Local relapse in head and neck cancer is a very difficult therapeutic situation. Chemotherapy rarely produces durable tumour control; therefore surgery as a potentially curative treatment option is the mainstay of salvage local therapy. For adenoid cystic carcinoma, the situation becomes even more complex: skull base invasion and perineural spread often prevent complete resections in the primary situation. Anatomical sites and relative radioresistance of the disease mandate initial high radiation doses for tumour control. In case of local relapse, treatment options are limited. First treatment of choice is salvage surgery. If this is not possible, patients’ options are limited: even highly aggressive chemotherapy regimens for adenoid cystic carcinoma achieve only objective response rates up to 30% , new agents such as EGFR- or tyrosine-kinase inhibitors mostly stabilize disease for some time [, , , ]. Only one tyrosine kinase inhibitor has been shown to produce objective responses .
Re-irradiation has rarely been used in the past for fear of considerable early and late toxicity; in addition, recurrent tumours could be shown to be more radiation-resistant than their initial clone . Neutron and charged particle therapy produced encouraging local control rates albeit at considerable side-effects [, ]. There is emerging evidence that re-irradiation can lead to long-term local control in a selected subset of patients, [, , , ] but local control remains strongly dependent on re-irradiation dose [, , ]. With the advent of modern radiotherapy techniques such as stereotactic radiotherapy and IMRT, proportion of severe treatment-related side-effects can be reduced [, , , ] and modern techniques for re-irradiation are increasingly offered to patients with locally recurrent head and neck cancer. While substantial data including prospective phase I and II trials have been reported for squamous cell head and neck cancer (SCCHN), there are little data on retreatment of malignant salivary gland tumours. Application of tumouricidal doses remains a challenge though due to the proximity of critical structures in the head and neck. Charged particle therapy in active beam application produces very sharp dose gradients  and hence may improve outcome in this desperate situation. We present our experience on re-irradiation of adenoid cystic carcinoma of the head and neck with scanned carbon ion beams.
Treatment decision in all patients was based on interdisciplinary consensus. Treatment-related toxicity was prospectively collected, patient data were retrospectively analysed.
Patients were immobilized using individual fixation devices (scotch cast or thermoplastic head masks). As a standard, treatment planning was carried out on 3 mm CT and contrast-enhanced MRI scans for 3D image correlation.
Carbon ion therapy was carried out at the Heidelberg Ion Beam Therapy Centre (HIT) in active beam application (raster-scanning method ) in 3 GyE per fraction (exception: 1 patient received 3.5 Gy/fraction) and 5–6 fractions per week under daily image guidance with orthogonal X-rays and position correction in six degrees of freedom .
Re-irradiation target volume included only the visible local relapse with a small safety margin (ca. 2 mm), no elective nodal irradiation was carried out. There were 4 exceptions in this cohort receiving combination treatment with IMRT and carbon ion boost. These patient’s prior radiotherapy interval was either very long or the prior radiotherapy dose was negligible in the area of relapse. In these cases, C12 volume included the visible tumour plus small safety margin, the IMRT volume included visible tumour, area at risk of microscopic spread and locoregional nodal levels.
In all cases, cumulative dose to the brain stem and spinal cord was kept below 60 Gy and 50 Gy respectively assuming around 50% recovery of the CNS . In the cases where optic nerves were involved in the tumour process, high probability of loss of vision was discussed with the patients prior to treatment start. Dose prescription was highly individual in each patient’s case taking into account time to prior radiotherapy as well as prior radiotherapy dose.
Patients received regular follow-ups including MRI scans and clinical exams 6–8 weeks post completion of radiotherapy, 3, 6, and 12 months thereafter. Regular follow-ups with their attending ENT or maxillofacial specialist were encouraged.
Response following re-irradiation was analysed using RECIST criteria , acute and late toxicity was evaluated according to NCI CTCAE v.4.03. Control and survival rates were estimated using Kaplan–Meier analysis  of the Addinsoft XLSTAT Life® package. Locoregional control was calculated from the first day of treatment to occurrence of locoregional failure, progression-free survival was measured from the first day of treatment to occurrence of locoregional failure, distant failure, or death.
Fifty-two patients with adenoid cystic carcinoma received re-irradiation using carbon ion therapy for local relapse between 04/2010 and 05/2013.
