Impact of aeration change and beam arrangement on the robustness of proton plans

Abstract This study determines the impact of change in aeration in sinonasal cavities on the robustness of passive‐scattering proton therapy plans in patients with sinonasal and nasopharyngeal malignancies. Fourteen patients, each with one planning CT and one CT acquired during radiotherapy were studied. Repeat and planning CTs were rigidly aligned and contours were transferred using deformable registration. The amount of air, tumor, and fluid within the cavity containing the tumor were measured on both CTs. The original plans were recalculated on the repeat CT. Dosimetric changes were measured for the targets and critical structures. Median decrease in gross tumor volume (GTV) was 19.8% and correlated with the time of rescan. The median change in air content was 7.1% and correlated with the tumor shrinkage. The median of the mean dose D mean change was +0.4% for GTV and +0.3% for clinical target volume. Median change in the maximum dose D max of the critical structures were as follows: optic chiasm +0.66%, left optic nerve +0.12%, right optic nerve +0.38%, brainstem +0.6%. The dose to the GTV decreased by more than 5% in 1 case, and the dose to critical structure(s) increased by more than 5% in three cases. These four patients had sinonasal cancers and were treated with anterior proton fields that directly transversed through the involved sinus cavities. The change in dose in the replanning was strongly correlated with the change in aeration (P = 0.02). We found that the change in aeration in the vicinity of the target and the arrangement of proton beams affected the robustness of proton plan.

shown to have very promising outcome with potential lower toxicity than photon therapy. [3][4][5][6][7] However, the same characteristics that make protons attractive may result in high sensitivity of the proton beam to anatomic changes occurring during treatment.
Malignancies of the nasal cavity, paranasal sinuses, and nasopharynx are located in areas of the skull base with variable amount of air and soft tissue densities. Variations in air and fluid content in the nasal cavity and paranasal sinuses during the course of radiotherapy could affect the proton dose distribution. This may lead to a change in dose to target coverage and/or normal structures.
Late toxicities such as brain injury, cerebrospinal fluid leakage, and vision loss has been reported for patients with head and neck cancer patients treated with proton or carbon therapy. 5,8,9 In photon treatments, dosimetric change during treatment is mainly caused by the change in position of the tumor and critical structures relative to planned treatment. 10,11 In contrast, the sharp fall-off of proton makes protons very sensitive to variations in treatment depths within patients. Reduction in tumor and clearing or opacification of sinuses may result in shift of the high dose deposition, potentially lead to change in dose to the targets and critical structures. Moreover, proton therapy is more susceptible than photon therapy to tissue density heterogeneities as proton range is density-dependent.
In this study, we investigated the impact of change in tumor size and aeration in the skull base as well as beam arrangement on the robustness of passive-scattering proton therapy plans and determined the importance of adaptive proton planning.

| MATERIALS AND METHODS
This study was performed with approval of the Institutional Review Board of Massachusetts General Hospital and in accord with an assurance filed with and approved by the U.S. Department of Health and Human Services.

2.B | Treatment
Patients were treated with proton radiotherapy and concurrent cisplatin-based chemotherapy, except for one patient who received radiotherapy only (patient #12 in Table 1). The total dose to the gross tumor volume (GTV) was 70 Gy(RBE) over 35 fractions except for one patient with NK/T cell lymphoma who received 62 Gy over 31 fractions (patient #2 in Table 1). Passively scattered treatment plans were generated using treatment planning system (XiO, version 2.4, Elekta AB Stockholm, Sweden) that is clinically commissioned for proton radiotherapy at our institution.
During treatment, the patients were immobilized with a thermoplastic mask and a head cup to assure a proper and repeatable posi- In this study, robustness of the proton plans was achieved by adding the following margins: a 3.5% +1 mm was added to the range, a compensator smearing of 3 mm was applied to account for setup uncertainty, and an aperture margin of 8 mm to account for the lateral penumbra and setup uncertainty. For each individual beam, a brass aperture was used to achieve lateral shaping whereas a compensator was used for distal shaping. A patch combination (split-target volume) technique was used to optimize the proton dose distribution within an irregular volume in close proximity to critical normal structures. 12 The target volume was divided into multiple segments that were optimized and treated by a separate radiation fields, referred to as "patch field" and "through field". Utilizing the sharp dose fall-off of the Bragg peak, distal fall-off 50% dose of the patch field was matched with the lateral penumbra's 50% dose of the through field. To ensure that the magnitude of the low and high dose along the path line is acceptable, a combination of patch fields with different junctions was used. Orthogonal kV setup films and kV field portals were acquired daily before treatment. Proton treatments were delivered at the MGH Francis H. Burr Proton Therapy Center using 230 MeV beams.

2.C | Anatomical analysis
High-resolution contrast-enhanced CT images of 1.25-2.5 mm thickness were used for delineation of targets and normal structures. Contouring was performed using commercial software (MIM Maestro,

2.D | Image registration
All planning CT scans were first registered rigidly with the replanning CT to match the spatial positions. The alignment was done with respect to bony structures of the skull. Deformable registration was then performed between the rigidly transformed repeat CT and the planning CT to transfer the delineated structures onto the replanning CT.

2.E | Statistical analysis
Tumor shrinkage was evaluated for association with increase in aeration in the skull cavities containing tumor. The doses to the tumor target and OARs calculated on the planning and repeat CT scans were compared to find statistically significant difference. The data were analyzed by means of Pearson's correlation coefficient. Onetailed t-test with P < 0.05 indicated statistical significance.

