Feasibility of lattice radiotherapy using proton and carbon ion pencil beam for sinonasal adenoid cystic carcinoma

the same plans were 3.42 (range: 3.15–3.79), 2.93 (range: 2.19– 3.74), and 3.58 (range: 3.09–4.68), respectively, with no significant differences in both. The brain stem, chiasm, optic nerve, parotids, spinal cord, and brain were better protected in proton and carbon-ion LRT plans than in photon LRT plans. Further, these plans did not introduce more doses to the OARs compared to the one-fraction clinical boost plan, with a significant difference. Conclusion: Despite the minimal difference in PVDR between proton and carbon-ion LRT plans and photon LRT plans, the former can better protect OARs than photon LRT plans. Therefore, PBS proton and carbon-ion LRT can be used for sinonasal ACC.


4
Lattice radiotherapy (LRT) is a technique that delivers highly inhomogeneous doses to tumours, increasing apoptosis by bystander effects 9 . Furthermore, LRT can modify the immunosuppressive tumour environment, potentially enhancing the benefit of antigen-specific immunotherapy [10][11] . As such, LRT is a good candidate to increase the radiosensitivity of ACCs.
By using collimated photon beams, LRT can generate dose distributions concentrated in vertices while rapidly falling off outside the vertices. Clinical trials using photon LRT with an ARC delivery method for ovarian carcinosarcoma and non-small cell lung cancer have been conducted since 2015. These trials reported that LRT would result in a significant reduction in tumour size, thus helping to ensure a longer overall survival without severe OAR complications [12][13][14][15] .
However, although these clinical trials have shown remarkable success, they still used a photon beam, which would deliver a high integral dose.
Proton and carbon ions delivered with a pencil beam scanning (PBS) technique can deposit Bragg peak to each vertex through inverse planning, thus delivering a high dose to the vertices while sharply decreasing the non-target doses. In addition, the scattering of proton beams is larger, resulting in carbon ion beams with a smaller full width at half maximum (FWHM). Less scattering may produce a higher peak-to-valley ratio (PVDR).
Gao et al 16 performed GRID therapy for deep and shallow targets using PBS proton beams, with a lateral space of 1.0 cm and 2.0 cm, in water phantoms.
Based on these settings, they compared the dosimetric parameters between both PBS dose distributions and the typical photon GRID technique. They found that the proton GRID therapy provided higher PVDR than the photon GRID technique 5 for the deep-seated phantom target. Snider et al 17 delineated 3-dimensional spherical vertices with a diameter and centre-to-centre (c-t-c) of 3 mm and 3 cm, respectively. Based on the targets, they delivered mono-energetic proton PBS beams to the five tumours at different sites. The results showed that the peak dose could be achieved at 15 Gy (RBE) with a confluent valley dose across the targeted tumour of approximately 1-2 Gy (RBE). In addition, the skin could be safely protected.
This study aimed to investigate the feasibility of proton or carbon ion LRT plans for the treatment of ACCs. Towards this goal, the spherical vertices of 10 patients were delineated. The PVDR and doses to the periphery targets and OARs were compared among photon, proton, and carbon-ion LRT to determine whether proton and carbon-ion beams could promote the efficacy of LRT. The LRT plans were also compared to the clinical plans to determine whether LRT plans would increase the normal tissue complications.

Study design and patient selection
This retrospective study evaluated ten patients with ACCs who underwent biopsy and were treated with proton and carbon ion radiotherapy at the Shanghai Proton and Heavy Ion Center between 2015 and 2019. Only those with the largest tumour size were included. The tumour diameter ranged from 6.34 cm to 9.58 cm (mean: 8.02 cm) in the transverse view, corresponding to a gross tumour volume (GTV) 6 of 72.64-178.09 cc (mean 120.76 cc). None of the patients underwent surgery.

Description of the lattice vertices
Based on the clinical contouring of OARs and targets, we additionally contoured the lattice vertices, which had a diameter of 1 cm, as suggested by Gholami et al 18 .
All the vertices were in solid parts of the GTVs and were ensured to not overlap in the beam eye views to avoid high entrance doses. Two to three vertices were contoured per patient. The mean c-t-c (i.e. average distance between each vertex in the three dimensions) was 3.51 cm (range: 2.94-4.73 cm). Further, they were at least 1.6 cm apart and 1 cm away from any OAR. Figure 1 shows an example of vertices and the GTV of one patient. In general, patients receive 15-17.5 Gy (RBE) in 5 fractions of carbon ion radiotherapy to CTVboost as boost plans. In this study, the first fraction of the boost plan was replaced by our LRT plan. Therefore, we assumed that the patient received one fraction of the LRT plan and four fractions of the boost plans (LRT+boost). The OAR constraints for LRT plans were the same for all three modalities. The Dmax for the brainstem was <1.5 Gy (RBE); chiasm, <0.6-3 Gy (RBE) (depending on the distance to CTVboost); optic nerves, <1-3 Gy (RBE); eyes, <1.8-2.6 Gy (RBE); lens, <1 Gy (RBE); skin, <5 Gy (RBE); and brain, <2.5 Gy (RBE). The patient characteristics and vertex details are listed in Table 1. Table 1. Characteristics of the patients and vertices 8 Abbreviations: GTV, gross tumour volume

Dose comparisons
The photon, proton, and carbon ion LRT plans were compared with respect to the peak-to-valley ratios (PVDRs), dose delivered to the OARs and vertices, and CTVboost. LRT plans were also compared to the one-fraction boost plans because the LRT plan should not increase the OAR toxicity.  Several variables including Dmax for the brain stem, chiasm, optic nerves, lens, spinal cord, skin, brain, and vertices; Dmean for the eyes, parotids, oral cavity and brain; and volume that received a minimum of 95% of the prescribed dose (V95) for CTVboost were assessed for all three types of LRT plans. These parameters were also compared between one fraction of boost plans and LRT plans to determine whether LRT plans would induce potential OAR complications.

