Parotid-sparing intensity modulated versus conventional radiotherapy in head and neck cancer (PARSPORT): a phase 3 multicentre randomised controlled trial

Summary Background Xerostomia is the most common late side-effect of radiotherapy to the head and neck. Compared with conventional radiotherapy, intensity-modulated radiotherapy (IMRT) can reduce irradiation of the parotid glands. We assessed the hypothesis that parotid-sparing IMRT reduces the incidence of severe xerostomia. Methods We undertook a randomised controlled trial between Jan 21, 2003, and Dec 7, 2007, that compared conventional radiotherapy (control) with parotid-sparing IMRT. We randomly assigned patients with histologically confirmed pharyngeal squamous-cell carcinoma (T1–4, N0–3, M0) at six UK radiotherapy centres between the two radiotherapy techniques (1:1 ratio). A dose of 60 or 65 Gy was prescribed in 30 daily fractions given Monday to Friday. Treatment was not masked. Randomisation was by computer-generated permuted blocks and was stratified by centre and tumour site. Our primary endpoint was the proportion of patients with grade 2 or worse xerostomia at 12 months, as assessed by the Late Effects of Normal Tissue (LENT SOMA) scale. Analyses were done on an intention-to-treat basis, with all patients who had assessments included. Long-term follow-up of patients is ongoing. This study is registered with the International Standard Randomised Controlled Trial register, number ISRCTN48243537. Findings 47 patients were assigned to each treatment arm. Median follow-up was 44·0 months (IQR 30·0–59·7). Six patients from each group died before 12 months and seven patients from the conventional radiotherapy and two from the IMRT group were not assessed at 12 months. At 12 months xerostomia side-effects were reported in 73 of 82 alive patients; grade 2 or worse xerostomia at 12 months was significantly lower in the IMRT group than in the conventional radiotherapy group (25 [74%; 95% CI 56–87] of 34 patients given conventional radiotherapy vs 15 [38%; 23–55] of 39 given IMRT, p=0·0027). The only recorded acute adverse event of grade 2 or worse that differed significantly between the treatment groups was fatigue, which was more prevalent in the IMRT group (18 [41%; 99% CI 23–61] of 44 patients given conventional radiotherapy vs 35 [74%; 55–89] of 47 given IMRT, p=0·0015). At 24 months, grade 2 or worse xerostomia was significantly less common with IMRT than with conventional radiotherapy (20 [83%; 95% CI 63–95] of 24 patients given conventional radiotherapy vs nine [29%; 14–48] of 31 given IMRT; p<0·0001). At 12 and 24 months, significant benefits were seen in recovery of saliva secretion with IMRT compared with conventional radiotherapy, as were clinically significant improvements in dry-mouth-specific and global quality of life scores. At 24 months, no significant differences were seen between randomised groups in non-xerostomia late toxicities, locoregional control, or overall survival. Interpretation Sparing the parotid glands with IMRT significantly reduces the incidence of xerostomia and leads to recovery of saliva secretion and improvements in associated quality of life, and thus strongly supports a role for IMRT in squamous-cell carcinoma of the head and neck. Funding Cancer Research UK (CRUK/03/005).


Introduction
Radiotherapy is the main non-surgical treatment for squamous-cell carcinoma of the head and neck (HNSCC). 1 High rates of local tumour control can be achieved with 5-year survival greater than 80% for stage 1 and 2 and 60-70% for stage 3 and 4 tumours; 2 however, long-term late sequelae of radiotherapy are highly prevalent and have severe adverse eff ects on quality of life (QoL). 3,4 Radiation-induced xerostomia is the most commonly reported late side-eff ect of radiotherapy to the head and neck. Lack of saliva aff ects speech and swallowing and can accelerate dental caries. 5 Intensity-modulated radiotherapy (IMRT) is a conformal radiotherapy technique that can spare the major salivary glands. Small phase 2 studies have shown that a reduction in radiation to the parotid glands (to 24-26 Gy) through parotid-sparing IMRT aids recovery of saliva fl ow. [6][7][8] We report results of the fi rst multicentre randomised controlled trial to assess parotid-sparing IMRT in patients with HNSCC.

