) < 60 years 60 (66.7%) 39 (53.4%) Gender 0.64 Male 74 (82.2%) 62 (84.9%) Female 16 (17.8%) 11 (15.1%) Primary tumor site 0.25 Oral cavity 9 (10.0%) 3 (4.1%) Oropharynx 49 (54.5%) 40 (54.8%) Larynx 21 (23.3%) 24 (32.9%) Supraglottis 20 (22.2%) 22 (30.1%) Glottis 1 (1.1%) 1 (1.4%) Subglottis 0 (0%) 1 (1.4%) Hypopharynx 11 (12.2%) 6 (8.2%) Tumor grade - Grade I 4 (4.4%) 5 (6.9%) Grade II 24 (26.7%) 24 (32.9%) Grade III 28 (31.1%) 22 (30.1%) Unknown 34 (37.8%) 22 (30.1%) T classification 0.10 T1 6 (6.7%) 4 (5.5%) T2 21 (23.3%) 13 (17.8%) T3 35 (38.9%) 24 (32.9%) T4a 24 (26.7%) 26 (35.6%) T4b 4 (4.4%) 6 (8.2%) N classification 0.03 N0 11 (12.2%) 20 (27.4%) N1 18 (20.0%) 16 (21.9%) N2a 1 (1.1%) 1 (1.4%) N2b 28 (31.1%) 12 (16.4%) N2c 30 (33.3%) 14 (19.2%) N3 2 (2.2%) 10 (13.7%) Clinical stage 0.25 Stage II - 6 (8.2%) Stage III 21 (23.3%) 16 (21.9%) Stage IV 69 (76.7%) 51 (69.9%) Total

49 A. Clinical application of IMRT Disease control and survival There was a 90.4% (n = 66) complete response (CR) rate to primary radiotherapy with the hybrid fractionation schedule. Salvage surgery of the primary site and/or neck was considered for 7 patients (9.6%) with residual disease after the end of treatment. Five patients did not undergo surgery because of the following reasons: extensive inoperable disease (n = 2), presence of distant metastasis (n = 1), or severe coexisting medical condition (n = 2). Two patients underwent salvage surgery: one patient had resection of both primary and neck disease, but developed extensive loco-regional and distant recurrence 3 months later. The other patient underwent salvage surgery for the primary site alone, although he nevertheless quickly relapsed locally. All 7 patients without CR died within 6 months after the end of RT. Of the remaining 66 patients who did show complete clinical and radiological remission after RT, 39% developed a loco-regional recurrence after 2 years. The majority of those recurrences (82.1%) were local relapses, at the primary tumor site. Loco-regional control (LRC) of the entire population of 73 patients was 55% at 2 years (Figure 4). Loco-regional control was significantly correlated with T-classification (p = 0.017) in univariate analysis. Distant metastasis in the overall population of 73 patients was 22% after 2 years. There was a significant correlation with clinical stage (p = 0.03) and N-classification (p = 0.03) in univariate analysis, although not in multivariate analysis. Overall survival (OS) was 59% at 2 years (Figure 5), significantly correlated with clinical stage (p = 0.004) and T-classification (p = 0.007) in univariate analysis. In multivariate analysis, clinical stage was the only statistically significant factor (relative risk (RR) 2.45 [95% confidence interval (CI) ], p = 0.03). Disease-specific survival (DSS) in the overall population was 63% at 2 years. A statistically significant correlation was found with clinical stage (p = 0.001), T-classification (p = 0.001), and N-classification (p = 0.04). In multivariate analysis, clinical stage (RR 3.80 [95% CI ], p = 0.01) was the only factor to independently influence disease-specific survival. Disease-free survival (DFS) was 46% after 2 years, significantly correlated with clinical stage (p = 0.03), T-classification (p = 0.004), and N- classification (p = 0.03) in univariate analysis. In multivariate analysis, only T-classification had a statistically significant impact on DFS (RR 2.13 [95% CI ], p = 0.03). 49

50 A. Clinical application of IMRT Figure 4. Loco-regional control. 100% Loco-regional control (LRC) in the 2 treatment arms. 90% 80% 70% LRC (%) 60% 50% 40% 30% 20% 10% RT (n = 73) CRT (n = 90) p = 0,03 0% Time (months) Numbers at risk (Loco-regional control) in the RT-only group (n = 73) Numbers at risk (Loco-regional control) in the CRT group (n = 90). 50

51 A. Clinical application of IMRT Figure 5. Survival in the RT group. 100% Survival 90% 80% 70% 60% 50% 40% 30% 20% 10% Overall survival (OS) Disease-specific survival (DSS) Disease-free survival (DFS) 0% O Time (months) Numbers at risk (Overall survival) Toxicity The most frequent acute toxicities were related to mucosa, skin and pharynx (Table 4). Mucositis was scored as grade 1 in 2.7% (n = 2) of patients, as grade 2 in 42.5% (n = 31), as grade 3 in 50.7% (n = 37), and as grade 4 in 4.1% (n = 3) of all patients. Dermatitis was grade 1 in 25.5% (n = 19), grade 2 in 40% (n = 29), and grade 3 in 34.5% (n = 25) of patients. Dysphagia was grade 1 in 17.8% (n = 13), grade 2 in 34.3% (n = 25), and grade 3 in 47.9% (n = 35) of patients. No grade 4 dermatologic toxicity or dysphagia was reported. All 73 patients could receive the full dose of 72 Gy without treatment interruption (100% compliance). Overall, 58.9% (n = 43) of the 73 patients were admitted to the hospital during the irradiation course. A few patients (n = 8) were hospitalized because they lived too far from the hospital or had too little support at home to allow out-patient treatment, although their medical 51

52 A. Clinical application of IMRT condition did not necessitate admission. The majority (n = 35) of hospitalized patients were admitted because of severe dysphagia, necessitating feeding support. The median duration of hospitalization because of feeding problems was 24 days (range, 2 78 days). Of all 73 patients, 21.9% (n = 16) lost less than 4% of their initial weight, 53.4% (n = 39) lost between 4 and 10% of their initial weight and 24.7% (n = 18) of patients lost more than 10 % of their initial weight. Overall, 47.9% (n = 35) of patients required feeding support during or immediately after treatment. Of those patients, 7 were treated with nasogastric tube placement and 16 patients received IV feeding. Of the 12 patients who required a percutaneous endoscopic gastrostomy, 2 had PEG tubes placed before treatment started but were admitted to initiate tube feeding. The median duration of the feeding support with PEG tubes was 31 weeks (range, 2 56 weeks); 8 patients required long-term feeding support because of ongoing swallowing difficulties. Late toxicity was available for 38 patients. No grade 4 toxicity was reported for any of the scored complications (Table 5), and, although late radiation morbidity of bone was not actively scored, no patients suffered from osteonecrosis of the jaw and/or spontaneous fracture (grade 4). The most frequent complication was severe xerostomia (13.2% of the scored patients had grade 3 xerostomia), although almost half (44.7%) had no xerostomia at all. Less frequent late morbidities were swallowing difficulties (5.3% grade 3), laryngeal edema and/or chondritis (7.9% grade 3), and atrophy of the mucosa (7.9% grade 3). No grade 3 toxicity was scored for either skin or subcutaneous tissue. 52

53 A. Clinical application of IMRT Table 4. Acute radiation toxicity. Adverse event No. (%) of CRT patients No. (%) of RT patients p n = 90 n = 73 Dermatitis 0.57 Grade Grade 1 18 (20.0%) 19 (25.5%) Grade 2 45 (50.0%) 29 (40.0%) Grade 3 27 (30.0%) 25 (34.5%) Grade Mucositis Grade Grade 1 3 (3.3%) 2 (2.7%) Grade 2 20 (22.2%) 31 (42.5%) Grade 3 67 (74.5%) 37 (50.7%) Grade 4-3 (4.1%) Nausea - Grade 0 57 (63.3%) not scored Grade 1 16 (17.8%) not scored Grade 2 11 (12.2%) not scored Grade 3 6 (6.7%) not scored Grade 4 - not scored Salivary glands < Grade 0-6 (8.2%) Grade 1 7 (7.8%) 40 (54.8%) Grade 2 81 (90.0%) 27 (37.0%) Grade 3 2 (2.2%) - Grade Dysphagia < Grade Grade 1 2 (2.2%) 13 (17.8%) Grade 2 14 (15.6%) 25 (34.3%) Grade 3 74 (82.2%) 35 (47.9%) Grade Pain Grade 0-5 (6.8%) Grade 1 16 (17.8%) 33 (45.2%) Grade 2 48 (53.3%) 28 (38.4%) Grade 3 26 (28.9%) 7 (9.6%) Grade

54 A. Clinical application of IMRT Table 5. Late radiation toxicity. Organ tissue No. (%) of CRT patients No. (%) of RT patients p n = 83 n = 38 Skin 0.28 Grade 0-21 (55.3%) Grade 1 72 (86.7%) 9 (23.7%) Grade 2 11 (13.3%) 8 (21.0%) Grade Grade Subcutaneous tissue 0.58 Grade 0-10 (26.3%) Grade 1 57 (68.7%) 18 (47.4%) Grade 2 26 (31.3%) 10 (26.3%) Grade Grade Mucous membrane 0.94 Grade 0 16 (19.3%) 20 (52.6%) Grade 1 50 (60.2%) 10 (26.3%) Grade 2 16 (19.3%) 5 (13.2%) Grade 3 1 (1.2%) 3 (7.9%) Grade Salivary glands < Grade 0-17 (44.7%) Grade 1 9 (10.8%) 9 (23.7%) Grade 2 56 (67.5%) 7 (18.4%) Grade 3 18 (21.7%) 5 (13.2%) Grade Larynx 0.63 Grade 0 40 (48.2%) 17 (44.7%) Grade 1 13 (15.7%) 9 (23.7%) Grade 2 30 (36.1%) 9 (23.7%) Grade 3-3 (7.9%) Grade Esophagus 0.96 Grade 0 39 (47.0%) 21 (55.3%) Grade 1 24 (28.9%) 8 (21.0%) Grade 2 16 (19.3%) 7 (18.4%) Grade 3 4 (4.8%) 2 (5.3%) Grade

55 A. Clinical application of IMRT Impact of adding concomitant chemotherapy From February 2004 to May 2007, 90 patients with biopsy-proven HNC were treated with a hybrid fractionation schedule and concomitant chemotherapy. Median follow-up was 24 months (range, 5 48 months) for all patients, and 29 months (range, 7 48 months) for patients still alive at the close-out date (January 2008). Results were retrospectively compared with the previous cohort of 73 patients who received primary RT with the same fractionation schedule, although without concomitant chemotherapy. Patient characteristics Patient and tumor characteristics are summarized in Table 3. There were 74 male and 16 female patients, with a median age of 56 years (range, years). Only patients with a squamous cell carcinoma of the oral cavity (10.0%), oropharynx (54.5%), larynx (23.3%), or hypopharynx (12.2%) were eligible for this protocol. All patients suffered from locally (> T2: 70.0%) and/or regionally (N+: 87.8%) advanced tumors. Disease control and survival The actuarial estimate of LRC was 70% after 2 years (Figure 4). Of the 90 patients, 20 (22.2%) developed a local failure during follow-up, with a median time to local recurrence of 9 months (range, 2 40 months). Actuarial local control (LC) was 76% at 2 years. Thirteen (14.4%) patients developed a regional recurrence, after a median interval of 10 months (range, 4 33 months). Regional control (RC) was 82% at 2 years. Salvage surgery was attempted in 8 patients with local (n = 4), regional (n = 1), or loco-regional (n = 3) persistent or recurrent tumor. Complete surgical resection was achieved in all 8 patients. Five of those patients remained disease-free after a median interval since re-treatment of 11 months (range, 3 36 months). One patient developed an isolated regional recurrence at 19 months after salvage surgery, for which he was re-irradiated to 60 Gy with concomitant cisplatinum, and remained disease-free at 7 months after re-irradiation. Two patients quickly developed a second loco-regional recurrence after salvage surgery and ultimately died from their disease. No unusual toxicity, other than what could normally be expected from the type of operation, was observed after salvage surgery, apart from one patient who developed a fistula from the neopharynx to the trachea after complete laryngectomy and another patient who had skin necrosis after neck dissection. Fifteen 55

56 A. Clinical application of IMRT (16.7%) patients were diagnosed with distant metastases, after a median time of 9 months (range, 4 22 months). The actuarial estimate of distant control (DC) was 77% at 2 years. The actuarial estimate of OS was 74% at 2 years (Figure 6). Twenty-six (28.9%) patients died during follow-up, after a median interval of 10 months (range, 5 41 months). Twenty patients died from HNC, 6 patients died from another cause. DSS was 78% at 2 years. Fifty-four (60%) patients were still alive and disease-free after a median interval of 21 months (range, 7 47). Actuarial DFS was 60% at 2 years. Comparing treatment outcome with the cohort of 73 patients treated without concomitant chemotherapy, an 15% improvement in both loco-regional control (70% vs. 55%, p = 0.03, Figure 4) and overall survival (74% vs. 59%, p = 0.09) was observed. Distant control was identical in both groups (77%, p = NS). Figure 6. Survival in the CRT group. 100% Survival 90% 80% 70% 60% 50% 40% 30% 20% 10% Overall survival (OS) Disease-specific survival (DSS) Disease-free survival (DFS) 0% Time (months) Numbers at Risk (Overall Survival) 56

57 A. Clinical application of IMRT Toxicity The prescribed dose of 72 Gy was delivered in all 90 CRT patients according to protocol; a significant treatment interruption (i.e. longer than 2 consecutive days) occurred in 1 patient, due to an acute myocardial infarction (median OTT was 42 days; range, 37 55). Chemotherapy could be administered according to protocol in 78 (86.7%) patients. In 3 patients, the second cycle had to be postponed for a week and in 4 patients, the second cycle had to be cancelled altogether because of severe ototoxicity with hearing loss (n = 1), patient refusal (n = 1), or local infection of the PEG tube (n = 2). In 1 patient, cisplatinum was administered weekly at 40 mg/m 2 and in 4 patients, cisplatinum was substituted by carboplatinum-fluorouracil because of renal function concerns. The most frequent acute toxicities were dysphagia (82.2% grade 3), mucositis (74.5% grade 3), and dermatitis (30.0% grade 3); no grade 4 or 5 acute toxicity was reported in the chemoradiotherapy group (Table 4). Ten patients (11.1%) also complained of hearing loss during or immediately after treatment. Compared to the previous cohort of patients treated with RT alone, the incidence of grade 3 mucositis (74.5% vs. 50.7%, p = 0.002), grade 2 3 xerostomia (92.2% vs. 37.0%, p < 0.001), grade 3 dysphagia (82.2% vs. 47.9%, p < 0.001), and grade 3 pain (28.9% vs. 9.6%, p = 0.002) was significantly higher in the CRT group. Thirty-eight (42.2%) patients were only hospitalized for the administration of chemotherapy at the beginning of week 1 and week 4 (median total hospital stay of 4.5 days; range, 2 7 days). Four (4.4%) patients were hospitalized during the entire treatment and 13 (14.5%) patients during the last two weeks (2 fractions per day) because they lived too far from the hospital or had too little support at home to allow out-patient treatment, although their medical condition did not necessitate admission. Thirty-five (38.9%) patients needed additional hospitalization because of medical reasons. Those patients were admitted because of severe mucositis and important weight loss (n = 21), neutropenic fever and infection (n = 9), severe nausea and vomiting (n = 3), or PEG-related problems (n = 2). The median duration of this additional hospitalization was 8.5 days (range, 2 35 days). Of all 90 patients, 28.9% (n = 26) lost less than 5% of their initial weight, 38.9% (n = 35) lost between 5 and 10% of their initial weight, and 32.2% (n = 29) of patients lost more than 10% of their initial weight. The first 3 patients did not receive a PEG tube, and consequently needed IV feeding support during treatment. Thereafter, a PEG tube was placed before the start of treatment in all patients. Sixteen patients (17.8%) did not use their PEG tube and completed RT with a modified diet and nutrition supplements. The PEG tube remained in place for a median 57

58 A. Clinical application of IMRT duration of 5 months (range, days). Of the 54 currently disease-free patients, 5 (9.3%) have a PEG tube in place, after a median time of 9 months (range, 8 11 months) since the end of treatment. Late toxicity could be evaluated in 83 (92.2%) patients with a minimum event-free followup of 6 months (Table 5). No grade 4 or 5 late toxicity was reported; 21.7% of patients suffered from grade 3 xerostomia, and 4.8% reported grade 3 dysphagia. Compared to the RT-only group, there was no significant difference in the incidence of late grade 2 3 mucositis (20.5% vs. 21.1%, p = 0.94), laryngitis (36.1% vs. 31.6%, p = 0.63), or esophagitis (24.1% vs. 23.7%, p = 0.96). However, there was a higher incidence of late grade 2 3 xerostomia (89.2% vs. 31.6%, p < 0.001). For 41 patients, event-free follow-up of more than 12 months was available. Of those patients, 25 (61.0%) complained of some degree of xerostomia and 6 (14.6%) patients complained of dysphagia. None of those patients had a PEG tube in place. There were no cases of bone or soft-tissue necrosis Intensity-modulated chemoradiotherapy From January 2006 to December 2008, 42 patients with stage IV, histologically proven head and neck cancer were treated with IMRT according to a hybrid fractionation schedule with concomitant chemotherapy. Median follow-up was 15 months (range, 5 40 months) for all patients, and 20 months (range, 6 40 months) for patients still alive at the close-out date (July 2009). Results were retrospectively compared with a previous cohort of 55 patients with stage IV head and neck cancer, treated with conformal radiotherapy without intensity-modulation to the same chemoradiotherapy schedule. Patient characteristics Patient and tumor characteristics are summarized in Table 6. There were 36 male and 6 female patients, with a median age of 56 years (range, years). Only patients with a squamous cell carcinoma of the oral cavity (14.3%), oropharynx (28.6%), larynx (21.4%), or hypopharynx (35.7%) were eligible for this protocol. All patients suffered from stage IVa or IVb disease. 58

59 A. Clinical application of IMRT Table 6. Patient and tumor characteristics. Characteristics No. (%) of IMRT patients No. (%) of 3D-RT patients p Median age (range) 56 (41 71) 55 (38 73) 0.53 Gender 0.47 Male 36 (85.7%) 44 (80.0%) Female 6 (14.3%) 11 (20.0%) Primary tumor site Oral cavity 6 (14.3%) 5 (9.1%) Oropharynx 12 (28.6%) 34 (61.8%) Larynx 9 (21.4%) 10 (18.2%) Hypopharynx 15 (35.7%) 6 (10.9%) Tumor grade - Grade I 3 (7.1%) 2 (3.6%) Grade II 13 (31.0%) 14 (25.5%) Grade III 9 (21.4%) 16 (29.1%) Unknown 17 (40.5%) 23 (41.8%) T classification 0.01 T1 4 (9.5%) 4 (7.3%) T2 6 (14.3%) 12 (21.8%) T3 5 (11.9%) 18 (32.7%) T4a 24 (57.1%) 18 (32.7%) T4b 3 (7.1%) 3 (5.5%) N classification 0.46 N0 1 (2.4%) 4 (7.3%) N1 3 (7.1%) 4 (7.3%) N2a 1 (2.4%) 1 (1.8%) N2b 18 (42.9%) 24 (43.6%) N2c 16 (38.1%) 21 (38.2%) N3 3 (7.1%) 1 (1.8%) Clinical stage 0.27 Stage IVA 36 (85.7%) 51 (92.7%) Stage IVB 6 (14.3%) 4 (7.3%) Total

60 A. Clinical application of IMRT IMRT treatment planning Mean dose to the CTV-boost was 73.2 Gy (range, Gy), with a mean 7.56% (range, %) of the CTV-boost volume receiving less than the prescribed dose (i.e. 72 Gy). Regarding the organs at risk, the mean dose was 48.9 Gy (range, Gy) to the ipsilateral parotid gland, 32.4 Gy (range, 12.4 Gy 54.8 Gy) to the contralateral parotid gland, and 42.3 Gy (range, Gy) to the mandible, respectively. For the patients were these structures were delineated, the mean dose to the oral cavity was 36.5 Gy (range, Gy), 59.6 Gy (range, Gy) to the larynx, and 63.7 Gy (range, Gy) to the pharyngeal constrictor muscles, respectively. Disease control and survival The actuarial estimate of LRC was 81% after 2 years (Figure 7). Of the 42 patients, 5 (11.9%) developed a local failure during follow-up, with a median time to local recurrence of 9 months (range, 5 17 months). All local failures were located within the high-dose (boost) CTV, i.e. in-field. All patients with a local recurrence were treated with palliative chemotherapy (n = 2) or best supportive care (n = 3). Actuarial LC was 87% at 2 years. Five (11.9%) patients developed a regional recurrence, after a median interval of 10 months (range, 6 18 months). All patients had a regional recurrence at the site of an initial adenopathy within the high-dose region (in-field regional recurrence). Four patients were treated with palliative chemotherapy (n = 3) or best supportive care (n = 1); one patient was eligible for salvage surgery (neck dissection), and remains disease-free at 11 months after surgery. Actuarial RC was 85% at 2 years. Twelve (28.6%) patients were diagnosed with distant metastases, after a median time of 8 months (range, 3 17 months). The actuarial estimate of DC was 61% at 2 years. The actuarial estimate of OS was 56% at 2 years (Figure 8). Fourteen (33.3%) patients died during follow-up, after a median interval of 11 months (range, 5 20 months). Thirteen patients died from HNC, 1 patient died from another cause. Twenty-three (54.8%) patients are still alive and disease-free after a median interval of 19 months (range, 6 40). Actuarial disease-free survival was 48% at 2 years. Comparing treatment outcome of the current study population with the previous cohort, no statistically significant differences in loco-regional control (81% vs. 66%, p = 0.38), distant control (61% vs. 73%, p = 0.13), overall survival (56% vs. 73%, p = 0.29), or disease-free survival (48% vs. 60%, p = 0.18) were observed. 60

61 A. Clinical application of IMRT Figure 7. Disease control in the IMRT group. 100% Disease control 90% 80% 70% 60% 50% 40% 30% 20% Loco-regional control (LRC) Local control (LC) Regional control (RC) Distant control (DC) Censored patient Disease recurrence 10% 0% Follow-up (months) 61

62 A. Clinical application of IMRT Figure 8. Survival in the IMRT group. 100% Survival 90% 80% 70% 60% 50% 40% 30% 20% 10% Overall Survival (OS) Disease-free Survival (DFS) Disease-specific Survival (DSS) Censored patient Death 0% Follow-up (months) Toxicity The prescribed dose of 72 Gy was delivered in all 42 patients according to protocol; no significant (i.e. longer than 2 consecutive days) treatment interruptions occurred (median OTT was 42 days; range, 36 44). Cx could be administered according to protocol in 37 (88.1%) patients. In 1 patient, the second cycle had to be postponed for a week and in 4 patients, cisplatinum was substituted by carboplatinum-fluorouracil because of renal function concerns. No grade 4 or 5 acute toxicity was reported in the IMRT group (Table 7). Compared to the previous cohort of patients, the incidence of acute grade 3 mucositis (54.7% vs. 72.7%, p = 0.07), grade 2 3 nausea (4.8% vs. 20.0%, p = 0.03), grade 2 3 xerostomia (81.0% vs. 92.7%, p = 0.08), and grade 2 3 pain (47.6% vs. 83.6%, p < 0.001) was lower in the IMRT group. 62

63 A. Clinical application of IMRT Table 7. Acute radiation toxicity. Adverse event No. (%) of IMRT patients No. (%) of 3D-RT patients p n = 42 n = 55 Dermatitis 0.49 Grade Grade 1 9 (21.4%) 13 (23.6%) Grade 2 18 (42.9%) 26 (47.3%) Grade 3 15 (35.7%) 16 (29.1%) Mucositis 0.07 Grade Grade 1 1 (2.4%) 3 (5.5%) Grade 2 18 (42.9%) 12 (21.8%) Grade 3 23 (54.7%) 40 (72.7%) Nausea 0.03 Grade 0 36 (85.7%) 34 (61.8%) Grade 1 4 (9.5%) 10 (18.2%) Grade 2 2 (4.8%) 7 (12.7%) Grade 3-4 (7.3%) Salivary glands 0.08 Grade Grade 1 8 (19.0%) 4 (7.3%) Grade 2 34 (81.0%) 49 (89.1%) Grade 3-2 (3.6%) Dysphagia 0.37 Grade Grade 1-2 (3.6%) Grade 2 10 (23.8%) 7 (12.7%) Grade 3 32 (76.2%) 46 (83.6%) Pain <0.001 Grade Grade 1 22 (52.4%) 9 (16.4%) Grade 2 20 (47.6%) 31 (56.4%) Grade 3-15 (27.3%) 63

64 A. Clinical application of IMRT Twenty (47.6%) patients were only hospitalized for the administration of chemotherapy at the beginning of week 1 and week 4 (median total hospital stay of 5 days; range, 4 10 days). Three (7.2%) patients were hospitalized during the entire treatment and 15 (35.7%) patients during the last two weeks (2 fractions per day) because they lived too far from the hospital or had too little support at home to allow out-patient treatment, although their medical condition did not necessitate admission. Only 4 (9.5%) patients needed additional hospitalization because of medical reasons, compared to 12 (21.8%, p = 0.10) in the previous cohort. Those patients were admitted because of severe mucositis and important weight loss (n = 1) or neutropenic fever and infection (n = 3). The median duration of this additional hospitalization was 18.5 days (range, 8 29 days). A PEG tube was placed before the start of treatment in all patients. Ten patients (23.8%) did not use their PEG tube and completed RT with a modified diet and nutrition supplements. The PEG tube remained in place for a median duration of 4 months (range, days). Of the 23 currently disease-free patients, 3 (13.0%) have a PEG tube in place, after a median time of 7 months (range, 6 8 months) since the end of treatment. Late toxicity could be evaluated in 34 (81.0%) patients with a minimum event-free followup of 6 months (Table 8). No grade 4 or 5 late toxicity was reported; 17.6% of patients suffered from grade 3 xerostomia, and 5.9% reported grade 3 dysphagia. Compared to the previous cohort, there was a significantly lower incidence of late grade 2 3 mucositis (8.8% vs. 25.5%, p = 0.06), and xerostomia (52.9% vs. 90.2%, p < 0.001). For 15 patients, event-free follow-up of more than 12 months was available. Of those patients, 5 (30.0%) complained of some degree of xerostomia and 3 (20.0%) patients complained of dysphagia. None of those 15 patients had a PEG tube in place. There were no cases of bone or soft-tissue necrosis. Looking at the preservation rates of the larynx, of the 24 patients with laryngeal (n = 9) or hypopharyngeal cancer (n = 15), 21 patients (87.5%) are in complete local response, with a functional larynx, and without major complications, e.g. PEG tube dependence, persistent grade 2-3 dysphagia, pharyngo-esophageal stenosis requiring dilatation, chronic lung aspiration, or permanent tracheostomy [108]. Only one disease-free patient still required a PEG tube, at 7 months after radiotherapy. In the previous cohort, 16 patients had laryngeal (n = 10) or hypopharyngeal (n = 6) cancer. Five of 16 patients had a local recurrence, treated with laryngectomy in 3 patients. Two disease-free patients remain PEG tube dependent, at 10 and 11 months after treatment, respectively. Overall, 9 (56.3%, p = 0.03) patients are in complete local response, with a functional larynx, and without major treatment complications. 64