3.A | Tumor shrinkage
The median decrease in the GTV of the primary site was 19.8% (range: 6.8-43.9%). We found a statistically significant correlation between the time of rescan and the extent of tumor shrinkage Fig. 1(a)]. For patients who were rescanned at or before 5 weeks, the median decrease in the primary GTV was 9.4% (P = 0.25) compared to 28.3% (P = 0.006) for patients who were rescanned after 5 weeks.

3.B | Change in aeration
In Fig. 2(a), we show schematically the change in relative sizes of the tumor, fluid, and air content within the composite cavity containing the tumor over the course of treatment. The chart below Fig. 2(b) shows that tumor in the cavity can be replaced by air, fluid, or combination of air and fluid.
We are particularly interested in the extent of increased air con- was observed (Fig. S2b). In both of these cases, dosimetric changes due to these anatomical cases were observed. | 17 mean dose D mean for the GTVs and CTVs and the maximum dose D max for optic and brain structures relative to the prescribed doses ( Fig. 3). The absolute doses for the targets and critical structures are given in Fig. S1. There was no statistically significant decrease in the dose to either GTV or CTV across the entire patient set (P = 0.99 and 0.86 for the GTV and CTV, respectively). The median change to the mean dose was 0.4% (range: −7.3 to 1.9%) to GTV and 0.3% (range: −2.8 to 1.8%) to CTV. In one outlier (patient #12 in Table 1 with sinonasal cancer), the mean dose of the GTV decreased by 7.3% due to clearing of the maxillary sinus. Figure 4(a)   There were two patients for whom the changes might be considered clinically significant. For one patient (patient #5 in the Table 1 with sinonasal cancer), the D max increased by 14.9% for the optic chiasm, 6.3% for the right optic nerve, and 13.6% for the retina. For this patient, the decrease in aeration of the right maxillary sinus resulted in over-shoot of the proton beam. Although the initial plan was created to spare the optic structures on the right side, the sparing was compromised due to the density variation around surgically removed maxillary sinus. Dose distributions for this case are shown in Fig. 4(b). The There was a case in which D max increased by 11.4% for the left frontal lone and by 9.4% for the right frontal lobe (patient #14 in the Table 1). The increase in frontal lobes was associated with complete clearing of the opacification of fluid in the frontal sinuses (see [ Fig. 4(d)].

3.D |
In two cases, D max for the right frontal lobe increased by 7.6% (patient #1 in the Table 1) and 7.4% (patient #5 in the Table 1  We identified four patients (#1, 5, 12, and 14 in the Table 1) for whom the dose distributions were compromised justifying a need for adaptive replanning. All these patients were treated for a sinonasal cancer with the primary tumor in nasal cavity (patient #1), maxillary sinus (patient # 5 and #12), and ethmoid sinus (patient #14). The change in aeration for these patients was 4.7, 10.2, 3.8, and 18.0%, respectively. All of these patients were treated with a combination of anterior, anterior oblique, ±lateral fields for the GTV boost that transversed through the nasal and sinus cavities.

| DISCUSSION
In this study, we investigated the impact of anatomical changes during proton radiation and beam arrangement on the robustness of proton plans. We focused on dosimetric changes that might justify a need for adaptive proton planning. As it is well known, the proton ranges are very susceptible to changes in air cavities within radiation fields. 13,14 Nevertheless, our study has shown that in most of the cases the coverage of target volumes was not compromised substantially even for the patients that were rescanned late in the course of treatment and for those who exhibited a pronounced change in aeration within the cavity. This indicates that the treatment planning technique that utilizes multiple beams and field patching technique often results in robust proton beam plans.
The major dosimetric consequences that affect critical structures occur when the amount of air within the irradiated cavity increases or decreases and the high dose deposition shifts forward or backward in the direction of the beam. In three cases, the dose to brainstem deviated from planned by more than 5%. For these patients, all were treated with proton beams that transversed through the sinus cavities-the most common for treatment of sinonasal cancers. The plans that did not involve beams that transverse through the sinus cavities were the most robust to change in aeration. In the case of posterior beams, a change in aeration affected only the exit dose and therefore the dose distribution was not substantially compromised. This may explain the fact that we observed alteration in dose distribution due to aeration change mainly in sinonasal cancer and not in nasopharyngeal carcinoma (Fig. 6). In the treatment of sinonasal cancer, a set of anterior or anterior oblique fields is predominantly used whereas in the treatment of nasopharyngeal carcinoma, posterior fields in addition to the anterior oblique fields are always employed.
Similar findings were recently published for the cohort of 20 patients treated with proton therapy for nasal cavity or paranasal sinus cavities. 14 The air content in the cavities increased in 18 out 20 cases. An increase in the dose to the brainstem beyond 60 Gy was observed for 3 patients, and for 10 patients the dose to the optic chiasm increased beyond 50 Gy. Our study, however, shows that proton beam arrangement is as important, if not more important, than change in aeration on the robustness of proton plan.

| CONCLUSION S
We found that there was no statistically significant decrease in the dose to either GTV or CTV across the entire patient set. For four out of fourteen patients the dose to the critical structures was compromised due to change in aeration near the target justifying a need for adaptive replanning. We also found that shrinkage of the surgical flap can compromise the planning dose distribution. In summary, adaptation during proton treatment may be needed for selected patients whose plans rely heavily on beams that transverse through the sinus cavities.
In our current clinical practice, patients with sinonasal and nasopharyngeal malignancies are routinely rescanned at least once during their course of proton treatment. Patients with change in aeration and tumor volume will be treated with the new proton plans. Prospective studies with a larger patient set and weekly CT rescans for each patient are necessary to determine the criteria for appropriate cause and time of replanning.

SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section at the end of the article.