Recurrence patterns of locally advanced head and neck squamous cell carcinoma after 3D conformal (chemo)-radiotherapy

Head-and-neck squamous cell carcinoma (HNSCC) is the sixth most common malignancy worldwide, responsible for approximately half a million new cases every year [1]. Approximately 60% of patients with HNSCC present with locally advanced, but non-metastatic disease (stage-III or IVA/B) at diagnosis. Based upon organ preservation studies [2, 3], radiotherapy is an accepted alternative to surgery. The results of radical radiotherapy regimens have been further improved by the use of induction chemotherapy [4], concurrent chemoradiotherapy [4], and concurrent epidermal growth factor inhibitors [5]. In parallel, radiotherapy techniques have developed rapidly; conformal radiotherapy (CRT), accelerated schedules [6] and intensity modulated radiotherapy (IMRT) [7] have been used to improve the therapeutic ratio between tumour control and normal tissue toxicity.

Historically, locoregional failure has been the predominant pattern of relapse following non-surgical treatment [8]. With the rapid advancement of non-surgical treatment strategies, it is critical to document the pattern of treatment failure in relation to, the radiotherapy dose distributions. These data are required to guide whether future improvements should be focused on improving local and/or regional control or, on reducing the development of distant metastases (DM). The former may involve modifications in target volume definition, delivery technique, or dose escalation. However, if DM is an increasing problem, consideration could be given to prioritizing the delivery of systemically active therapy. Therefore, the aim of this retrospective study is to determine recurrence patterns among HNSCC patients treated with radical three-dimensional (3D) CRT with or without chemotherapy, and to correlate the sites of local-regional recurrence (LRR) to previously treated radiotherapy fields and dose distribution.

After institutional review board approval, we retrospectively reviewed the medical records of patients with locally advanced stage III/IV HNSCC treated with 3D-CRT with curative intent at the Yorkshire Cancer Centre between January 2004 and December 2005. Patients with nasopharynx carcinomas were excluded. Patients who had undergone initial therapeutic surgery to the primary tumour site were excluded.

A critical aspect of the management of HNSCC is to understand the patterns of treatment failure, in order to guide future attempts to optimize radiotherapy planning and improve the therapeutic ratio. If treatment failures are predominantly distant, intensification of systemic therapy may be needed to improve outcomes. By contrast, if failures are loco-regional, this may emphasize the need to identify patients with a radiation resistant tumour subpopulation and underlie the rationale for studies of dose escalation to the highest risk regions.

In this study, 82.8% of the patients achieved a complete tumour response 4 months after completion of therapy. The 4 month timepoint used here is intended to allow adequate time for the response to radiotherapy. Pacagnelli et al. showed that an 8 week response assessment is too early, with more complete responses being seen at 8 months than 8 weeks post-treatment [10]. We have demonstrated that 3-year local and regional control was high in these patients with rates of 86.8% and 89.5%, respectively. In 125 complete responders, there were 20 LRR. Amongst the 117 patients who achieved a CR to (chemo)-radiotherapy without salvage surgery, there were only 14 LRR. Six of the 8 patients who had undergone salvage surgery after a failure to achieve a complete response had a locoregional recurrence and 5 of these cases failure was infield. One third of cases of LRR were associated with the development of synchronous DM. DM in the absence of LRR was uncommon, occurring in only 9 of 125 patients.

Analysis of patterns of failure in relation to the dosimetry of the radiotherapy plan, demonstrated that 76% of LRR occurred within the PTV (12 of 14 local recurrences and 7 of 11 regional recurrences). There were no cases of isolated recurrences within the volume treated by the lower dose prophylactic anterior neck fields. 12 of the patients with LRR had been treated with hypofractionated 55Gy in 20 fractions. Based upon this data, and national guidance [11], hypofractionated treatment for locally advanced HNSCC has been abandoned in our centre. Five patients had an out-of-field LRR. Four of them were treated with radical radiotherapy only. Two were found to have simultaneous out-of-field recurrences and DM. Out-of-field relapse was mainly associated with deliberate omissions of elective lymph node regions at the time of radiotherapy planning. This omission of elective nodal regions reflects the nature of the patient population with locally advanced HNSCC, in which treatment has to be individualized on the basis of age, comorbidity, treated volume, disease extent. A marginal recurrence at the edge of the PTV was seen in only one patient.