65 A. Clinical application of IMRT Table 8. Late radiation toxicity. Organ tissue No. (%) of IMRT patients No. (%) of 3D-RT patients p n = 34 n = 51 Skin 0.88 Grade 0 17 (50.0%) - Grade 1 14 (41.2%) 46 (90.2%) Grade 2 3 (8.8%) 5 (9.8%) Grade Subcutaneous tissue 0.02 Grade 0 17 (50.0%) - Grade 1 13 (38.2%) 34 (66.7%) Grade 2 4 (11.8%) 17 (33.3%) Grade Mucous membrane 0.06 Grade 0 19 (55.9%) 12 (23.5%) Grade 1 12 (35.3%) 26 (51.0%) Grade 2 3 (8.8%) 12 (23.5%) Grade 3-1 (2.0%) Salivary glands < Grade 0 6 (17.6%) - Grade 1 10 (29.4%) 5 (9.8%) Grade 2 12 (35.3%) 37 (72.5%) Grade 3 6 (17.6%) 9 (17.6%) Larynx 0.17 Grade 0 14 (41.2%) 30 (58.8%) Grade 1 15 (44.1%) 7 (13.7%) Grade 2 5 (14.7%) 14 (27.5%) Grade Esophagus 0.42 Grade 0 16 (47.1%) 26 (51.0%) Grade 1 8 (23.5%) 14 (27.5%) Grade 2 8 (23.5%) 9 (17.6%) Grade 3 2 (5.9%) 2 (3.9%) 65

66 A. Clinical application of IMRT 1.3. Discussion In the Leuven University Hospitals, a hyperfractionated accelerated radiotherapy schedule, initially without concomitant chemotherapy, was implemented for loco-regionally advanced disease in An initial retrospective analysis of the 73 consecutive patients treated with this schedule confirmed its practicability. Loco-regional control after 2 years was 55% and 2-year overall survival was 59%. These results are comparable to what was reported in most other studies with altered fractionation, and much better than historical series where conventional fractionation was used. In successive RTOG trials for example, 2-year LRC rates for conventional RT have improved from 29% in the late 70s over 40% in the late 80s to 46% in the latest study [30, 109, 110]. Hyperfractionation has been tested in two large randomized trials (EORTC and RTOG 90-03). In the EORTC trial, HF resulted in a 5-year LRC rate of 59%, compared to 40% for conventional RT [111]. The RTOG trial reported a smaller advantage, with 2-year LRC of 54% for hyperfractionation and 46% for conventional RT [30]. Both trials confirmed the biological expectation that local control can be significantly improved by increasing the dose through hyperfractionation. Several studies have shown the success of accelerated fractionation in minimizing tumor proliferation during radiotherapy. The EORTC trial was the first to compare a fractionation schedule with three fractions per day to conventional RT [112, 113]. Although no significant differences between treatment arms were seen (LRC was 27% at 5 years), this study provided a starting platform for further studies in fractionation [28, 114, 115]. In EORTC 22851, with a shorter overall treatment time, acceleration to a median of 33 days resulted in a 13% increase in LRC (59% vs. 46% at 5 years) [26]. In both studies, late toxicity was significantly increased [26, 28]. Bourhis et al. from the French groupe oncologie radiothérapie tête et cou (GORTEC 94-02) tested a very accelerated schedule (63 Gy in 23 days), leading to significantly improved tumor control (58% vs. 34% 2-year LRC rates, respectively), although at the cost of intolerably high acute toxicity [116, 117]. Poulsen and colleagues from the trans-tasman radiation oncology group (TROG) found no significant difference in loco-regional control at 5 years (52% for AF and 47% for conventional RT) [118]. The DAHANCA 6 & 7 trial demonstrated that by using 6 instead of 5 fractions per week, 5-year LRC improves with 10% [29]. In the study by Dische et al., CHART achieved similar LRC and survival as conventional fractionation, although with a total dose of only 54 Gy [119]. The concomitant-boost schedule was tested in RTOG and RTOG 90-03, and led to significantly improved 2-year LRC rates 66

67 A. Clinical application of IMRT (55% vs. 46% for conventional RT) [30, 120]. As expected, the most common sites of acute side-effects were the mucous membranes and the pharynx, with 54.8% developing grade 3 or worse acute mucositis and nearly half of patients (47.9%) developing grade 3 dysphagia, needing hospitalization for feeding support. This is not dissimilar from the most recent alternative fractionation trials. For instance, DAHANCA 6 & 7 found mucositis grade 3 or worse in 53% of patients receiving a modest acceleration of one week [29]. In RTOG 90-03, 42% of patients treated with HF and 47% of patients treated with concomitant-boost RT suffered acute mucositis grade 3 or worse [30]. Dermatological morbidity was mild, with only 34.5% developing grade 3 acute toxicity. Generally, the schedule was well tolerated: the prescribed radiotherapy dose could be administered to all patients without treatment interruption. These results suggested that it was possible to add concomitant chemotherapy to the hybrid fractionation schedule, in an effort to further improve results. In 2004, a concomitant chemoradiotherapy protocol for loco-regionally advanced disease was initiated, opting for a cisplatinum-based regimen, based on the literature [32 34]. With the combination of hyperfractionated accelerated radiotherapy and concomitant chemotherapy, a 2-year locoregional control rate of 70% and a 2-year overall survival rate of 74% was achieved in a group of 90 consecutive patients treated from 2004 to Although it is obviously precarious to compare these retrospective results with those of prospective trials, using several different fractionation and chemotherapy schedules, it is nonetheless clear that these percentages fall well within the upper range of most other studies (Table 9) combining intensified radiotherapy with concomitant chemotherapy [35 38, ]. For example, Ang et al. found 2-year LRC and OS rates of 65% and 72%, respectively [36]. The protocol used in that study, the MD Anderson concomitant-boost schedule (72 Gy in 6 weeks) with cisplatinum 100 mg/m 2 on day 1 and 22, was somewhat similar to ours. In comparison with the patients treated with RT alone, a significant improvement of 15% in both 2-year LRC and OS rates was noted in the CRT group. This could roughly correspond to the 5-year overall survival benefit of 6.5 to 8% with concomitant chemotherapy observed in the MACH-NC meta-analyses [32 34]. Obviously, all the limitations of a retrospective, singleinstitute analysis on a relatively small number of patients (n = 163) apply, and this percentage is most likely an overestimation of the real benefit. First of all, stage migration because of more accurate imaging techniques (e.g. distant metastasis screening) might be responsible for a certain proportion of the perceived benefit. Secondly, RT techniques have also improved since 2001, with the introduction of IMRT at our department. Nonetheless, the study period was short 67

68 A. Clinical application of IMRT ( ) and the two patients groups were very similar, since our institutional guidelines remained constant throughout. It is in any case apparent that important progress in the nonsurgical treatment of advanced HNC was made in recent years. Table 9. Selected literature review of altered fractionation CRT schedules. Authors Treatment RT Dose Cx LRC OS Brizel DM, 1998 [35] HF 75 Gy - 44% 45% HF + CT 70 Gy CDDP/5-FU 70% 60% Wendt TG, 1998 [121] AF - 17% 35% 70.2 Gy AF + CT CDDP/5-FU/LV 40% 60% Jeremic B, 2000 [122] HF - 42% 49% 77 Gy HF + CT CDDP 60% 68% Dobrowsky W, 2000 [123] AF - 32% 55.3 Gy AF + CT MMC 50% Staar S, 2001 [124] HAF - 45% 39% 69.9 Gy HAF + CT CDDP/5-FU 51% 48% Huguenin P, 2004 [125] HF - 42% 49% 74.4 Gy HF + CT CDDP 55% 59% Budach V, 2005 [126] HAF 77.6 Gy - 42% 39% HAF + CT 70.6 Gy 5-FU/MMC 58% 50% Ang KK, 2005 [36] HAF + CT 72 Gy CDDP 65% 72% Beckmann GK, 2005 [127] HAF + CT 69.9 Gy CDDP 67% Bensadoun RJ, 2006 [37] HF - 28% 20% Gy HF + CT CDDP/5-FU 59% 38% Leuven experience HAF - 55% 59% 72 Gy HAF + CT CDDP 70% 74% Abbreviations: RT: radiotherapy; Cx: concomitant chemotherapy; LRC: 2-year loco-regional control; OS: 2-year overall survival; HF: hyperfractionation; AF: accelerated fractionation; HAF: hyperfractionated accelerated radiotherapy; CDDP: cisplatinum; 5-FU: fluorouracil; LV: leucovorin; MMC: mitomycin C. One of the important questions that currently remain unanswered is whether altered fractionation is still necessary in the setting of concomitant chemotherapy. The few studies that have been performed to date appear to confirm that a combination of altered fractionation and concomitant Cx is superior to conventionally fractionated CRT [128, 129]. However, the results of large randomized trials, e.g. the RTOG trial, need to be awaited before definite conclusions can be drawn. It should be noted that the GORTEC trial, randomizing 840 patients to 70 Gy in 7 weeks with concomitant chemotherapy (arm A), 70 Gy in 6 weeks with concomitant chemotherapy (arm B), or 64.8 Gy in 3.5 weeks without chemotherapy (arm C), 68

69 A. Clinical application of IMRT demonstrated no apparent benefit of accelerated fractionation in a concomitant chemotherapy setting. Indeed, conventional radiotherapy with chemotherapy provided the best efficacy/tolerance ratio, as compared to the other 2 arms [130]. However, improvement in outcome through the addition of chemotherapy comes at the cost of increased acute as well as late toxicity [131]. In our analysis, the increase in both acute mucositis and dysphagia (leading to higher pain rates) since the use of concomitant chemotherapy was highly significant, albeit on retrospective analysis, corresponding to our clinical observations. Soon after the introduction of this regimen, it became evident that feeding support was required in almost all patients (82%) and that PEG tube placement before the start of treatment was advantageous. Nonetheless, the weight loss was still considerable: 32% of patients lost more than 10% of their initial weight. Although initial late toxicity appeared mild and largely comparable to what was observed before the use of concomitant Cx, follow-up was too short to allow any definite conclusions. However, a recent analysis of severe late toxicity observed in three RTOG trials on chemoradiotherapy suggested that with this approach, the limits of acceptable long-term toxicity have been reached [132]. No further treatment intensification is possible without effective techniques to prevent serious late complications, whether through the use of highly conformal radiotherapy or the addition of concomitant radioprotectors. At the Leuven department, IMRT has been implemented for the radiotherapy treatment of head and neck cancer since Initially, experience was developed in sinonasal cancer and post-operative irradiation. Since 2006, IMRT was used for the primary treatment of HNC, starting with patients with large laryngeal or hypopharyngeal tumors or with extensive nodal disease, since sparing the parotid glands with three-dimensional conformal RT without intensitymodulation is difficult in those patients [102]. Moreover, because laryngeal and hypopharyngeal carcinomas often lie in close proximity to the spinal cord, adequate coverage without exceeding spinal cord tolerance is difficult with conventional photon/electron matching techniques. Ultimately, given its clear dosimetric advantages, IMRT was implemented for the routine treatment of loco-regionally advanced HNC in the setting of concurrent chemotherapy. In a third analysis, we evaluated our preliminary experience with IMRT at our institution. To date, only a limited number of clinical studies have reported on the use of IMRT in concurrence with chemotherapy for HNC at sites other than the nasopharynx [ ]. Awaiting results of prospective trials, such as RTOG 00-22, all available publications consist of retrospective, single-institute reviews (Table 10). 69

70 A. Clinical application of IMRT Table 10. Selected literature review on intensity-modulated CRT. Reference n RT Disease Dose LRC (%) OS (%) Lee et al. [135] IMRT 3D-RT Stage III/IV OP 70 Gy 82% 92% 81% 91% Lee et al. [136] 31 IMRT Stage III/IV LA&HP 70 Gy 84%* 63%* Huang et al. [137] 71 IMRT Stage III/IV OP 70 Gy 90% 83% Saba et al. [138] 80 IMRT Stage III/IV LA&OP 70 Gy - 81% Leuven experience 42 IMRT Stage IV HNC 72 Gy 81%* 56%* Abbreviations: n: number of patients; RT: radiotherapy; LRC: loco-regional control; OS: overall survival; OP: oropharynx; LA: larynx; HP: hypopharynx; : 3-year rates, *: 2-year rates. With the combination of hyperfractionated accelerated IMRT and concomitant chemotherapy, a 2-year loco-regional control rate of 81% was achieved, well within the range of what can be expected in stage IV HNC with a majority (57.1%) of patients suffering from laryngeal and hypopharyngeal cancer [139]. No improvement in LRC was observed compared to 3D-RT (66%, p = 0.38), which was to be expected since no radiation dose-escalation was undertaken. Nonetheless, it is clear that IMRT is associated with at least equal tumor control rates as conventional radiotherapy techniques. Moreover, because the possibility of geographic miss is a possible concern with IMRT, the fact that there were no marginal failures is reassuring and suggests that both the definition and the coverage of the target volumes were adequate. As LRC is consistently improving, distant metastasis becomes a more important cause of death for HNC patients. In our IMRT analysis, distant metastasis (n = 12) was a much more frequent site of disease recurrence than either primary tumor (n = 5) or lymph nodes (n = 5), resulting in disappointing overall survival rates (56% at 2 years) despite excellent loco-regional control (81% at 2 years). Current concomitant chemotherapy schedules are meant to increase the efficacy of RT, and do not decrease the incidence of distant metastases [32 34]. Despite the clear efficacy of a combined modality approach in advanced HNC, toxicity can be considerable. Long-term xerostomia and dysphagia are the 2 most debilitating side effects from a quality of life perspective [91]. After conventional radiotherapy, 60% 75% of patients are expected to have grade 2 or higher xerostomia [92]. Recently, a prospective longitudinal study has shown that parotid-sparing IMRT can reduce long-term xerostomia and 70

71 A. Clinical application of IMRT therefore broadly improve a patient s quality of life [140]. The findings in the current study substantiate the benefits of IMRT with respect to salivary function sparing, since patients had significantly less grade 2 3 xerostomia in the IMRT group (52.9% vs. 90.2%, p < 0.001). Regarding late dysphagia, recent studies have indicated the possibility of sparing the relevant swallowing structures with IMRT [93]. In our study, the dose to these structures was informative and no actual dose-constraints were applied. Consequently, no improvement in the incidence of late dysphagia was observed. However, all disease-free patients with more than 12 months follow-up no longer had a PEG tube in place. In the literature, long-term feeding tube dependence rates in disease-free patients have been reported of up to 17% [141]. Also, the majority of patients with laryngeal or hypopharyngeal cancer were in complete local control and without major complications, suggesting a certain larynx-sparing potential of intensity-modulated chemoradiotherapy [108, 142]. Some important limitations of this retrospective, single-institute analysis on a relatively limited number of patients (n = 97) need to be discussed. First, IMRT was initially implemented for large tumors and for laryngeal and hypopharyngeal tumors, resulting in a statistically significant disparity in T-classification (p = 0.01) and tumor site (p = 0.005) between the two groups. However, since both recognized negative prognostic factors were worse for patients treated with IMRT, this confirms that IMRT results in at least equal LRC rates as conventional techniques. It could also explain the higher distant metastasis rate and resulting worse overall survival in the IMRT group. Second, because of the short median follow-up time, additional follow-up is warranted to fully assess the long-term incidence and severity of the complications. In conclusion, the use of concomitant chemotherapy significantly increases loco-regional tumor control as well as treatment-related toxicity, especially in the setting of intensified fractionation schedules. Prevention of late complications is paramount to further improve the therapeutic index of chemoradiotherapy. The use of IMRT significantly reduces long-term toxicity, especially regarding late xerostomia, without compromising loco-regional tumor control. Late dysphagia remains a problem and strategies using the ability of IMRT to limit the dose delivered to the swallowing structures without compromising tumor coverage could be useful to further minimize this late complication. 71

72

73 CHAPTER 2: IMRT FOR SINONASAL CANCER Published as: Dirix P., Nuyts S., Geussens Y., Jorissen M., Vander Poorten V., Fossion E., Hermans R., Van den Bogaert W. Malignancies of the nasal cavity and paranasal sinuses: long-term outcome with conventional or three-dimensional conformal radiotherapy. Int J Radiat Oncol Biol Phys 2007; 69(4): Dirix P., Nuyts S., Vanstraelen B., Nulens A., Hermans R., Jorissen M., Vander Poorten V., Van den Bogaert W. Post-operative intensity-modulated radiotherapy for malignancies of the nasal cavity and paranasal sinuses. Radiother Oncol 2007; 85(3): Dirix P., Vanstraelen B., Jorissen M., Vander Poorten V., Nuyts S. Intensity-modulated radiotherapy for sinonasal cancer: improved outcome compared to conventional radiotherapy. Int J Radiat Oncol Biol Phys; In press.

74

75 A. Clinical application of IMRT 2. IMRT for sinonasal cancer Malignancies of the nasal cavity and paranasal sinuses represent a particular sub-site of HNC, where IMRT is expected to become the standard treatment option. First, we performed a retrospective analysis of our long-term results with conventional or conformal RT. Secondly, tumor control and toxicity of patients treated with IMRT was prospectively evaluated Material & methods Patient selection All patients presenting with cancer of the nasal cavity or paranasal sinuses at the Leuven University Hospitals between 1976 and 2008 were evaluated by a multidisciplinary team consisting of ENT head & neck surgeons, maxillofacial surgeons, head and neck radiologists, medical oncologists, and radiation oncologists. Between January 1976 and December 2008, 167 consecutive SNC patients were treated with radiotherapy Pre-treatment evaluation Pre-treatment evaluation consisted of complete history and physical examination, routine blood counts, liver function tests, ultrasound of the abdomen, and chest X-ray. All patients were imaged with X-ray films, CT, or MRI of the sinuses and head & neck region. All tumors were staged or restaged according to the 2002 TNM classification system of the AJCC [101] Treatment A treatment proposal was put forward by the multidisciplinary team according to institutional guidelines at the time of evaluation: 61 (36.5%) patients received pre-operative RT ( 70s and 80s), 91 (54.5%) patients were treated with post-operative RT (from the early 90s onwards), and 15 (9.0%) patients received primary RT. None of the patients received chemotherapy during primary treatment. The majority of patients (n = 152, 91.0%) were treated with a combination of surgery and radiotherapy. The type of surgery consisted of external resection via a trans-facial or cranio-facial approach (n = 62, 40.8%) or, since 1992, endoscopic resection (n = 90, 59.2%) for patients without brain invasion [24]. Only 4 patients, all of whom 75

76 A. Clinical application of IMRT had node-postive disease, underwent neck dissection as part of their surgical treatment Conventional and conformal radiotherapy One hundred and twenty-seven patients underwent RT using 60 Co γ-rays (n = 63, 49.6%) or, since 1990, using 6 MV photon beams from a linear accelerator (n = 64, 50.4%). Patients were treated in supine position using a variety of different immobilization devices, depending on the preferred method at the time. Radiotherapy was delivered once daily, 5 days a week, using 2 Gy per fraction, to a prescribed dose of Gy (median dose was 50 Gy, mean dose was 57.4 Gy ± 8.7 Gy). No elective irradiation of the cervical lymph nodes was performed. Five patients received bilateral neck irradiation because of clinical lymph node involvement: 3 patients received pre-operative RT to 66 Gy, 1 patient was treated post-operatively to 66 Gy, and 1 patient received primary RT to a total dose of 80 Gy. In the majority of patients (n = 74, 58.3%), RT was performed with conventional planning using a three-field technique with a heavily weighted anterior portal and two lateral, wedged portals. From 1992 to 2003, 53 (41.7%) patients were treated with forward-planned 3D-RT, using multibeam techniques with the assistance of beam s eye view reconstruction, 3D dose distributions, and dose-volume histogram (DVH) analysis. Since March 2003, patients with sinonasal cancer were treated with IMRT IMRT Forty patients were treated with an inverse-planning IMRT technique (sliding window), based on a non-coplanar 6-field arrangement. Treatment was delivered with 5 fields of 6 MV and 1 field of 10 or 18 MV photon of a linear accelerator. At the time of simulation, patients were immobilized with a thermoplastic mask (three-point fixation), and treatment planning CT scans were taken. Serial CT scan slices, 3 mm thick, from the head down to the clavicles were obtained. The planning CT was co-registered with the pre-operative MRI to accurately estimate the pre-operative tumor volume or GTV and to facilitate the delineation of the organs at risk. The CTV was defined as the GTV plus a margin to account for possible spread of microscopic disease. Usually, this meant that the CTV consisted of the resection cavity plus all paranasal sinuses that were (partially) invaded. In 19 patients, an additional boost volume was delineated because of incomplete surgical resection (either microscopic or macroscopic). No elective irradiation of the cervical lymph nodes was performed. An automated isotropic expansion of 5 76

77 A. Clinical application of IMRT mm was used for the expansion from the CTV to the PTV to account for patient set-up error. The following structures were contoured as OAR, with the dose constraints, used for optimization, between brackets: both optic nerves with the retina ( 60 Gy/2 Gy), optic chiasm ( 60 Gy/2 Gy), brain ( 60 Gy/2 Gy), spinal cord ( 50 Gy/2 Gy), and brainstem ( 54 Gy/2 Gy). Both lenses were contoured bilaterally and the dose was kept as low as possible, although no strict limit was applied. The prescribed total dose was 60 Gy in 30 daily fractions of 2 Gy (5 fractions per week); in the 19 patients with positive surgical margins, the boost volume received an additional 6 Gy (3 daily fractions of 2 Gy) to a total dose of 66 Gy. In regions where the PTV and the OAR overlapped, an underdosage of the PTV was tolerated to fulfill the constraints for the optic structures Evaluation of toxicity and response All patients were evaluated by their treating radiation oncologist at least once a week, or more frequently if required, during the RT course. Patients then returned for follow-up visits until regression of acute toxicity (approximately 8 weeks after the end of treatment). Afterwards, patients were regularly followed at the multidisciplinary outpatient clinic: every 2 months the first 2 years after treatment, every 3 months the third year, every 4 months the fourth year, every 6 months the fifth year, and then every year. Follow-up consisted of history, routine physical examination, and nasosinusal endoscopy. A baseline post-treatment scan was obtained within 2 to 3 months after completion of RT. Thereafter, imaging was done yearly or at the discretion of the treating oncologist or ENT head & neck surgeon. For the IMRT patients, acute toxicity was prospectively scored according to the CTCAE version 3.0. Late toxicity was prospectively graded according to the RTOG/EORTC late radiation morbidity scoring schema, with special emphasis on ocular toxicity, as the highest score at 6 months since radiotherapy. In the other patients, acute and late normal tissue effects as described in the RT and/or ENT charts were retrospectively collected Statistical analysis Patient and tumor characteristics were recorded at the start of treatment. Follow-up data were retrospectively collected: information was sought on the date of first recurrence, distant metastasis, and/or death. The data were analyzed using the software package Statistica 8 (StatSoft Inc, Tulsa, OK, USA). Cumulative survival and tumor control rates were calculated 77

78 A. Clinical application of IMRT using the Kaplan-Meier product-limited (actuarial) method. Patient groups were compared for age, gender, primary site (ethmoid sinus vs. other), histology (adenocarcinoma vs. other), T- classification (T2 3 vs. T4), the presence of intracranial or intraorbital invasion, surgical procedure, and radiation dose (60 vs. 66 Gy) using the non-pared student t-test. Survival and recurrence rates were compared using the log-rank test. Incidence of toxicity was compared using the non-pared student t-test. A p value < 0.05 was considered statistically significant Results Conventional and conformal radiotherapy From January 1976 to February 2003, 127 patients with histologically proven sinonasal cancer were treated at the Leuven department with conventional or conformal radiotherapy. Median follow-up was 67 months (range, months) for all patients, and 88 months (range, months) for patients alive at the close-out date (December 2006). Patient characteristics Patient, tumor, and treatment characteristics of the 127 patients are summarized in Tables 11 and 12. There were 107 (84.3%) male and 20 (15.7%) female patients, with a median age of 58 years (range, years). The primary site was determined as the region harboring the predominant bulk of disease. The majority of patients presented with a tumor of the ethmoid (ES, 55.1%) or maxillary (MS, 35.4%) sinus. In the other patients, the tumor originated from the nasal cavity (6.3%), the sphenoid sinus (2.4%), or the frontal sinus (0.8%). Adenocarcinoma was the most frequent histologic type (51.9%), followed by squamous cell carcinoma (37.8%), undifferentiated carcinoma (7.9%), and adenoid cystic carcinoma (2.4%). Distribution of T-classification was as follows: 2 (1.6%) patients presented with a T1 tumor, 6 (4.7%) patients with a T2 tumor, 44 (34.6%) patients with a T3 tumor, and 75 (59.1%) patients with a T4 tumor. Twenty-seven (21.3%) patients presented with intracranial invasion; in 53 (41.7%) patients the orbit was infiltrated. Five (3.9%) patients presented with cervical lymph node metastasis at the time of diagnosis, including 4 patients with a SCC (3 maxillary sinus and 1 ethmoid sinus) and one patient with an adenocarcinoma of the MS. 78

79 A. Clinical application of IMRT Table 11. Patient and tumor characteristics. Characteristics No. of patients % Age 60 years > 60 years Gender Male Female Primary site Ethmoid Sinus Maxillary Sinus Nasal Cavity Sphenoid Sinus Frontal Sinus Histology Adenocarcinoma Squamous Cell Carcinoma Undifferentiated Carcinoma Adenoid Cystic Carcinoma Tumor grade Grade I Grade II Grade III No grade available T-classification T T T T4a T4b N-classification N N N N The reported symptoms at the time of diagnosis were nasal obstruction in 94 (74.0%) patients, epistaxis in 64 (50.4%) patients, headache in 51 (40.2%) patients, ocular symptoms in 34 (26.8%) patients, swelling of the cheek in 21 (16.5%) patients, olfactory disturbance in 13 79

80 A. Clinical application of IMRT (10.2%) patients, and neurological symptoms in 5 (3.9%) patients. There was a mean delay of 6.5 months (range, 1 72 months) between the onset of symptoms and the diagnosis. A history of occupational wood exposure was present in 74 (58.3%) patients. The majority (n = 75, 59.1%) had a history of prior or current tobacco use. Table 12. Treatment characteristics. Treatment No. of patients % Surgery Procedure External approach Endoscopic resection Neck Dissection Yes No Total Radiotherapy Sequencing and Dose Pre-operative RT Gy Gy Post-operative RT Gy Gy Gy Primary RT Gy Gy Gy Neck irradiated Yes No Modality Conventional D-RT

81 A. Clinical application of IMRT Disease control and survival Of the 127 patients, 54 (42.5%) developed a local failure during follow-up, with a median time to local recurrence of 13 months (range, months). In 52 cases, local recurrence was a first site of failure, and in 2 patients it was a subsequent site of failure. The actuarial estimate of LC was 53% after 5 years (Figure 9). In univariate analysis, LC was significantly correlated with T-classification (73% 5-year LC for T1 3 vs. 38% for T4, p < 0.001) and the presence of intracranial invasion (30% 5-year LC when present vs. 59% when not present, p = 0.02). There was a trend towards significance with the presence of intra-orbital infiltration (39% 5-year LC when present vs. 61% when not present, p = 0.06). Post-operative RT resulted in better LC than pre-operative RT (66% 5-year LC for post-op RT vs. 40% for pre-op RT, p = 0.06); no significant difference was seen with respect to RT technique, RT dose, or tumor histology (Table 13). In multivariate analysis, only T4 (RR 3.63 [95% CI ], p < 0.001) independently predicted LC. Regional lymph node failure occurred in 6 (4.7%) patients, after a median interval of 11 months (range, 5 51 months). In 4 of these 6 patients, the site of origin was the MS, while in 2 patients the ES was the primary site. Three patients had a SCC, 2 patients an undifferentiated carcinoma and one patient an adenocarcinoma. Two of the 5 patients with N+ necks at presentation developed a regional recurrence. Actuarial RC in the entire population was 93% after 5 years of follow-up. In all 6 patients, the regional recurrence was the first site of relapse, so they were all treated with a salvage neck dissection, followed by post-operative RT. Three patients were free from disease at the time of last follow-up, at respectively 2, 5, and 13 years after regional recurrence. Three patients ultimately died from their disease, because they subsequently developed distant metastases. Twenty-six (20.5%) patients were diagnosed with distant metastases during follow-up, after a median time of 11 months (range, 3 61 months). Distant metastasis was a first site of failure in 20 patients and a subsequent site in 6 patients. Distant control in the overall population was 75% after 5 years of follow-up. There was a significant correlation with T-classification (83% vs. 72% 5-year DC, p = 0.04) and the presence of intracranial invasion (48% vs. 83% 5- year DC, p = 0.003) in univariate analysis. In multivariate analysis, only the presence of intracranial invasion (RR 2.91 [95% CI ], p = 0.04) independently predicted the likelihood of distant metastasis. 81

