Gross tumor volume determines toxicity and quality of life for patients with nasopharyngeal carcinoma treated by concurrent chemoradiotherapy with simultaneously integrated boost technique

Background: To evaluate the impact of gross tumor volume (GTV) on toxicities and quality of life (QoL) for patients with nasopharyngeal carcinoma (NPC) treated by concurrent chemoradiotherapy with simultaneously integrated boost technique (chemo-SIB-IMRT). Methods: A total of 278 NPC patients with stage II-IVb treated by chemo-SIB-IMRT were enrolled. Toxicities were evaluated according to CTCAE version 4.03. QoL outcomes (n=219) were measured by using the EORTC QLQ-C30 and HN35 questionnaires at the time point of 12 months after chemo-SIB-IMRT. Results: A higher GTV was observed to be signicantly associated with a higher mean or maximal dose in most organs at risk, together with more severe acute (mucositis, dermatitis, weight loss, and use of analgesic) and late toxicities (xerostomia, neck brosis, and radiation neuropathy). A linear regression model revealed that a higher GTV was signicantly associated with a decline in role functioning and an increment in taste/smell, speech, social eating, opening mouth, dry mouth, and sticky saliva. Conclusion: GTV is the determining factor of some acute and late toxicities and QoL scales for NPC patients treated by chemo-SIB-IMRT. techniques

coplanar arcs of 360° and was simultaneously optimized to be delivered with opposite rotation (clockwise and counter-clockwise). There was a total of 182 control points, with a collimator angle of 10-15°. Continuous gantry motion, dose-rate variation and the motion of multi-leaf collimators (MLC) were approximated by optimizing individual beams at 2-4° gantry angle increments. The collapsed cone convolution algorithm was used for dose calculations and the SmartArc module was adopted for dose optimization. For both techniques, the beam delivery was generated with 6-MV photons by the Linac machines equipped with dynamic MLC. All treatment plans were evaluated to ensure that 95% of all the PTVs received the prescription dose. The dose volume histograms of PTVs and OARs were quantitatively assessed and the isodose curves on axial CT slices were qualitatively inspected for each plan.
The regimen of concurrent chemotherapy was weekly cisplatin 30 mg/m 2 administered during SIB-IMRT courses. Adjuvant chemotherapy with cisplatin 70-80 mg/m 2 on day 1 and 5-uorouracil 700-800 mg/m 2 /d on days 1-4 were administered every 3-4 weeks was given for 2-4 cycles to those patients (n=138) with advanced nodal status or persistent elevation of plasma EBV-DNA titer after SIB-IMRT.

Toxicity assessment
The severity of acute toxicities (mucositis, dermatitis, weight loss, use of analgesic, and presence of tube feeding) and late toxicities (xerostomia, dysphagia, otitis media, neck brosis, and neuropathy) were retrospectively collected from the medical charts. Acute toxicities were considered peak grade during the period of chemo-SIB-IMRT. The grade of late toxicities was considered as the maximal grade that persisted or occurred during the period from three months after chemo-SIB-IMRT to the date of the last visit. The objective assessment of pain severity was regarded as the strongest analgesic used, which was graded according to the WHO analgesic ladder (0: no analgesic use 1: use of non-narcotic analgesics, 2: use of weak narcotics, 3: use of strong narcotics). The toxicities were graded mainly based on the criteria of Common Terminology Criteria for Adverse Events version 4.03 (CTCAE v.4.03).

QoL assessment
The QoL outcomes were periodically assessed by the Taiwan Chinese versions of EORTC QLQ-C30 and HN35 questionnaires at the following time points: before RT, during RT (about 40 Gy), at 3 months and 12 months after RT, and then annually thereafter in the institute. Data measured at the time point of 12 months after chemo-SIB-IMRT was used for analysis in the study. The questionnaires have been tested in Taiwanese NPC patients and excellent reliability and validity were obtained [12]. The EORTC QLQ-C30 is a widely used questionnaire, incorporating a range of QoL issues relevant to a broad range of cancer patients. It contains ve functional scales (physical, role, cognitive, emotional, and social), three symptom scales (fatigue, pain, and nausea/vomiting), a global QoL scale, and six single items (dyspnea, insomnia, appetite loss, constipation, diarrhea, and nancial di culties) [13]. The HN35 is a module used for assessing the QoL for head-and-neck cancer patients. It incorporates seven multiple-item scales that assess the symptoms of pain in the head and neck, swallowing ability, senses (taste /smell), speech, social eating, social contact, and sexuality. Also included are six single-item scales, which survey the presence of symptomatic problems associated with teeth, mouth-opening, dry mouth, sticky saliva, coughing, and feeling ill [14]. All scales pertaining to the EORTC QLQ-C30 and HN35 range from zero to 100. A high score for a functional or global QoL scale represents a relatively high/healthy level of functioning or global quality of life, whereas a high score for a symptom scale represents the presence of a symptom or problem(s).

