Multi‐isocentric 4π volumetric‐modulated arc therapy approach for head and neck cancer

Abstract Objectives To explore the feasibility of multi‐isocentric 4π volumetric‐modulated arc therapy (MI4π‐VMAT) for the complex targets of head and neck cancers. Methods Twenty‐five previously treated patients of HNC underwent re‐planning to improve the dose distributions with either coplanar VMAT technique (CP‐VMAT) or noncoplanar MI4π‐VMAT plans. The latter, involving 3–6 noncoplanar arcs and 2–3 isocenters were re‐optimized using the same priorities and objectives. Dosimetric comparison on standard metrics from dose‐volume histograms was performed to appraise relative merits of the two techniques. Pretreatment quality assurance was performed with IMRT phantoms to assess deliverability and accuracy of the MI4π‐VMAT plans. The gamma agreement index (GAI) analysis with criteria of 3 mm distance to agreement (DTA) and 3% dose difference (DD) was applied. Results CP‐VMAT and MI4π‐VMAT plans achieved the same degree of coverage for all target volumes related to near‐to‐minimum and near‐to‐maximum doses. MI4π‐VΜΑΤ plans resulted in an improved sparing of organs at risk. The average mean dose reduction to the parotids, larynx, oral cavity, and pharyngeal muscles were 3 Gy, 4 Gy, 5 Gy, and 4.3 Gy, respectively. The average maximum dose reduction to the brain stem, spinal cord, and oral cavity was 6.0 Gy, 3.8 Gy, and 2.4 Gy. Pretreatment QA results showed that plans can be reliably delivered with mean gamma agreement index of 97.0 ± 1.1%. Conclusions MI4π‐VMAT plans allowed to decrease the dose‐volume‐metrics for relevant OAR and results are reliable from a dosimetric standpoint. Early clinical experience has begun and future studies will report treatment outcome.


| INTRODUCTION
IMRT for head and neck cancer (HNC) has been the standard practice for the last decade as it has shown to reduce xerostomia and improve associated quality of life (although such improvement did not resulted statistically significant). 1 Since its introduction in 2008, volumetric-modulated arc therapy (VMAT) has been extensively evaluated for treating HNC. [2][3][4][5][6][7][8] The literature suggests that both treatment efficiency and sparing of organs at risk (OAR) are superior with VMAT compared to conventional static field IMRT (SF-IMRT), although questioned by some authors. 2,3 Early clinical outcome reports showed comparable toxicity and local control with respect to IMRT. [4][5][6] Advanced planning methods, like knowledge-based automated planning strategies have also been explored to further improve the level of OAR sparing and the harmonization of the results at an interpatient and an interplanner level. 9 Nevertheless, due to the anatomical complexity and several tradeoffs between target coverage and OAR sparing, the use of a simple coplanar approach to the arc geometry setting being currently used seems to leave space for improvement. More recently, some groups explored the possibility to deliver SF-IMRT with conventional c-arm linear accelerators using most of the 4p space, i.e., making extensive use of noncoplanar beam arrangements and creating complex delivery trajectories for the couch-gantry-collimator system around the patient. [10][11][12][13][14][15] These investigators focused on stereotactic irradiation in the brain, lungs, and prostate and have shown that significantly sharper dose gradients can be achieved with this approach. These studies concluded that the 4p technique reduced mean or maximum doses to all OAR and may allow for safe dose escalation. The original investigations published provided evidence of benefit and proof of principle for smaller tumors. A study involving the 4p approach to SF-IMRT for HNC was also attempted but it was for small and recurrent cancers. 16 The question of applicability of 4p techniques to truly large target volumes and its feasibility for conventional IMRT/VMAT treatments remains unaddressed for HNC. The aim of this study was to explore multi-isocentric 4p volumetric-modulated arc therapy (MI4p-VMAT) plans in terms of dosimetry and delivery and comparing it with best coplanar VMAT(CP-VMAT) plans for the irradiation of HNC patients characterized by large targets and the presence of several organs at risk. Deliverability was addressed in terms of dosimetric accuracy. In the absence of an automated collision avoidance engine, this aspect was qualitatively addressed with the pretreatment quality assurance procedures performed with a body phantom.