Median age of these patients was 55 years [35–78 years], median follow-up to date is 14 months [1–39 months], 13 patients have deceased thus far. Seven out of 52 patients underwent surgery for local relapse, 45 patients (86.5%) had macroscopically visible tumour prior to re-irradiation. Tumour stages were mostly advanced (T4: 76.9%) while nodal and distant metastases (N+: 6 pts/11.5%; M1: 15 pts/28.8%) were less common. Tumours mostly originated from the paranasal sinuses (36.5%), the base of skull (21.2%), or the parotid (19.2%). Forty-eight patients (92.3%) received carbon ions only, 4 patients (7.7%) combined treatment with IMRT plus carbon ion boost (Table 1: patient characteristics and radiotherapy).
|Base of skull/intracranial||11||21.2|
|Macroscopic tumour at re-RT||45||86.5|
|IMRT + C12-boost||4||7.7|
|Median (Gy/GyE)||Range (Gy/GyE)|
|Cumulative life-time dose|
|Interval between RT courses||61 mo||9–620 mo|
|CTV (C12)||93 ml||6–618 ml|
|CTV (IMRT); 4 pts only!||334 ml||211–344 ml|
Patients received a median re-irradiation dose of 51 GyE [36–74 GyE]. Using an alpha/beta of around 2 for adenoid cystic carcinomas, this corresponds to a median biologically equivalent dose (BED) of 63 Gy [45–82 Gy]. Target volumes were 93 ml (median) for carbon ions [range: 9–618 ml] and 334 ml (median) for IMRT [211–344 ml]. Median prior radiotherapy dose of treated patients was 66 Gy [20–115 Gy] corresponding to a BED of 66 Gy [20–133 Gy] as 14 patients had undergone prior carbon ion therapy. Interval between prior radiotherapy and re-irradiation ranged between 9 and 620 months (median: 61 months) (Table 1: patient characteristics and radiotherapy).
We observed no grade III acute toxicity; none of the reported patients had treatment interruptions. There was only °II mucositis (11.5%), dermatitis (9.6%) and xerostomia (5.8%). Eight patients had trismus, which was a pre-existing condition or due to tumour process. Observed acute toxicity quickly resolved, at first follow-up (6–8 weeks post completion of radiotherapy), only one patient still had grade II xerostomia (1.9%), grade I residual toxicity was 5.8% (xerostomia, dysphagia) and 3.8% (hyperpigmentation). Trismus had resolved in 2 patients, leaving 6 patients with residual trismus (11.5%) (Table 2: toxicity).
Higher-grade late toxicity was rare, however, we observed 8 cases of temporal lobe blood–brain-barrier changes (CNS necrosis °I: 15.4%). Two patients underwent surgery (CNS necrosis °III: 3.8%), one developed symptomatic epilepsy (1.9%). Two patients developed tissue necrosis in the nasopharynx following re-irradiation (3.8%) consequently leading to carotid artery haemorrhage (CTC°IV) (cumulative applied BED: 149 Gy and 182 Gy). Fortunately, both underwent coiling of the vessel without any consecutive symptoms. Osteoradionecrosis was observed in 3 patients (5.8%) resulting in surgical procedures in one patient (1.9%). Another patient acquired a chronic corneal ulcer (1.9%) requiring permanent ophthalmological treatment. In total, 8 patients developed serious radiation late effects (6.5%) (Table 2: toxicity).
|Acute toxicity (CTCAE v. 4.03)|
|I (pts/%)||II (pts/%)|
|Late toxicity (CTCAE v. 4.03)|
|CNS necrosis °I||8||15.4|
|CNS necrosis °III||2||3.8|
|ICA haemorrhage (°IV)||2||3.8|
|Cranial nerve palsy||1||1.9|
|Symptomatic epilepsy °I||1||1.9|
|Difficulty in concentration||1||1.9|
Objective response rates (CR/PR) were 38.5% (CR: 3.8%, PR: 34.6%; SD: 46.2%, dna: 7.7%, unknown: 5.8%) at first follow-up and 53.8% best response (CR: 5.8%, PR: 48.1%; SD: 36.5%, dna: 7.7%, unknown: 1.9%). Fig. 1 shows a 3-field carbon ion IMPT plan (54 GyE in 3 Gy per fraction) of a 50-year old patient with extensive locally recurrent adenoid cystic carcinoma 2 years after initial diagnosis of his disease. Fig. 2a depicts the initial tumour extension on treatment planning and Fig. 2b a very good partial remission 8 weeks post completion of re-irradiation with carbon ions.