Follow-up
Patients were regularly followed up until death or their last appointment. They were scheduled to visit the clinics at 3 months, and 4-to 6-month intervals in the rst two, and third to fth years, respectively. The median follow-up months were 36.6 months (range, 2 to 83.6 months). Physical and nasopharyngoscopic examinations were routinely performed on every visit. Head and neck MRI scans were performed within 2 months after RT and then annually or when there were clinical indications.

Statistical analysis
The mean scores of QoL scales were calculated according to the EORTC QLQ scoring manual. To deal with the missing data, the missing items were assumed to have values equal to the average of those items that were present for the respondent, if at least half of the items from the scale had been answered (i.e., mean imputation). The Pearson correlation was used to assess the correlation between GTV and the dosimetric values of OARs and the scores of QoL. The Chi-square test was used to assess the associations between category variables. A binary logistic regression model was used to perform the multivariate analysis to evaluate the prognostic factors associated with the outcome of acute and late toxicity, and a linear regression model (backward) was used to explore the prognosticators associated with the scores of the scales of QoL. The cumulative incidence of the occurrence of late toxicity was estimated by using the Kaplan-Meier method. Microsoft SPSS-20.0 was used for data processing, with a p-value <0.05 from the two-sided test regarded to be statistically signi cant.

GTV and dosimetry of OARs
The mean (SD), median and range of GTV in the study cohort were 84.6 (69.4) ml, 63 ml, and 2-372 ml, respectively. GTV was observed to be signi cantly and positively correlated with the mean and/or maximal dose of OARs concerned, as revealed in Table 2, including the brain stem, eyeballs, optic nerves, parotid glands, oral cavity, larynx, posterior pharyngeal wall, and cervical esophagus (p<0.05, correlation coe cients: 0.227-0.481). The mean dose of parotids was observed to be correlated with GTV with the highest correlation coe cients (right parotid: 0.481, left parotid: 0.472).
Almost every patient experienced some degree of acute toxicities, including mucositis, dermatitis, weight loss, or use of analgesic during the treatment period. As shown in Table 3, the distributions of acute toxicities with grade 2 or more were 214 (76.9%) for mucositis, 259 (57.2%) for dermatitis, 78 (28.0%) for weight loss, and 206 (74.1%) for the use of analgesic, respectively. Tube feeding by nasogastric tube or through gastrostomy was observed in 30 patients (10.8%). Entering all the socio-demographic and clinical variables simultaneously with each of the ve dependent variables of acute toxicities at a time into the binary logistic regression models, we found that GTV was signi cantly associated with four acute toxicities (mucositis, dermatitis, weight loss, and use of analgesic). Compared with the counterpart, those with GTV>63ml (median value) had a higher probability of acute toxicities, with a difference of 2.6-fold (95% CI: 1.3-5.1) in mucositis≧grade 2, 7.2-fold (95% CI: 3.8-13.6) in dermatitis≧grade 2, 1.7-fold (95% CI: 1.1-3.1) in weight loss≧grade 2, and 4.3-fold (95% CI: 1.6-12.1) in use of analgesic≧grade 2 (Table 4). In contrast, those with T3-4 classi cation were observed to have a higher probability of mucositis≧grade 2, with a difference of 3.0-fold (95% CI: 1.1-8.3) compared with those with T1-2; those with N2-3 classi cation were observed to have a higher probability of dermatitis≧grade 2, with a difference of 2.0-fold (95% CI: 1.0-3.8) compared with those with N0-1.

