Radiation therapy plays a substantial role in the treatment of patients with head and neck cancer. Although there has been remarkable advances in radiation therapy techniques over the years, many patients still experience significant treatment toxicities, and several will unfortunately develop cancer recurrence. As head and neck cancers are widely heterogeneous, one approach that may improve outcomes is treatment adaptation to the anatomical or functional changes that occur during the course of therapy. Researchers conclude the current and upcoming literature on radiation therapy treatment adaptation for patients with head and neck cancer, with an emphasis on the added value of magnetic resonance imaging.
1. Computed Tomography-Based Treatment Adaptation
As head and neck cancer patients undergo their 6 to 7 weeks of daily radiation treatments, weight loss and tumor response often lead to notable anatomical changes. These changes may lead to the migration of organs at risk such as the parotids and pharyngeal constrictor muscles to high dose regions, resulting in an unplanned overdosage and potentially worse predicted chronic xerostomia
. Intended target volumes may also be at risk of inadequate coverage, possibly jeopardizing local control
. In a small prospective trial evaluating the need for replanning at weeks 3 and 6 during the course of radiation therapy (RT), computed tomography (CT)-based adaptive RT reduced doses to the spinal cord, avoided overdosage of the parotids and improved target coverage and homogeneity
. A second prospective trial by Schwartz et al. confirmed the reduction in mean parotid dose with CT-based adaptive RT
. One properly timed replanning seemed to deliver most of these dosimetric improvements. CT-based replanning is now standard for head and neck cancers and is usually done once the anatomical changes noted on the daily cone-beam CT seem important enough to significantly alter the initial dosimetry.
2. Magnetic Resonance Imaging-Based Treatment Adaptation
2.1. Anatomical Adaptation
Although magnetic resonance imaging (MRI) has now been widely adopted during the treatment planning of many head and neck cancers, its superiority in describing local extension and nodal disease has potential uses for RT beyond initial baseline evaluation, notably the opportunity for adaptive planning.
For its many advantages in the head and neck region, repeated MRI during the course of radiotherapy could further improve treatment adaptation by allowing better visualization of the tumor response and optimal target volume delineation. Subesinghe et al. reported a significant tumor shrinkage as early as by the 11th treatment fraction for head and neck cancer patients treated with RT, while Ding et al. found a complete response rate as high as 50% by mid-treatment for patients with human papillomavirus (HPV)-related oropharyngeal cancer (OPC)
. The latter study also demonstrated that the volume change happens early for the primary tumor, whereas the trajectory of nodal disease shrinkage is more linear
. Such important tumor responses, especially early during the treatment course, raise the question of whether maintaining the initially planned gross tumor volume (GTV) is necessary, or if these early-responding patients are being overtreated with a consequent unnecessary risk of toxicity. The accuracy of the initial dosimetry plan may also be questioned given the anatomical changes. In that sense, a prospective in silico study from the MD Anderson MRLinac Development Working Group showed that response-based biweekly MR-guided adaptive RT for HPV-related OPC patients could reduce the probability of chronic dysphagia by 11% and the mean parotid dose by 3.3 Gy in comparison to standard intensity-modulated RT (IMRT). They also showed that the GTV volume decreased on average by 44, 90 and 100% at weeks 2, 4 and 6, respectively, while the nodal volumes decreased by 25, 60 and 80% during the same period
Although these results are very promising for the reduction of toxicities for these patients, randomized data are needed to prove the safety and clinical benefit of MRI-based adaptive RT. In particular, the impact on local control of reducing target volumes to only the shrinking MRI-visible disease must be assessed, since tumors could potentially be dissolving instead of shrinking, leaving behind microscopic disease in areas previously occupied by tumor. Fortunately, several prospective randomized and single-arm clinical trials investigating the benefit and safety of MR-guided RT (MRgRT) are currently ongoing. MR-ADAPTOR is an R-IDEAL stage 2a–2b/Bayesian phase II trial led by the MD Anderson Cancer Center. The first stage of the trial will enroll 15 low-risk HPV-positive OPC patients to MRgRT dose adaptation with a weekly MRI-based RT replan before enrolling 60 additional patients randomized to either MRgRT or standard IMRT, with their primary outcome being locoregional control (LRC)
. ART-OPC is another phase II randomized trial currently recruiting and aiming to enroll 120 patients with advanced OPC who will be randomized to either standard IMRT or MRgRT with a single MRI-based RT replan on week 3 of treatments. This trial is assessing the rates of chronic dysphagia, toxicities, LRC, overall survival (OS) and quality of life (QoL) associated with this novel technique
. The MARTHA trial is a phase II single arm trial recruiting patients with locally advanced head and neck cancers to assess the rate of xerostomia associated with daily imaging for MRgRT and once weekly offline plan adaptation
. The INSIGHT-2 trial is a phase I/II single arm trial also recruiting advanced head and neck cancer patients, aiming to assess the feasibility and safety of MR-guided adaptive RT at weeks 2 and 4 and, for HPV-negative patients, dose escalation at week 2 according to response identified on diffusion-weighted imaging (DWI)
2.2. Functional Adaptation
This last trial highlights another potential role for MRI in adaptive planning beyond the assessment of anatomical changes during treatments, which is the opportunity for functional adaptation. It is now well known that head and neck cancers are widely heterogeneous with an obvious example being HPV-associated OPC, which is associated with a significantly better prognosis and a greater response to therapy
. For this reason, in the past decade, much effort has extended towards tailoring treatments to individual patients to enhance control of more resistant tumors (treatment escalation), and to reduce treatment toxicities for patients with better responding tumors (treatment de-escalation). DWI-guided treatment adaptation was also investigated in a phase II randomized clinical trial for locoregionally advanced nasopharyngeal cancers, where patients received either standard IMRT or dose-painted dose-escalation to tumor regions showing lower apparent diffusion coefficient (ADC) on DWI. This experimental technique showed promising results with an improved 2-year disease-free survival of 93.6% vs. 87.5% for standard IMRT (p
= 0.015), with no change in acute treatment toxicities
MRgRT is also promising in the era of personalized treatments by allowing precise delineation of target and normal tissue volumes for every delivered fraction. This provides an opportunity to investigate and use hypofractionation and stereotactic ablative RT (SABR), which was not previously possible in head and neck cancers due to the presence of highly vulnerable organs in immediate proximity to tumor targets. Hypofractionation is currently being investigated in the phase I DEHART trial, including head and neck cancer patients unfit for chemoradiation who will receive 40 Gy in 15 fractions to elective low-risk regions and 50–60 Gy in 15 fractions to the GTV, with MR-guided volume adaptation at fractions 6 and 11, followed by adjuvant immunotherapy
. SHORT-OPC is an ongoing phase II randomized trial for HPV-associated OPC patients, who will receive either standard chemoradiation over 7 weeks or 40 Gy in 20 fractions to elective regions with a SABR boost of 14 Gy in 2 fractions to the GTV. The study is investigating the LRC, toxicity, OS, progression-free survival and QoL of this novel technique