| MATERIALS AND METHODS
Institutional scientific and ethics board approved this study. Twentyfive previously treated HNC patients with two coplanar volumetric arcs (CP-VMAT) were included in a retrospective preclinical planning study. Patient characteristics are summarized in Table 1. For each patient, the gross tumor volume (GTV) was defined as the macroscopic tumor seen on imaging, while the clinical tumor volumes (CTVs) were defined as per standard institutional practice for HNC.
CTV nodal volumes were defined as per standard RTOG protocol. 17 Planning target volume (PTV) was generated by isotropic expansion of CTV by 0.5 cm. Each PTV was defined as the mutual subtraction of each other, so they were not mutually including each other. For all the patients, the following organs at risk were defined: parotids, oral cavity, esophagus, trachea, larynx, pharyngeal muscles, mandible, temporomandibular joint, middle ear, spinal cord, and brain stem. For the spinal cord, the near-to-maximum dose constraint was set to 45 Gy to 1% of its volume (50 Gy for the brain stem). For the parotids, the mean dose was aimed to be lower than 32 Gy. For the other structures, the planning strategy was to minimize as much as reasonably possible their involvement. Standard dose prescription was used (PTV-high: 70 Gy, PTV-mid: 60/63 Gy and PTV-low: 56 Gy). ccording to the patients anatomy. The CP-VMAT plans were reoptimized starting from the clinically accepted ones to improve the reference dose distributions. Aims were to achieve the highest possible dose conformity to the target with the least involvement of the organs at risk. Multiple planners calculated the CP-VMAT plans but the selection of the final plan was made on a shared consensus. The optimal plans were selected in terms of numerical plan quality metrics, these were the ones with the "best" results for each of the planning dose-volume objectives. No knowledge-based planning tools were applied since not available at the clinic.
MI4p-VMAT plans were optimized using the same objectives and priorities as the co-planar ones. Plan geometry consisted of 3-6 arcs with 2-3 isocenters which were manually selected to avoid any risk of collisions (and verified qualitatively during the pretreatment dosimetric verification with the body phantom) which might occur during the noncoplanar arc trajectory. Extreme care was taken to assign isocenters such that there was a minimum of 10 cm clearance between the patient and gantry-collimator system as well as between the collimator and couch surfaces to avoid risks of collisions. For each patient, MI4p-VMAT plan geometry was validated for delivery by simulating the planned field geometry with their immobilization system and actual isocenters at place. This simulation also ruled out the possibility of collision during noncoplanar arc trajectories. The typical field geometry for an example case is illustrated in Fig. 1. The average total arc length for twenty-five MI4p-VMAT plans was 1115 AE 228 degrees with a maximum of 1358 degrees.
All the plans had one full coplanar arc in addition to the noncoplanar arcs. Typical field geometry consisted of one full-arc with couch angle 0 degree, two partial arcs (arc length of AE 210°) with average couch rotation of AE 45°, and two more partial arcs (arc length of AE 250°) with couch rotation of AE 15°. The arc selection in general was determined according to the following strategy: when the overlap between PTV and the parotid glands (or other relevant structures in addition) was smaller than 20% of the glands, then three arcs were selected (one coplanar and two noncoplanar with average couch rotation of AE 45°). In the other cases, two additional arcs (with couch rotation of AE 15°) were added. For a few patients with intracranial extension, one more arc (60°arc length) with a couch rotation of 90°was added.
For all plans, fixed jaws setting was applied with the restriction of a maximum x-field size smaller than 15-16 cm to prevent loss in modulation power induced by insufficient over-travel movement of the MLC leaves in that direction. of 24.4 9 24.4 cm 2 with a 7.62 mm center-to-center distance between chambers) in a multicube phantom (IBA dosimetry, GmbH, Germany) without resetting couch, gantry and collimator and isocenters. Omnipro IMRT QA software (IBA dosimetry, GmbH, Germany) was used to perform global gamma agreement index (GAI) analysis with criteria of 3 mm distance to agreement (DTA) and 3% dose difference (DD). GAI was used to quantify the agreement between the predicted and measured dose distribution at the isocenter plane. Figures 2 and 3 show the DVH parameters for PTV and OAR depicting the best CP-VMAT and MI4p-VMAT plan comparison. Figure 4 shows the typical dose distributions of both the techniques in axial, coronal, and sagittal planes for a patient. The color-wash display is set to 5-70 Gy to display the dose bath and to 48-70 Gy to display dose conformality.