GTV and late toxicity
Xerostomia, dysphagia, otitis media, and neck brosis were the common late toxicities concerned. The distributions of the four late toxicities with grade 2 or more were 78 (28.4%) for xerostomia, 53 (19.0%) for dysphagia, 82 (29.5%) for otitis media, and 9 (3.2%) for neck brosis. In addition, radiation neuropathy was also observed in 15 (5.4%) patients, including temporal lobe necrosis in 10 (3.6%) patients, and cranial nerve palsy in ve (1.8%) patients (four hypoglossal nerve palsy and one optic neuropathy). All radiation neuropathy was mild and did not interfere in the activities of patients' daily life (grade 1). The 5-year cumulative incidence of developing at least one of the ve late toxicities with grade 3 or more was 7.8%. Entering all the socio-demographic and clinical variables simultaneously with each of the ve dependent variables of late toxicities at a time into the binary logistic regression models, we found that GTV was signi cantly associated with three late toxicities (xerostomia, neck brosis, and radiation neuropathy). Compared with the counterpart, those with GTV>63ml had a higher probability of late toxicities, with a difference of 2.2-fold (95% CI: 1.2-4.1) in xerostomia≧grade 2, 2.4-fold (95% CI: 1.2-4.3) in neck brosis≧grade 1, and 8.9-fold (95% CI: 2.1-16.6) in radiation neuropathy≧grade 1 ( Table 5). In contrast, those with age>52 years (median value) were observed to have a higher probability of xerostomia≧grade 2, with a difference of 1.8-fold (95% CI: 1.1-3.8) compared with those with age≦52 years; those with N2-3 classi cation were observed to have a higher probability of neck brosis≧grade 1, with a difference of 1.5-fold (95% CI: 1.0-3.0) compared with those with N0-1; and those with T3-4 classi cation were observed to have a higher probability of radiation neuropathy≧grade 1, with a difference of 1.9-fold (95% CI: 1.0-4.0) compared with those with T1-2.