Thirty-three patients relapsed: 18 patients (34.6%) developed local recurrences: 13 patients within the re-irradiation field (72.2%), 1 at the field edge (5.6%), 1 in the dose gradient towards spared structures (optic chiasm) (5.6%), and 3 patients out of field (16.7%). Fifteen patients developed distant disease progression in the lung (26.7%), bone (20%), liver, and brain (each 13.3%). Six patients developed recurrences on both local and distant sites (18.2%).
Local control at one year is 70.3% (2-year estimate: 47.4%) and overall survival at one year is 81.8% (2-year estimate: 63.3%) (Fig. 3). Median local control is 19 months in this cohort. As expected, higher T-stages show lower local control rates, however, differences do not reach statistical significance in this cohort (Supplement Fig. 1s). There is also a trend towards improved local control in adenoid cystic carcinoma with increased re-irradiation doses (Supplement Fig. 2s).
While T-stage and cumulative life-time dose and short interval between the two courses of radiotherapy showed a trend towards less favourable outcome, neither of these factors reached statistical significance (Table 3s).
Faced with the often long natural history of the disease, obtaining local control is a very important issue in adenoid cystic carcinoma of the head and neck even if cure may ultimately not be possible due to the presence of distant disease. While there is emerging data on improved local control using re-RT in recurrent or second primary head and neck cancer, radiation oncologists have so far eyed re-irradiation with caution for fear of high-grade late toxicity. In excessively pre-treated patients as in the presented cohort, there are no valid treatment alternatives once surgical salvage is no longer an option. Chemotherapy was often given in a situation of local recurrence where neither surgery nor standard radiotherapy is possible with the aim of stopping/delaying local progression or ameliorate symptoms of local tumour growth. Combination regimen such as CAP (cyclophosphamide, adriamycin, platin ) can lead to objective tumour response but outcome in terms of long-term control is dismal. Much effort has gone into improving toxicity and efficacy with new substances: while toxicity could be reduced with molecular targeted agents such as imatinib and sorafenib, these substances may lead to a prolonged disease stabilization, but reported objective response rates are only around 10–15% [, ]. Data on re-irradiation in malignant salivary gland tumours of the head and neck are scarce, however, there is emerging evidence on re-irradiation in squamous cell carcinoma of the head and neck. While toxicity in potentially curative treatment regimen was sometimes unacceptably high, new treatment techniques such as FSRT/SBRT and IMRT have improved toxicity profiles of re-irradiation dramatically.
Acute toxicity in our cohort was very mild: there was no higher-grade acute toxicity; toxicity grade II reached a maximum of under 12% (mucositis) supporting our initial findings . Unfortunately, available literature mostly concentrates on squamous cell carcinoma of the head and neck (SCCHN) and re-irradiation with photons (Supplement Table 1s). More recent publications also explore the additional use of concomitant chemoradiation in re-treatment of head and neck cancer (Supplement Table 1s). Comparison of our experience with available data is limited and difficult. Lacking data on re-irradiation of ACC with modern radiotherapy techniques, comparison of the toxicity profiles in this cohort with results of re-irradiation in squamous cell head and neck cancer in similar anatomical sites and with comparable dose levels is appropriate. Most common treatment sites in our cohort are paranasal sinus and skull base (57.7%), only few groups report on these re-treatment sites in substantial proportion [, , , , , ]. Treatment techniques have evolved over the last two decades, while techniques available to Feehan and Errington were necessarily based on 2D treatment planning [, ], Roh, Lee, and Sulman later reported on results achieved by more sophisticated treatment techniques such as IMRT and fractionated stereotactic treatment [, , ]. There is currently no data on re-irradiation with high-precision particle therapy for head and neck cancer. Information on treated volumes is rarely available, if reported, median re-treated volumes range between 25 and 50 ml, while our median treatment volume (CTV) is 93 ml with a range up to more than 600 ml. Treatment-related side effects may therefore differ substantially between available literature and data presented here. Despite all the caveats, large treatment volumes and high re-irradiation and cumulative life-time doses, we have found no grade III acute toxicity which is in contrast to available photon series with grade III toxicity rates ranging between 20% and 46% [, , , ].
Late toxicity has thus far been moderate though our median follow-up with 14 months may yet be short. We did observe 2 cases (3.8%) of °IV ICA haemorrhages following tissue necrosis in the nasopharynx after cumulative doses of 149 Gy and 182 Gy BED. Fortunately, both patients underwent successful coiling of the ICA and recovered well and without sequelae of the occlusion. One of these patients also developed a swallowing disorder (dysphagia °III) following soft tissue necrosis and has undergone PEG placement. Asymptomatic blood–brain barrier changes on MRI (CNS necrosis °I) are not uncommon in high-dose radiotherapy of reported anatomical sites . Two patients underwent surgery following diagnosis of these MRI changes, hence these had to be classified CNS necrosis grade III. One patient developed symptomatic epilepsy and is currently asymptomatic under anticonvulsants.