Participants
We undertook a phase 3 randomised controlled trial at six UK radiotherapy centres (recruitment between Jan 21, 2003, andDec 7, 2007). Eligible patients had histologically confi rmed HNSCC that arose from the oropharynx or hypopharynx and were to be treated by radiotherapy either primarily or postoperatively without concomitant chemotherapy. These patients were at high risk of radiationinduced xerostomia-ie, if they were treated with conventional radiotherapy the estimated mean dose to both parotid glands would be greater than 24 Gy. Patients had WHO performance status 0 or 1 and any stage of disease except M1. Patients were required to attend regular follow-up, undergo salivary fl ow measurements, and complete self-assessed QoL questionnaires.
Exclusion criteria included previous head or neck radiotherapy; previous malignancy except non-melanoma skin cancer; pre-existing salivary gland disease; tumour involvement of the parotid glands; or previous or concurrent illness that would compromise completion of treatment or follow-up. Prophylactic amifostine or pilocarpine was not permitted. Patients who had received neoadjuvant chemotherapy were eligible.
All patients provided written informed consent. PARSPORT (CRUK/03/005) was approved by the national South-West Multicentre Research Ethics Committee (MREC 03/6/79) and the local ethics committees of all participating centres. Our trial was sponsored by the Royal Marsden NHS Foundation Trust and undertaken in accordance with the principles of Good Clinical Practice.

Randomisation and masking
Patients were randomly assigned in a 1:1 ratio to parotidsparing IMRT or conventional radiotherapy (control). Independent randomisation was via telephone to the Clinical Trials and Statistics Unit at the Institute of Cancer Research (ICR-CTSU). Computer-generated random permuted blocks were used; stratifi cation was by treatment centre and tumour site. Treatment allocation was not masked; however, the patient was not informed of the treatment until they had completed the baseline QoL questionnaires.

Procedures
Staging investigations included examination under anaesthetic, tumour biopsy, diagnostic CT or MRI of head and neck, chest radiograph, full blood count, and biochemistry. In postoperative patients, histology reports that documented the extent of surgical resection were required.
The protocol for target volume defi nition and treatment planning has been previously described. 9 All patients underwent CT-planned radiotherapy with either threedimensional conformal radiotherapy with parallel opposed lateral fi elds (conventional radiotherapy) or parotid-sparing IMRT. The conventional radiotherapy regimen was the national standard of care in the UK and most other countries at the time our trial was designed. In both treatment groups, the primary tumour and involved lymph nodes were treated with 65 Gy in 30 daily fractions given Monday to Friday. 60 Gy in 30 fractions was delivered to postoperative patients unless there was macroscopic residual disease in which case 65 Gy in 30 fractions was given. Nodal groups at risk of harbouring occult metastatic disease received a biologically equivalent dose of either 50 Gy in 25 daily fractions (conventional radiotherapy) or 54 Gy in 30 fractions (IMRT). For IMRT patients a planning constraint of less than 24 Gy to the whole contralateral parotid gland was used. 9,10 For quality assurance, plans were assessed from all centres for protocol compliance and dosimetric consistency. 10,11 Acute side-eff ects were graded weekly with National Cancer Institute Common Toxicity Criteria (version 3) 12 during radiotherapy and until 8 weeks after treatment. Late radiotherapy side-eff ects were assessed with the Late Eff ects of Normal Tissues Subjective-Objective Management Analytic (LENT SOMA) 13,14 and the Radiation Therapy Oncology Group (RTOG) 15  measurements were done before radiotherapy, at week 4 of radiotherapy, and at 2 weeks, 3, 6, 12, 18, and 24 months after radiotherapy. Unstimulated and sodium-citratestimulated parotid saliva from each parotid duct orifi ce and fl oor of mouth saliva were collected by standard methods. 7,8 After treatment, clinical follow-up was monthly in year 1, every 8 weeks in year 2, then every 3-6 months until the end of year 5. Assessments were not blinded to treatment allocation. Patient-reported QoL was collected with questionnaire booklets that contained the European Organization for Research and Treatment of Cancer (EORTC) QLQC30 quality-of-life instrument 16 (which measures generic cancer-related QoL), the associated head and neck specifi c module HN35, 17 and the modifi ed xerostomia questionnaire. 8 Patients completed the baseline booklet in the clinic before randomisation. Follow-up booklets were sent directly to the patients' homes at 2 weeks, 3, 6, 12, 18, and 24 months after radiotherapy.
Our primary objective was to assess late side-eff ects. Our primary endpoint, agreed in discussion with the independent trial steering committee, was the proportion of patients with xerostomia of grade 2 or worse by the LENT SOMA subjective side-eff ect scale 1 year after treatment. This endpoint was chosen because it assesses an abnormal symptom (ie, "partial but persistent or complete dryness" or worse) measured by a reliable and sensitive method for scoring late side-eff ects in HNSCC. 18 We decided on 12 months a priori as a clinically appropriate time at which to make a valid assessment of late eff ects.
Secondary endpoints were the proportion of patients with any measurable salivary fl ow after radiotherapy, acute and other late radiation side-eff ects, QoL that included xerostomia-related QoL as measured by the modifi ed xerostomia questionnaire, locoregional progression-free survival (PFS), and overall survival. We defi ned locoregional PFS as time from randomisation to locoregional recurrence or progressive disease as defi ned by Response Evaluation Criteria in Solid Tumours. 19 We defi ned overall survival as time from randomisation to death from any cause.