Discussion
The dosimetry of OARs is usually tumor site and volume dependent. It has been reported that the tumor site is a prognostic factor for morbidities and QoL for patients after receiving head and neck irradiation by conventional 2D-RT [15]. Unlike other head and neck cancer, NPC usually presents with high propensity of tumor in ltration into the skull base and bilateral neck lymph nodes; thus, elective irradiation with large elds from the level of the skull base to the infra-clavicular nodal area is mandatory, even in patients with an early clinical stage. Using SIB-IMRT composed of three xed dose levels, with a higher dose per faction to GTV and two lower doses to cover the uniform subclinical or elective anatomic sites, we observed the major differences of dose-volume received by OARs in NPC patients depending on the GTV. GTV was observed to be signi cantly and positively correlated with the mean and/or maximal dose in most OARs concerned. Furthermore, as far as we know from the literature, this is the rst report to show GTV to be the determining variable for acute/late toxicity and QoL of NPC patients treated by SIB-IMRT combined with chemotherapy.
The location and size of the primary tumor, lymph node staging, tumor laterality, treatment techniques, and anatomical changes during treatment have all been shown to be signi cant factors in terms of the actual dose received by OARs [16]. Mucositis is common for NPC patients receiving chemo-RT. The problems associated with mucositis include oral pain, odynophagia, impaired taste, and weight loss. Mucositis was known to have a dose-effect relation with RT; however, the ability of clinical or dosimetric parameters in predicting severe mucositis was still controversial, using IMRT technique [17,18]. In the study of Li et al [19]., weight loss and V30Gy predict severe oral mucositis in NPC patients treated by concurrent chemo-IMRT. In contrast, the mean dose in the oral cavity was found to impact the duration of oral mucositis by Orlandi et al [20]. In treating NPC, the oral cavity was usually outside the CTV elds, but its mean dose still ranges 35-42 Gy using the IMRT or VMAT technique [21][22][23]. Without using the detailed dosimetric analysis for oral mucositis in current study, we observed GTV positively correlated with the mean dose in the oral cavity and was predictive of mucositis and its related problems, e.g. weight loss and analgesia used.
Xerostomia remains one of the most common late toxicities for NPC patients treated by chemo-SIB-IMRT. The severity of xerostomia is largely dependent on the dose/volume in the salivary gland in the radiation eld [24,25]. The incidence of grade 2 or greater xerostomia in our cohort was 28.4%, which was higher than that reported in the literature [11,21]. The parotid mean dose was around 30 Gy in the study, which is compatible with other reports [26,27]. We observed the mean dose in the parotids most signi cantly correlated with GTV. This is mainly due to the major overlap of the PTV with the deep lobes of the parotids. An overzealous effort in sparing the parotids can result in the under-dosing of part of the GTV which is at risk for local recurrence, especially in those with parapharyngeal space invasion or upper neck lymph node enlargement. Growing reports have demonstrated there is no direct dose-effect relationship between the parotid dose and xerostomia for NPC patients [28,29]. The reasons might be that the parotid dose revealed in the pre-treatment plan was different from the actual dose received because of the positional change of the gland during the treatment process and the mean dose of parotid in most NPC patients beyond the dose constraints recommended from the QUANTEC guideline [30,31]. In contrast, GTV was observed to be predictive of xerostomia in our study. In this scenario, the other major salivary glands, the submandibular glands, were taken into consideration because they were usually exposed to high-dose irradiation if large nodes existed at level II.
Progressive neck brosis, which might compromise neck movement and increases the risk of vasculopathy of the carotid artery or neuropathy, was common in NPC patients treated by conventional 2D-RT but has been markedly reduced by IMRT. 32 The incidence of symptomatic neck brosis and cranial neuropathy was 3.2% and 1.8%, respectively, in our series, which is compatible with other reports [33,34]. However, our patients with larger GTV beneath the sternocleidomastoid muscle still had a signi cantly higher risk of suffering from progressive neck brosis.
The determinants of radiation neuropathy in previous reports included T classi cation, total RT dose, fraction size, overall treatment time, receiving chemotherapy, and GTV. In these studies, the patients analyzed underwent heterogeneous RT techniques, varying from conventional 2D-RT, 3D conformal RT, to IMRT [34][35][36]. In contrast, with a consistent technique of SIB-IMRT in our series, we observed the pre-treatment variable of advanced T classi cation and larger GTV were signi cantly associated with the occurrence of radiation neuropathy.
Socio-demographic variables and RT techniques were observed to be the determinants of post-treatment QoL outcome for NPC patients in the literature [15,37,38]. Most of the reports were in cohorts either treated by conventional 2D-RT or a mixture of different RT techniques. Huguenin et al. observed that head and neck cancer patients with large target volumes (e.g. NPC) treated by 2D-RT suffered from more QoL problems, including dry mouth, sticky saliva, trismus, problems with teeth, and trouble eating, than those with small target volumes (e.g. glottis ca) [15]. They also observed that these symptoms did not have a high impact on global QoL or functional scales on the QLQ-C30 core questionnaire. In the current study, we also observed NPC patients with a large GTV had more symptom problems, including dyspnea, constipation, taste/smell, speech, social eating, opening mouth, dry mouth, and sticky saliva.
Meanwhile, global QoL and most functional scales (except for role functioning) were not observed to be signi cantly correlated with GTV in our study.
There are growing reports supporting the therapeutic bene ts in dosimetry, tumor control and QoL for NPC patients if re-planning could be conducted either for patients with GTV reduction or the geometric change of OARs during the treatment course of SIB-IMRT [39,40]. Using SIB-IMRT techniques, the major bene t of re-planning, especially for those without signi cant body shape change, is to reduce the GTV during the treatment course, which might potentially reduce the dose to the OARs. In the study by Yang et al. [39], the authors reported NPC patients treated by SIB-IMRT would have better QoL in global QoL, role and social functioning and some symptom scales if re-planning was done. We believe our study could echo the need for re-planning, especially for those with GTV reduction during the course of SIB-IMRT in NPC patients.
It should be noted that there are some limitations in the study. First, as mentioned above, geometric changes of GTV and OARs during treatment were not determined and the actual dose received might have varied from the dose shown on the pre-treatment plans. Second, it was di cult to determine whether the toxicities measured were the result of RT, chemotherapy or both, or the result of the existing cancer.
In conclusion, for NPC patients treated by chemo-SIB-IMRT, we observed GTV was signi cantly associated with mean or maximal dose in most OARs concerned, and patients with higher GTV presented with more severe acute and late toxicities and more symptomatic problems as measured by QoL scales.

Consent for publication: not applicable
Availability of data and materials: The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
Availability of data and materials: The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.