Potential Impact of Daily Setup Variation on Pencil-Beam Scanning for Head and Neck Cancer

The epidemiology of head and neck cancer is evolving, and a growing number of younger patients diagnosed with human papillomavirus–positive oropharyngeal squamous cell carcinoma [1, 2] have been treated and have achieved excellent long-term disease outcomes [3, 4]. Given that standard radiation techniques can cause severe late effects [5–11] and affect quality of life [7], it is increasingly important to consider novel approaches to minimizing long-term treatment-related complications.

In this context use of proton therapy holds significant promise in the treatment of head and neck malignancies. The unique physical characteristics of a proton beam allow for potential therapeutic gains, particularly by decreasing exposure and toxicity for healthy tissue. However, issues such as daily variation in patient positioning [12], anatomic changes that occur during the course of treatment [13], and the inherent uncertainties associated with pencil-beam scanning (PBS) [14–16] warrant particular attention when implementing proton therapy. Here we present a case report of the dosimetric impact of random variation of soft tissue positioning when treating a patient with PBS for locally advanced head and neck cancer.

A 79-year-old woman presented with a mass on the right side of the neck. Initial workup with computed tomography (CT) scan of the neck revealed a 2-cm node on the right side of the upper neck, and positron emission tomography–CT revealed increased fludeoxyglucose uptake in the right node and tonsil, with no evidence of systemic metastases. Biopsy of the neck node and right tonsil revealed squamous cell carcinoma, positive for p16 on immunohistochemistry.

She underwent initial treatment with a staged selective (levels I though IV) dissection of the right side of the neck, followed by a transoral robotic resection of the primary tumor. Pathologic examination revealed that 3 of 41 were nodes involved with cancer (2 nodes in level II, 1 node in level IV), with focal microscopic extracapsular extension in a level II node and a tonsillar tumor, 2.5 cm in greatest dimension, resected with negative margins, without evidence of lymphovascular involvement or perineural invasion.

The treating physician recommended that the patient then receive adjuvant radiotherapy with concurrent weekly cetuximab. Based on the dosimetric advantages of a pencil-beam proton plan compared with a volumetric modulated arc therapy intensity-modulated radiation therapy (IMRT) photon plan (Figure 1), the decision was made to deliver her radiotherapy with protons. Immobilization was achieved with thermoplastic mask, neck rest, and a foot box attached to ropes of indexed length held by the patient for shoulder retraction. The planning technique used a single-field uniform dosing approach with 2 posterior oblique beams and a right lateral beam. Targets were defined as follows: the highest-risk target (planning target volume [PTV] 1) comprised the region of extracapsular extension, PTV2 comprised the primary tumor bed and the ipsilateral portion of the neck (levels II through IV), and PTV3 comprised second echelon nodal levels (ipsilateral levels Ib and V, ipsilateral retropharyngeal and contralateral levels II through IV). Prescribed doses were 63 Gy (relative biological effectiveness) to PTV1, 60 Gy to PTV2, and 54 Gy to PTV3, all over 30 fractions. Treatment instructions were given for online daily kilovolt alignment (to bony anatomy) with offline weekly verification CT imaging surveillance.

During the first week of treatment, daily kilovolt imaging was obtained and demonstrated consistent setup to bony anatomy, with daily physician approval. However, the first weekly verification CT imaging scan demonstrated variation in the skinfolds and soft tissues of the posterior portion of the neck (Figure 2), which was not visualized on online kilovolt imaging. The dose distribution from the original proton plan calculated on the verification CT showed a change in coverage in the targets of the anterior portion of the neck (Figure 2) and reflected a significant compromise in coverage of the clinical targets as observed on dose-volume histogram (Figure 3). Given concern over these findings, a repeat CT verification scan was performed a few days after, which again demonstrated an unpredictable variation in the skinfolds and soft tissues of the posterior portion of the neck compared with the original CT simulation as well as the first CT verification scan, with corresponding loss of clinical targets.

As a result, the decision was made to switch the remainder of the radiation therapy course to IMRT because the IMRT plan was much less sensitive to the random variation of setup in the soft tissues of the patient's neck (Figure 4), and there was no appreciable compromise in target coverage. She tolerated the remainder of treatment well, with mild mucositis and minimal other toxicity. She remained disease free 20 months after completion of radiation therapy.

Several studies have suggested potential dosimetric advantages associated with the use of proton therapy for head and neck cancers [17–27], and recent clinical implementation of PBS has demonstrated encouraging early outcomes [28]. Efforts such as these can and should be continued and extended for other patients, as the unique dosimetric advantages available with PBS may lead to significant improvements in the therapeutic ratio for these patients; however, these unique qualities and uncertainties [14–16] subsequently require attention to such issues as patient positioning [12] and anatomic changes [13], for which proton therapy may be more sensitive than IMRT.

Our case reveals how random variation in the setup and positioning of soft tissues can affect the robustness of a PBS-based plan. In the process, it brings up several areas that warrant further study and development. Current daily imaging for patients undergoing proton therapy is largely limited to kilovolt imaging, which is optimally suited for matching to bony anatomic structures but inadequate to assess soft tissue anatomy. In our study, we performed offline CT verification, and because of the issues identified, we advocate for the continued development and implementation of online cone-beam CT imaging for proton therapy. CT-based imaging is critical for evaluating daily setup and response of soft tissue, and we believe it has a greater importance for proton-based therapy than for photon-based radiation. Another topic to consider is patient and planning technique selection. Our patient was an elderly woman with significant soft tissue and skinfolds in the posterior portion neck, treated with most of the dose delivered via posterior oblique fields. Considerations such as the detailed evaluation of patient-specific anatomy at the time of consultation and simulation, as well as selection of beam directions to avoid areas of potential greatest variability in setup of the soft tissues, is critical in a proton-based treatment setting. Additionally, the continued development of customized immobilization devices will help reduce setup variability. In our patient, the thermoplastic mask and neck rest reliably immobilized the head and upper neck but did not adequately address the tissues of the mid and lower neck. As a result, we have developed and implemented a new immobilization device where patients have customized molds that cover and immobilize posterolaterally the extent of the anatomic region being treated with an overlying 5-point thermoplastic mask extending from the vertex of the scalp to below the shoulders and clavicles and to the upper thorax. The patient documented in this report was treated early in our institutional experience with PBS for head and neck cancer, when our currently used immobilization device was not available. Therefore, we believed the safest clinical decision at the time was to switch her to IMRT. However, since we have incorporated our new immobilization device as our standard approach, there has been far less setup variability in soft tissues, largely reducing the need for routine CT verification and eliminating the need to switch to IMRT.

Random setup variability of soft tissues is an important issue to consider for patients undergoing proton therapy for head and neck cancer. We advocate for the continued development of online CT-based imaging for routine setup verification and for the continued use of the most appropriate customized immobilization devices to minimize soft tissue setup variation. Doing so will help maximize the opportunities that proton therapy can have to improve the outcomes for our patients.

Conflict of Interest: The authors have no conflicts of interest to disclose.