Statistical analysis
Phase 2 studies had reported reduction in salivary fl ow rates of 90% at 1-3 months compared with preradiotherapy rates with conventional therapy and of 40% with IMRT. 7,[20][21][22] If we assume a 1-year xerostomia rate of 90% in the conventional radiotherapy group 23 a sample size of 84 patients is needed to detect a 30% absolute diff erence in LENT SOMA of grade 2 or worse xerostomia between the study groups (90% power, 5% two-sided signifi cance). In March, 2007, the independent data monitoring committee and the trial steering committee approved an increase in the target sample size to 84 evaluable patients (ie, alive 1 year after the end of radiotherapy) that was anticipated to be achievable with 100 randomly assigned patients. In December, 2007, both committees approved closure of recruitment after 94 patients had been randomly assigned to the study groups with the expectation that this would provide suffi cient evaluable patients to allow robust statistical analysis. Our trial was not powered to reliably assess small diff erences in locoregional PFS or overall survival, although these are reported for completeness.
Our analysis was done on an intention-to-treat basis, with all patients who had a 12-month xerostomia assessment included. We compared the proportion of patients with grade 2 or worse xerostomia between groups with a χ² test. We assessed the sensitivity of results by repeating analyses of the primary endpoint with patients Data are mean (SD; range), n (%), or median (IQR; n). IMRT=intensity-modulated radiotherapy. *American Joint Committee on Cancer-groupings based on TNM staging data collected. †Mann-Whitney test p<0·0001. Table 1: Baseline characteristics and treatment details censored 1 month before any disease recurrence, in case recurrence could adversely aff ect salivary fl ow, and by excluding patients whose side-eff ect assessment was not within 2 months either side of its expected date. We have not presented these sensitivity analyses because they gave similar results to the main analysis. Odds of grade 2 or worse xerostomia at 12 and 24 months were calculated with a logistic-regression model. We present unadjusted odds ratios (ORs) and ORs adjusted for tumour site (oropharynx or hypopharynx), stage of disease (1 and 2 or 3 and 4), and radiotherapy indication (radical or postoperative). All other analyses are unadjusted. We compared the proportions of patients with any measurable saliva fl ow and proportions ever reporting grade 2 or worse acute and late side-eff ects between treatment groups with Fisher's exact tests. For LENT SOMA scales, we used the maximum of the subjective, objective, management, and analytic component scores. We calculated CIs for diff erences in proportion between groups with a normal approximation. To make some adjustment for multiple testing we used a signifi cance level of 1% for all secondary side-eff ects, sialometry, and QoL endpoints and accordingly we provide 99% CIs. Acute and late side-eff ects in our report were those where sideeff ects of grade 2 or worse were experienced by at least 20% of patients in either group or those where proportions were signifi cantly diff erent between treatment groups.
We calculated QoL scores with standard algorithms with a higher score suggesting poorer QoL on all scales except EORTC global health status, where a higher score suggests better QoL. 24 We deemed diff erences in EORTC QoL scores of 10 points or more clinically signifi cant in line with EORTC guidelines. 25 The primary QoL analysis included all completed questionnaires. We did a sensitivity analysis after censoring at 1 month before disease recurrence or progressive disease. We compared mean changes in EORTC QoL and xerostomia questionnaire item scores from baseline between groups by two-sample t tests.
We used generalised estimating equations (GEE), adjusting for the correlations in multiple measurements from the same patient (with an exchangeable correlation matrix) to account for the longitudinal nature of the xerostomia and QoL data. A pragmatic approach to modelling was taken, with treatment-by-time interaction terms included if they were identifi ed in advance as clinically relevant or they were statistically signifi cant. A GEE logistic regression model was fi tted with xerostomia (grades 0 and 1 vs grades 2-4) as the response and allocated treatment, days since the completion of radiotherapy, and the interaction between the two as covariates. QoL GEE models also included terms for baseline score for the item of interest.
For survival-related endpoints, alive and disease-free patients were censored at date of last follow-up. We compared treatment groups with the log-rank test. Hazard ratios (HRs) with 95% CIs were obtained from Cox proportional hazards regression models with HRs of less than one favouring IMRT. The proportionality assumption of the Cox model held when tested with Schoenfeld residuals. Our analyses were based on a database snapshot frozen on May 14, 2010, and were done in STATA version 10. ICR-CTSU had overall responsibility for trial coordination. Data collation, central statistical monitoring of data, and all interim and fi nal analyses were performed at ICR-CTSU. The trial management group was responsible for the day to day running of the trial. The trial was overseen by an independent trial steering committee. The independent data monitoring committee regularly reviewed emerging safety and effi cacy data in confi dence. This study is registered as an International Standard Randomised Controlled Trial, number ISRCTN48243537.