Carotid blow-out and subsequent fatal haemorrhage was and is one of the most feared radiation late effects but rarely seen in the primary situation. In re-irradiation however, many series have reported the incidence of vascular complications [, , , , , , , , , , ]. In a recent review including data on more than 1500 patients, McDonald et al. found an incidence of 2.6% carotid artery blow-out with 76% of these complications being fatal . Factors influencing vascular complication risks include cumulative dose, fractionation and volume of irradiation. Sparing these structures in re-irradiation may not be possible though if the tumour encases or infiltrates the vessel . While the rate of vascular complications in our cohort is higher than the pooled 2.6%, our treatment volumes and re-irradiation doses are significantly above reported series. Total observed serious late toxicity (6.5%) in comparison to other published series is still moderate and less than reported in other series maybe owing to the sharper dose-gradients achievable by scanned particle beams. Comparison with available particle data is hampered by different planning techniques in the 1980s and today; serious late effects were reported at 6–25% in these early series [, , ].
Tumour response went as high as 53.8% (CR/PR), which compares favourably with reported results including patients with other histologies (Adenoid cystic carcinoma: 27% of patients) . Saroja et al. reported complete responses of 83% in patients with adenocarcinoma, while reported local control at 2 years is only 44% in this series  compared to 47.4% (median local control: 19 months) in our heavily pre-treated patient cohort. Feehan et al. reported a similar median local control in their series albeit at higher severe late toxicity . While available re-irradiation series have reported local control rates of up to 68%  at 2 years for mainly patients with SCCHN, achieving long-term local control in adenoid cystic carcinoma has remained a challenge even in the primary situation [, ] owing to the relative radioresistance of the disease. Long-term local control in re-irradiation of adenoid cystic carcinoma can rarely be achieved in our series despite aggressive treatments. Based on the neutron experience from the late 1980s, one of the main parameters for local control in recurrent ACC is re-irradiation dose [, ]. Hence, escalation of re-irradiation dose should have resulted in higher probability of local control. Despite high re-irradiation doses even to large treatment volumes, this does not seem to be the case in our cohort as demonstrated by the cohort’s local control even despite the short follow-up (Fig. 3, Fig. 2s). However, compared to available chemotherapy data [, , ], objective treatment response (56.6%) and response duration (median local control: 19 months) using re-irradiation is still impressive, especially considering the moderate toxicity profile and treatment alternatives: with the most effective CAP regime including cisplatin already given, expected objective response rates of second or third line treatment will be less than 20% (Supplement Table 2s). Re-irradiation with carbon ions may therefore be a valuable tool in palliative treatment of ACC in order to avoid systemic treatment as long as possible and prevent local complications due to aggressive local tumour growth.
Despite high re-irradiation doses, the most common site of relapse is still within the re-treatment field, which is consistent with observations by other groups [, , , , , , , , , ] and further supports our target volume definition concept. In contrast to Duprez, Lee, Haraf, Salama, Hoebers, De Crevoisier we have found no significant correlation of either local control or overall survival with re-irradiation dose, time interval or target volume [, , ,, , ]. While relapse tumour stage and prior surgery may show a trend towards influencing local control and overall survival, this was reaching nowhere near statistical significance. The same was reported by Sher and colleagues in their retrospective analysis . In both series this finding may be attributable to the retrospective nature of the analysis, low patient number and therefore lacking statistical power.
Tumour response rates and local control with this new treatment technique are promising. While long-term local control continues to be rare, re-irradiation in adenoid cystic carcinoma remains essentially a palliative treatment. Still, patients may achieve a median of 19 months local control without severe toxicity and without continuous chemotherapy, therefore re-irradiation with carbon ions is a valid treatment alternative.
As most local failures following re-irradiation occur within the re-RT high-dose area, dose escalation may be an option. In view of observed vascular toxicity, this may be accompanied by increasing severe late effects and should be viewed with caution.
Early results were presented in part as an abstract at ECHNO meeting 2014 in Liverpool, UK.
JD is CEO of the Heidelberg Ion Beam Therapy Centre (HIT).
Conflict of interest
JD is CEO of the Heidelberg Ion Beam Therapy Centre (HIT), all other authors declare they have no conflict of interests.