Dose distribution of intensity-modulated proton therapy with and without a multi-leaf collimator for the treatment of maxillary sinus cancer: a comparative effectiveness study

Cancers of the nasal cavity and paranasal sinuses are uncommon and comprise about approximately 3–5% of all head and neck cancers. Standard treatment methods for locally advanced sinonasal carcinomas include surgical resection and adjuvant radiation therapy [1, 2]. In addition, preoperative chemoradiotherapy (CRT) has been successful in achieving a negative surgical margin [3]. At the Okayama University Hospital, preoperative CRT is performed for T3–4 large maxillary sinus tumors, with a total dose of 66 Gy. However, after conventional irradiation, three-dimensional radiation therapy, and intensity-modulated radiation therapy (IMRT), serious (grade 3–4) complications, such as eye damage and salivary gland dysfunction, have been reported at a rate of 1–24% [4‐6]. In addition, the use of proton beam, heavy particle beam therapy, or intensity-modulated proton therapy (IMPT)—instead of three-dimensional radiation therapy—reduces the rate of complications and improves the local control rate [7].

Treatment with proton beam therapy involves two types of irradiation methods: passive scattering proton therapy and IMPT, which is among the more recently developed irradiation methods. To cover a target during IMPT, each beam is scanned laterally across the target using magnetic fields in a technique called pencil beam scanning (PBS). Similar to IMRT, PBS enables state-of-the-art IMPT that optimizes all beams to deliver a sufficient dose to the target. However, serious adverse events such as visual impairment have also been reported after treatment with proton beams using PBS for head and neck cancers [8]. On the other hand, during PBS, collimation can reduce the penumbra of the surrounding normal tissue in the low energy region of the proton beam [9, 10]. Winterhalter et al. referred to the possibility of improving the dose distribution in the head region using a multi-leaf collimator (MLC) [11]. In addition, Moignier et al. reported that IMPT for head tumors improves the marginal dose while using clinical data [12]. Moreover, Yasui et al. showed that the dose distribution of peripheral organs at risk (OARs) was improved when IMPT was administered via an aperture, even in the region including the head and neck area [13]. However, their study was not limited to a single site of tumor disease. To the best of our knowledge, no previous study has examined the extent by which the dose to the normal tissue is reduced by when IMPT with an MLC is used for a single tumor site in the head and neck region. Therefore, in this study, we examined by how much the dose to the normal tissue was reduced by when IMPT with an MLC is used for maxillary sinus cancer.

Typical dose distribution and its difference in dose planning with and without an MLC are shown in Fig. 1. The IMPT plan using an MLC resulted in a sharp lateral penumbra, and the dose to the ipsilateral optic nerve was reduced. The OARs and PRV dose metrics are shown in Figs. 2 and 3 as boxplots for each patient overlaid with boxplots summarizing the data. The central colored line in the boxplots represents the median, with the edges representing the 25th and 75th percentiles. The whiskers show the range of data excluding outliers. The central, dashed black line represents the mean. Wilcoxon signed rank tests were performed between each pair of treatment modalities.

The parameters of OARs and PRVs in dose planning with or without an MLC are shown in Table 4. As shown in Table 4, the mean of ROI-_D2% and ROI-_mean of the ipsilateral optic nerve were significantly reduced in the group with an MLC. The mean of ROI-_mean of the optic chiasm was also significantly reduced in the MLC group, as was the mean of ROI-_D2% of the ipsilateral optic nerve, which reduced from 50.06 Gy to 49.58 Gy, with a difference of 0.48 Gy. Similarly, the mean of ROI-_mean of the ipsilateral optic nerve was reduced from 27.66 Gy to 26.62 Gy, with a difference of 1.04 Gy. In the optic chiasm, the mean of ROI-_mean was reduced from 8.20 Gy to 7.50 Gy, indicating a difference of 0.70 Gy. Other OARs and PRVs are shown in Table 4. A dose reduction was observed in most OARs and PRVs, with the exception of the D2% of the optic chiasm, pituitary gland, ipsilateral posterior retina, and eyeball, as well as the ROI-_mean of the ipsilateral posterior retina, ipsilateral eyeball, and contralateral parotid gland.

As shown in Fig. 4, the irradiation volumes of the lacrimal gland with and without an MLC were as follows: V5 Gy, 0.23 cm³ vs. 0.20 cm³, difference = 0.03 cm³, p = 0.080; V10 Gy, 0.11 cm³ vs. 0.08 cm³, difference = 0.03 cm³, p = 0.003; and V15 Gy, 0.05 cm³ vs. 0.20 cm³, difference = 0.01 cm³, p = 0.007. In general, these data show significant reductions associated with the use of an MLC in the low dose area only, while no significant differences were found above V20 Gy.

The beam energy was automatically selected by the treatment planning device and was 70.70–126.50 MeV. The mean beam monitor unit (MU) used in the plan was 21,82,193/fraction (range, 12,30,730–38,85,439) without an MLC and 23,34,309/fraction (range, 13,53,181–40,76,774) with an MLC. The use of an MLC resulted in an increase in the MU in all cases, increasing on average by 7% (range, 2–17%).

The results of the current study demonstrate that, compared to IMPT without an MLC, IMPT using an MLC resulted in a significantly lower the mean of ROI-_D2% to the ipsilateral optic nerve, as well as lower the mean of ROI-_mean to the ipsilateral optic nerve and optic chiasm, while providing an optimal dose to the treatment targets for maxillary sinus cancer.