Role of the funding source
The funding source provided peer-reviewed approval for the trial but had no other role in study design, collection, analysis, interpretation of data, or writing of the report.

Conventional radiotherapy IMRT
We recorded baseline sialometry in 80 patients, all of whom had measurable salivary fl ow. At 12 months unstimulated saliva fl ow from the contralateral parotid gland was noted in 16 (47%) of 34 patients in the IMRT group compared with none of 25 in the conventional radiotherapy group (p<0·0001). Corresponding data at 24 months were seven (44%) of 16 in the IMRT group versus none of 15 in the conventional radiotherapy group (p=0·0068). Signifi cant diff erences were also noted in stimulated saliva fl ow from the contralateral parotid at 12 months (p<0·0001). No signifi cant diff erences between the random assigned groups were seen in proportions with unstimulated or stimulated fl ow from either the ipsilateral parotid or fl oor of mouth. Strong concordance was noted between measurable contralateral saliva fl ow and grade 2 or worse xerostomia (table 3).
The xerostomia questionnaire was only completed by 39 patients at baseline and 12 months and by 33 patients at baseline and 24 months (compared with 73 reporting the primary endpoint at 12 months and 55 at 24 months). In both treatment groups all eight xerostomia questionnaire items were signifi cantly worse at 12 and 24 months than at baseline and although the changes were smaller in the IMRT group, no statistically signifi cant diff erences between group changes were noted (webappendix p 1).
Overall, there were seven locoregional recurrences in the conventional radiotherapy group: fi ve in the highdose volume and two in both the high-dose volume and electively irradiated neck. In the IMRT group there were 12 locoregional recurrences: 11 in the high-dose volume and one in the electively irradiated neck. No patients had a recurrence in the spared parotid tissue. 2-year locoregional PFS was 80% (95% CI 65 to 90) in the conventional radiotherapy group and 78% (62 to 87) in the IMRT group (absolute diff erence 3%, 95% CI -15 to 20; HR 1·53, 95% CI 0·63 to 3·70; log-rank p=0·34; fi gure 4). 32 deaths have been reported so far (18 in the conventional radiotherapy group vs 14 in the IMRT group; HR for overall survival 0·68, 95% CI 0·34 to 1·37). Of these deaths, 20 were due to head and neck cancer (ten in the conventional radiotherapy group vs ten in the IMRT Time from randomisation (months) Proportion with no locoregional recurrence or progressive disease (%) 30 36