Uncertainty in the demarcation of the target volumes is one of the major limitations to improving outcomes of HNSCC with radiotherapy. The extent of gross tumour determined clinically and using the planning CT scans is not straightforward. A major difficulty is the delineation of lymph node groups at risk of subclinical disease. It has been shown that there is significant variation in nodal volumes when defined by different radiation oncology specialists [12, 13]. Our data, demonstrating that LRR are predominantly occurring in-field, is reassuring in terms of the quality of target volume definition and elective nodal targets.

There is limited published literature regarding patterns of disease recurrence after radiotherapy for HNSCC. This has been best documented in several IMRT series. Recently the IMRT technique has replaced conformal 3D-planning in many centres. It is important to document the recurrence patterns of IMRT series which may carry a higher risk of geographical tumor misses and compare with 3-D conformal techniques. Most IMRT series have focused on decreased rates of xerostomia in HNSCC with IMRT or investigated the pattern of failure in patients treated with surgery and/or chemotherapy in addition to IMRT [14‐16]. In one of the larger series, Chao et al. analyzed 126 head and neck cancer patients treated with IMRT delivered with radical intent without surgery in 41% of patients, and post-operatively in 59% of patients [17]. After a median follow up of 26 months, 17 LRR were noted, of which 9 were in-field, 3 were marginal failures and 5 were outside of the IMRT field. Eisbruch et al. reported results of 133 patients treated with parotid-sparing 3D-CRT or IMRT and of the 21 LRR, 17 were in-field and 4 were marginal [18]. Studer et al. reported a 80% and 87% local and regional control rate, respectively and 95% of failures occurred in-field following IMRT [19]. Therefore, the majority of failures are in-field in reported IMRT series. However, given the heterogeneity of disease sites, stage, treatment types and the number of patients in these studies, it is not possible to accurately compare the rate of recurrence in our patients to patients treated with IMRT. In addition, longer follow-up is needed to validate IMRT findings, particularly with regards to LRC and late complications.

The use of 3D-CRT radiotherapy allowing increased GTV doses within normal tissue tolerances, and the increasing utilization of combined chemotherapy with radiotherapy, is likely to lead increasing rates of LRC. This may alter failure patterns on HNSCC. Indeed, some studies have shown that DM have become an increasingly important site of recurrence and mortality [20, 21]. In our series, the 3-year DMFS rate was 87.3%. By contrast, 26 of 151 patients never achieved LRC, and a further 20 of 125 complete responders subsequently developed LRR. These data highlight the importance of continued efforts to improve therapy to enhance LRC; the large majority of treatment failures remain locoregional failures.

HNSCC are heterogenous in their aetiology and behaviour. HPV DNA has been found in approximately 25% of HNSCC and HPV-associated tumours tend to arise in oropharynx but not in the larynx [22]. In addition, HPV-positive tumours are associated with a better prognosis [23]. This may explain better treatment results in oropharyngeal tumours in our study. In addition, patients with tumour stage T3-T4 had a higher risk of PFS and CSS both with univariate and multivariate analysis. However, we did not observe a statistically significant effect of N-stage. Hence, our series suggests that T-stage is the predominant prognostic factor. Accurate stratification of patients in terms of prognosis is likely to be important in identifying subgroups that may benefit from an intensification of treatment.

Although this is a series in a single cancer centre, under the supervision of a dedicated head and neck oncology team, the treatment related death rate was 4%. The most common cause of death was aspiration pneumonia. However, co-morbidities, limited performance status, poor social support and heavy alchohol consumption may have had an impact upon toxicity (Table 6). Based upon these data, we now routinely offer prophylactic gastrostomy-tube placement for patients receiving concurrent chemoradiotherapy or have heavy alcohol consumption, and poor social support. We believe that better patient selection and provision of intensive, experienced, multidisciplinary support during treatment and beyond, decreases the mortality rate.

In summary, this series demonstrates that conformal (chemo-)radiotherapy offers high rates of LRC and OS. Systemic targeting may improve outcomes, however the majority of LRR after a CR to treatment occur within the PTV and isolated distant metastatic recurrence was uncommon. This may relate to intrinsic radioresistance or factors such as tumour hypoxia. These data provide a clear rationale for efforts aimed at improving locoregional tumour control. Useful approaches may include induction chemotherapy regimens, biological therapies, radiosensitisers, altered fractionation, and dose escalation. It is likely that the future will involve the identification of predictive markers of treatment response, identifying patients likely to fail locoregionally. This may allow the selection of an individually tailored treatment regimen.