82 A. Clinical application of IMRT Figure 9. Local (LC), regional (RC) and distant (DC) control for all patients. Fifty-eight patients were still alive at the time of this analysis. Of the 69 patients who died during follow-up, the majority (n = 53) died as result of disease recurrence and 16 patients died from other causes. Actuarial OS at 5 years was 54% (Figure 10). This was significantly correlated with tumor histology (37% 5-year OS for SCC vs. 62.5% for non-scc, p = 0.002), T- classification (72% 5-year OS for T1 3 vs. 40% for T4, p < 0.001), and the presence of intracranial invasion (32% 5-year OS when present vs. 59% when not present, p = 0.001) as well as intraorbital infiltration (46% 5-year OS when present vs. 58% when not present, p = 0.03) in univariate analysis. In multivariate analysis, SCC (RR 2.06 [95% CI ], p = 0.004), T4 classification (RR 2.23 [95% CI ], p = 0.005), and intracranial invasion (RR 1.97 [95% CI ], p = 0.03) were significant factors. Disease-specific survival in the overall population was 58% at 5 years and 49% at 10 years. Of the 5 N+ patients, 4 died of their disease while 1 patient was free of disease at last 82

83 A. Clinical application of IMRT follow-up. In univariate analysis, DSS was significantly correlated with histology (40% vs. 67.5% 5-year DSS, p = 0.007), T-classification (77% vs. 44% 5-year DSS, p < 0.001), intracranial invasion (32% vs. 65.5% 5-year DSS, p = 0.002), and intraorbital infiltration (49.5% vs. 63.5% 5- year DSS, p = 0.03). In multivariate analysis, SCC histology (RR 2.08 [95% CI ], p = 0.01), T4 classification (RR 2.87 [95% CI ], p = 0.003), and the presence of intracranial invasion (RR 2.15 [95% CI ], p = 0.02) predicted worse DSS. Among the 58 living patients, 38 were free of disease at the time of last follow-up. Actuarial DFS for the entire population was 37% at 5 years. A significant correlation with tumor histology (29% vs. 42% 5-year DFS, p = 0.05), T classification (55% vs. 26% 5-year DFS, p < 0.001), and the presence of intracranial invasion (19% vs. 43% 5-year DFS, p = 0.01) as well as intraorbital infiltration (25.5% vs. 46% 5-year DFS, p = 0.03) was found in univariate analysis. Post-operative RT resulted in better DFS than pre-operative RT (53% vs. 28% 5-year DFS, p = 0.05); no significant difference was seen with respect to RT technique, RT dose, or primary tumor site. Both T4 classification (RR 2.59 [95% CI ], p < 0.001) and the presence of intracranial invasion (RR 1.67 [95% CI ], p = 0.05) significantly and independently predicted DFS in multivariate analysis. Table 13. Significant factors on univariate analysis. Variable LC DC OS DSS DFS Age ( 60 vs. > 60) Gender (male vs. female) T classification (T 1 3 vs. T4) p < p = 0.04 p < p < p < IC invasion (present vs. not) p = 0.02 p = p = p = p = 0.01 IO infiltration (present vs. not) p = p = 0.03 p = 0.03 p = 0.03 RT technique (2D vs. 3D) RT dose (< 60 Gy vs. 60 Gy) RT sequence (pre-op vs. post-op) p = p = 0.05 Site (ethmoid vs. other) Histology (SCC vs. other) - - p = p = p = 0.05 Abbreviations: LC = local control; DC = distant control; OS = overall survival; DSS = diseasespecific survival; DFS = disease-free survival; IC = intracranial; IO = intraorbital; RT= radiotherapy; 2D = conventional, two-dimensional RT; 3D = three-dimensional conformal RT; SCC = squamous cell carcinoma. 83

84 A. Clinical application of IMRT Figure 10. Disease-specific (DSS), overall (OS) and disease-free (DFS) survival for all patients. Toxicity The prescribed dose could be delivered in all patients; a treatment interruption of more than 2 days occurred in 21 (16.5%) patients (median OTT was 43 days; range, days). The presence of acute side-effects during treatment was systematically recorded in the RT charts, although their severity was not always graded. Generally, almost all patients suffered from a varying degree of mucositis (n = 124, 97.6%), dermatitis (n = 122, 96.1%), and local alopecia in the exit area of the anterior beam (n = 120, 94.5%). Other frequent toxicities were fatigue (n = 116, 91.3%), xerostomia (n = 114, 89.8%), and headache (n = 108, 85.0%). Dysphagia (n = 40, 31.5%), nausea (n = 10, 7.9%), and otitis (n = 7, 5.5%) were relatively rare and, when present, usually mild. The majority of patients complained of altered taste (n = 121, 84

85 A. Clinical application of IMRT 95.3%) and sense of smell (n = 115, 90.6%), although this often resulted from tumor invasion and/or surgery. Acute ocular and/or visual toxicity typically consisted of conjunctivitis (n = 71, 55.9%). Fewer patients treated with 3D-RT suffered from conjunctivitis (25/53, 47.2%) than patients who received conventional RT (44/74, 59.5%), but this difference was not statistically significant on chi-square testing. Late toxicity could be evaluated in 117 (92.1%) patients with a minimum event-free follow-up of 6 months. Permanent xerostomia was present in 25 (21.4%) patients; late mucositis was recorded in 46 (39.3%) patients. The incidence of permanent xerostomia was much lower in patients treated with 3D-RT (5/50, 10.0%) than in patients treated with conventional RT (20/67, 29.9%; p = 0.08). Similarly, there was no significant difference in late mucositis between the two treatment groups. Hearing loss was reported by 10 (8.5%) patients; 3 (2.6%) patients complained of tinnitus. Thirteen (11.1%) patients suffered from permanent diminished sense of taste and/or smell. For 72 patients, follow-up of 2 years or more was available. There were no documented cases of bone or soft-tissue necrosis. One patient, who received primary RT to 80 Gy in 1984, developed symptomatic focal brain necrosis 1 year later, which was palliatively treated with steroids because of significant co-morbidity and dismal prognosis. In two patients, cataract was diagnosed at 2 and 3.5 years after RT, respectively. Dry eye symptom, requiring chronic treatment with artificial tears, was present in 18 patients; no significant correlation with RT technique was seen. Fifteen patients suffered from radiation-induced retinopathy, resulting in mild to moderate visual impairment. Two patients developed optic neuropathy: both patients complained of rapid and severe visual impairment in both eyes, at 6 and 11 years follow-up respectively Initial results with post-operative IMRT From March 2003 to February 2006, 25 patients with histologically proven sinonasal cancer were treated at the Leuven department with post-operative IMRT. Median follow-up was 24 months (range, 7 47 months) for all patients, and 27 months (range, months) for patients alive at the close-out date (March 2007). 85

86 A. Clinical application of IMRT Patient characteristics There were 20 male and 5 female patients, with a median age of 65 years (range, years). The primary site of the tumor was surgically determined. The majority of patients presented with a tumor of the ethmoid sinus (n = 20). In the other patients, the tumor originated from the nasal cavity (n = 4), or the maxillary sinus (n = 1). Since patients were treated postoperatively, definitive histology was available in all cases. Adenocarcinoma was most common (n = 17), followed by neuro-endocrine carcinoma (n = 4), olfactory esthesioneuroblastoma (n = 2), and squamous cell carcinoma (n = 2). Distribution of T-classification was as follows: 4 patients presented with a T2 tumor, 11 with a T3 tumor, and 10 with a T4 tumor. Three patients presented with intracranial invasion; in 2 patients the orbit was infiltrated. None of the patients had evidence of lymph node or distant metastasis at the time of diagnosis. Fifteen (60%) patients were treated to 60 Gy, 10 (40%) patients received an additional boost of 6 Gy to a total dose of 66 Gy. Disease control and survival Five patients developed a local recurrence at a median of 8 months (range, 3 39 months) after the start of RT. Of those 5 patients, 2 could be successfully salvaged by surgery and are currently without evidence of disease, at respectively 9 and 22 months after retreatment. The other 3 patients were considered inoperable due to the presence of distant metastasis (n = 2) or the extent of the relapse (n = 1), and were treated with palliative chemotherapy. The actuarial 2-year estimate of LC was 81%. All failures were considered to have been within the CTV. There were no regional failures. A total of 3 patients developed distant metastasis during the evaluation period, after a median of 7 months (range, 3 9 months). In 2 patients, the distant failure developed simultaneously with or almost immediately after a local failure; 1 patient developed diffuse metastases as a sole site of failure. All 3 patients were treated with palliative chemotherapy. The actuarial 2-year rate of distant control was 88%. Twenty-two patients were still alive at last follow-up. All 3 patients who died during the follow-up period, died as a result of local (n = 2) and/or distant (n = 2) tumor recurrence. Overall survival of the entire population was 88% at 2 years. Among the 22 living patients, 19 were free of disease at the time of last follow-up visit (Table 14). DFS at 2 years was 77%. 86

87 A. Clinical application of IMRT Table 14. Local and distant disease recurrence characteristics. Histology T Dose LR Time to Event DM Time to Event Treatment Alive 1. Adenocarcinoma T3 66 Gy yes 39 mo no Surgery yes 2. Adenocarcinoma T4 60 Gy yes 22 mo no Surgery yes 3. Adenocarcinoma T4 60 Gy yes 6 mo no CT no 4. Adenocarcinoma T4 66 Gy no yes 7 mo CT no 5. Adenocarcinoma T4 66 Gy yes 3 mo yes 3 mo CT no 6. Neuro-endocrine Carcinoma T3 66 Gy yes 8 mo yes 9 mo CT yes Abbreviations: T: T-classification; LR: local recurrence; DM: distant metastasis; CT: palliative chemotherapy. Toxicity The prescribed dose could be delivered in all patients without noticeable treatment interruptions (mean OTT was 45 days; range, days). No acute toxicity grade 3 and/or 4 was seen. The most frequent toxicities were dermatitis (88%), fatigue (84%), mucositis (72%), xerostomia (60%), and headache (60%). Dysphagia was relatively rare (32%) and, when present, generally mild. Ocular and/or visual acute toxicity was similarly mild. Most frequent complications were conjunctivitis (80%) and tearing (72%). No keratitis, double vision, photophobia, or other ocular acute toxicity was present in any of the patients. Late toxicity was scored after more than 6 months of follow-up in all patients; no grade 3 or 4 late toxicity was seen. At that time point, 10 patients reported taste disturbance and 5 patients complained of diminished sense of smell. Although not scored, the majority of patients had nasal crusts, which were treated by daily nasal irrigations with a physiologic solution. For 12 patients, follow-up of 2 years or more was available (mean follow-up: 37 months). Very little chronic toxicity was observed in those patients: no skin, mucous membrane, subcutaneous tissue, bone, or spinal cord toxicity was reported. Two patients had grade 1 xerostomia, 2 patients complained of mild headaches (grade 1), and 2 patients suffered from moderate headaches (grade 2). Two patients complained of a dry eye, requiring chronic treatment with artificial tears (grade 2), and 1 patient suffered from radiation-induced retinopathy, resulting in visual impairment (grade 2). To date, no radiation-induced blindness was seen. 87

88 A. Clinical application of IMRT Dose-volume Analysis Table 15. Median (± standard deviation) of the maximum and mean doses to the OAR. Critical Structure Maximum Dose (Gy) Mean Dose (Gy) Optic Chiasm 53.3 ± ± 11.4 Optic Nerve + Retina (ipsilateral) 59.6 ± ± 10.3 Optic Nerve + Retina (contralateral) 34.9 ± ± 7.5 Lens (ipsilateral) 14 ± ± 5.6 Lens (contralateral) 6.2 ± ± 1.8 Brainstem 28 ± ± 4.7 Brain 62.7 ± ± 2.7 Table 16. PTV coverage. Structure Mean Median ± SD Range Mean Dose (Gy) ± PTV Median Dose (Gy) ± Volume (cm 3 ) ± Mean Dose ± PTV boost Median Dose (Gy) ± Volume (cm 3 ) ± Update of IMRT and comparison to conformal radiotherapy Between March 2003 and December 2008, 40 consecutive patients were treated with post-operative IMRT for malignancies of the nasal cavity and paranasal sinuses at the Leuven department. Median follow-up was 30 months (range, 4 74 months) for all patients, and 31 months (range, 8 74 months) for patients alive at the close-out date (June 2009). 88

89 A. Clinical application of IMRT Patient characteristics Patient and tumor characteristics are summarized in Table 17. There were 34 male and 6 female patients, with a median age of 66 years (range, years). The primary site of the tumor was surgically determined. The majority of patients (n = 33) presented with a tumor of the ethmoid sinus. In the other patients, the tumor originated from the nasal cavity (n = 6) or the maxillary sinus (n = 1). Since patients were treated post-operatively, definitive histology was available in all cases. Adenocarcinoma was the most common type (n = 31), followed by neuroendocrine carcinoma (n = 4), olfactory esthesioneuroblastoma (n = 2), squamous cell carcinoma (n = 2), and undifferentiated carcinoma (n = 1). Distribution of T-classification was as follows: 9 patients presented with a T2 tumor, 19 patients with a T3 tumor, and 12 patients with a T4 tumor. Five patients presented with intracranial invasion; in 5 patients the orbit was infiltrated. None of the patients had evidence of lymph node or distant metastasis at the time of diagnosis. Disease control and survival The actuarial estimate of LC was 76% after 2 years (Figure 11). Of the 40 patients, 9 developed a local failure during follow-up, with a median time to local recurrence of 14 months (range, 3 29 months). Salvage surgery was attempted in 5 of 9 patients with a local recurrence. Four of those 5 patients currently remain disease-free after a median interval since re-treatment of 16 months (range, 6 63 months). One patient developed distant metastases shortly after salvage surgery. Four patients with a local recurrence were considered inoperable due to the presence of distant metastases (n = 2) or the extent of the relapse (n = 2), and were treated with palliative chemotherapy. All local failures were within the original CTV (in-field recurrences); there were no regional failures. The actuarial estimate of DC was 89% at 2 years. Five patients were diagnosed with distant metastases during follow-up, after a median time of 9 months (range, 3 55 months). In 3 patients, the distant failure developed simultaneously with or shortly after a local failure; in 2 patients the distant metastases were the sole site of failure. The actuarial estimate of overall survival was 89% at 2 years. Six patients died during follow-up, after a median interval of 11 months (range, 4 57 months). Five patients died as the result of a disease recurrence, 1 patient died from gastric cancer. Twenty-eight patients are still alive and disease-free after a median interval of 33 months (range, 8 74). Actuarial diseasefree survival was 72% at 2 years. 89

90 A. Clinical application of IMRT Table 17. Patient and tumor characteristics. Characteristics No. (%) in IMRT group No. (%) in 3D-RT group p Age 0.42 Mean age (range) 63 (37 84) 61 (37 85) Gender 0.80 Male 34 (85.0%) 34 (82.9%) Female 6 (15.0%) 7 (17.1%) Tumor site 0.32 Ethmoid sinus 33 (82.5%) 30 (73.2%) Nasal cavity 6 (15.0%) 2 (4.9%) Maxillary sinus 1 (2.5%) 7 (17.1%) Sphenoid sinus - 1 (2.4%) Frontal sinus - 1 (2.4%) Histology 0.11 Adenocarcinoma 31 (77.5%) 25 (61.0%) Neuro-endocrine carcinoma 4 (10.0%) - Esthesioneuroblastoma 2 (5.0%) - Squamous cell carcinoma 2 (5.0%) 9 (22.0%) Undifferentiated carcinoma 1 (2.5%) 5 (12.2%) Adenoid cystic carcinoma - 2 (4.8%) T classification 0.28 T2 9 (22.5%) 10 (24.4%) T3 19 (47.5%) 23 (56.1%) T4a 7 (17.5%) 5 (12.2%) T4b 5 (12.5%) 3 (7.3%) Intracranial invasion 5 (12.5%) 6 (14.6%) 0.78 Intraorbital invasion 5 (12.5%) 11 (26.8%) 0.11 Surgical procedure External approach 2 (5.0%) 12 (29.3%) Endoscopic resection 38 (95.0%) 29 (70.7%) Radiation dose Gy 21 (52.5%) 27 (65.9%) 66 Gy 19 (47.5%) 14 (34.1%) Total

91 A. Clinical application of IMRT Figure 11. Disease control for the IMRT group. 100% Disease control 90% 80% 70% 60% 50% 40% 30% 20% 10% Local control (LC) Disease-free survival (DFS) Overall survival (OS) censored patient event 0% Follow-up (months) Numbers at risk (OS) Comparing treatment outcome of the current study population with the previous cohort of 41 patients treated with conformal radiotherapy without intensity-modulation, an improvement in 2-year local control (76% vs. 67%, p = 0.06), 2-year disease-free survival (72% vs. 60%, p = 0.02), and 2-year overall survival (89% vs. 73%, p = 0.07) was observed (Figure 12). Distant control was identical in both groups (89%, p = 0.68). 91

92 A. Clinical application of IMRT Figure 12. Disease-free survival according to radiotherapy technique. 100% Disease-free survival (DFS) according to radiotherapy technique Disease-free survival (%) 90% 80% 70% 60% 50% 40% 30% 20% 10% IMRT group (n = 40) 3D-RT group (n = 41) p = 0,02 censored patient disease recurrence 0% Follow-up (months) Numbers at risk (DFS) in the 3D-RT group (n = 41) Numbers at risk (DFS) in the IMRT group (n = 40) Toxicity The prescribed dose could be delivered in all patients without noticeable treatment interruptions (mean OTT was 45 days; range, days). No acute toxicity grade 3 or 4 was seen. The most frequent non-ocular toxicities were dermatitis (75.0%), mucositis (62.5%), fatigue (50.0%), and headache (45.0%), as shown in Table 18. Dysphagia was relatively rare (22.5%) and, when present, generally mild. The majority of patients complained of slightly (grade 1, 42.5%) or markedly (grade 2, 30.0%) altered taste during treatment. Also, half of patients reported slightly (grade 1, 22.5%) or markedly (grade 2, 22.5%) changed sense of smell. However, both loss of taste and smell was often already present before the start of radiotherapy, due to the extent of the disease or as a result of the surgery. Less frequent complications were local alopecia and nausea. Five patients (12.5%) had mild hair loss (grade 92

93 A. Clinical application of IMRT 1), and 1 patient (2.5%) had pronounced hair loss (grade 2). Five patients (12.5%) developed mild nausea during treatment, although all were still able to eat (grade 1). One patient developed tinnitus (grade 2), and 1 patient developed serous otitis (grade 2), treated with topical antibiotics. Table 18. Acute radiation toxicity in both groups. Adverse No. (%) in IMRT group No. (%) in 3D-RT group p Event Grade 0 Grade 1 Grade 2 Any grade Dermatitis 10 (25.0%) 27 (67.5%) 3 (7.5%) 40 (97.6%) Mucositis 15 (37.5%) 20 (50.0%) 5 (12.5%) 40 (97.6%) < Dysphagia 31 (77.5%) 8 (20.0%) 1 (2.5%) 14 (34.1%) 0.25 Salivary Glands 25 (62.5%) 15 (37.5%) 0 (0%) 37 (90.2%) < Pain (Headache) 22 (55.0%) 14 (35.0%) 4 (10.0%) 34 (82.9%) 0.02 Sense of Smell 22 (55.0%) 9 (22.5%) 9 (22.5%) 36 (87.8%) < Taste Disturbance 11 (27.5%) 17 (42.5%) 12 (30.0%) 38 (92.7%) 0.02 Fatigue 20 (50.0%) 11 (27.5%) 9 (22.5%) 32 (78.0%) Conjunctivitis 12 (30.0%) 19 (47.5%) 9 (22.5%) Not scored - Dry Eye 28 (70.0%) 11 (27.5%) 1 (2.5%) Not scored - Tearing 11 (27.5%) 22 (55.0%) 7 (17.5%) Not scored - Compared to the previous cohort of patients treated with 3D-RT, the incidence of dermatitis (75.0% vs. 97.6%, p = 0.003), mucositis (62.5% vs. 97.6%, p < 0.001), xerostomia (37.5% vs. 90.2%, p < 0.001), headache (45.0% vs. 82.9%, p = 0.02), and fatigue (50.0% vs. 78.0%, p = 0.008) was significantly lower in the IMRT group. Acute ocular and visual toxicity was similarly mild: again no grade 3 or 4 toxicity was reported. The most frequent ocular complications were conjunctivitis (70.0%) and tearing (62.5%). Twelve patients (30.0%) suffered from a dry eye, requiring artificial tears. Blurred vision was reported by 2 patients; no keratitis, double vision, photophobia, or other acute ocular toxicity was present in any of the patients. Late toxicity could be evaluated in 39 patients. No grade 3 or 4 late toxicity was reported. Compared to the previous cohort of patients treated with 3D-RT (38 patients with at least 6 months event-free follow-up), the incidence of late skin toxicity (7.7% vs. 23.7%, p = 0.05), late 93

94 A. Clinical application of IMRT mucositis (30.7% vs. 73.7%, p < 0.001), and permanent xerostomia (12.8% vs. 34.2%, p = 0.03) was significantly lower in the IMRT group (Table 19). Table 19. Late radiation toxicity in both groups. Organ/Tissue No. (%) in IMRT group (n = 39) No. (%) in 3D-RT group (n = 38) p Grade 0 Grade 1 Grade 2 Any grade Skin 36 (92.3%) 2 (5.1%) 1 (2.6%) 9 (23.7%) 0.05 Mucous Membrane 27 (69.2%) 10 (25.6%) 2 (5.1%) 28 (73.7%) < Salivary Glands 34 (87.2%) 5 (12.8%) 0 (0%) 13 (34.2%) 0.03 Brain (Headache) 25 (64.1%) 11 (28.2%) 3 (7.7%) Not scored - Dry Eye Syndrome 36 (92.3%) 2 (5.1%) 1 (2.6%) 12 (31.6%) Concerning visual toxicity, 3 patients (7.7%) complained of a dry eye, requiring chronic treatment with artificial tears. In the 3D-RT group, 12 of 38 patients (31.6%, p = 0.007) developed the dry eye syndrome. To date, no radiation-induced visual impairment was observed in the IMRT group. In the 3D-RT group, 6 patients (15.8%) developed radiationinduced retinopathy, resulting in mild to moderate visual impairment. No radiation-induced neuropathy was observed in either treatment group Discussion The management of sinonasal cancer continues to represent a significant treatment challenge. Because of the paucity of these tumors, randomized trials are impossible to perform and several uncertainties remain about the optimal management of this disease. The vast majority of published data consists of small retrospective reviews by single institutions with a substantial diversity in both patient selection and treatment. In a first analysis, the long-term Leuven experience with conventional or conformal radiotherapy was reviewed. The patient distribution in our population was consistent with that of other publications: sinonasal cancer is much more common in men than in women, and the ES and MS are the most affected sites [10 14]. Worldwide, SCC is the most frequent pathology [10 14]. In our series however, adenocarcinoma was more frequent than SCC, probably because Leuven is situated in a region with a flourishing furniture industry, and professional exposure to wood dust (present in 58.3% of our patients) is a significantly greater risk factor for adenocarcinoma (x 896) 94

95 A. Clinical application of IMRT than for SCC (x 20) [10, 12]. As the localization of these tumors permits inconspicuous growth, most patients present with locally advanced disease [10 14, 143, 144]. In our series, 93.7% of patients were diagnosed with T3 4 disease. The presenting symptoms result from the invasion of surrounding structures, but are generally non-specific and often mimic benign conditions such as sinusitis, resulting in a further delay of diagnosis. In our patients, the most common reported symptoms were nasal obstruction, epistaxis, and headache, and there was a mean interval of 6.5 months between the onset of symptoms and the diagnosis. Because of the considerable variation in patient selection and therapeutic strategy in most published studies, it is difficult to compare treatment outcome between series. Local control, overall survival, and disease-free survival with conventional or conformal radiotherapy were respectively 53%, 54%, and 37% at 5 years follow-up. These findings are more or less consistent with those found in the literature, as reviewed in Table 20. The predominant failure pattern in our study was local recurrence, further reflecting the problem of achieving adequate local control and the necessity of multidisciplinary treatment [ ]. Distant metastases, and certainly regional relapses, were much less common sites of failure. Only 6 patients developed a nodal failure, and 3 of those 6 patients could be successfully salvaged. Two other large series, both published in 2001, analyzed clinical outcome of SNC treated by several different approaches. Dulguerov et al. reviewed 220 patients treated between 1974 and 1994, and reported 5-year LC and OS rates of 57% and 63% respectively [153]. Grau et al. examined all patients (n = 315) treated in Denmark between 1982 and 1991; LC was 41% and OS was 35% at 5 years [154]. One recently published study with a comparable population to our analysis evaluated the outcome of 106 patients with paranasal sinus cancer treated by preoperative (n = 28), post-operative (n = 41), or primary (n = 37) RT between 1960 and 1998, and reported 5-year LC and OS rates of 58% and 27%, respectively [158]. Although our long-term results lie within the published range, the high local failure rate was disappointing and suggested a need for improved local treatment. In our analysis, no statistically significant association was observed between RT dose or technique and outcome. Although a meta-analysis suggested improved outcomes in recent decades, this was not evident from our results [153]. Indeed, the relatively low rates of disease control and survival in sinonasal cancer, treated with conventional or conformal techniques, remained virtually consistent in all published studies since the early 90s. 95

96 A. Clinical application of IMRT Table 20. Selected literature review of treatment results with conventional or conformal radiotherapy in sinonasal cancer. Authors No. Site Treatment LC OS DFS Jiang et al., 1991 [147] 73 MS Surgery + post-op RT 78% - 51% Paulino et al., 1997 [148] 42 MS RT ± Surgery - 43% 33% Waldron et al., 1998 [149] 29 ES Primary RT 41% 39% - Jiang et al., 1998 [150] 34 ES RT ± Surgery 71% 55% 58% Tiwari et al., 1999 [151] 50 ES Surgery + post-op RT - 65% - Jansen et al., 2000 [152] 50 PNS Surgery + post-op RT 65% 60% 53% Dulguerov et al., 2001 [153] 220 NC/PNS Surgery and/or RT ± Cx 57% 63% - Grau et al., 2001 [154] 315 NC/PNS Surgery and/or RT ± Cx 41% 35% - Claus et al., 2002 [155] 47 ES Surgery + post-op RT 59% 60% 36% Katz et al., 2002 [156] 78 NC/PNS RT ± Surgery 60% 50% - Porceddu et al., 2004 [157] 60 PNS Surgery and/or RT ± Cx 49% 40% 34% Blanco et al., 2004 [158] 106 PNS RT ± Surgery 58% 27% 33% Hoppe et al., 2006 [159] 85 NC/PNS Surgery + post-op RT 62% 67% 55% Chen et al., 2007 [160] 127 NC/PNS Surgery and/or RT ± Cx 62% 52% 54% Snyers et al., 2009 [161] 168 NC/PNS RT ± Surgery 62% 35% 30% Leuven experience 127 NC/PNS RT ± Surgery 53% 54% 37% Abbreviations: No.: number of patients included in the analysis; NC: nasal cavity; ES: ethmoid sinus; MS: maxillary sinus; PNS: paranasal sinuses; RT: radiotherapy; Cx: chemotherapy; LC: 5- year local control; OS: 5-year overall survival; DFS: 5-year disease-free survival. Several studies have investigated potential prognostic factors in sinonasal cancer. T- classification is an evident prognostic factor for all head and neck cancers and its significance was also consistently demonstrated in our series, as in most other studies. Often, N- classification is similarly found to have an adverse effect on disease control and, consequently, survival. Although the incidence of N+ patients (n = 5) was too low to include nodal involvement into the statistical analysis, all but one N+ patient ultimately died from the disease. As in other series, both local control and survival were adversely affected by invasion in surrounding structures. In our analysis, intracranial invasion was a more consistent and significant prognostic factor than orbital infiltration, like for instance in the analysis by Blanco et al. [158]. However, because of the retrospective design of our study and the lack of high-quality CT scans for the patients who were treated in the 70s, results should be interpreted with caution. Nonetheless, it is clear that both primary tumor extent and lymph node involvement are important prognostic factors for sinonasal cancer. In contrast to other publications, we found no significant difference in disease control between tumor localizations [153, 154]. However, the negative prognostic impact of SCC histology was confirmed [149, 153, 159]. Outcome after single-modality therapy is generally poor [143, 146, 149, 153]. For 96