| RESULTS
From the qualitative inspection of the DVHs, the target coverage was similar between the best CP-VMAT and MI4p-VMAT plans with respect to conformity, homogeneity, and percentage doses. However, when the OAR doses were compared, the MI4p-VMAT plans delivered significantly less dose to various relevant OAR. There was significant reduction in average mean doses with MI4p-VMAT plans with respect to bilateral parotids by 3 Gy, oral cavity by 5 Gy, pharyngeal constrictors by 4.3 Gy, larynx by 4 Gy, and upper esophagus by 3.3 Gy. There was also significant reduction in the average maximum doses to the brain stem that was reduced by 6.0 Gy, to the spinal cord by 3.7 Gy and to the oral cavity by 2.4 Gy. For the brain, there was a statistically significant increase in the mean and low dose involvement due to a somehow broader dose bath delivery for the not coplanar arcs. The dose bath is represented by the dose to the healthy tissue (conventionally defined as the body's volume in the CT minus the encompass of all target volumes). Tables 2 and 3 show the results of the quantitative analysis conducted on the DVH for the various parameters considered for PTV and the subset of OAR where remarkable differences were observed.

| DISCUSSION
There is a perception that current IMRT and VMAT techniques have hit a plateau with respect to physically achievable dose distributions.
One of the techniques that challenge this perception is the 4p approach, which involves the use of multiple noncoplanar beams using robotic couch and gantry on modern C-arm linear accelerators.
The studies from the University of California (Los Angeles, UCLA) have elegantly described this technique and have demonstrated significant sparing of OAR and discussed the potential for dose escalation in patients of lung, liver, prostate, HNC, and brain tumors. [9][10][11][12][13] But these studies have gives the optimizer additional room to reduce the dose to OAR without loosing the PTV coverage even for larger and complex targets.
In our opinion, doubling the arc length with multiple isocenter has significantly improved the quality of the plan.
Our MI4p-VMAT plans showed uniformly superior sparing for studied OAR compared to best CP-VMAT plans without compromising dose conformity to planning target volumes. Concerning organs at risk, significant sparing was observed for brain stem, esophagus, larynx, parotids, oral cavity, and pharyngeal muscles, either for the mean or the near-to-maximum doses with an obvious potential benefit in terms of reduced risk of normal tissue complication probability. Reduction in mean and/or maximum doses to structures such as  IMRT. The total beam on time with MI4p-VMAT appears to be similar to that of conventional SF-IMRT for HNC, which also ranges from 8-12 min. 3,4 There are certain limitations to this approach. The foremost is TrueBeam system) might have sub-mm accuracy in their rotational axes which could mitigate these risks. 19  dose-volume metrics of relevant OAR decreased significantly without altering dose conformity for relatively large and complex PTV volumes.
These plans are clinically deliverable with acceptable quality assurance.
The improvements in hardware and availability of MI4p-VMAT optimization algorithm with automated delivery can further improve the quality of plans as well as enhance treatment efficiency and thereby making it possible for this technique to be adopted for routine day-today clinical practice. Early clinical experience has begun and future studies will aim to report treatment outcomes.

L. Cozzi acts as Scientific Advisor to Varian Medical Systems and is
Clinical Research Scientist at Humanitas Cancer Center. All other coauthors have no conflicts of interest.