Discussion
Our trial showed a signifi cant reduction of radiationinduced xerostomia for patients treated with IMRT compared with conventional radiotherapy by use of both LENT SOMA and RTOG scales. Furthermore, we showed recovery of saliva fl ow by quantitative measurements, and improvements on QoL measures associated with xerostomia. To our knowledge our trial is the fi rst to show that parotid-sparing IMRT reduces xerostomia in HNSCC (panel). A consistently higher QLQC30 Global and HN35 dry mouth QoL score was reported in patients who received IMRT; between group diff erences at 24 months were clinically but not statistically signifi cant. Xerostomia questionnaire results showed changes in favour of IMRT in all eight questions but these diff erences were not large enough to reach statistical signifi cance, probably because of the small number of patients that completed this questionnaire. Although an association between measurable saliva fl ow and presence of grade 2 or worse xerostomia was recorded, there was not perfect concordance. We postulate that this could be because of diff erences in patient perception of the xerostomia symptom or because of other factors such as submandibular gland or oral cavity dose or comorbidity.
Detailed analyses of the distribution of dose to the salivary tissue including parotid glands and other minor salivary glands, and its correlation with clinical outcomes are ongoing. Initial results suggest that there is no correlation between submandibular gland dose and xerostomia. A limitation of our trial was that it was not possible to mask the treatments from patients or clinicians because of diff erences in treatment delivery. However, results that relate to multiple secondary endpoints support the primary analysis and the size of the observed eff ect is unlikely to be due entirely to assessment or reporting bias. After our trial was designed, several small nonrandomised studies [29][30][31][32][33][34][35] and one case-control study 36 of parotid-sparing IMRT have been published with a range of endpoints including saliva fl ow rate, patient-reported symptoms, and QoL. These studies reported apparent improvements for IMRT over conventional radiotherapy. Two small single-institution randomised phase 3 trials of IMRT in nasopharyngeal cancer have also reported benefi ts of IMRT over conventional radiotherapy. Pow and colleagues 37 reported an increase in stimulated whole saliva fl ow rate in patients receiving IMRT in a randomised trial of 51 patients with early-stage nasopharynx cancer. QoL was assessed with EORTC QLQC30, HN35, and the SF36 health survey and although QoL scores for some domains were better for IMRT patients, no improvements in patient-reported dry mouth symptoms on the HN35 questionnaire were noted. Kam and colleagues 38 reported a reduction in observer-rated severe xerostomia (RTOG grade 2 or worse) with IMRT (39% vs 82%; p=0·001) in 60 patients with early-stage nasopharyngeal cancer. The results of the PARSPORT trial are thus likely to be generalisable to all head and neck tumours for which conventional radiotherapy is used.
In our study, fewer cases of acute dermatitis were recorded in patients treated with IMRT than in those treated with conventional radiotherapy, although diff erences were not statistically signifi cant at the 1% level, probably because of reduced dose to skin. The proportions of patients that reported grade 2 or worse acute xerostomia and grade 2 or worse salivary gland changes also showed reductions, albeit not statistically signifi cant (table 2). Late xerostomia side-eff ects thus accord with acute side-eff ects; this suggests that late radiation-induced xerostomia is a consequential eff ect. We did not attempt to spare the submandibular or mucosal minor salivary glands within the planning target volume in our trial. It is possible that further reductions in severe xerostomia can be achieved by sparing these tissues, but this might risk underdosing crucial target tissues. Unexpectedly, acute fatigue was greater in patients treated with IMRT, which could be due to the greater radiation dose to non-tumour tissues.