97 A. Clinical application of IMRT instance, Amendola et al. reported 5-year OS rates of 31% and 35% for surgery and RT alone, respectively (p = NS) [162]. At Leuven, a relatively constant treatment philosophy was followed troughout the years and, generally, multidisciplinary treatment was preferred whenever possible. Therefore, primary RT was not included as a factor in the statistical analysis because inherent selection bias: only unresectable or medically inoperable patients received RT alone. Nowadays, post-operative RT is generally considered to be the optimal sequencing of RT relative to surgery [147, 151, 152, 155, 159]. As mentioned before, no randomized-controlled trials could be conducted and currently, it is unclear whether there is any true superiority of one approach over the other [163, 164]. In our analysis, the switch from pre-operative to postoperative RT was made due to changed surgeon s preference. Post-operative RT resulted in better local control (not statistically significant) and disease-free survival (only just statistically significant). No chemotherapy was used in this analysis. Recently, however, the incorporation of (neo-)adjuvant Cx in the treatment of SNC, as in other head and neck cancers, has produced encouraging results, and would now be considered for certain indications [165, 166]. Another controversial topic in the management of sinonasal tumors is the precise role of elective neck irradiation. Although there were few (4.7%) regional failures in our series, despite the fact that none of the patients received elective neck irradiation, it has been correctly suggested that the risk of subclinical involvement in the neck can be considered sufficiently high in some patients to warrant elective treatment [147, 148, 167]. Jiang et al. found that histological type (SCC and undifferentiated carcinoma) was the most important prognostic factor for nodal relapse (rates of 33% and 50%, respectively) in 73 patients with tumors of the MS [147]. Paulino et al. reported 11 (28.9%) nodal recurrences in 38 patients with N0 tumors (all SCC) of the MS [148]. Recently, Le et al. saw a 5-year actuarial nodal relapse risk of 12% in 97 patients with a MS tumor, significantly correlated with SCC histology. However, none of the patients presenting with SCC histology and N0 necks had a nodal relapse after elective neck irradiation, highlighting the efficacy of this treatment [167]. Still, even without elective neck irradiation a 5-year regional control rate of 93% was achieved in our series. Only 4.7% (6/127) of the entire population, and 3.3% (4/122) of all originally N0 patients, developed a regional recurrence during follow-up. Of those 4 N0 patients who relapsed in the neck, 3 could be successfully salvaged. Elective neck irradiation should be evaluated in each patient on an individual basis, carefully considering the local extent and the lymphatic drainage of the primary tumor. Despite considerable variations in patient selection and therapeutic strategy, nearly all published studies analyzing outcome after conventional or conformal RT for sinonasal cancer have reported a high rate of complications [ ]. Particularly chronic toxicity of the optic 97

98 A. Clinical application of IMRT system is of major concern, and RT-induced blindness rates of up to 37% have been reported, depending on radiation dose and irradiated volume [ ]. Seventeen (23.6%) of the 72 patients with longer ( 2 years) developed moderate to severe visual impairment as a result of retinopathy or optic neuropathy. This number is probably even somewhat underestimated because of the retrospective nature of this analysis. Clearly, the outcome of conventional and conformal radiotherapy is suboptimal in terms of both disease control and radiation-induced morbidity. Results from several published studies in sinonasal cancer suggest that the use of IMRT reduces treatment-related toxicity, although it is not yet clear whether local control or survival is also improved [ ]. Obviously, the supreme conformality of IMRT theoretically increases the risk of a marginal miss [42 44]. On the other hand, target coverage is undoubtedly superior with IMRT [179, 180]. In our initial experience with IMRT, the almost complete lack of acute or late toxicity and the excellent tumor control rates confirmed its potential for SNC, and IMRT became the standard treatment modality in our department. In a second analysis, we presented an update of our prospective results with IMRT in a larger patient group with longer follow-up, as well as a retrospective comparison with patients treated with three-dimensional conformal radiotherapy without intensity-modulation. In the IMRT group, local control at 2 years was 76% and overall survival at 2 years was 89%, well within the range of earlier studies [ ]. Compared to the 3D-RT group, a significant improvement in disease-free survival was observed (60% vs. 72%, p = 0.02). A possible explanation is the underdosage of the target volume that often had to be tolerated with 3D-RT, in an effort to prevent radiation-induced blindness or temporal lobe necrosis. Obviously, all the limitations of a retrospective, single-institute analysis on a relatively small number of patients (n = 81) apply, and this percentage is most likely an overestimation of the real benefit. First of all, stage migration because of more accurate imaging techniques might be responsible for a certain proportion of the perceived benefit. For instance, all IMRT-treated patients received a pre-operative MRI, which was also used for GTV delineation. Secondly, surgical techniques have improved as well, with increasing experience in the use of endoscopic sinus surgery at our department [24]. Nonetheless, the two patients groups were largely similar, since our institutional guidelines remained constant throughout. It is in any case apparent that important progress in the treatment of SNC was made in recent years and that IMRT is associated with at the very least equal tumor control rates as conventional techniques [153, 160]. Moreover, because the possibility of geographic miss is a possible concern with IMRT, the fact that there were no marginal failures is reassuring. Acute toxicity in the IMRT group was, if present, always mild: no grade 3 or 4 side-effects 98

99 A. Clinical application of IMRT were reported. Not surprisingly, IMRT significantly reduced the incidence of acute toxicity compared to the 3D-RT group. Again, the retrospective design could have introduced a significant bias in the analysis and these results must be interpreted with caution. However, the actual toxicity in the 3D-RT group is likely to have been even higher (rather than lower) due to incomplete documentation. In longer follow-up, only 3 patients started to suffer from a dry, red, and itchy eye, requiring chronic treatment with artificial tears. This confirms the ability of IMRT to prevent the so-called dry eye syndrome, which usually starts during the first year after RT [173]. With relatively mature follow-up, it is reassuring that no patient to date developed radiation-induced visual impairment after treatment with IMRT, although the follow-up period is still too short to allow definitive conclusions. Optic neuropathy usually develops 2 to 4 years after treatment, but can occur as late as 14 years after treatment. Radiation-induced retinopathy develops earlier, within a period of 1.5 to 5 years after RT. Therefore, although these results are certainly encouraging, it should be stressed that additional follow-up is necessary before final conclusions can be reached. In conclusion, it is clear that IMRT for sinonasal cancer significantly improves diseasefree survival compared to three-dimensional conformal radiotherapy without intensitymodulation, and reduces acute as well as late toxicity. In light of these results, IMRT should be considered the standard treatment modality for SNC. 99

100

101 B. TARGET DEFINITION: SELECTION OF THE NODAL TARGET VOLUME

102

103 CHAPTER 1: RETROPHARYNGEAL LYMPH NODES IN OROPHARYNX CANCER Published as: Bussels B., Hermans R., Reijnders A., Dirix P., Nuyts S., Van den Bogaert W. Retropharyngeal nodes in squamous cell carcinoma of oropharynx: incidence, localization, and implications for target volume. Int J Radiat Oncol Biol Phys 2006; 65(3): Dirix P., Nuyts S., Bussels B., Hermans R., Van den Bogaert W. Prognostic influence of retropharyngeal lymph node metastasis in squamous cell carcinoma of the oropharynx. Int J Radiat Oncol Biol Phys 2006; 65(3):

104

105 B. Target definition: selection of the nodal target volume 1. Retropharyngeal lymph nodes in oropharynx cancer 1.1. Retropharyngeal lymph nodes The retropharyngeal (RP) lymph nodes, also called nodes of Rouvière, lie within the retropharyngeal space (RPS), a thin virtual space bordered anteriorly by the pharyngeal constrictor muscles and posteriorly by the prevertebral fascia. The RPS extends from the base of skull to approximately the T4 vertebral level, although no RP nodes are found below the level of C3 caudally [181]. Although hidden from clinical examination behind the posterior pharyngeal wall, the existence of these nodes was mentioned by numerous authors long before radiological images were available [182]. Historically, the RP lymph nodes are divided into 2 anatomically distinct subgroups: medial and lateral. The medial RP nodes are located near the midline and anterior to the prevertebral musculature; the lateral RP nodes are situated between the internal carotid artery and the pharyngeal airspace, anterolateral to the prevertebral muscles [183, 184]. As pathological changes in the RP nodes are very difficult to detect clinically and their surgical dissection is usually quite complicated, radiological assessment is essential for diagnosis [185, 186]. In 1983, Mancuso et al. published a two-part study describing the normal and pathological anatomy of RP nodes [187, 188]. They observed that abnormal RP nodes can be as small as 7 to 15 mm, so a size limit for normal RP lymph nodes of 10 mm (axial diameter) was suggested. Radiological diagnosis of RP adenopathy by CT or MRI is very effective [189]. Of 42 patients with advanced carcinoma of the oropharynx and hypopharynx who underwent RP node dissection, 6 (14,3%) had pathologically proven RP node metastasis. Of these 6, preoperative diagnosis was possible in 5 patients on either CT or MRI. Sensitivity, specificity, and accuracy of CT scan were all 100%. Recently, there has been some interest in the use of ultrasound for the imaging of RP nodes, which could be valuable for monitoring response during radiotherapy or for follow-up purposes [190]. Ballantyne, in a landmark study of 1964, was the first to draw attention to the RP nodes as a possible site of metastasis [191]. He performed a nodal dissection of the RPS in 34 patients with SCC involving the pharyngeal wall and found pathologically confirmed disease in 15 (44%) patients. During follow-up it became clear that patients with RP adenopathy had very poor survival. Although very few studies have been published since Ballantyne s initial observation, it is sometimes presumed that patients who present with RP adenopathy are virtually incurable [192]. Indeed, some authors suggest that patients in whom the disease has 105

106 B. Target definition: selection of the nodal target volume reached the RP nodes will sooner or later present with distant metastasis and that loco-regional treatment can no longer help these patients [193, 194]. Most radiological studies concentrated on nasopharyngeal cancer [ ]. Concerning non-nasopharyngeal cancer, only a few reports were published [200, 201]. Recently, consensus guidelines were published concerning CT-based delineation of lymph nodes levels in the node-negative neck [106]. The panel proposed to delineate the RP nodes from the base of skull to the cranial edge of the body of the hyoid bone. They proposed to use the fascia posteriorly to the pharyngeal mucosa as the anterior limit and the prevertebral muscle as the posterior limit. Laterally, the retropharyngeal nodes are limited by the medial edge of the internal carotid artery and medially extend at or near the midline. As recurrences in this space are uncontrollable and result in a dismal outcome, a careful delineation of the region most likely to be involved as well as a selection of the patients in which this space should be part of the target volume is of the utmost importance. This analysis presents the first radiological series concentrating on the incidence pattern and anatomical localization of RP nodes in oropharyngeal cancer. The final aim was to evaluate the implications of our findings for recommendations concerning target volume definition and delineation of the RP nodes in preparation for intensity-modulated radiotherapy Material & methods Patient population All patients presenting between 1984 and 2003 with oropharyngeal SCC at the University Hospitals Leuven were included in this study (n = 344, after exclusion 208 CT studies remained evaluable). The patient and tumor characteristics are represented by Table 21. There were 182 male (87.5%) and 26 female (12.5%) patients. The majority of the patients (n = 139, 67%) presented with tonsillar tumors. In the other patients, the tumor was located at the base of tongue (n = 31, 15%), the posterior pharyngeal wall (n = 13, 6%), the vallecula (n = 13, 6%), the soft palate (n = 9, 4%), and the lateral pharyngeal wall inferior to the tonsil (n = 3, 2%). Staging was performed according to the TNM classification staging system of the International Union Against Cancer (1997) and was only based on radiological criteria. The majority of the patients presented with locally (> T2 tumors: 63.5%) and regionally (N+: 71.2%) advanced tumors; 86.1% had stage III or IV disease. 106

107 B. Target definition: selection of the nodal target volume CT studies Patients without pre-therapeutic CT studies or with incompletely available or low quality imaging studies were excluded. Also, patients in whom a biopsy was obtained just before the CT study were excluded. Finally, 208 pre-therapeutic CT studies remained for further analysis. CT examinations were obtained with different CT machines in different hospitals. All imaging studies were performed during IV injection of a contrast agent. All studies had a slice thickness of 3 5 mm. One radiologist, experienced in head and neck oncology, reviewed all imaging studies. Lymph nodes were considered pathological when showing a minimal axial diameter of 1 cm or when showing a central hypodensity [202, 203]. The localization of the nodal neck disease was registered using the definition of the neck nodal levels according to the recently published international consensus guidelines concerning neck nodal delineation in the node-negative neck [106]. Special attention was given to the RP nodes. For each patient with pathological RP nodes, the area involved by the pathological RP nodes was manually drawn on a schematic anatomical neck, in a sagittal and a transverse plane, considering both localization and volume of these nodes. The nodal center was determined by the crossing of their longest and shortest axis. The outer anatomical extension of the pathological RP nodes was determined by the margin of the superimposed drawings Treatment Treatment was decided by a multidisciplinary team according to institutional guidelines: 175 (84.1%) patients received radiotherapy alone, 24 (11.5%) underwent surgery followed by radiotherapy, and 9 (4.4%) had concomitant chemoradiotherapy. If patients were treated with primary radiotherapy, they received a total dose of Gy in daily fractions of Gy (84.6% of patients). A minority (15.4%) received the hybrid fractionation schema (currently standard) consisting of 20 daily fractions of 2 Gy followed by 20 fractions of 1.6 Gy twice daily to a total dose of 72 Gy. Post-operative radiotherapy consisted of a dose of 50 Gy in 25 daily fractions of 2 Gy followed by an additional boost of 16 Gy if appropriate (i. e. extranodal spread of lymph node metastasis and/or positive resection margins). In all patients a conventional radiation technique was used with 2 opposing lateral beams and 1 lower neck field for the supraclavicular regions. The upper border of the lateral fields was at the base of skull to include the RPS, so all patients received a minimal dose of 50 Gy on the RP lymph nodes. 107

108 B. Target definition: selection of the nodal target volume Table 21. Patient and tumour characteristics. Characteristics RP+ patients RP- patients All patients GENDER Male 29 (85.3%) 153 (87.9%) 182 (87.5%) Female 5 (14.7%) 21 (12.1%) 26 (12.5%) SUBSITE Tonsil 19 (55%) 120 (69%) 139 (67%) Base of tongue 4 (12%) 27 (16%) 31 (15%) Posterior pharyngeal wall 5 (15%) 8 (4%) 13 (6%) Vallecula 1 (3%) 12 (7%) 13 (6%) Soft palate 5 (15%) 4 (2%) 9 (4%) Lateral pharyngeal wall 0 (0%) 3 (2%) 3 (2%) T-CLASSIFICATION T1 3 (8.8%) 22 (12.6%) 25 (12.0%) T2 1 (2.9%) 50 (28.7%) 51 (24.5%) T3 14 (41.2%) 45 (25.9%) 59 (28.4%) T4 16 (47.1%) 57 (32.8%) 73 (35.1%) N-CLASSIFICATION N0 0 (0%) 60 (34.5%) 60 (28.8%) N1 5 (14.7%) 37 (21.3%) 42 (20.2%) N2 26 (76.5%) 69 (39.6%) 95 (45.7%) N3 3 (8.8%) 8 (4.6%) 11 (5.3%) CLINICAL STAGE Stage I 0 (0%) 9 (5.2%) 9 (4.3%) Stage II 0 (0%) 20 (11.5%) 20 (9.6%) Stage III 5 (14.7%) 47 (27.0%) 52 (25.0%) Stage IV 29 (85.3%) 98 (56.3%) 127 (61.1%) TREATMENT RT 27 (79.4%) 148 (85.1%) 175 (84.1%) Surgery + RT 5 (14.7%) 19 (10.9%) 24 (11.5%) RT + Cx 2 (5.9%) 7 (4.0%) 9 (4.4%) TOTAL 34 (16%) 174 (84%) 208 Concomitant chemotherapy consisted of cisplatin 100 mg/m² on days 1, 22, and 43 of the radiotherapy treatment. Surgery involved resection of the primary tumor with an elective neck dissection in the N0 neck or a radical neck dissection in the N+ neck. Resection of the RP lymph nodes is not performed in a standard radical neck dissection, so in none of the patients the RP nodes were removed. 108

109 B. Target definition: selection of the nodal target volume Of the 34 patients with RP node metastasis, 22 (65%) patients were irradiated to 70 Gy (one patient had concomitant cisplatinum), 7 (20%) patients received the hybrid fractionation schedule to 72 Gy (in one patient with concomitant chemotherapy), and 5 (15%) patients had surgery with post-operative radiotherapy Statistical analysis Patient characteristics (age, gender, primary tumor site, TNM classification, clinical stage, and treatment modality) were previously recorded, as well as the presence of RP node metastasis. Follow-up data were retrospectively collected: information was sought on the dates of first recurrence (local or regional), distant metastasis, or death. Local or regional relapse, distant metastasis, death from disease, and death from another cause were used as endpoints. The close-out date for survival analysis was August, The potential follow-up time for each patient was the time from treatment start to the close-out date. The data were analysed using the software package Statistica 7 (StatSoft Inc, Tulsa, OK, USA). An univariate analysis was performed using the Fisher s exact test, a multivariate analysis was performed using logistic regression. Cumulative survival and tumor control rates were calculated using the Kaplan-Meier product-limited (actuarial) method. Survival and recurrence rates were then compared between patients with and without RP lymph node metastasis using the log-rank test. A p value of less than 0.05 was considered significant. The significance and independence of each parameter (age, primary tumor site, T-classification, N- classification, clinical stage, treatment modality, and RP lymph node involvement) was tested using a Cox regression method Results Incidence of RP nodal involvement Of all 208 patients, 34 (16%) presented with pathological RP nodes. In patients with positive nodal neck disease other than RP, 23% (31/134) presented with RP nodal involvement. In patients with pathological RP nodes, the primary tumor was located in 55% (n = 19) of the patients at the tonsil, in 12% (n = 4) at the base of tongue, in 15% (n = 5) at the posterior pharyngeal wall, in 3% (n = 1) at the vallecula, and in 15% (n = 5) at the soft palate. In 88% of the patients with pathological RP nodal involvement, level II ipsilateral to the 109

110 B. Target definition: selection of the nodal target volume tumor was involved; the ipsilateral levels I, III, IV, and V were involved in respectively 20%, 53%, 9%, and 15%. Patients with involved RP nodes presented with contralateral involved nodes in level II (44%) and III (29%); none of the patients presented with involvement of level I and level V, 3% with involvement of level IV. This is shown in Table 22. Table 22. Involvement of neck nodal levels in patients with RP involvement. Level Ipsilateral neck Contralateral neck I 20 % 0 % II 88 % 44 % III 53 % 29 % IV 9 % 3 % V 15 % 0 % A solitary ipsilateral RP node without involvement of other neck nodal levels was found in 9% (3/34) of patients; in 2 of these 3 patients the primary tumor was located at the posterior pharyngeal wall. Solitary contralateral pathological RP nodes were not observed. Retropharyngeal nodes contralateral to the tumor occurred in 11.7% (4/34) of all patients with pathological RP nodes, 2 of those four patients had bilateral RP nodes. Patients with bilateral RP nodes had a primary tumor located at the tonsil (1 patient) or at the base of tongue (1 patient). Patients with contralateral RP nodes without ipsilateral retropharyngeal nodes had a primary tumor located at the base of tongue (1 patient) or at the soft palate (1 patient). Of all patients with a tonsillar tumor, 14% (19/139) had pathological RP nodes. This number is similar for tumors located at the base of tongue (13%, 4/31). Patients with a tumor located at the posterior pharyngeal wall (38%, 5/13) or soft palate (56%, 5/9) had more frequent involvement of the RP nodes. In tumors of the vallecula, the RP nodes were only involved in 8% (1/13) of cases and patients with a tumor of the lateral pharyngeal wall had no pathological RP nodes. This is shown in Table 23. In univariate analysis, involvement of the RP nodes was significantly correlated with T- classification (p = 0.016), N-classification (p = 0.001), and with involvement of level Ib (p = 0.01), II (p = 0.002), III (p = 0.013), IV (p = 0.048), and V (p = 0.022) ipsilateral to the tumor and level II (p = 0.002) and III (p < 0.001) contralateral to the tumor. In multivariate analysis, involvement of level II in the ipsilateral neck (p = 0.028) and level III in the contralateral neck (p < 0.001) were statistically significant predictive factors for RP nodal involvement. 110

111 B. Target definition: selection of the nodal target volume Table 23. Primary tumors with respect to involvement of RP nodes. Tonsil BOT PPW Vallecula SP LPW Total RP- 120 (86%) 27 (87%) 8 (62%) 12 (92%) 4 (44%) 3 (100%) 174 (84%) RP+ IL 18 (13%) 2 (7%) 5 (38%) 1 (8%) 4 (44%) 0 (0%) 30 (14%) CL 0 (0%) 1 (3%) 0 (0%) 0 (0%) 1 (12%) 0 (0%) 2 (1%) BL 1 (1%) 1 (3%) 0 (0%) 0 (0%) 0 (0%) 0 (0%) 2 (1%) All 19 (14%) 4 (13%) 5 (38%) 1 (8%) 5 (56%) 0 (0%) 34 (16%) Total Abbreviations: BOT: base of tongue; PPW: posterior pharyngeal wall; SP: soft palate; LPW: lateral pharyngeal wall; RP-: negative retropharyngeal lymph nodes; RP+: positive retropharyngeal lymph nodes; IL: ipsilateral; CL: contralateral; BL: bilateral Anatomic localization of pathologic RP nodes In the sagittal plane (Figure 13), the centers of all pathological RP nodes were localized at the level of C1 and C2. The maximal extension of the nodes was up to the base of skull cranially and down to the caudal border of C2. In the transverse plane (Figure 14) all but one center (of one RP node in one patient) were located in the space bordered laterally by the internal carotid artery and medially by the prevertebral muscles. The maximal extension from the nodes was anteriorly to the lateral pharyngeal wall, posteriorly to the vertebral transverse process, medially to the lateral margin of the prevertebral muscle and laterally into the carotid space (level of internal jugular vein). Only one patient (0.5% of the entire patient population) presented with a node located anterior to the prevertebral muscle. This patient had also a pathological contralateral RP node localized between the prevertebral muscle and the carotid artery. The primary tumor of this patient was located at the posterior pharyngeal wall. None of the patients presented with a node within the space between the prevertebral muscles. 111

112 B. Target definition: selection of the nodal target volume Figure 13. Sagittal plane. 112

113 B. Target definition: selection of the nodal target volume Figure 14. Transverse plane Prognostic influence of pathologic RP nodes Two patients (1.0%) were lost to follow-up immediately after treatment and 5 patients (2.4%) had eventless follow-up of less than a year, so these 7 patients were excluded from statistical analysis. Only one of them had radiological evidence of RP lymph node metastasis on the pre-therapeutic CT. This left a total of 201 patients for survival analysis, of which 33 (16%) had involvement of the RP nodes. Median follow-up was 47 months (range, months). Local recurrence after 5 years was 37% in the overall population. A significant correlation with T-classification (p = ) and N-classification (p = ) was found, although not with clinical stage (p = 0.27). In multivariate analysis both T-classification (RR 2.15 [95% CI ], p = 0.002) and N-classification (RR 1.67 [95% CI ], p = 0.03) independently predicted the likelihood of local failure. Patients with involvement of the RP nodes had more local recurrences than those without RP metastasis (47% vs. 35%, p = 0.54). Of the 201 patients, 15% had regional recurrence. This rate was significantly correlated with T-classification (p = 0.02), N-classification (p = 0.02), and clinical stage (p = 0.02). Patients with RP nodes had significantly more regional recurrences than those without RP involvement (45% vs. 10%, p = 0.004), as shown in Figure 15. Only involvement of RP lymph nodes (RR 113

114 B. Target definition: selection of the nodal target volume 4.29 [95% CI ], p = 0.01) significantly and independently predicted regional recurrence in multivariate analysis. Interestingly enough, in only one of the patients with initial RP node involvement who later presented with regional recurrence, the RP lymph nodes themselves were also involved in the recurrence. Figure 15. Regional recurrence. 100% Regional Recurrence Regional Recurrence (%) 90% 80% 70% 60% 50% 40% 30% RP - RP + All patients p = 0,004 20% 10% 0% Time (years) Distant metastasis in the overall population of 201 patients was 22% after 5 years. There was a significant correlation with T-classification (p = 0.03) and clinical stage (p = 0.01), but not with N-classification (p = 0.11). Patients with involvement of the RP nodes developed more metastases than patients without RP disease (26% vs. 22%, p = 0.34). Disease-specific survival in the overall population was 54% at 5 years. A large difference between the two groups (Figure 16) was found. Of patients with RP nodal disease, 62% had died from the disease after 5 years. In the group without RP disease this was only 42%. This 114

115 B. Target definition: selection of the nodal target volume difference is statistically significant (p = 0.03). In multivariate analysis, T-classification (RR 1.69 [95% CI ], p = 0.001) and N-classification (RR 1.68 [95% CI ], p = 0.02) were the only factors to independently influence disease-specific survival. Figure 16. Disease-specific survival. 100% Disease-specific Survival Disease-specific Survival (%) 90% 80% 70% 60% 50% 40% 30% RP - RP + All patients p = 0,03 20% 10% 0% Time (years) Overall survival at 5 years was 45%. There was no significant difference between the two groups in overall survival (36% vs. 46% respectively, p = 0.3). As patients with head and neck cancer usually suffer from other health problems, overall survival is affected by comorbidities. Indeed, 46 of the 201 patients (23%) died from other causes in 10 years time. 115