Although postoperative CRT is the standard treatment for T3–4 maxillary sinus cancer, there are also reports of inoperable cases, surgical refusal, and the use of preoperative CRT [4, 6, 19]. Multidisciplinary treatment with surgery, chemotherapy, and radiation therapy is performed for maxillary sinus cancer [2, 3]. Visual impairments have been reported as a complication of surgery and photon therapy for maxillary sinus cancer on both the ipsilateral and contralateral sides [19]. In recent years, particle beam therapy has been attempted for paranasal sinus cancers to reduced sever complications. However, Fukumitsu et al. reported brain necrosis and optic dysfunction, even with proton beam therapy for sinus cancer [8]. During proton beam treatment for ethmoid sinus cancer, which is relatively similar to maxillary sinus cancer, adverse events include optic neuropathy and cerebral necrosis [20]. Thus, dose reduction to normal organs is necessary to avoid complications such as optic neuropathy and cerebral necrosis.

On phantom experiments, a reduction in penumbra due to low energy regional collimation with PBS was reported [21]. In the current study, the beam energy was 70.70–126.50 MeV. We planned to use low energy and the use of an MLC should have improved the penumbra more clearly, and the marginal dose was better and the dose distribution was improved.

In current study, the dose reductions were small with and without the use of an MLC. OARs and PRVs dose in PBS proton therapy will depend on the parameters used for optimization. These include patient distance, tumor depth, irradiation angle, PBS beam quality, algorithms for optimization of treatment planning systems, and the use of optimization and an MLC. The cases included in our study were consecutive cases of maxillary sinus cancer at the Okayama University Hospital. Also, the size, degree of invasion, and shape of each tumor are different, all plans are prepared under the same conditions, except for the use of an MLC, to eliminate planning bias. We speculate that the plan under this condition was the reason for the small difference of our result. Better dose distribution may be achieved if IMPT constraint optimization and gantry angle are fully considered.

According to Yasui et al. [13], the use of an MLC for lesions in the head and neck region has led to cases where the dose to the optic nerve decreased by a few percent and the dose to the optic chiasm decreased by up to 32%. However, that study was not focused on a single disease and the evaluated OARs were different for each irradiation site. In contrast, the current study showed that the dose to the optic nerve and lacrimal gland can be reduced using an MLC during IMPT among patients with a single disease. In the overall comparison of the mean ROI-_D2% and ROI-_mean with and without an MLC, the mean of ROI-_D2% for the ipsilateral optic nerve was reduced by 0.48 Gy and the mean of ROI-_mean was reduced by 1.04 Gy. The mean of ROI-_mean to the optic chiasm was reduced by 0.70 Gy.

In previous studies regarding the effect of radiation on the lacrimal gland, no significant dry eye symptoms were observed in patients who received a mean dose of < 30 Gy to the lacrimal gland during photon radiation therapy and these studies have shown symptoms in the all patients treated with doses > 45 Gy and with doses > 57 Gy during photon radiation therapy [22, 23]. Although the lacrimal gland did not receive a high enough dose to cause complications when the mean doses were administered with and without an MLC (5.7 Gy and 6.5 Gy, respectively) in the current study, the dose of lacrimal gland was decreased significantly. This finding may be useful when irradiation is performed near the lacrimal gland.

Yasui et al. showed the utility of the patient-specific aperture system. In their study, the OAR dose was reduced by several percent to several tens of percent owing to the patient-specific aperture system [13]. We believe this was the reason why the dose reduction effect to the surrounding OAR was low in our study compared to that in the study by Yasui et al. First, it is possible that the spot size used in our treatment system was smaller (5–6 mm at the isocenter) than that of their machine, which was 7.2–11.8 mm at the isocenter. According to Moteabbed et al., a smaller spot size is associated with a less effective MLC [24]. Second, in the current study, PRVs were not modified even if the CTV and PRV overlapped to prevent planned bias, which was observed in several plans. Third, to prevent bias owing to the plan, the gantry angle and dose constraints were fixed in all plans.

In the current study, we used the collimating MLC technique along the largest edge of the tumor. Daniel et al. reported a dynamic collimation system in which collimation is performed by moving two blades per layer with the beam spot [25, 26]. In their method, collimation can be performed on all layers of the tumor. The penumbra of the beam on the distal side and on the proximal side of the tumor is reduced; it is thought that the concentration of the proton beam is improved. Therefore, the dose distribution of the surrounding normal tissue is significantly improved by approximately 13.65% compared to cases that were treated without the dynamic collimation system [12]. Further advances in therapeutic devices may show potential for improved dose distribution.

Even if an MLC is used, there are no additional costs for each patient. In contrast, the use of MLC plans increased the MU by mean of 7% (range, 2–17%). Therefore, the increase in MU slightly increases the treatment time. The dose of the ipsilateral eyeball was increased by the use of an MLC. Doses to the eye and surrounding normal tissue are issues of optimization and gantry angle, suggesting that optimization and examination of angles are necessary for each case.

This study has some limitations, including the relatively small sample size of 26 cases. Moreover, it was not possible to obtain MR fusion images, and we did not re-plan treatments as might be performed during adaptive therapy. Finally, the dose calculation algorithm used in this study did not include Monte-Carlo simulations.

During IMPT, the use of an MLC for the treatment of maxillary sinus cancer reduces the dose to the ipsilateral optic nerve. Also, the dose of most OARs and PRVs were reduced in the MLC group. In the future, a large-scale prospective study should be conducted to determine the effectiveness of the collimation, the results of which can be used to improve IMPT dose distribution and thus reduce adverse events.