In an unplanned dosimetry review in a subset of patients, mean radiation doses to the posterior fossa were 20-30 Gy in the patients treated with IMRT compared with about 6 Gy in patients treated with conventional Panel: Research in context Systematic review Intensity modulated radiotherapy (IMRT) allows focused radiation delivery to tumours. In patients with head and neck cancer it has been used to reduce the irradiation of salivary tissue to prevent radiation-induced xerostomia. Before the design of our randomised trial, a few small single centre experiences had been published and a review of the published work on IMRT had been done. 26 No randomised trials were identifi ed. During the recruitment period of the PARSPORT trial two smaller randomised trials were reported in nasopharyngeal cancer from centres in Asia, and several other single institutional experiences were reported. 27,28 Interpretation Our trial is the largest randomised trial of IMRT in head and neck cancer, and the only trial addressing squamous-cell carcinoma, the predominant form seen worldwide. Our trial shows that IMRT reduces patient-reported xerostomia, allows recovery of salivary fl ow, and improves quality of life after treatment compared with conventional radiotherapy. radiotherapy, which could account for the recorded diff erence in acute radiation fatigue. Late fatigue data were not collected because lethargy is not a recognised long-term side-eff ect of radiotherapy. There was no signifi cant association between the giving of neoadjuvant chemotherapy and either acute fatigue or xerostomia (data not shown). The addition of concurrent chemotherapy to altered fractionation radiotherapy remains experimental and was not used in our study. Further research is needed to establish the eff ect of concurrent chemotherapy on xerostomia. Apart from salivary gland changes and radiation-induced xerostomia, other late side-eff ects of conventional radiotherapy were not altered by IMRT.
Our trial was too small to detect small diff erences in, or conclude non-inferiority of, locoregional PFS or overall survival. Although patients continue to be followed up for long-term survival, to show non-inferiority in overall survival to no more than 5% at 2 years (80% power, onesided 5% signifi cance) would need a randomised controlled trial of more than 900 patients. In this, and other, head and neck IMRT studies most tumour recurrences happen within the high-dose volume. Recurrences have not been noted in the spared parotid tissue in patients treated with IMRT or surgery, 21,39 suggesting that a large study to show non-inferiority in this tumour type is probably both impractical and inappropriate. Our trial has shown a clinically and statistically signifi cant reduction in xerostomia, improved salivary fl ow, and improved QoL, and thus strongly supports a role for IMRT in HNSCC.

Contributors
CMN and EH were responsible for the trial design, trial management, data interpretation, and writing of the report. CMN was chief investigator of the trial and contributed to trial recruitment. EH oversaw all statistical analyses. JPM did the main analyses and contributed to data interpretation and writing of the report. RPA'H contributed statistical support, trial management, data interpretation, and writing of the report. MAS, ME, and MT were responsible for the trial coordination and data collection and contributed to data interpretation and writing of the report. CC and EAM were responsible for the design and conduct of the quality assurance programme and contributed to the trial management, data interpretation, and writing of the report. KJH, KN, CS, SJJ, FA, and BKY all contributed to trial recruitment, trial management, data interpretation, and writing of the report. TGU, SAB, and ABM were clinical coordinators for the trial, contributed to trial management, data collection, data interpretation, and writing of the report. All authors reviewed and approved the fi nal version of the paper.

Confl icts of interest
The authors declared no confl icts of interest.