116 B. Target definition: selection of the nodal target volume 1.4. Discussion The incidence of metastatic nodal involvement of different neck nodal levels in head and neck cancer patients has since long been the subject of a multitude of studies. The first papers were published in the 70 s based on clinical examination; later most series were based on surgical findings [ ]. The results of these studies had a tremendous influence on the treatment strategy of head and neck cancer by introducing the concept of selective nodal treatment both in the surgical and the radiotherapeutic management. Obviously, interest in selective nodal irradiation is increasing since the introduction of IMRT. Correct delineation and definition of the target volume is a critical step towards highly conformal radiotherapy, as missing the tumor inevitably leads to an increased risk of persisting or recurring tumor. The incidence and anatomical presentation of RP nodes is not well represented in the literature because of their localization hidden from clinicians and often also from surgeons. Very few data are available on the incidence of retropharyngeal nodal involvement, especially in oropharyngeal cancer, although undertreatment can very well lead to recurrence, which in this area often escapes local treatment modalities. In the present study, a high incidence of retropharyngeal nodal involvement was observed: 16% in the total patient population and 23% in patients with involvement of neck node levels other than RP. These results are in accordance to the literature; reported incidences of RP nodes in oropharyngeal cancer vary between 0% and 63% [200, 201]. Given the high incidence of RP nodal involvement in oropharyngeal cancer obtained in our series, inclusion of this nodal level in the target volume must be strongly considered in node-positive necks. The inclusion of the RP nodes in the elective target volume in node-negative necks should be mandatory in posterior pharyngeal wall tumors as these tumors were most likely to present with solitary involved RP adenopathies. In the multivariate analysis, involvement of level II at the ipsilateral side of the neck and level III contralateral to the tumor were found to be significant predictors for RP nodal involvement. Whereas level II ipsilateral to the tumor is generally accepted as the first nodal level involved in oropharyngeal nodes, level III contralateral to the tumor reflects advanced disease, likely correlated with the presence of pathological RP nodes. The observed anatomical localisation of the RP nodes in the sagittal plane, at the level of C1 and C2 with maximal extension up to the base of skull cranially and down to the caudal border of C2, is in total accordance with the recent published international consensus guidelines concerning target volume delineation in the node negative neck [106]. In the transverse plane all but a single retropharyngeal node were located in the space bordered laterally by the internal 116

117 B. Target definition: selection of the nodal target volume carotid artery and medially by the prevertebral muscles. The maximal extension from the nodes was anteriorly to the posterolateral pharyngeal wall, posteriorly towards the vertebral processus transversus, medially towards the prevertebral muscle, and laterally towards the carotid artery and vein. Only one patient (0.5% of the population) presented with a node located anterior to the prevertebral muscle, the primary tumor was located at the posterior pharyngeal wall. Although posterior pharyngeal wall tumors and soft palate tumors are rather rare in our series, in agreement with the general patient population, they represented together one third of the primary locations in patients with pathological RP nodes. Posterior pharyngeal wall tumors were most likely to be linked with solitary RP nodes and were the only tumor site presenting with a medially located RP nodes. As they are a midline structure we would therefore recommend including the bilateral medial and lateral RP nodal area in the target volume of posterior pharyngeal wall and soft palate tumors. Concerning the other tumor localizations, the value of extending the delineation of the lateral RP nodal area to the space between and anteriorly to the prevertebral miscles, as proposed in the international consensus guidelines, can be questioned. The discussion on the incidence of the medial retropharyngeal nodes was initiated with the work of Rouvière mentioning that RP nodes were found in 1 out of 5 patients [182]. However, the radiological study by Mancuso et al. mentioned that medial RP nodes located near the midline anterior to the prevertebral muscle were never identified [187, 188]. Chong et al. detected enlarged RP nodes in 52% (59/114) of the studied patients, all nasopharyngeal carcinoma, which is the most common site of primary malignancy with RP metastatic disease. None of the patients presented with medial RP lymfadenopathies [196]. More recently, Poon et al. detected most of the RP nodes clustered laterally away from the midline, and RP nodes medially approaching the midline were seen infrequent [209]. In view of all the above mentioned findings and our own results only 0.5% of the entire studied population presented with a medial retropharyngeal node it can be proposed that the medial RP nodes are to be regarded as a separate rarely involved neck nodal group, which does not need to be systematically included with the lateral RP nodes in one target volume. The prognostic importance of involvement of RP lymph nodes has received little attention. Ballantyne described in 1964 how RP nodal metastasis had a negative impact on prognosis in patients with pharyngeal wall lesions (66% vs. 84% 3-year overall survival). This observation led him to hypothesise that removal of positive retropharyngeal nodes may increase chances for survival in cancers of selected sites [191]. Very few studies have been published since. Most series concentrated on nasopharynx cancer (NPC), in which retropharyngeal lymph node metastasis is present in most patients and has no clear adverse 117

118 B. Target definition: selection of the nodal target volume effect on survival [ ]. This is probably because the RP lymph nodes are one of the first echelons of lymphatic drainage for the nasopharynx and cervical node involvement of NPC spreads orderly down the neck with very few skip metastases. In contrast, cancer arising from the oropharynx or the hypopharynx usually spreads first to the upper jugular (level II) lymph nodes and subsequently to the middle jugular (level III) lymph nodes [207]. One would therefore expect the presence of RP lymph node metastasis to be of greater prognostic significance in those tumors. Indeed, it has been demonstrated that the presence of retropharyngeal lymph node involvement in hypopharyngeal carcinoma leads to a dismal outcome [210, 211]. Three studies have been published regarding the prognostic impact of RP lymph node metastasis in oropharyngeal cancer, all of them including tumours from other primary sites as well. Hasegawa and Matsuura treated 24 patients with stage III/IV SCC of the hypopharynx and oropharynx with standard resection and RP node dissection, followed by post-operative radiotherapy of 50 Gy to the RPS if any RP nodes were involved [200]. In the largest study to date, by McLaughlin et al., pre-treatment CT scans of 774 patients with SCC of the nasopharynx, oropharynx, hypopharynx, and supraglottic larynx were reviewed to determine the presence of RP adenopathy [201]. Almost all patients were treated with radiotherapy alone (82%) and only a few had surgery (18%). A study by Gross et al. included 51 patients who were surgically treated, including RP node dissection, for clinically advanced SCC of the oral cavity, oropharynx, hypopharynx, or supraglottic larynx [212]. Most patients received post-operative radiotherapy (55 65 Gy).In our series, retropharyngeal lymph node involvement did not influence local control. In multivariate analysis, both T-classification and N-classification were independent prognostic factors for local recurrence, but not clinical stage or RP lymph node metastasis. Similarly, in the study by McLaughlin et al. no statistically significant difference in local recurrence was found between the two groups [201]. However, the impact of RP lymph node metastasis on regional recurrence was quite substantial: RP involvement is an independent prognostic factor for regional recurrence in multivariate analysis. Rates of neck relapse in our series (45% vs. 10%, p = 0.004) are comparable to those reported by McLaughlin et al. (40% vs. 15%, p = ) [201]. Because of the small and statistically insignificant difference in distant metastasis (4%), the higher incidence of regional recurrence (35%) is the cause of worse disease-specific survival (20%). Very few patients who develop regional recurrence can be cured by salvage treatment. In our series for example, all but two patients who developed regional recurrence died from their disease. A significant correlation between RP node metastasis and disease-specific survival was found. Similarly, in two of the three other studies RP adenopathy was found to have a 118

119 B. Target definition: selection of the nodal target volume detrimental impact on survival. For example, in the study by McLaughlin et al., the rates of 5- year disease-free survival (35% vs. 58%, p = ) and absolute survival (29% vs. 44%, p = ) were significantly lower in the RP node positive group [201]. Hasegawa and Matsuura reported 5-year overall survival rates of 53% vs. 66 % respectively [200]. Only in the study by Gross et al. an inferior outcome for patients with RP nodal disease was not found: no statistically significant difference in loco-regional recurrence, distant metastasis, disease-free survival, and overall survival was seen between patients with and without RP involvement [212]. However, in this series only 43 patients were included for survival analysis, with an unknown proportion of oropharyngeal tumours, limiting the statistical power to find a significant difference. Since the most important predictors of loco-regional control in oropharyngeal cancer are T-classification as well as N-classification, it is logical that RP lymph node metastases imply a more advanced clinical stage, with a higher risk of disease recurrence. Inevitably, this difference in regional recurrence rate translates into a difference in mortality because of the low success rate of salvage procedures. On the other hand, 38% of the patients with initially involved RP nodes were still alive and without any evidence of disease after 5 years, demonstrating that a substantial proportion of these patients can be cured. Since radiotherapy can achieve long-term cure in gross retropharyngeal disease, it will most certainly succeed in eradicating microscopic disease and it is therefore imperative to take the possible involvement of these nodes into account when designing treatment strategy, especially in node-positive necks. In node-negative necks, inclusion of RP nodes into the target volume is advised in posterior pharyngeal wall tumors. As only one patient presented with a medially located RP node, it is proposed to regard this subgroup as a separate nodal region in target volume delineation. 119

120

121 CHAPTER 2: DW-MRI FOR LYMPH NODE STAGING & IMPACT ON RADIOTHERAPY PLANNING Published as: Dirix P., Vandecaveye V., De Keyzer F., Op de Beeck K., Vander Poorten V., Delaere P., Verbeken E., Hermans R., Nuyts S. Diffusion-weighted MRI for nodal staging of head and neck squamous cell carcinoma: impact on radiotherapy planning. Int J Radiat Oncol Biol Phys 2009; Jun 17. [Epub ahead of print].

122

123 B. Target definition: selection of the nodal target volume 2. DW-MRI for lymph node staging & impact on radiotherapy planning DW-MRI is an imaging technique able to detect the diffusion of water molecules in biologic tissues, which is quantified as the apparent diffusion coefficient. A prior investigation on the use of DW-MRI for nodal staging in HNC found a sensitivity of 84% and a specificity of 94% per LN in 33 surgically treated patients [72]. We performed a radiotherapy planning study on these patients. One set of target volumes was designed with conventional imaging (CT and T1/T2-MRI) guidance only, and another with the corresponding DW-MRI images. A third, reference set was contoured solely based on pathology results Material & methods Patient population In this prospective study, 22 patients underwent both a CT and a MRI scan before upfront surgical treatment for biopsy-proven, previously untreated HNC. Patients and tumor characteristics are summarized in Table 24. There were 13 male and 9 female patients, with a median age of 60 years (range, years). The study protocol was approved by the local ethics committee; informed consent was obtained from all patients Imaging The CT examinations were performed in a clinical routine setting: with slice thickness of 3 mm and after IV injection of a contrast agent. All MRI examinations were performed on a 1.5-T scanner (Magnetom SONATA, Siemens, Erlangen, Germany). First, transverse T2- and T1-weighted turbo spin-echo (TSE) sequences were acquired. Then, transverse DW sequences were acquired with b-values of 0, 50, 100, 500, 750, and 1000 sec/mm 2. Finally, the T1-weighted sequence was repeated after IV administration of a single-dose of Gadolinium-BOPTA (Multihance, Bracco, Milan, Italy). The MRI data were transferred to a Linux workstation with dedicated software (Biomap, Novartis, Basel, Switzerland). On the native DW images, acquired using a b-value of 0 sec/mm², regions of interest were drawn freehand by two experienced radiologists in consensus around all identifiable lymph nodes. These were then copied and pasted on the images acquired with other b-values. The ADC values were calculated by combining b0 and b1000 images. 123

124 B. Target definition: selection of the nodal target volume Table 24. Patient and tumor characteristics. Characteristics No. of patients % Gender Male Female Primary site Oral cavity Larynx Hypopharynx Unknown primary Pathologic AJCC clinical stage II III IVa Pathologic T-classification Tx T T T4a Pathologic N-classification N N N2b N2c Surgery, topographic correlation, and histopathologic analysis After imaging, patients went on to surgical treatment consisting of resection of the primary tumor in the 20 patients with known disease localization, and neck dissection in all 22 patients. Patients underwent either a selective (n = 4) or a (modified) radical (n = 7) unilateral dissection for clinically N0 or N1 disease, or bilateral neck dissections, i.e. bilateral selective (n = 5), ipsilateral (modified) radical and contralateral selective (n = 5), or bilateral (modified) radical (n = 1), in case of more advanced disease. Retropharyngeal nodes were not sampled during surgery. All neck dissection specimens were removed en bloc by the study s surgeons and divided into anatomical neck levels according to the American academy of otolaryngology head and neck surgery (AAO-HNS) criteria [213]. 124

125 B. Target definition: selection of the nodal target volume Particular attention was placed on standardized processing of the sampled tissues. All dissection specimens were initially examined grossly, in comparison with the contrast-enhanced T1-images, and the position of each LN was mapped in correlation to its surrounding anatomical structures. This allowed each LN to be assigned to its corresponding position on the T1-images. Afterwards, all LN were put in separate recipients, and carefully labeled. All LN were sectioned in 3 mm slices, from which 5 μm sections were prepared for histological examination. In addition to hematoxylin-eosin staining, prekeratine (Clone KL1 from AbD Serotec, Kidlington, UK) immunohistochemical staining was applied to improve the detection of small metastases. An experienced pathologist oversaw all pathologic evaluations Nodal GTV and CTV delineation The CT images were transferred to a dedicated treatment planning system (Eclipse, Varian, Palo Alto, CA) and used for a RT delineation study. First, 3 separate sets of nodal GTV were contoured: 1) based on conventional (CT and T1/T2-MRI) imaging, by a radiologist experienced in head and neck oncology who was unaware of the DW-MRI and the pathology results. The diagnosis of LN malignancy conformed to accepted morphologic criteria: a) any node with a minimal axial diameter > 1 cm; b) any node with internal central or peripheral attenuation suggestive of necrosis; c) extracapsular extension; and/or d) obliteration of fat or perivascular soft tissue planes [202, 203]. 2) based on DW-MRI, according to objective ADC calculations and blinded to pathology results. An ADC cut-off of mm 2 /sec, established in a previous study, was used for differentiation between malignant and benign LN [72]. 3) based on immuno-pathology results, and considered the standard of reference nodal GTV. Thereafter, 3 separate sets of nodal CTV were delineated by 2 radiation oncologists in consensus. The elective CTV was defined according to our institutional guidelines for primary RT of HNC (Table 2), based on the validated proposals by Eisbruch and Chao [ ]. Nodal levels were delineated according to the consensus guidelines [106]. For the nodenegative neck, the retrostyloid space was not included in the CTV and the upper limit of level II was placed at the caudal edge of the lateral process of the first vertebra [106]. For the nodepositive neck, the CTV was extended to the base of skull (jugular foramen) and included the retrostyloid space [107]. 125

126 B. Target definition: selection of the nodal target volume Statistical analysis The data were analyzed using the statistical package StatXact.3 (Cytel, Cambridge, MA). We used kappa measures to compare the agreement in nodal staging between each modality (conventional imaging and DW-MRI) and the reference pathology results. Differences in agreement with the pathology findings were tested for significance with an exact McNemar s test. For both imaging modalities, the absolute differences in RT volumes with those based on pathology results were calculated. An exact paired Wilcoxon test was used to compare these differences between both modalities. Since we assumed the superiority of DW-MRI, one-sided p-values were calculated; a p-value of less than 0.05 was considered statistically significant Results Neck dissection results A total of 33 heminecks were dissected: 11 patients underwent unilateral dissection, and 11 patients underwent bilateral dissections. From these dissections, a total of 433 lymph nodes (45 positive and 388 negative on pathology) were isolated from 128 nodal levels (32 positive and 96 negative on pathology). Of the 433 dissected lymph nodes, 198 could be identified on imaging, including all 45 pathologically involved nodes. These 198 lymph nodes had a median diameter of 6 mm (range, 4 30 mm) in the longest axis, and the majority (n = 171, 86.4%) were subcentimetric on conventional imaging. The 45 pathologically involved LN had a median diameter of 8 mm (range, 4 30 mm), also with more than half of the nodes (n = 28, 62.2%) measuring less than 1 cm Nodal staging Sensitivity, specificity, accuracy, negative predictive value (NPV), and positive predictive value (PPV) were calculated for both conventional imaging (CT and T1/T2-MRI) and DW-MRI (Table 25). Nodal staging agreement between imaging results and pathology findings was significantly stronger for DW-MRI (κ 0.97, 95% CI: ) than for conventional imaging (κ 0.56, 95% CI: ; p = by McNemar s testing). There was complete agreement for nodal staging between both imaging modalities and pathology in 9 of all 22 patients (5 pn0 patients, 3 pn2b patients, and 1 pn2c patient). Ten of the 22 patients were correctly staged by 126

127 B. Target definition: selection of the nodal target volume DW-MRI, but incorrect by conventional imaging alone: DW-MRI detected nodal disease in 4 patients considered N0 on conventional imaging, and bilateral disease in 2 patients considered to have unilateral disease. Two patients were correctly changed from N1 to N2b disease on DW-MRI. Also, 2 patients considered to have nodal disease on conventional imaging were correctly downstaged to N0 by DW-MRI. Three of the 22 patients were incorrectly staged by DW-MRI: 2 pn0 patients were wrongly staged as N1, and one pn2b patient was staged as N2c. Table 25. Nodal staging accuracy results for conventional imaging and DW-MRI. Per LN Per Level Per Neck Site Modality CT/MRI DW-MRI CT/MRI DW-MRI CT/MRI DW-MRI Sensitivity (%) Specificity (%) Accuracy (%) PPV (%) NPV (%) 19/45 (42.2%) 143/153 (93.5%) 162/198 (81.8%) 19/29 (65.5%) 143/169 (84.6%) 40/45 (88.9%) 149/153 (97.4%) 189 /198 (95.5%) 40/44 (90.9%) 149/154 (96.8%) 15/32 (46.9%) 92/96 (95.8%) 107/128 (83.6%) 15/19 (78.9%) 92/109 (84.4%) 30/32 (93.8%) 93/96 (96.9%) 123/128 (96.1%) 30/33 (90.9%) 93/95 (97.9%) 10/16 (62.5%) 14/17 (82.4%) 24/33 (72.7%) 10/13 (76.9%) 14/20 (70.0%) 16/16 (100.0%) 14/17 (82.4%) 30/33 (90.9%) 16/19 (84.2%) 14/14 (100%) Abbreviations: LN = lymph node; PPV = positive predictive value; NPV = negative predictive value Nodal GTV delineation The mean nodal GTV based on pathology was 2.43 cm 3 (range, cm 3 ). While DW-MRI overestimated the nodal GTV (mean: 2.49 cm 3 ; range, cm 3 ), conventional imaging underestimated it (mean: 2.39 cm 3 ; range, cm 3 ). The nodal GTV based on DW-MRI was identical to the reference GTV in 17 patients. There was an overestimation in 3 patients with a false-positive LN on DW-MRI (patients 1, 6, and 22). In 2 patients, the nodal GTV was underestimated. In patient 10, an ipsilateral level I (diameter: 6 mm) and a contralateral level II (diameter: 7 mm) adenopathy were not detected on DW-MRI. In patient 21, two contralateral level I (diameter: 6 mm in both) and one contralateral level II (diameter: 5 mm) adenopathies were not detected on DW-MRI, while an ipsilateral level IV lymph node was falsepositive on DW-MRI (Table 26). 127

128 B. Target definition: selection of the nodal target volume Table 26. Details on nodal staging and target volume delineation of all patients. No. Primary tumor pt pn DW CT/MRI Lymph Node Impact on elective CTV 1. supraglottis FP in level III IL RSP & level V IL 2. floor of mouth 4a tongue 4a 2b 2b 1 TP level II IL no impact on elective CTV 4. floor of mouth tongue 2 2b 2b 2b TP level I, IV & V IL no impact on elective CTV 6. piriform sinus 4a 2b 2c 2b FP level II CL TP level II & III IL RSP & level V CL no impact on elective CTV 7. retromolar trigone 4a 2c 2c 2b TP levels I & II CL RSP & level V CL 8. glottis 3 2c 2c 0 TP level II CL RSP, level Ib & level V CL 9. glottis floor of mouth 4a 2c 2c 2c FN level II CL FN level I IL TN level I IL level V CL no impact on elective CTV no impact on elective CTV 11. unknown primary X 2b 2b 2b TP level I IL no impact on elective CTV 12. glottis glottis b TN level II IL RSP, level Ib & level V IL 14. glottis 4a TP level III IL RSP & level V IL 15. tongue 2 2b 2b 2b TP level II IL no impact on elective CTV 16. tongue glottis 4a TP level II IL RSP, level Ib & level V IL 18. floor of mouth 4a TN level I IL RSP IL 19. tongue 2 2b 2b 0 TP level I IL RSP IL + IV CL 20. unknown primary X 2b 2b 1 TP level II IL no impact on elective CTV 21. floor of mouth 4a 2c 2c 2b TP level II CL FN level II CL FN level I CL FN level I CL FP level IV IL RSP & level V CL no impact on elective CTV no impact on elective CTV no impact on elective CTV no impact on elective CTV 22. supraglottis FP level III IL RSP & level V IL Abbreviations: pt = pathologic T-classification; pn = pathologic N-classification; DW = N- classification according to DW-MRI; CT/MRI = N-classification according to conventional imaging; FP = false-positive lymph node on DW-MRI; TP = true-positive lymph node on DW- MRI; FN = false-negative lymph node on DW-MRI; TN = true-negative lymph node on DW-MRI; IL = ipsilateral to the primary tumor; CL = contralateral to the primary tumor; RSP = retrostyloid space. The absolute differences between the GTV based on conventional imaging and the reference GTV (mean: 0.45 cm 3 ; median: 0.24 cm 3 ; range, cm 3 ) were significantly larger than those between the GTV based on DW-MRI and the reference GTV (mean: 0.11 cm 3 ; median: 0 cm 3 ; range, cm 3 ; p = ), as shown in Figure

129 B. Target definition: selection of the nodal target volume Nodal CTV delineation The mean nodal CTV based on pathology (reference CTV) was cm 3 (range, cm 3 ). DW-MRI overestimated the nodal CTV (mean: cm 3 ; range, cm 3 ), while conventional imaging underestimated it (mean: cm 3 ; range, cm 3 ). The nodal CTV based on DW-MRI was identical to the reference CTV in 18 patients. There was an overestimation in 3 patients with a false-positive LN on DW-MRI (patients 1, 6 and 22). In 1 patient, the nodal CTV was underestimated (patient 10). The absolute differences between the CTV based on conventional imaging and the reference CTV (mean: cm 3 ; median: 0.22 cm 3 ; range, cm 3 ) were significantly larger than those between the CTV based on DW-MRI and the reference CTV (mean: 3.49 cm 3 ; median: 0 cm 3 ; range, cm 3 ; p = ). Figure 17. Nodal GTV delineated on DW-MRI and CT/MRI, together with the reference GTV, in the 17 patients with discrepancies between the imaging modalities and/or pathology Nodal GTV (cm³) Reference GTV DW-MRI CT/MRI 2 0 pt 1 pt 3 pt 5 pt 6 pt 7 pt 8 pt 10 pt 11 pt 13 pt 14 pt 15 pt 17 pt 18 pt 19 pt 20 pt 21 pt

130 B. Target definition: selection of the nodal target volume 2.3. Discussion It is clear that accurate disease localization is critical to spare organs at risk and direct escalated doses to the GTV. Indeed, the full potential of highly conformal radiotherapy can only be realized by the exact definition of tumor in individual patients. Obviously, accurate nodal staging is essential, since underestimation of nodal involvement could lead to regional recurrences, while the elective irradiation of at-risk nodal levels complicates the sparing of salivary glands and swallowing structures [ ]. A prior investigation by our group demonstrated the added value of DW-MRI for nodal staging in HNC [72]. DW-MRI showed special promise in the detection of subcentimetric nodal metastases, where it still had a sensitivity of 76% and a specificity of 94% per LN. These results encouraged us to evaluate the impact of DW-MRI on RT planning. First, the superiority of DW-MRI over conventional imaging with CT or T1/T2-MRI was confirmed. Formal testing of agreement between the nodal staging results of both imaging modalities and the reference pathology results showed an important advantage towards DW- MRI at a significance level of p = by McNemar s testing. This is a robust finding given the limited sample size of the study. Although the sensitivity and specificity of conventional imaging in our study fall well within the range of what has previously been reported, they are somewhat lower than expected, especially regarding sensitivity [58]. This is probably due to the fact that a majority of the lymph nodes were subcentimetric. The size-related and morphological criteria for nodal staging that have to be applied on anatomical imaging have well known limitations for detecting subcentimetric nodal metastases [203]. Also, pathologic examination was more comprehensive than in clinical routine, or indeed most other studies, since we used prekeratine immunohistochemical staining to improve the detection of small metastatic deposits [214]. Secondly, DW-MRI allowed the radiation oncologists to very closely approach the true nodal GTV and CTV, as based on the standard of reference, i.e. pathology. Currently, prophylactic coverage of large portions of potentially normal neck tissue is necessary to avoid undesirable marginal failure, negating some of the potential organ-sparing advantages of IMRT [ ]. In this respect, the high NPV (97% per LN and 98% per level) of DW-MRI, in a setting of predominantly subcentimetric nodal disease, appears very promising. Obviously, these preliminary results require confirmation, preferably on a larger patient group with more clinically N0 necks. It should also be noted that DW-MRI did underestimate the nodal GTV in 2 patients, which could have resulted in a geographical miss. A possible explanation for these false-negative results could be an insufficient metastatic volume in the LN: dispersed small 130

131 B. Target definition: selection of the nodal target volume deposits in an otherwise normal nodal architecture are less likely to build up sufficient tissue boundaries to restrict water diffusion. In the end, no imaging technique to date allows the complete sparing of at-risk, clinically negative nodal levels from prophylactic radiation. But, as functional imaging becomes increasingly able to detect very small tumor deposits in lymph nodes, the question arises whether disease too small for combined imaging may not be effectively sterilized with decreased or so-called de-escalated doses [215]. Another promising application of IMRT is the targeting of high-risk gross disease for dose-escalation [79, 87]. The high PPV of DW-MRI suggests that it could reliably direct dose-escalation on the nodal GTV. Some limitations of this small institutional series need to be discussed. First of all, this study was inevitably performed on surgically treated patients, some of whom would not normally be treated with primary RT. More importantly, no oropharyngeal primary tumors could be included, since almost all of these patients are treated with primary RT in our institution. It is therefore open to debate as to what point these results can be generalized to the RT population as a whole. Secondly, pathology was only applicable to those nodal levels that were actually dissected, and half of our patients underwent unilateral dissection. This might have limited our power to accurately define specificity and NPV, which depends heavily on thorough sampling of all denominator values. Furthermore, no pathology data were available for the retropharyngeal lymph nodes, although this was unavoidable since this area is not sampled during neck dissection. Besides, none of the patients had evidence of retropharyngeal disease on any of the imaging modalities. Thirdly, although FDG-PET cannot yet be considered the standard imaging technique for nodal staging in HNC, a direct comparison between DW-MRI and FDG-PET would have been valuable [50]. 131

132

133 C. TARGET DEFINITION: DELINEATION OF THE PRIMARY TUMOR

134

135 CHAPTER 1: VALUE OF REPEATED FUNCTIONAL IMAGING WITH FDG-PET, FMISO-PET, DW-MRI, AND DCE-MRI Published as: Dirix P., Vandecaveye V., De Keyzer F., Stroobants S., Hermans R., Nuyts S. Dose painting in radiotherapy for head and neck squamous cell carcinoma: value of repeated functional imaging with 18F- FDG PET, 18F-fluoromisonidazole PET, diffusion-weighted MRI, and dynamic contrast-enhanced MRI. J Nucl Med 2009; 50(7):

136

137 C. Target definition: delineation of the primary tumor 1. Value of repeated functional imaging with FDG-PET, FMISO-PET, DW-MRI, and DCE-MRI The most extensively researched functional imaging modality for tumor delineation in HNC is undoubtedly FDG-PET. Some data indicate that pre-treatment FDG uptake is correlated with outcome, and could therefore signify a valuable target for dose-escalation. Another promising PET tracer is FMISO, providing quantitative measurements of hypoxia, one of the main factors affecting treatment resistance in HNC. Some reports also suggest that DW-MRI could indicate lesion aggressiveness and response to therapy. DCE-MRI assesses changes in signal intensity after the injection of a contrast agent. These changes are related to the perfusion of the tumor, known to influence treatment response. We prospectively included patients treated with chemoradiotherapy in a bio-imaging protocol: sequential PET/CT (FDG and FMISO) and MRI (T1-, T2-, DCE- and DW-sequences) scans were performed before, during, and after RT Material & methods Study design In this prospective study, 15 patients with locally advanced HNC scheduled for primary radiotherapy underwent repeated imaging with PET/CT before, during and, after treatment between January 2004 and May Since April 2005, when the MRI acquisition protocol became fully validated, the last 8 patients (patients 8 15) also received MRI scans. Patients and tumor characteristics are summarized in Table 27. There were 13 male and 2 female patients, with a median age of 57 years (range, years). Patients had a SCC of the oropharynx (n = 6), larynx (n = 5), hypopharynx (n = 3), or oral cavity (n = 1). Treatment was decided by a multidisciplinary team according to institutional guidelines, and consisted of concomitant chemoradiotherapy in all 15 patients. Radiotherapy was delivered according to the hybrid fractionation schedule. Patients also received cisplatinum (100 mg/m 2 ) intravenously, on the first day of week 1 and 4. The study protocol was approved by the local ethics committee; informed consent was obtained from all patients. 137

138 C. Target definition: delineation of the primary tumor Table 27. Patient and tumor characteristics. Pat. Primary Age Sex ct cn GTV CT Disease No. tumor site [ml] recurrence 1. Hypopharynx 59 M 4a no 2. Hypopharynx 58 F 1 2b no 3. Larynx 48 M no 4. Oropharynx 57 M 3 2b DM 5. Oropharynx 48 M 2 2c LR + DM 6. Oropharynx 60 M 2 2c 19.9 LR + DM 7. Oropharynx 48 M 4a no 8. Oropharynx 55 M 4a LR + DM 9. Oropharynx 58 M 2 2b LR + DM 10. Larynx 58 M 3 2b no 11. Larynx 60 M 4a 2c LR 12. Hypopharynx 46 M 4a no 13. Larynx 51 F 4a 2c no 14. Larynx 61 M 1 2c DM 15. Oral cavity 55 M 4a 2c no Abbreviations: M = male; F = female; ct = clinical T-classification; cn = clinical N-classification; GTV CT = anatomical gross tumor volume delineated on planning CT; DM = distant metastases; LR = loco-regional recurrence. The treatment and imaging protocol is shown in Figure 18. Patients received an FDG- PET/CT scan before the start of radiotherapy (median interval: 5 days; range, 0 18 days), and 8 weeks after the end of treatment (median interval: 58 days; range, days). All patients also received an FMISO-PET/CT scan before the start of treatment (median interval: 7 days; range, 1 14 days), although three scans could not be interpreted because of technical reasons (patients 3, 7, and 12). The FMISO-PET/CT scan was repeated during the fourth week after the start of radiotherapy (median interval: 28 days; range, days) to evaluate residual hypoxia. The last 8 patients (patients 8 15) received an MRI scan before the start of treatment (median interval: 10 days; range, 7 21 days), during the second (median interval: 15 days; range, days) and the fourth (median interval: 29 days; range, days) week after the start of radiotherapy, and 3 weeks after the end of treatment (median interval: 26 days; range, days). The DW-MRI and DCE-MRI scans of patient 14 could not be interpreted due to technical reasons. After treatment, patients were regularly followed at the multidisciplinary outpatient clinic: every 2 months the first 2 years after treatment, every 3 months the third year, every 4 months the fourth year, every 6 months the fifth year, and then every year. Response evaluation 138

139 C. Target definition: delineation of the primary tumor consisted of physical examination and CT scan at 2 to 3 months after treatment. Thereafter, a CT scan was done yearly or at the discretion of the treating oncologist. Figure 18. Treatment and imaging protocol Image acquisition A planning CT was performed for conformal treatment planning purposes in a clinical routine setting: serial slices, 3 mm thick, from the head down to the clavicles were obtained after IV injection of a contrast agent. All investigational PET/CT scans were performed on a Biograph integrated PET/CT scanner (Siemens, Erlangen, Germany), with an intrinsic resolution of 4 mm full-width at half maximum. Patients were scanned in supine position on a flat-top couch insert. The head, neck, and shoulders were immobilized using the thermoplastic mask (Posicast, Sinmed, Reeuwijk, The Netherlands) prepared during the RT simulation session and used throughout treatment. Fluorodeoxyglucose was prepared and a mean 358 MBq (range, MBq) was administered intravenously to patients who had been in a fasting state for 6 hours. At minutes after injection, FDG-PET/CT scanning was performed. The FMISO-PET/CT data were acquired at minutes after IV injection with 370 MBq of FMISO; no fasting period was required. The FMISO emission data were corrected for attenuation, scatter, and random counts, and then iteratively reconstructed using the standard FDG parameters used in clinical routine. Venous blood samples were obtained immediately after the FMISO-PET/CT session. Measured aliquots of each blood sample were counted in triplicate, and the net count rates were converted to activity concentrations (Bq/ml), and then decay corrected to the time of injection. The MRI examinations were performed on a 1.5-T scanner (Magnetom SONATA, Siemens, Erlangen, Germany) using a combination of a standard head coil and a two-channel phased-array neck coil. First, transverse T2- (TR/TE = 3080/106 ms) and T1-weighted (TR/TE = 139

140 C. Target definition: delineation of the primary tumor 775/8.3 ms) turbo spin-echo sequences were acquired, in an imaging block of 48 slices with 4 mm slice thickness and a field of view of 20x25 cm. An echo-planar DW sequence was then acquired using identical geometry, a TR/TE of 7400/84 ms, and b-values of 0, 50, 100, 500, 750, and 1000 sec/mm 2. DCE sequences were acquired using a three-dimensional T1-weighted gradient-echo sequence with fat saturation (volumetric interpolated breath-hold examination). The DCE sequence consisted of 48 slices, 4 mm slice thickness, a 22.5x30 cm field of view, a TR/TE of 4.3/1.6 ms, and each run lasted for 8.4 seconds. After 5 baseline DCE runs, a singledose IV bolus of Gadolinium-BOPTA (Multihance, BRACCO, Milan, Italy) was injected, and the sequence was continuously repeated for 20 more DCE runs. Coronal or sagittal T1-weighted TSE sequences were added depending on tumor localization Image analysis Planning CT images were transferred to a commercial planning workstation (Eclipse, Varian Inc, Palo Alto, CA) for treatment planning. The anatomical gross tumor volume (GTV CT ) was delineated by the treating radiation oncologist, and these volumes were retrospectively retrieved for comparison with the investigational imaging techniques. The GTV FDG was the result of automatic segmentation of the FDG-PET images, based on the source-to-background ratio [51, 57]. The standardized uptake value (SUV), an index of glucose metabolism, which equals the FDG uptake in each pixel normalized by the injected dose and body weight, was calculated in each tumor volume. A region of interest (ROI) was centered in the area of highest tumor FDG uptake (primary tumor and/or lymph nodes) and the maximal SUV value (SUV max ) was recorded. The hypoxic volume (HV) on FMISO-PET was defined as the pixels with a tissue to blood (T/B) FMISO ratio of 1.2 [216]. In each tumor volume, we also located the pixel with the maximum T/B ratio and recorded this variable (T/B max ) for statistical analysis. While the HV evaluates the volume of tumor that has crossed the threshold for significant hypoxia, the T/B max depicts the magnitude of hypoxia. In order to transfer contours of structures drawn on the PET/CT to the planning CT, the CT data from the PET/CT study were registered with the planning CT using the software available in the radiotherapy planning workstation. Both T2- and contrast-enhanced T1-weighted sequences were transferred to the radiotherapy planning system and corresponding tumor volumes (GTV T2 and GTV T1, respectively) were manually delineated by a radiation oncologist and a radiologist in consensus, while blinded to the other imaging data. 140

141 C. Target definition: delineation of the primary tumor The DW-MRI and DCE-MRI data were transferred to an independent Linux workstation with dedicated software (Biomap, Novartis, Basel, Switzerland). On the DW images with b- value of 0 sec/mm², the primary tumor and all identifiable lymph nodes (GTV DW ) were contoured. These were then copied and pasted on the images acquired with other b-values. From the signal intensity averages per ROI and per b-value, ADC values were calculated [67, 72]. For evaluation of the DCE images, pixel-wise signal-intensity curves (SIC) were calculated from the perfusion image series by using the following formula: SIC i = (I i I 0 )/I 0, for all time points i, where I i is the signal intensity at perfusion imaging at time point i and I 0 is the signal intensity at baseline perfusion imaging. Further evaluation was performed by calculating the slope of the contrast enhancement time curve at the time point of maximal contrast agent inflow, which was defined as the initial slope (IS): IS = max[d(sic)/dt]. Each local and/or regional recurrence volume (V r ) was delineated on the diagnostic CT demonstrating the relapse. That CT was then registered to the planning CT scan to characterize the failure as in-field, marginal, or out-of-field. This also allowed to determine the exact correlation of the V r with the initial GTV FDG and the hypoxic volume Statistical analysis Patient characteristics were recorded at the start of treatment. Follow-up data were retrospectively collected: information was sought on the date of first loco-regional recurrence, distant metastasis, and/or death. The close-out date for survival analysis was August 1, The statistical data were analyzed using the software package Statistica 8 (StatSoft Inc, Tulsa, OK, USA). Cumulative disease-free survival rates were calculated using the Kaplan-Meier product-limited (actuarial) method. To compare ADC values between recurring and controlled regions of interest (primary tumor and lymph nodes), an unpaired two-tailed student t-test was used. A p value of less than 0.05 was considered statistically significant Results Imaging results Before radiotherapy, the mean GTV CT was 33.6 ml (range, ml), the mean GTV T1 was 40.3 ml (range, ml), and the mean GTV T2 was 34.6 ml (range, ml). There was an excellent correlation between GTV CT and GTV T1 (R² = 0.98), and 141

142 C. Target definition: delineation of the primary tumor between GTV CT and GTV T2 (R 2 = 0.96) before the start of radiotherapy. During treatment, there was considerable shrinkage of the anatomical tumor volume. During the second week of radiotherapy, the mean GTV T1 was 28.7 ml (range, ml) and the mean GTV T2 was 24.9 ml (range, ml): a decrease of 28.8% and 28.1%, respectively. During the fourth week of radiotherapy, the mean GTV T1 was 19.2 ml (range, ml) and the mean GTV T2 was 16.4 ml (range, ml), so that 47.6% and 47.4% of the initial volume remained, respectively. There was an excellent correlation between GTV T1 and GTV T2 before RT (R 2 = 0.99), and during the second (R 2 = 0.98) as well as the fourth (R 2 = 0.98) week of treatment. On baseline FDG-PET, the mean GTV FDG was 18.7 ml (range, ml) and the mean SUV max was 9.81 (range, ). The GTV FDG was significantly smaller than the GTV CT on a paired two-tailed student t-test (p = ). On baseline FMISO-PET, the mean HV was 4.1 ml (range, ml) and the mean T/B max was 1.5 (range, ). There was very little residual hypoxia on the second FMISO-PET scans (Table 28): the mean HV was 0.3 ml (range, ml) and the mean T/B max was 1.2 (range, ). Table 28. Positron emission tomography data. FDG-1 FMISO-1 FMISO-2 FDG-2 Pat. GTV No. FDG HV HV SUV [ml] max T/B [ml] max T/B [ml] max Response CR CR CR CR CR RD CR RD CR CR CR CR CR CR CR Abbreviations: GTV FDG =gross tumor volume delineated on FDG-PET/CT; SUV max = maximal standardized uptake value on FDG-PET/CT; HV = hypoxic volume on FMISO-PET/CT; T/B max = maximum tissue to blood FMISO ratio on FMISO-PET/CT; CR = complete response; RD = residual disease. 142

143 C. Target definition: delineation of the primary tumor On baseline DW-MRI, the mean GTV DW was 14.8 ml (range, ml), significantly smaller than the GTV CT on a paired two-tailed student t-test (p = 0.004). During RT, there was considerable shrinkage of the tumor volume on DW-MRI. During the second week of radiotherapy, the mean GTV DW was 5.1 ml (range, ml) and during the fourth week, the mean GTV DW was 1.4 ml (range, ml). At all time-points before and during treatment, the GTV DW was significantly smaller than the GTV T1 (p < 0.01) and the GTV T2 (p < 0.01). At 3 weeks after treatment, the mean GTV T1 was 5.9 ml (range, ml) and the mean GTV T2 was 3.3 ml (range, ml). On DW-MRI, ADC calculations showed residual disease in 3 patients (mean GTV DW : 0.1 ml; range, ml), all of whom developed a locoregional recurrence during follow-up. At 8 weeks after the end of treatment, FDG-PET detected residual disease in 2 patients; both developed a loco-regional recurrence Patterns of failure Median follow-up of all patients was 30.7 months (range, months). Seven patients developed a disease recurrence after a median time of 9.4 months (range, months). At the time of analysis, 8 patients remained alive and disease-free, after a median follow-up time of 42.6 months (range, months). Actuarial disease-free survival was negatively correlated with T/B max on the baseline FMISO scan (p = 0.04, Figure 19), the size of the HV on the baseline FMISO scan (p = 0.04), and T/B max on the second FMISO scan during radiotherapy (p = 0.02, Figure 20). No significant correlation was observed between disease-free survival and the size of the initial GTV CT, SUV max on the baseline FDG scan, the size of the initial GTV FDG, or the size of the HV on the second FMISO scan. Five patients developed a local and/or regional recurrence during follow-up: 1 patient had a local relapse (patient 5), 2 patients had a regional relapse (patients 8 and 9), and 2 patients developed a loco-regional relapse (patients 6 and 11). There were 9 recurrent lesions (3 local and 6 regional) in total, with a mean V r of 12.7 ml (range, ml). All recurrences were in-field, i.e. within the GTV CT and thus the high-dose region. Similarly, all recurrences occurred within the initial GTV T1 and GTV T2, and within the pre-treatment GTV FDG on the baseline FDG- PET. Three recurrences mapped outside the pre-treatment HV on the baseline FMISO-PET. Lesions (i.e. primary tumor as well as lymph nodes) where a loco-regional recurrence developed during follow-up had significantly lower ADC values, both during the fourth week of radiotherapy ( vs mm 2 /sec, p = 0.01) as well as at 3 weeks after the end of 143

144 C. Target definition: delineation of the primary tumor treatment ( vs mm 2 /sec, p = 0.01), than lesions that remained controlled. Also, recurring lesions had a higher IS (26.2/sec vs. 17.5/sec, p = 0.03) on baseline DCE-MRI. Figure 19. Actuarial disease-free survival according to T/B max before radiotherapy (n = 12). 100% 90% Actuarial disease-free survival according to T/B max before RT (n = 12). censored patient disease recurrence 80% Disease-free survival (DFS) 70% 60% 50% 40% 30% 20% 10% High T/B max Low T/B max p = 0,04 High T/B max is defined as a value above the median of 1,45. 0% Time since treatment (months) 144

145 C. Target definition: delineation of the primary tumor Figure 20. Actuarial disease-free survival according to T/B max during radiotherapy (n = 15). 100% 90% Actuarial disease-free survival according to T/B max during RT (n = 15). censored patient disease recurrence 80% Disease-free survival (DFS) 70% 60% 50% 40% 30% 20% 10% High T/B max Low T/B max p = 0,02 High T/B max is defined as a value above the median of 1,17. 0% Time since treatment (months) 1.3. Discussion In this prospective feasibility study, a multimodality imaging protocol with FDG-PET/CT, FMISO-PET/CT, DW-MRI, and DCE-MRI was applied to patients before, during, and after concomitant chemoradiotherapy for locally advanced head and neck cancer. Regarding the anatomical imaging modalities, a very good correlation was observed between volumes based on CT, T2-weighted TSE-MRI, or contrast-enhanced T1-weighted TSE- MRI, corroborating results from earlier reports [51, 217]. Although TSE-MRI provides a clear benefit in the delineation of tumors located in the oral cavity and the nasopharynx or at the base of skull, its routine use in pharyngo-laryngeal tumors remains arguable [47 49]. Reassessment with MRI during treatment did nonetheless demonstrate clear shrinkage of the tumor volume to approximately half its initial size at the fourth week, emphasizing the need for truly adaptive 145

146 C. Target definition: delineation of the primary tumor radiotherapy. Our data confirmed the potential value of FDG-PET/CT for target volume delineation in patients with head and neck cancer [50 55]. Although the GTV FDG, based on a validated source-to-background algorithm, was significantly smaller than the anatomical tumor volume on CT, all loco-regional failures occurred within the initial FDG-avid volume. This would suggest that the GTV FDG within the GTV CT warrants escalated radiation doses [79]. However, we did not find any correlation between the amount (size of the GTV FDG ) or the level (SUV max ) of FDG uptake and disease-free survival, although this could be due to the limited number of patients. In fact, while the prognostic value of FDG-PET appears intuitively appealing, recent trials comparing SUV data to disease control in larger patient groups have reported conflicting results [52, 76 78]. In any case, it appears prudent not to use FDG-PET as the sole instrument for GTV delineation, since loco-regional failures can occur outside the GTV FDG [52, 79]. Pathology studies have convincingly demonstrated that no single imaging modality can depict all tumor extensions [51]. However, MRI, CT, and FDG-PET can, when used together, each add complimentary data to improve tumor delineation. Tumor hypoxia has been shown to be one of the major factors affecting treatment resistance in head and neck cancer [80]. Non-invasive FMISO-PET imaging evaluating the gross disease can provide serial quantitative measurements of hypoxia, with potential prognostic implications [81, 82]. In our study, both the pre-treatment amount (size of the HV) and level (T/B max ) of hypoxia was negatively correlated with disease-free survival. This is consistent with earlier studies evaluating the significance of pre-treatment FMISO uptake in HNC [85, 218, 219]. We are also the first to report that the level of hypoxia (T/B max ) during radiotherapy is significantly correlated with disease control. In 2 of 12 patients (patients 8 and 9, Table 28) with FMISO-PET data at the 2 time-points available, the T/B max values actually increased. Both patients developed a loco-regional recurrence as well as distant metastases (Table 27) during follow-up, apparently confirming the expected negative prognostic impact of this finding. Similarly, Eschmann et al. observed increased FMISO uptake after 30 Gy in 2 of 14 patients, both of whom developed a relapse [220]. A planning study by Popple et al. proposed that a modest boost dose ( % of the primary dose) on regions of permanent hypoxia could significantly increase tumor control probability. However, only tumors with a geometrically stable hypoxic volume will have an improved control rate following dose-escalation [221]. Our data showed that considerable reoxygenation takes place during radiotherapy, corresponding with observations from other groups [220, 222, 223]. In our opinion, more work is required to elucidate the spatial-temporal 146

147 C. Target definition: delineation of the primary tumor evolution in intra-tumor distribution before single time-point FMISO-PET images can be used as a basis for hypoxia-targeting dose-escalation protocols. Our preliminary results with DW-MRI showed that lesions where a loco-regional recurrence developed during follow-up had significantly lower ADC values on the scans during week 4 of radiotherapy and at 3 weeks after treatment. Because treatment-induced loss of tumor cells results in an increase in water mobility at the microscopic level, treatment response corresponds to an increase in ADC values. On the other hand, remaining tumor cells can be detected as residually decreased ADC levels [88, 224]. Hypothetically, DW-MRI could thus be used as a non-invasive tool to select patients who have a high risk of not achieving or maintaining loco-regional control and therefore would benefit the most from treatment intensification, e.g. boost dose-escalation on those subvolumes that harbor radioresistant clonogens (Figure 21). Obviously, other approaches such as the addition of targeted agents or more toxic drugs could also be considered. Figure 21. (a) Pre-treatment FDG-PET showed a metastatic adenopathy in level 2 (arrow), (b) which remained hypoxic on FMISO-PET scan during RT week 4. (c) This lymph node still had a restricted ADC on DW-MRI at 3 weeks after RT, and (d) remained suspect on FDG-PET at 8 weeks post-rt. (e) Residual disease was confirmed in this node (patient 8). While the DCE-MRI scans during and after radiotherapy did not provide any information regarding treatment response, we did observe a significant correlation between initial slope on the baseline scan and disease control. Although preliminary, these results suggest that DCE- MRI could be used as a predictor of tumor response to radiotherapy in HNC. Several studies have assessed the predictive role of pretreatment tumor enhancement, quantified as the initial slope or the peak enhancement, especially in gynecological cancer. Most reports have shown that tumors with better enhancement are associated with better tumor regression and local 147

148 C. Target definition: delineation of the primary tumor control [90]. This is attributed to better perfused tumors having less hypoxia-related resistance. However, other studies have reported that high enhancement was associated with poor clinical outcome, which could be attributed to the increased aggressiveness of tumors with a high angiogenic activity [90, 225]. These apparently contradictory findings are probably due to differences in patient population, tumor type, treatment, and clinical endpoints. It should be noted that our results appear at odds with an earlier study using perfusion-ct in a similar setting. Hermans et al. stratified 105 HNC patients according to the median perfusion value and found that the patients with the lower perfusion rate had a significantly higher local failure rate after RT [226]. In conclusion, these results confirm the added value of FDG-PET and FMISO-PET for radiotherapy treatment planning of head and neck cancer, and suggest the potential of DW-MRI and DCE-MRI for dose-painting and early response assessment. 148

149 CHAPTER 2: PREDICTIVE VALUE OF DW-MRI FOR DOSE- PAINTING Published as: Vandecaveye V., Dirix P., De Keyzer F., Op de Beeck K., Vander Poorten V., Roebben I., Nuyts S., Hermans R. Predictive value of diffusion-weighted magnetic resonance imaging during chemoradiotherapy for head and neck squamous cell carcinoma. Eur Radiol; In press.

150

151 C. Target definition: delineation of the primary tumor 2. Predictive value of DW-MRI for dose-painting Because of the extremely promising nature of the results with DW-MRI in the previous study, we validated these results in a much larger patient group Material & methods Study design Thirty-one patients, 29 men and 2 women with a mean age of 53 years (range, years), with histologically proven HNC were consecutively included in this prospective study. One patient was excluded due to detection of metastatic disease. The primary tumor location and clinical stage of the remaining 30 patients are summarized in Table 29. Treatment consisted of CRT in 27 patients and primary radiotherapy in 3 patients. RT was delivered according to the hybrid fractionation schedule in 29 patients. In the patient with a T1 glottic cancer, a total dose of 55 Gy was delivered. All patients completed the planned treatment. A time line shows the different methodological steps in the study (Figure 22). Routine pre-treatment examinations included diagnostic contrast-enhanced CT in all patients, FDG- PET/CT in 14 patients, and panendoscopy with biopsy in all patients. MRI with TSE and DW sequences was added prior to and at 2 and 4 weeks after the start of RT, using the same scanner and methodology as described above. All MRI studies were of sufficient technical quality to allow analysis. The study was approved by the local ethics committee and all patients gave informed consent prior to inclusion. Figure 22. Timeline illustrating the consecutive methodological steps. 151

152 C. Target definition: delineation of the primary tumor Table 29. Tumor localization and loco-regional stage of disease. Patient Primary tumor localization Staging Patient 1 Tonsil T2N3 Patient 2 Tonsil T2N2a Patient 3 Piriform sinus T4a2b Patient 4 Supraglottic T3N1 Patient 5 Piriform sinus T4aN2b Patient 6 Supraglottic T4aN2c Patient 7 Glottic T1N0 Patient 8 Base of tongue T3N1 Patient 9 Oropharynx T4aN2c Patient 10 Supraglottic T3N2b Patient 11 Oropharynx T4aN2c Patient 12 Supraglottic T3N2b Patient 13 Base of tongue T3N2b Patient 14 Oropharynx T4N1 Patient 15 Piriform sinus T2N1 Patient 16 Supraglottic T4aN2c Patient 17 Glottic T3N0 Patient 18 Supraglottic T4aN2c Patient 19 Oropharynx T2N1 Patient 20 Oropharynx T3N2c Patient 21 Piriform sinus T1N2b Patient 22 Oropharynx T4aN2c Patient 23 Tonsil T3N0 Patient 24 Piriform sinus T4aN1 Patient 25 Base of tongue T3N2c Patient 26 Base of tongue T4 Piriform sinus T3N1 Patient 27 Supraglottic T3N2b Patient 28 Piriform sinus T3N2c Patient 29 Tonsil T2N2b Patient 30 Tonsil T4aN1 152

153 C. Target definition: delineation of the primary tumor Image analysis The primary tumors and nodal metastases were identified on the baseline TSE-MRI and DW-MRI based on the clinical information and the pre-therapeutic routine imaging results. Primary tumors were topographically annotated by their slice position and anatomical position in the upper aero-digestive tract. Metastatic adenopathies were topographically annotated by their position in correlation to anatomical structures in the neck, i.e. the submandibular gland, the sternocleidomastoid muscle, and the jugular vein, and level position according to the AAO-HNS [213]. The primary tumors and metastatic adenopathies were identified on the 2 and 4 weeks follow-up TSE-MRI and DW-MRI examinations by visual correlation and correlation of slice positions with the baseline examination. The DW images were analyzed on a PACS-station (Agfa-Gevaert, Mortsel, Belgium) by a radiologist with 6 years of experience in head and neck DW-MRI blinded from the TSE-MRI. Regions of interest were placed on the primary lesions and adenopathies on the images acquired by a b-value of 0 sec/mm 2, and automatically copied to the other b-value images by the software. For solid primary tumors and adenopathies, the ROIs were placed over the entire lesion. In case of obvious solid and necrotic components on the DW images, ROIs were placed on the solid tumoral components according to their signal heterogeneity on the images acquired by a b-value of 1000 sec/mm 2 [67]. From the signal intensity averages per ROI and per b-value, ADC values were calculated [67, 72]. The ADC changes (ΔADC N ) in % for each lesion between the baseline and the 2 and 4 weeks time point were calculated using the formula: ΔADC N = [(ADC N ADC B )/ADC B ] x 100 where ADC B represents pre-treatment ADC value and ADC N represents the ADC on the N th time point during RT. The TSE-MRI images were analyzed on an off-line work station (Eclipse, Varian, Palo Alto, CA) blinded from the DW-MRI results. For each lesion, contours were drawn around the lesion border at each slice position. Subsequently, the volume of each lesion was calculated with the following equation: (Σsurface at each slice position) x (slice thickness + interslice gap). The volume changes (ΔV N ) in % for each lesion between the baseline and the 2 and 4 weeks time point were calculated using the formula ΔV N = [(V N V B )/V B ] x 100 where V B represents pre-treatment lesion volume and V N represents the volume on the N th time point during RT. 153

154 C. Target definition: delineation of the primary tumor Correlation to treatment outcome Imaging was primarily compared to 2-year patient follow-up. Complete response was defined as a persistent complete regression of the primary lesions at 2 years follow-up, clinically and on imaging. Tumor recurrence was defined as persisting or recurrent tumor within the first two years after the end of RT, consisting of either a volume increase of 65% on serial imaging or histopathological proof of HNC on biopsy and/or surgical specimen [227] Statistical analysis The software package Statistica 8 (StatSoft Inc, Tulsa, OK, USA) was used for the statistical analysis. A p-value < 0.05 was considered statistically significant. The ΔADC N and the ΔV N of primary lesions and adenopathies showing complete response versus those with tumor recurrence were compared with a Mann-Whitney U test. Receiver-operating characteristics (ROC) analysis with the area under the curve (AUC) was employed to investigate the discriminatory capability of the ΔV N and ΔADC N. For calculation of the sensitivity, specificity, accuracy, PPV, and NPV of the ΔV N and ΔADC N, the optimal threshold was determined by giving equal weighting to sensitivity and specificity on the ROC curve. Cumulative LRC was examined with the Kaplan-Meier product-limited (actuarial) method and the statistical significance of differences between curves was compared with the log-rank test. In univariate analysis the following clinical variables were entered: tumor localization (laryngeal vs. pharyngeal), age, T-classification (T1 2 vs. T3 4), N-classification (N1 2a vs. N2b 3), initial primary tumor volume, and initial nodal volume. All clinical variables with a p-value < 0.1 were included in a multivariate analysis comparing the predictive value of the ΔV N and ΔADC N. 2.2 Results Treatment outcome Complete LRC at 2 years after the end of treatment was achieved in 15 of 30 patients (50%). Six of 30 patients (20%) developed an isolated local recurrence (at a median of 129 days post-rt; range, 27 to 189 days) for which salvage surgery could be performed in 5 patients. Seven of 30 patients (23%) developed a regional recurrence without primary tumor recurrence (at a median of 224 days post-rt; range, 91 to 364 days), of whom 4 were eligible 154

155 C. Target definition: delineation of the primary tumor for neck dissection. Two of 30 patients (7%) developed a simultaneous loco-regional tumor recurrence (at 126 and 161days after RT, respectively): one patient received salvage surgery with bilateral neck dissection, the other was inoperable because of the extent of disease. Nine of 30 patients (30%) deceased during the 2 years follow-up: 7 patients because of a tumor recurrence. The death of two patients was unrelated to the HNC DW-MRI versus volumetric assessment: lesion-based analysis Table 30. Comparison of accuracy of % ADC versus % V at (a) 2 weeks and (b) 4 weeks. a) Primary lesion % ADC % V LN % ADC % V %Threshold %Threshold True positive 7 7 True positive 8 9 False positive 2 10 False positive 5 23 True negative True negative False negative 1 1 False negative 2 1 Sensitivity 88% 88% Sensitivity 80% 90% Specificity 91% 57% Specificity 89% 48% Accuracy 90% 65% Accuracy 87% 56% PPV 78% 41% PPV 62% 28% NPV 96% 93% NPV 95% 96% AUC 94% 78% AUC 83% 68% b) Primary lesion % ADC % V LN % ADC % V %Threshold %Threshold True positive 8 6 True positive 8 7 False positive 2 10 False positive 2 17 True negative True negative False negative 0 2 False negative 2 3 Sensitivity 100% 75% Sensitivity 80% 70% Specificity 91% 57% Specificity 96% 61% Accuracy 94% 61% Accuracy 93% 63% PPV 80% 38% PPV 80% 29% NPV 100% 87% NPV 96% 90% AUC 97% 64% AUC 90% 68% 155

156 C. Target definition: delineation of the primary tumor The ΔADC 2w was significantly lower for lesions with post-rt recurrence than with CR (primary tumor: -1.6% ± 9.7 vs. 40.5% ± 25.6; LN: 10.9% ± 26.0 vs. 52.1% ± 36.5, both p < 0.001). The ΔV 2w was significantly lower for primary tumors with post-rt recurrence than with CR (-28.1% ± 34.5 vs. 31.3% ± 32.1, p = 0.03) but no significant difference was found for LN (- 15.4% ± 58.9 vs. 19.9% ± 39.7, p = 0.1). The ΔADC 4w was significantly lower for lesions with post-rt recurrence than with CR (primary tumor: -1.0% ± 24.6 vs. 65.1% ± 36.0; LN: 15.8% ± 25.6 vs. 67.6% ± 41.4, p < ). The ΔV 4w showed no significant difference for either lesions (primary tumor: 46.4% ± 27.0 vs. 56.3% ± 30.8; LN: 15.1% ± 67.4 vs. 49.3% ± 41.3, both p > 0.1). Based on the calculated optimal threshold, the ΔADC N showed higher AUC and accuracy than the ΔV N during RT (Table 30) Correlation of DW-MRI and volumetric assessment with LRC Univariate analysis of clinical pre-treatment variables showed no significant correlation between volume of the primary tumor, T-classification, age, or tumor localization and LRC (p > 0.2). A substantial correlation of 2-year LCR was found with nodal volume ( 18.6 cc vs. < 18.6 cc) and N-classification (p = 0.06). In multivariate analysis, N-classification and nodal volume no longer showed any correlation with LRC (p > 0.5), while ΔADC 2w (p = 0.003) and ΔADC 4w (p < ) correlated significantly to 2-year LRC (Figure 23). The ΔV 2w and ΔV 4w showed no significant correlation with LRC (p > 0.05). 156

157 C. Target definition: delineation of the primary tumor Figure 23. Multivariate analysis shows significant correlation of ΔADC at (a) two and (b) four weeks with loco-regional control. High ΔADC indicates a ΔADC above the threshold as stated in Table 30. a) 100% 90% 80% Loco-regional Control (LRC) 70% 60% 50% 40% 30% 20% 10% High Δ ADC Low Δ ADC p = % Follow-up (months) 157

158 C. Target definition: delineation of the primary tumor b) 100% 90% 80% Loco-regional Control (LRC) 70% 60% 50% 40% 30% 20% 10% High Δ ADC Low Δ ADC p < % Follow-up (months) 2.3. Discussion Highly conformal radiotherapy techniques, such as IMRT, provide the ability for individually tailored treatment intensification in case of poor response, provided that nonresponding lesions are early recognized. In this study group, DW-MRI significantly correlated to 2-year LRC at 2 and 4 weeks during RT and showed a substantially higher accuracy than volumetric assessment for differentiation between responding and non-responding lesions. The ability to probe the tissue microstructure and the sensitivity of DW-MRI for low levels of intratumoral cell loss may explain the feasibility of DW-MRI for early response assessment [88]. Lesions with CR after RT showed a significantly higher ΔADC than lesions that recurred. Historadiological correlation during treatment follow-up in prior studies has related the ADC increase to the disorganized microstructure in necrosis and apoptosis in response to cytotoxic and radiation treatment [ ]. Conversely, absent ADC increase during RT corresponded 158

159 C. Target definition: delineation of the primary tumor to lesions with a high likelihood of recurrence, probably correlating to diffusion restriction in the dense microstructure of persistent HNC [67]. These findings agree with a recent study by Kim et al., which also found a significant ADC increase in responding compared to non-responding metastatic adenopathies [231]. In the current study, near identical accuracy was reached for the evaluation of nodal metastases at 2 weeks, compared to 1 week in the study by Kim et al. [231]. This suggests the robustness of DWI as a potential biomarker for early prediction of treatment response. Additionally to the findings by Kim et al., the ΔADC 2w and ΔADC 4w correlated significantly to 2-year LRC in this study. Recently, DW-MRI was shown to predict final imaging and clinical response at 2 weeks during CRT for cervical cancer [232]. In patients with primary cerebral neoplasms, the combination of DW-MRI before and at 3 weeks after the initiation of treatment has been demonstrated to accurately predict later radiographic response, time to progression, and overall survival [ ]. The underlying causative link between ADC changes during RT and ultimate LRC remains unclear due to lack of historadiological correlation. The relation between the observed ADC changes and the underlying pathophysiological processes, such as apoptosis, hypoxia, proliferation index, or tumor grade, are to be further investigated [236]. Moreover, it is clear that considerable further research and development is necessary before DW-MRI could be routinely used in radiotherapy planning. Interpretation of DW-MRI in the head and neck region is not straightforward, making the use of (semi-)quantitative measurements for target volume delineation absolutely necessary. However, this is also one of the strengths of the technique, since a straightforward cut-off value would allow simple and reliable differentiation, potentially eliminating both intra- and inter-observer variability [46]. Also, DW images have a non-negligible non-affine distortion, which makes image fusion with planning CT images difficult. Because of the low signal-to-noise ratio and the high level of deformation in DW-MRI, automatic non-rigid co-registration algorithms based on mutual information do not have enough information for accurate co-registration [237]. Therefore, semi-automatic alterations of this algorithm have to be developed to provide the additional information required. However, despite current limitations of standardization and image interpretation, DW-MRI with the ΔADC at 2 and 4 weeks during RT for HNC allowed early response prediction. Awaiting larger studies that prospectively evaluate the thresholds obtained in this study, DW-MRI may be useful as a response biomarker during RT of HNC, and could potentially guide radiation doseescalation on persistent tumor deposits. 159

160

161 D. REDUCING TOXICITY

162

163 CHAPTER 1: RADIATION-INDUCED XEROSTOMIA IN HNC PATIENTS A LITERATURE REVIEW Published as: Dirix P., Nuyts S., Van den Bogaert W. Radiation-induced xerostomia in patients with head and neck cancer: a literature review. Cancer 2006; 107(11):

164

165 D. Reducing toxicity 1. Radiation-induced xerostomia in HNC patients a literature review 1.1. Pathophysiology Radiation-induced xerostomia starts early during treatment: in the first week a 50% to 60% decrease in salivary flow occurs, and after 7 weeks of conventional radiotherapy, salivary flow diminishes to approximately 20% [238]. In 1911, the French radiobiologist Jean Bergonie described this apparent radiosensitivity of the salivary glands as an enigma [239]. This is because the functional (i.e. excretory, acinar) cells of the salivary glands are highly differentiated and have a slow turnover, but behave like acute responding tissues to radiation. Classically of course, tissues with a slow mitotic rate should not be particularly radiosensitive [240]. The first explicatory concept that was suggested was the so-called granulation hypothesis: the membranes of secreting granules in acinar cells become damaged by radiationinduced lipid peroxidation and consequently proteolytic enzymes begin to leak from these granules, causing immediate lysis of the cells [241]. However, a clinical study using salivary gland scintigraphy (SGS) early after RT showed that trapping of Technetium [Tc] pertechnetate was not affected, although saliva excretion was severely reduced [242]. This finding seemed to indicate that the gland volume remains intact, while the excretory function is impaired, questioning massive cell loss as the cause of early loss of function. Apparently, irradiated saliva-producing cells do not disappear but lose their function during the first days after irradiation. Konings et al. proposed two separate mechanisms to explain radiation-induced salivary gland dysfunction. First, there is a defect in cellular functioning because of selective membrane damage, confounding the receptor-mediated signaling pathways of water excretion. No immediate cell death or lysis takes place. Late damage is explained by classical cell killing of progenitor cells and stem cells, thus inhibiting proper cell renewal and by damage to the cellular environment, causing a shortage of properly functioning secretory cells [243]. Salivary function continues to decline for up to several months after RT [244]. Thereafter, some recovery is possible until 12 to 18 months after RT, depending on the dose received by the salivary glands and the volume of the gland tissue included in the irradiation fields, but generally, xerostomia develops into an irreversible life-long problem [245, 246]. Recently, Braam and colleagues reported that salivary output could still recover many years after RT, with an approximately 32% increase in salivary flow from 1 year to 5 years after treatment [247]. This is by no means generally accepted and most longitudinal studies found 165

166 D. Reducing toxicity very little recovery over time in patients that did not receive some sort of salivary gland-sparing radiation technique [248]. Although altered fractionation schedules are increasingly used, it is not yet clear what impact this will have on the incidence of xerostomia. There is evidence to suggest that when multiple daily treatments with small (less than 1.8 to 2 Gy) fractions are given, this does not increase the incidence of xerostomia, although more aggressive regimens can exacerbate late toxicity, including xerostomia [249] Measurement Measuring and reporting the severity of xerostomia is not straightforward. However, standardized measurements are necessary to compare the efficacy of preventive or curative interventions. Generally, both objective methods such as salivary flow measurements or salivary gland scintigraphy, and subjective measurements with observer-based toxicity grading or patient self-reported scoring are used. It is not clear which method most accurately reflects the impact of xerostomia on patient well-being and health. Measurements of salivary flow rate are currently the most commonly applied objective measures of salivary gland function. In healthy persons, the salivary glands produce between 1 to 1.5 L of saliva per day. The major glands (parotid, submandibular, and sublingual) produce up to 90% of saliva. Typically, about 60 65% of the total salivary volume is produced by the parotid glands, 20 30% by the submandibular glands, and 2 5% by the sublingual glands. The minor glands are distributed throughout the oral cavity and pharynx, their number is variable. Usually, saliva is collected selectively from each major gland: the output from the parotid gland is measured by placing a suction cup (Lashley cup) to the orifice of Stensen s duct; for the submandibular and sublingual glands, gentle suction with a micropipette at the orifices of Wharton s duct is necessary. Collection can be either unstimulated or stimulated (e.g. with 2% citric acid or by chewing). Saliva production by all the glands collectively can be measured by spitting, drainage, or weighing cotton rolls inserted in the mouth [250]. However, results are not always comparable between studies because of differences in the nature and length of application of stimulants, differences in the method and duration of collection, and neglect of other factors that may affect salivary output [251]. There is also a weak correlation between salivary flow measurements and xerostomia symptom scores, probably due to the variation in normal salivary flow rates and discrepancies between the salivary output and the hydration status of the mucosa [252]. This seriously impedes the definition of a threshold of saliva output 166

167 D. Reducing toxicity which characterizes xerostomia. Arbitrarily, a reduction of salivary flow to 25% of the preradiotherapy flow is considered a relevant threshold [253]. Several imaging techniques, such as SGS, can also be used for evaluating the effect of radiation on salivary gland function [254]. Scintigraphy is especially useful when combined with single photon emission computed tomography (SPECT), because of the additional spatial information [255]. The ability of MRI sialography to depict radiation-induced changes to the salivary glands and ducts has recently been demonstrated [256]. Another valuable technique is DW-MRI, which can be used to non-invasively demonstrate functional changes in the salivary glands [257, 258]. Observer-based toxicity scoring is generally based on the RTOG/EORTC grading scale [259]. However, since xerostomia is defined as a symptom, it is equally important to estimate the subjective appreciation of oral dryness by the patient. Several xerostomia questionnaires have been developed to permit patient self-reporting, most notably by the University of Michigan. It was suggested that this questionnaire is more accurate in estimating the severity of xerostomia, compared to the RTOG/EORTC grading system [260]. More recently, the LENT- SOMA scoring system offers a detailed evaluation of xerostomia [261]. The system is a combination of observer-based grading of patient-reported degree of mouth dryness, oral moisture, necessary frequency of saliva substitutes, and objective measurement of the salivary flow rate. It correlates well with patient-reported xerostomia and could prove to be a valuable tool for the correct assessment of xerostomia [102]. The National Cancer Institute has lately developed the common toxicity criteria to replace the RTOG grading system [262]. Its use in the estimation of radiotherapy-induced xerostomia has not yet been reported Impact on quality of life Quality of life in patients treated for HNC is strongly influenced by xerostomia and all its ramifications. A survey of 65 patients who survived for longer than 6 months after RT found that 91.8% complained of a dry mouth, 43% had difficulty chewing, 63.1% had dysphagia, 75.4% had taste loss, 50.8% had altered speech, 48.5% had difficulty with dentures, and 38.5% reported increased tooth decay. Pain was common (58.4%) and interfered with daily activities in 30.8%. More than half of the patients (58.3%) had mood complaints and 60% had interference by their physical condition on their social activities [263]. Most patients with xerostomia experience difficulty eating dry or hard food, which forces them to adjust their diet, albeit sometimes unconsciously [264]. Mastication and oral 167

168 D. Reducing toxicity manipulation of food becomes uncomfortable or even painful, most patients need frequent sips of water while they eat and food gets stuck in their mouth or throat [265]. Not only chewing, but also swallowing of food becomes a problem. A generalized decrease in the mobility of pharyngeal structures is demonstrated after RT, with prolonged pharyngeal transit and a delay of laryngeal closure [266, 257]. A study compared swallowing function between patients (1 year after RT) and healthy volunteers. Patients showed a significant degree of abnormality in the bolus transport. Elevation of the hyoid bone began too late and it was held in an elevated position for too long. As a result, the upper esophageal sphincter opened too early relative to the arrival of the bolus [268]. Other changes include reduced contact of the base of the tongue to the pharyngeal wall, restricted laryngeal motion, impaired closure of the laryngeal vestibule and true vocal folds, resulting in aspiration [269, 270]. As oral and pharyngeal mucosa is exposed to radiation, taste receptors become damaged and taste discrimination becomes increasingly compromised [271]. Decreased saliva output may affect taste, often contributing to the slow return of taste perception after radiotherapy. This is most pronounced after 2 months, when bitter and salt qualities are generally the most impaired. Although gradual recovery of taste is seen during the first year, partial loss still persists 1 to 2 years after treatment [272]. Difficulty with speech is another common complaint of patients with radiation-induced xerostomia [273, 274]. Even after 5 years, patients still report self-perceived speech problems, difficulty being understood and diminished intelligibility [275]. Risk of dental caries increases secondary to a number of factors including shifts to a cariogenic flora (e.g. increased colonization with Streptococcus mutans and Lactobacillus), reduction of the salivary ph, altered immunoglobulin composition, and loss of mineralizing components [ ]. The reduction in the salivary flow may also contribute to the risk of osteonecrosis of the mandible and to esophageal injury by decreasing acid clearance by salivary bicarbonate [279, 280]. Dryness of the oral mucosa creates a predisposition to mucosal fissures and ulcerations [281]. These secondary effects contribute to the so-called xerostomia syndrome. In the end this combination of factors can result in decreased nutritional intake and weight loss, posing a major health problem for some patients [282, 283]. 168

169 D. Reducing toxicity 1.4. Prevention Cytoprotectants Several agents have been developed to protect normal tissue against the toxic effect of radiotherapy and/or chemotherapy. Amifostine (WR-2721, Ethyol ), a spin-off of the nuclear warfare program, has long been recognized as a potential radioprotector. When amifostine enters the bloodstream, it is rapidly hydrolyzed by alkaline phosphatases of the endothelium and converted to its active form, WR This active form enters cells and nuclei, where it acts as a potent scavenger against free radicals, thus preventing radiation damage to DNA. It has been suggested that both the lack of alkaline phosphatases in the endothelium and the acidic conditions in the micro-environment prevent the activation of amifostine in the tumor, assuring a selective protection of normal tissues [284]. However, various pre-clinical data are conflicting, and the issue continues to divide the scientific community [ ]. On the other hand, there is clinical evidence to support the use of amifostine. The largest study to date, a phase III trial by Brizel et al., randomized 303 patients, treated for HNC with conventional RT (both post-operative and as primary treatment), to receive amifostine daily before each fraction (200 mg/m²; IV). Amifostine significantly reduced the incidence of grade 2 acute xerostomia from 78% to 51 % and grade 2 chronic xerostomia from 57% to 34%, without altering disease control or survival [289]. The use of amifostine was consequently approved by the Food and Drug Administration. Recently, a follow-up to that study was published, suggesting that the administration of amifostine during RT reduces the severity of xerostomia until 2 years after treatment. No difference after 2 years was found in LRC or survival [290]. No trial to date, however, is sufficiently powered to detect small differences in survival. A recent meta-analysis tried to overcome this problem. They found that amifostine significantly reduced the risk of developing acute grade 2 xerostomia by 76% and late grade 2 xerostomia by 67% in patients receiving RT [291]. There was no evidence that this would in any way weaken the effectiveness of treatment. The use of amifostine during concomitant chemoradiotherapy is controversial [292]. No randomized-controlled trial to date has shown that this might be an indication for the use of amifostine, so it should probably not be applied outside a clinical study [293]. Another important issue is the toxicity of amifostine. Nausea and vomiting are common side-effects, but they are generally mild and can be effectively controlled with standard antiemetic medication. There is a risk of transient hypotension when intravenously administered, 169

170 D. Reducing toxicity although not with subcutaneous administration [294] Salivary gland-sparing radiotherapy The extent of the damage caused by RT depends both on the volume of tissue that is irradiated and on the dose of radiation that is delivered. Therefore, a sound approach to prevent radiation-induced xerostomia is to focus the radiation beams better to the target volume and to avoid unnecessary irradiation of salivary gland tissue. It has recently become possible to spare a portion of the parotid gland by the implementation of highly conformal radiotherapy techniques in clinical practice. A high dose is administered to a small part of the parotid, positioned close to the tumor, while the rest of the gland receives a low dose or no dose at all. Several centers have started to use parotid-sparing protocols in order to prevent permanent xerostomia. At the Leuven department for example, a comparatively straightforward 3D-conformal technique (without intensity modulation) has been implemented in clinical practice since September 1999, aimed at sparing the parotid gland contralateral to the tumor. Patients thus treated have gained partial sparing of the salivary output from the parotid gland [102]. Data regarding the doses and irradiated volumes that permit preservation of the salivary flow following RT are slowly emerging. Usually, three-dimensional dose distributions in parotid glands are compared with residual saliva production. Correlation of dose with salivary flow measurements allows production of dose/volume-response relationships for parotid gland function. It became clear that there is an exponential relationship between saliva flow reduction and mean parotid dose for each gland, suggesting that it is essential to respect a certain threshold for mean parotid dose in order to preserve gland function [295]. A mean gland dose of 26 Gy was initially proposed as a planning goal for substantial sparing of the gland function by Eisbruch and colleagues from the University of Michigan [296]. Researchers from Washington University reported very similar results: in their analysis, a mean parotid dose of more than 25.8 Gy was likely to reduce salivary flow to 25% of its pre-treatment value and the incidence of xerostomia was significantly decreased when the mean parotid dose of at least one gland was kept 25.8 Gy [297]. Investigators using SGS to evaluate parotid function after RT found similar results: a mean parotid dose of less than 26 to 30 Gy allows preservation of the salivary gland function [298]. Gradually, a consensus was formed that a significant reduction of xerostomia can be achieved by holding a mean parotid dose of less than 26 to 30 Gy as a planning criterion [299]. However, the use of a mean dose as a threshold for the control of normal tissue 170

171 D. Reducing toxicity tolerance is only helpful when an organ consists of independent functional units that are organized in parallel. In such an organ, irradiation of a small part results in less loss of function than in an organ that has a serial anatomic organization. In the latter case, damage to one substructure disables the entire organ (e.g. the spinal cord). For the healthy parotid gland, it is generally assumed that a homogenous distribution of saliva production takes place over the entire volume. Some interesting data with rats, however, show that late radiation damage after partial irradiation of the parotid glands might be region-dependent [300, 301]. This means that partial irradiation leads to varying late radiation damage, depending on the region that has been exposed: irradiation of the cranial half resulted in more late damage than irradiation of the caudal half. It was suggested that spatial information should be included in the comparison of different plans and that the mean dose concept has limited use for the prediction of late radiation damage. However thought provoking these results are, the delineation of anatomical regions within the parotid is highly theoretical and without firm anatomical basis. Similarly, it remains to be seen whether regional differences in gland radiosensitivity will prove to be equally important in humans as they appear to be in rats. At the Leuven University Hospitals, a combination of SGS with SPECT was used to determine the salivary function of different regions within the gland after parotid-sparing RT [302]. Each of the 8 to 12 transverse slices within the parotid gland were considered as a functional subunit and a salivary excretion fraction (SEF) was measured for each. Before RT all slices contributed equally; seven months after an average dose of 22.5 Gy (D50), the subunit had lost 50% of its SEF. There was a high inter-patient variability in D50 and low doses (10 15 Gy) could induce serious loss of function as well, casting doubt on the utility of a general, standard mean gland dose threshold. Probably, the mean parotid dose should be kept as low as possible Salivary gland transfer Another, although less wide-spread, approach is salivary gland transfer. Colleagues Jha and Seikaly were the first to propose surgical transfer of one submandibular gland to the submental space, outside the proposed radiation field [303, 304]. This is only practicable in patients who will are planned to receive post-operative RT, as the transfer is done as part of the surgical intervention. Obviously, it is not always straightforward to predict which patients will need post-operative RT and some patients may refuse further treatment. Then again, in some 171

172 D. Reducing toxicity patients the submental space cannot be shielded due to proximity of the disease. These are important limitations, and in their largest study to date, 28.3% (17/60) of patients underwent salivary gland transfer without subsequent RT or without sparing of the relocated gland [305]. However, all the glands survived transfer and functioned well; the surgical technique had no complications and added an average of 45 minutes to the surgical protocol. The results in preventing xerostomia are convincing: 81% of patients had none or minimal xerostomia and 19% had moderate to severe xerostomia. Long-term follow-up data have recently been published, 83% of patients reported normal amounts of saliva 2 years after RT [306]. Other centers attained similar results, but it should not be considered as a standard procedure [307, 308]. Institutional experience with this technique is essential, and it is probably inevitable that a substantial percentage of patients will be operated on without real benefit Treatment Current therapies for the management of radiation-induced xerostomia include stringent oral hygiene with fluoride agents and antimicrobials to prevent dental caries and oral infection, saliva substitutes to relieve symptoms, and sialogogic agents to stimulate saliva production from remaining intact gland tissue. A variety of artificial saliva substitutes have been developed to supplement the reduced production of saliva, although very few have been appropriately evaluated. Obviously, since saliva is a complex substance with many functions, it is difficult to replace, so saliva substitutes are rarely effective and some patients find regular sips of water equally useful [309, 310]. Moreover, they do not replace the antibacterial and immunological protection of saliva and therefore do not exclude the need for regular dental care and appropriate oral hygiene. Antimicrobial mouthwashes such as chlorhexidine and hexitidine play a central role in reducing the bacterial load and inhibiting cariogenesis [311, 312]. Studies have shown that when there is still some residual salivary function, saliva stimulants produce greater relief than saliva substitutes [313]. Pilocarpine is currently the sole sialogogic agent approved by the FDA for radiation-induced xerostomia. Pilocarpine is a naturally occurring alkaloid that functions primarily as a muscarinic-cholinergic agonist with mild β-adrenergic activity. As a parasympathomimetic agent it causes stimulation of cholinergic receptors on the surface of exocrine glands, resulting in diaphoresis, salivation, lacrimation, and pancreatic secretion. The first trials with pilocarpine in radiation-induced xerostomia were performed in the early 90 s and showed significant improvement of oral dryness in approximately 172

173 D. Reducing toxicity half of patients [314, 315]. For optimal results, it is necessary to treat the patient during 8 to 12 weeks with doses greater than 2.5 mg three times per day [316]. Pilocarpine can also be safely used as maintenance therapy during longer periods of time [317]. Pilocarpine is obviously contra-indicated in patients with asthma, acute iritis, or glaucoma and should be used with extreme caution in patients with chronic obstructive pulmonary disease and cardiovascular disease. The side-effects of pilocarpine are due to a generalized parasympathetic stimulation, causing mild to moderate sweating in almost half of patients and, less frequent, urinary frequency, lacrimation, and rhinitis. Some work has been done with topical, i.e. oral application of pilocarpine and this seems to have similar results as systemic delivery methods, but with improved patient tolerance [318]. It has been suggested that pilocarpine given during RT could in some way salvage salivary gland function and prevent xerostomia, but results were disappointing and this indication warrants further investigation [319, 320]. Cevimeline is a newer muscarinic agonist that has been found safe and effective in treating xerostomia in patients with Sjögren s disease, and could also have its merits for radiation-induced xerostomia [321]. Other drugs, including bethanechol, metacholine, and carbachol have also been investigated, although results have generally been poor [322, 323]. Some promising pre-clinical results have been obtained by gene transfer, although clinical studies have yet to be initiated [324]. Recently, it was suggested that the expanding field of stem cell research could also yield results in the treatment of radiation-induced xerostomia. Apparently, mobilized bone marrow cells could home to salivary glands and induce repair by secreting stimulatory factors, causing improved salivary gland function [325] Conclusion Xerostomia is an almost ubiquitous, long-term complication of radiotherapy for head and neck cancer. Recently, significant progress has been made in the prevention of xerostomia through salivary gland-sparing radiation techniques such as 3D-RT or IMRT, and, more controversially, by the use of concomitant pilocarpine or surgical transfer of a submandibular gland to the submental space. However, it still impossible to successfully prevent radiationinduced xerostomia in all patients, and a large percentage of HNC survivors continue to suffer from it. Therefore, further research, particularly regarding treatment, is urgently warranted. Moreover, discrepancies in threshold for parotid-sparing RT suggest the need for objective, noninvasive ways to evaluate salivary gland function. 173

174

175 CHAPTER 2: DW-MRI TO EVALUATE MAJOR SALIVARY GLAND FUNCTION BEFORE AND AFTER RADIOTHERAPY Published as: Dirix P., De Keyzer F., Vandecaveye V., Stroobants S., Hermans R., Nuyts S. Diffusion-weighted magnetic resonance imaging to evaluate major salivary gland function before and after radiotherapy. Int J Radiat Oncol Biol Phys 2008; 71(5):

176

177 D. Reducing toxicity 2. DW-MRI to evaluate major salivary gland function before and after radiotherapy The prevention of xerostomia is paramount. However, our technical ability to spare the salivary glands exceeds our knowledge of the underlying mechanism of radiation-induced xerostomia. We evaluated DW-MRI as a non-invasive tool to measure salivary gland function Material & methods Study design In this prospective study, 8 HNC patients underwent a DW-MRI examination, as well as a SGS, before and at a mean of 9 months after RT. Clinical xerostomia was also assessed, according to the RTOG/EORTC late morbidity scoring system. Patients and tumor characteristics are summarized in Table 31. There were 7 male and 1 female patients, with a median age of 57 years. All patients had a SCC of the oropharynx, oral cavity, or lymph node metastases of an unknown primary. Table 31. Patient and tumor characteristics. PT No. Age Gender Primary site Stage Surgery Dose RT 1 54 M tonsil ct2n0 no 72 Gy 3D-RT 2 65 F tonsil pt3n1 yes 66 Gy IMRT 3 52 M tonsil ct2n2 no 72 Gy 3D-RT 4 58 M tonsil ct3n2 no 72 Gy 3D-RT 5 48 M tonsil pt2n0 yes 66 Gy 3D-RT 6 58 M floor of mouth pt2n1 yes 66 Gy 3D-RT 7 76 M unknown primary ptxn2 yes 60 Gy 3D-RT 8 56 M floor of mouth pt2n0 yes 50 Gy IMRT Treatment was decided by a multidisciplinary team according to institutional guidelines. Primary radiotherapy consisted of the hybrid fractionation schedule to a total dose of 72 Gy. Post-operative RT consisted of a dose of 50 Gy in 25 daily fractions of 2 Gy, followed by an additional boost of 16 Gy in case of extranodal spread of lymph node metastasis and/or positive resection margins. Surgery involved resection of the primary tumor with an elective neck dissection in the N0 neck or a radical neck dissection in the N+ neck. Consequently, the ipsilateral submandibular glands were removed in 2 patients (patients 2 and 6) and could not be assessed with DW-MRI or SGS. One patient (patient 7) was treated for LN metastases of an 177

178 D. Reducing toxicity unknown primary by radical neck dissection with superficial parotidectomy, followed by postoperative RT to 60 Gy. In this patient, both ipsilateral parotid and submandibular glands could not be evaluated. No concomitant or (neo-)adjuvant Cx was prescribed in any of the patients. All patients were treated according to a parotid-sparing protocol with the intent of sparing the contralateral parotid gland (planning goal: mean dose < 26 Gy), using 3D-RT or IMRT [102]. The electively treated node levels are shown in Table 32. Node levels were delineated according to the consensus guidelines [106]. Because none of the patients had contralateral lymph node metastasis or tumors crossing the midline, the contralateral junctional or high jugular (high level II) nodal area was not included in the elective CTV [104]. So, the cranial limit of level II was defined at the base of skull in the ipsilateral neck and at the crossing of the internal jugular vein with the posterior belly of the digastric muscle in the contralateral neck. This allowed relatively straightforward sparing of the contralateral parotid gland [102]. No dose constraints were applied for the ipsilateral parotid gland or any of the submandibular glands. Dose-volume histograms of both parotid and submandibular glands were retrieved from the planning system (Eclipse, Varian Medical Systems, Palo Alto, CA, USA). Table 32. DVH analysis of parotid glands, electively treated lymph node levels, and clinical evaluation of xerostomia. PT No. V PI [cm 3 ] V PC [cm 3 ] Mean PI [Gy] Mean PC [Gy] Max PI [Gy] Max PC [Gy] Levels Ipsilateral Levels Contralateral 1 10,8 15, II-III-IV II-III-IV ,2 14, RP-Ib-II-III-IV-V RP-II-III-IV RP-Ib-II-III-IV-V RP-II-III-IV RP-Ib-II-III-IV-V RP-II-III-IV RP-Ib-II-III-IV RP-II-III-IV RP-I-II-III-IV-V RP-I-II-III-IV RP-Ib-II-III-IV-V RP-II-III-IV RP-I-II-III-IV RP-I-II-III-IV 1 Mean RTOG/ EORTC Abbreviations: V PI = volume of the ipsilateral parotid gland; V PC = volume of the contralateral parotid gland; Mean PI = mean dose to the ipsilateral parotid gland; Mean PC = mean dose to the contralateral parotid gland; Max PI = maximum dose to the ipsilateral parotid gland; Max PC = maximum dose to the contralateral parotid gland; RP = retropharyngeal lymph nodes; EORTC/RTOG = radiation therapy oncology group/european organization for research and treatment of cancer late radiation morbidity scoring scheme. 178

179 D. Reducing toxicity None of the patients suffered from Sjögren's syndrome or any other medical cause of xerostomia. None of the patients received any treatment for xerostomia (e.g. saliva substitutes or stimulants) during or after treatment. The study protocol was approved by the local ethics committee; informed consent was obtained from all patients DW-MRI examinations All patients were asked not to eat or drink for at least one hour prior to the MRI examination. The imaging protocol used in this study was previously described in detail [257, 258]. All MRI examinations were performed on a 1.5-T scanner (Magnetom SONATA, Siemens, Erlangen, Germany) using a standard head coil. For morphologic evaluation of the salivary glands, T1-weighted and T2-weighted series were acquired in the transverse plane. The images extended from the base of skull to the undersurface of the submandibular glands, including the full volume of both the parotid and the submandibular glands.. Thereafter, transverse DW images were obtained with b-values of 400, 500, 600, 700, 800, and 1000 sec/mm 2. The acquisition time was 2 minutes 22 seconds for every DW sequence that covered the parotid and submandibular glands. One DW series was acquired at rest, after which one 500mg tablet of ascorbic acid (Redoxon, Roche, Basel, Switzerland) was given orally. Patients were instructed to let the tablet melt and not to chew on it. During the salivary stimulation, the DW sequence was repeated 8 times (with intervals of 30 seconds between consecutive repeats). The data were transferred to an independent Linux workstation with dedicated software (Biomap, Novartis, Basel, Switzerland). On the native DW images, regions of interest were drawn freehand by two observers in consensus around both parotid and submandibular glands on all sections. Regions containing large vessels such as the retromandibular vein and external carotid artery were excluded. The ROIs were delineated in a similar way on all repeated sequences over the entire stimulation period. Calculation of ADC values was performed using the average value of each ROI (ipsilateral parotid, contralateral parotid, ipsilateral submandibular, and contralateral submandibular glands) for all the applied b-factors. To reduce the influence of noise on the calculations, diffusion images with six different b-factors were used for each ADC calculation. 179

180 D. Reducing toxicity Salivary gland scintigraphy A pre-rt SGS was performed in all patients, but one scintigraphy (patient 3) could not be interpreted due to technical failure. Two patients (patients 1 and 5) refused a follow-up SGS. The acquisition procedure was previously described in detail [102, 255, 302]. 99m Tcpertechnetate (370 MBq) was injected intravenously and a planar dynamic acquisition was obtained for a period of 12 minutes, using a gamma camera. A second acquisition was performed between minute 18 and 34. One minute (minute 19) after the start of the second dynamic acquisition, a commercially available α- and β-sympatheticomimetic agent (0.25mg carbachol) was administered subcutaneous. The dynamic geometric mean planar images were used to calculate salivary excretion fraction, i.e. the percentage of activity that disappeared within 15 minutes following the administration of carbachol. The SEF was calculated for all glands pre- and post-rt. A high SEF value corresponds with a good excretion function of the salivary gland. A SEF value of 26.5% was considered as the lowest fraction still compatible with a normal function [302]. To evaluate the percentage loss of salivary excretion fraction after RT, with respect to the baseline salivary excretion fraction, dsef was calculated: dsef (%) = (SEF pre SEF post )/SEF pre x 100 [254]. A low dsef signifies a small loss of function and a negative dsef signifies an improvement of function after RT Statistical analysis The data were analyzed using the software package Statistica 7 (StatSoft Inc, Tulsa, OK, USA). For statistical analysis, paired two-tailed student t-tests were used to compare ipsilateral and contralateral parotid and submandibular glands in patients before and during stimulation, both before and after RT. The ADC values of submandibular glands were compared to the parotid glands before stimulation using unpaired two-tailed student t-tests. A p value of less than 0.05 was considered statistically significant Results Radiotherapy The mean total dose to the contralateral parotid glands was 20.0 Gy (range, Gy). Generally, a mean parotid gland dose < Gy is considered to be the dose constraint 180

181 D. Reducing toxicity for functional sparing [299]. The ipsilateral parotid glands received a mean dose of 53.9 Gy (range, Gy). The submandibular glands received a mean dose of 58.5 Gy to the ipsilateral and of 47.5 Gy to the contralateral glands. Xerostomia, measured at the same timepoints as the DWI and SGS imaging by the RTOG/EORTC late morbidity score, was, if present, generally mild (Table 32). Salivary scintigraphy data confirmed the successful sparing of the contralateral parotid glands: mean post-rt SEF was 35.2% (range, 17 48%) and mean dsef was 37.6% (range, %). The loss of function in the ipsilateral parotid gland and both submandibular glands was similarly confirmed by the SGS data, as shown in Table 33. Table 33. Salivary gland scintigraphy data. PT No. SEF1 [%] PAROTID IPSI PAROTID CONTRA SM IPSI SM CONTRA SEF2 dsef SEF1 SEF2 dsef SEF1 SEF2 dsef SEF1 SEF2 [%] [%] [%] [%] [%] [%] [%] [%] [%] [%] radical neck dissection radical neck dissection superficial parotidectomy radical neck dissection dsef [%] Abbreviations: SEF = salivary excretion fraction (a SEF value of 26.5% is considered as the lowest fraction still compatible with a normal function); SEF1 = SEF before RT; SEF2 = SEF at a mean of 9 months after RT; dsef = loss of salivary excretion fraction; SM ipsi = ipsilateral submandibular glands; SM contra = contralateral submandibular glands DW-MRI In patients before RT, the mean ADC value at rest was significantly lower in the parotid (0.85 (± 0.11) x 10-3 mm 2 /sec) than in the submandibular (1.14 (± 0.15) x 10-3 mm 2 /sec) glands (p < ). No significant difference was observed between ipsilateral (0.84 (± 0.08) x 10-3 mm 2 /sec) and contralateral (0.85 (± 0.14) x 10-3 mm 2 /sec) parotid (p = 0.15) or ipsilateral (1.12 (± 0.13) x 10-3 mm 2 /sec) and contralateral (1.16 (± 0.18) x 10-3 mm 2 /sec) submandibular (p = 0.52) glands at rest. A similar shaped biphasic response to gustatory stimulation was found in parotid and 181

182 D. Reducing toxicity submandibular glands of all patients before RT. An initial drop in ADC value to a minimum value, reached after a median of 5.44 minutes, was followed by a slow increase to a peak, reached after a median of minutes. In the parotid glands, the ADC value first showed a decrease from 0.85 (± 0.11) x 10-3 mm 2 /sec to a minimum of 0.80 (± 0.10) x 10-3 mm 2 /sec (p = 0.002), followed by an increase to a maximum of 0.92 (± 0.11) x 10-3 mm 2 /sec (p < ), higher than the baseline value (p = ), as shown in Figure 24. In the submandibular glands, a decrease from 1.14 (± 0.15) x 10-3 mm 2 /sec to a minimum of 0.93 (± 0.15) x 10-3 mm 2 /sec (p < ) was seen, followed by an increase to a peak of 1.18 (± 0.16) x 10-3 mm 2 /sec (p < ), higher than the baseline value (p = 0.35). Figure 24. Evolution of ADC values during stimulation before RT for (a) contralateral parotid glands and (b) ipsilateral parotid glands. a) 182

183 D. Reducing toxicity b) Abbreviations: Min = time-point at which the minimum ADC value is reached; Max = time-point at which the peak ADC value is reached. For the ipsilateral parotid glands, the baseline ADC values at rest were significantly higher after RT (1.08 (± 0.21) x 10-3 mm 2 /sec) than before RT (p = 0.01). Contrary, in the contralateral parotid glands, the baseline ADC values after RT (0.84 (± 0.11) x 10-3 mm 2 /sec) were not significantly different compared to before RT (p = 0.75). When comparing the ADC values at exactly the same time-points as before RT, a comparable response to stimulation as before RT was observed in the contralateral, but not in the ipsilateral parotid glands (Figure 25). In the contralateral glands, a decrease from 0.84 (± 0.11) x 10-3 mm 2 /sec to 0.80 (± 0.09) x 10-3 mm 2 /sec (p = 0.08) was seen, followed by an increase to 0.84 (± 0.09) x 10-3 mm 2 /sec (p = 0.23). This response to stimulation was absent in the ipsilateral glands. 183

184 D. Reducing toxicity Figure 25. A comparable response to stimulation as before RT was observed in the contralateral, but not in the ipsilateral parotid glands. 1,20 Response to stimulation of the parotid glands of all patients before and after RT. 1,10 ADC (x 10-3 mm 2 /sec) 1,00 0,90 0,80 0,70 0,60 Contralateral parotid glands before RT Contralateral parotid glands after RT Ipsilateral parotid glands before RT Ipsilateral parotid glands after RT Start Min Max Looking at the individual patient data, 4 patients (patients 4, 5, 6 and 8) still showed the typical biphasic response with an initial decrease, followed by an increase, in the contralateral parotid gland after RT (Figure 26). One patient (patient 2) showed an increase, followed by a plateau. In this patient, SGS confirmed the loss of function in both parotid glands after RT and this patient also suffered from manifest clinical xerostomia (grade 3). Three patients (patients 1, 3, and 7) showed an initial decrease, although without subsequent increase in ADC values. In those patients, SGS data are either not available (patient 1) or do not support function loss (patients 3 and 7), and no clinically relevant xerostomia was reported, suggesting at least some residual salivary gland function. 184

185 D. Reducing toxicity Figure 26. Evolution of ADC values during stimulation after RT for (a) contralateral parotid glands and (b) ipsilateral parotid glands. a) 185

186 D. Reducing toxicity b) Abbreviations: Min = time-point at which the minimum ADC value was reached in the DW-MRI scan before RT; Max = time-point at which the peak ADC value was reached in the DW-MRI scan before RT. For the submandibular glands, the baseline ADC value was higher after RT compared to before RT in both the ipsilateral (1.36 (± 0.07) x 10-3 mm 2 /sec; p = 0.004) and the contralateral submandibular gland (1.32 (± 0.30) x 10-3 mm 2 /sec; p = 0.2). At exactly the same time-points as before RT, the ADC values in the ipsilateral glands changed first to 1.30 (± 0.05) x 10-3 mm 2 /sec (p = 0.31) and then to 1.38 (± 0.11) x 10-3 mm 2 /sec (p = 0.33). In the contralateral glands, first to 1.13 (± 0.37) x 10-3 mm 2 /sec (p = 0.02) and then to 1.25 (± 0.03) x 10-3 mm 2 /sec (p = 0.18). 186

187 D. Reducing toxicity 2.3. Discussion In this prospective feasibility study, a DW-MRI protocol was applied to patients before and after parotid-sparing radiotherapy for head and neck cancer. First, these results confirm previous findings and add additional information on the normal response of functioning salivary glands to gustatory stimulation [257, 258]. A significant decrease in ADC could be observed during the first 5 minutes of stimulation. This is most likely attributable to the emptying of stored saliva and the consequent reduction of free water in the extracellular space. During the next 12 minutes, an increase in ADC was observed, corresponding with the active production of new saliva, which correlates to increase of free water in the extracellular space. As in other studies, ADC values at rest were significantly lower for the parotid glands than for the submandibular glands [258, ]. In the unstimulated state, about two-thirds of all saliva is produced by the submandibular glands, compared to less than one-third by the parotid glands. Thus, the expected higher amount of water in the extracellular space of the submandibular glands can explain the higher ADC at rest. The difference may also be explained by the different histologic composition of the glands: the parotid gland is purely serous, whereas the submandibular gland is a mixed serous and mucous gland. Furthermore, the higher amount of adipose tissue in the parotid gland compared to the submandibular gland is another possible contributing factor to the lower ADC of the parotid gland. The DW-MRI technique used relies on fat-suppression, which can result in an overall decreased signal intensity of fatty tissues for all b-factors. Second, this study is the first to demonstrate that DW-MRI can assess radiation-induced loss of function in salivary glands. Baseline ADC values were significantly higher after RT compared to the ADC values prior to RT in non-spared glands, possibly corresponding with the presence of fibrosis and/or necrosis [67]. In contralateral parotid glands, no significant difference in baseline ADC value was seen before and after RT. After RT, a preserved biphasic response to stimulation, similar to the response before RT, was seen in the contralateral parotid glands, suggesting a preserved function. This was confirmed by the normal SEF values and the low dsef values in those glands, and by the low xerostomia symptom scores, implying residual salivary gland activity. However, in the ipsilateral glands, no significant changes were seen in ADC values during stimulation and the typical response to stimulation was absent. This suggests functional loss of these salivary glands, no longer capable of production and excretion of saliva. The loss of function was confirmed with a validated imaging technique, SGS, which was shown to correlate well with salivary flow as measured by Lashley cups in several studies, and can be used to evaluate salivary gland function [251]. 187

188 D. Reducing toxicity Xerostomia is one of the most frequent and serious complications of conventional radiotherapy for head and neck cancer. Highly conformal radiotherapy techniques, allowing a significant decrease of the doses to the major salivary glands, are increasingly used to prevent permanent xerostomia. However, the technical ability to spare the salivary glands from highdose irradiation currently exceeds our knowledge of the underlying biology. For parotid glands, an exponential relationship between function loss and mean parotid dose has been clearly demonstrated, suggesting that it is essential to respect a certain mean dose threshold [296, 297]. Recently, however, it was suggested that late damage after partial irradiation of the parotid gland might be region-dependent [300, 301]. This corresponds with earlier research from our group, whereby a high inter-patient variability in D 50 for different regions within each parotid gland was found and it was also noted that low doses (10 15 Gy) could nonetheless result in serious loss of function [302]. These data suggest that spatial information should be included in the comparison of different plans and that the mean dose concept might have limited use for the prediction of late radiation damage. For submandibular glands, it was recently demonstrated that sparing of the contralateral submandibular gland (mean dose < 25 Gy) is feasible with IMRT and results in prevention of permanent xerostomia, although evidence-based data on thresholds are still lacking [329]. Therefore, an imaging technique, allowing physicians to non-invasively assess functional changes in both parotid and submandibular glands at the same time, and with sufficiently high spatial resolution, could significantly improve our current models of dose-response relationships. The combination of SGS with SPECT is at the moment considered as the standard imaging technique [251, 255, 298]. By providing three-dimensional information on salivary gland function, it offers the opportunity to evaluate the dysfunction after irradiation of different areas within the parotid glands [302]. However, because of the limited spatial resolution of SPECT analysis (in-plane spatial resolution of about 10 mm), this technique is unable to reproduce sharp dysfunction gradients within one gland. The combination of anatomic and functional information provided by DW-MRI may allow better correlation of function and dysfunction of partial salivary gland volumes in correlation to radiation dose distributions within each gland. Indeed, since ADC values can be theoretically calculated for each pixel, functional three-dimensional maps of the parotid and submandibular glands before and at different time-points after RT could be compared with the isodoses to evaluate dose-response relationships within the glands and between patients. This depends to a large extent on sufficiently accurate co-registration of different imaging modalities (planning CT and DWI), taken at different time-points (before and after RT). Since DW-MRI has a 188

189 D. Reducing toxicity considerable distortion and a low signal-noise ratio, non-rigid registration methods have to be developed [237]. Some limitations of this feasibility study should be pointed out. First, patient numbers are small and some data are lacking, most notably the follow-up SGS after RT was not performed in 2 patients. Moreover, looking at individual patient data, the post-rt contralateral parotid glands of 3 patients did show an initial decrease in ADC values, but not the subsequent increase. This could correspond with normal emptying of stored saliva (initial decrease), but delayed production of new saliva (no increase). Longer scanning time is necessary to evaluate this hypothesis. However, results are sufficient to provide the proof of principle that DW-MRI can assess salivary gland function before and after RT. Second, as this is the first study performing DW-MRI with gustatory stimulation after RT, or indeed any other pathological condition causing functional loss, there are currently no guidelines on how the results after RT should be interpreted. We choose to evaluate ADC values at corresponding time-points after RT as before RT, expecting an identical response. This methodological approach allows maximal reproducibility and indeed yields results that are consistent with the standard imaging technique (SGS) and clinical reality. Notwithstanding the potential added information that is gained by DW-MRI, the technique provides little or no information on the salivary ducts, which can also undergo important changes after RT. Recently, MRI sialography was developed as a relatively straightforward method to image radiation-induced changes in the major salivary ducts [246]. A combination of both imaging modalities could provide valuable insights into the mechanism of saliva production and excretion. In conclusion, DW-MRI depicts changes in the major salivary glands during gustatory stimulation based on changes in extracellular water content and shows a biphasic response with an initial decrease of ADC followed by a slow increase. In this limited study population, we demonstrated that DW-MRI can non-invasively assess functional changes in salivary glands after RT. Non-rigid registration of those measured radiation-induced effects to the 3D dose distribution is being developed in order to evaluate spatial heterogeneity of glandular functionality. This makes DW-MRI a promising tool for investigating RT-induced xerostomia. 189

190

191 CHAPTER 3: DYSPHAGIA AFTER CRT FOR HNC: DOSE- EFFECT RELATIONSHIPS FOR THE SWALLOWING STRUCTURES Published as: Dirix P., Abbeel S., Vanstraelen B., Hermans R., Nuyts S. Dysphagia after chemoradiotherapy for head and neck squamous cell carcinoma: dose-effect relationships for the swallowing structures. Int J Radiat Oncol Biol Phys 2009; 75(2):

192

193 D. Reducing toxicity 3. Dysphagia after CRT for HNC: dose-effect relationships for the swallowing structures Radiotherapy-induced dysphagia is correlated with compromised quality of life, anxiety and depression, and can lead to life-threatening complications such as aspiration pneumonia. It is to be expected that limiting the dose to the critical swallowing structures will reduce the incidence of dysphagia. However, several questions regarding which swallowing structures are essential, and what volume and dose constraints should be applied, remain to be answered Material & methods Patient selection and characteristics Medical records were reviewed for all 121 patients treated with primary CRT for previously untreated loco-regionally advanced HNC between January 2004 and June 2007 at the Leuven radiotherapy department. Seventy-three patients (60.3%) were still alive and free from disease at the time of analysis (January 2008). These patients were then invited to participate to the study and to fill in 4 QoL questionnaires during their next routine follow-up visit at the multidisciplinary outpatient clinic. Fifty-three patients (72.6% of eligible patients) agreed to partake, forming the primary study population of this study. The protocol was approved by the local ethics committee; informed consent was obtained from all patients. Patient and tumor characteristics are summarized in Table 34. There were 42 male and 11 female patients, with a median age of 57 years (range, years) at the time of diagnosis. The primary tumor was localized in the oropharynx (45.3%), larynx (26.4%), hypopharynx (18.9%), oral cavity (7.5%), or nasopharynx (1.9%). Staging was performed according to the 2002 TNM classification system of the AJCC at the time of diagnosis [101]. All patients suffered from locally (> T2: 66.0%) and/or regionally (N+: 88.7%) advanced tumors. Treatment consisted of primary radiotherapy with concomitant chemotherapy. The majority of patients (75.5%) was treated according to the hybrid fractionation schedule. A minority of patients (24.5%) were treated with a conventional fractionation schedule, to 70 Gy in 7 weeks. A 3D-RT technique was used in 44 (83.0%) patients. In 14 of those 44 patients, a 4- field technique was used to spare the contralateral parotid gland [102]. The other 30 patients were treated with 2 opposing lateral beams and 1 lower neck field for the supraclavicular regions. Nine (17.0%) patients were treated with IMRT. PEG tubes were placed before the start of CRT in all patients, although feeding by mouth was encouraged for as long as possible. 193

194 D. Reducing toxicity Table 34. Patient and tumor characteristics. Characteristics No. of patients % Gender Male % Female % Age < 60 years % 60 years % Primary tumor site Oral cavity 4 7.5% Buccosal mucosa 2 Tongue 1 Floor of mouth 1 Oropharynx % Anterior wall 9 Lateral wall 11 Posterior wall 4 Larynx % Supraglottis 13 Subglottis 1 Hypopharynx % Pharyngo-oesophageal junction 1 Piriform sinus 7 Posterior pharyngeal wall 2 Nasopharynx 1 1.9% T-classification T % T % T % T4a % T4b 1 1.9% N-classification N % N % N2a 2 3.8% N2b % N2c % N % Clinical stage Stage III % Stage IV % 194

195 D. Reducing toxicity Evaluation of late dysphagia Patients were assessed at the outpatient clinic for evaluation of late swallowing problems after CRT at a mean 20.4 months since the end of treatment (range, months). During the patient s visit, late dysphagia was scored by the treating physician according to the RTOG/EORTC late radiation morbidity scoring schema. At the same time, patients independently filled in 4 QoL questionnaires: 1) the EORTC core quality of life questionnaire (QLQ) C30 version 3.0, consisting of 30 items; 2) the EORTC head and neck cancer module QLQ-H&N35, consisting of 35 items; 3) the performance status scale (PSS) of List, with the functions eating in public and normalcy of diet; and 4) the MD Anderson dysphagia inventory (MDADI), consisting of 20 questions with global, emotional, functional, and physical subscales Clinical and dosimetric factors The clinical variables examined for correlation with late dysphagia included age (< 60 vs. 60 years), gender, presence of swallowing problems at diagnosis, acute swallowing toxicity (still able to eat or drink vs. unable to swallow), primary site (larynx, hypopharynx, posterior oropharyngeal wall vs. oral cavity, nasopharynx, anterior or lateral oropharyngeal wall), T- classification (T1 2 vs. T3 4), N-classification (N0 1 vs. N2 3), use of IMRT, fractionation schedule (70 Gy vs. 72 Gy), and follow-up after CRT (< 1 year vs. 1 year). In addition to these clinical variables, we also analyzed the radiation dose to the swallowing structures and parotid glands. The original planning CT scans together with the 3D- RT or IMRT treatment plans of the 53 previously irradiated patients were retrieved from the planning system (Eclipse, Varian Inc, Palo Alto, CA) for retrospective delineation and DVH calculation of the swallowing OAR. The mean and maximum doses to each structure as well as the partial volumes of a structure receiving a specific dose (V D S) were calculated. For example, the V 50 of a structure was defined as the volume of a structure receiving 50 Gy. Based on a literature search, anatomic text books and radiological data, 8 swallowing structures were identified: 1) superior pharyngeal constrictor (SPC) muscle; 2) middle pharyngeal constrictor (MPC) muscle; 3) inferior pharyngeal constrictor (IPC) muscle; 4) base of tongue (BOT); 5) supraglottic larynx (SGL); 6) glottic larynx (GL); 7) upper esophageal sphincter (UES), including the cricopharyngeus muscle; 8) esophagus (ES) [ ]. Definitions of the swallowing OAR are provided in Table 35, and illustrated in Figure

196 D. Reducing toxicity Figure 26. Swallowing structures: superior pharyngeal constrictor muscle (cyan blue), middle pharyngeal constrictor muscle (red), inferior pharyngeal constrictor muscle (green), upper esophageal sphincter (yellow), esophagus (dark blue), base of tongue (white), supraglottic larynx (orange), glottic larynx (magenta) Statistical analysis Since the endpoints of all scoring manuals were continuous, the presence or absence of a dysphagia event was established by clustering for the purposes of this study [ ]. For RTOG/EORTC scores, grades 0 and 1 (no swallowing problems), as well as 2 and 3 (swallowing problems), were combined [333]. C30, H&N35, PSS, and MDADI responses were categorized according to the median score [334]. Logistic regression analysis was used to assess the relationships between dosimetric (i.e. radiation dose and volume) factors and 196