Statistical analysis
The PVDRs and doses to the vertices and targets of the three types of LRT plans were analysed using the SPSS 20.0 software (V20, IBM, America). Data between two types of plans were compared using the Wilcoxon rank-sum test.
P<0.05 was considered statistically significant.

Comparison among LRT plans and between one fraction of boost plans and LRT plans
The dose statistics of the variables from photon, proton, and carbon ion LRT plans are summarised in Table 2. There was no difference between proton and carbon-ion LRT plans, and they showed advantages over photon LRT plans in protecting the brain stem, chiasm, optic nerve, parotids, spinal cord, and brain.
With respect to the dose to targets, photon LRT plans showed a higher V95 of CTVboost than proton LRT plans. Further, photon LRT plans showed the highest Dmax of vertices among the three LRT plans.
Except for the Dmax of the brain, there was no significant difference in the OAR dose and CTVboost dose between one fraction of boost plans and proton or 13 carbon-ion LRT plans. Meanwhile, the photon LRT plans would result in apparently higher doses to normal tissues. The results are shown in Table 3.

Discussion
In this study, we delineated vertices for radioresistant ACC patients. Photon, proton and carbon-ion LRT plans were generated based on the vertices and peripheral target volumes. This is the first study to our knowledge to investigate the difference of dosimetry between photon, proton and carbon-ion LRT plans.
Our results provide compelling evidence for proton and carbon-ion LRT as one Further research is needed to validate the optimal characteristics of the vertices in proton and carbon-ion LRT plans.
High PVDR is the key objective of successful LRT for large tumours. The small difference in PVDR among scattered photons, protons, and carbon ions is a result of the large PBS spot size caused by low energy when treating shallow tumours and the fewer fields in proton and carbon-ion plans than photon plans.
This leads to the comparable PVDRs between photon, proton, and carbon-ion LRT plans. Concurrently, less lateral scattering of particle beams, especially carbon-ion beams, enables lower doses to the OARs in proton and carbon-ion plans.
LRT can be used to induce immune responses through its high-dose component, including abscopal effects and bystander effects. The presence of a bystander in GRID has been experimentally documented by a significant decrease in clonogenic survival and an increase in the expression of DNA damage genes in bystander cells [22][23] . Larger tumours may release more antigens in response to irradiation, which potentially intensifies the abscopal effects 24 . Furthermore, the combination of radiation and immunotherapy can enhance T cell infiltration and inhibition of myeloid-derived suppressor cells and regulatory T cells 25 . As such, many factors, including the characteristics of vertices and radiation sequence, need to be further explored to augment the immunogenic responses.
Carbon-ion beams with high linear energy transfer values have been shown to possess higher RBE and induce more complex DNA damage.
Importantly, they cause a significant reduction in radioresistance [26][27][28][29][30] . Meanwhile, carbon-ion beams may lead to radiation-induced bystander effects and abscopal effects in high-dose irradiation of partial tumours. These effects may be induced in a different manner as compared to that of photons, as reported in an animal model study 31 . However, the advantage of carbon-ion is yet to be investigated. We believe that it is imperative to conduct this dosimetric comparison and thus conducted it as part of this study.
Sinonasal ACC is a bulky tumour that is not tractable by conventional radiation due to the sensitive surrounding OARs. Moreover, the large hypoxic tumour volumes within the tumour may induce hypoxia, thus reducing apoptosis 9 .
In contrast to conventional photon beam radiotherapy, LRT with carbon ions can better deliver high-dose radiation to the target tumour. Further, it can induce bystander effects and abscopal effects to increase apoptosis in the peripheral tumour cells while having no adverse effects on adjacent organs, as indicated in our study. Therefore, LRT with carbon ions is an advantageous option as one fraction of the boost plans to treat sinonasal ACCs.
This study has some limitations. The evaluation of LRT plans needs to take more factors into consideration, especially when combined with proton and carbon-ion beams. Range uncertainty is a main problem in particle radiotherapy.
Due to the limited number of beams and the complicity of the patient's anatomy, this problem may be more severe. Therefore, the geometry design of the vertex and beam angle selection should be more cautious. In this study, LEM was used to calculate the RBE-weighted carbon-ion LRT. Whether this model can accurately predict the RBE of carbon ions at such a high dose level is still a question. Thus, the prescribed doses may not be clinically equivalent for the three modalities.

Conclusion
Dosimetric comparisons of photon, proton, and carbon-ion LRT plans showed no significant difference in PVDR. However, proton and carbon-ion LRT plans can protect the OARs better than photon LRT plans. This was also confirmed by comparing doses delivered to OARs and the GTV between one fraction of boost plans and LRT plans. Therefore, a prospective clinical study is warranted to evaluate the efficacy of proton or carbon-ion LRT plans for the treatment of ACCs, particularly in terms of tumour control and toxicity of normal tissues.

Ethics approval and consent to participate
As a retrospective study, the work was approved by the Shanghai Proton and Heavy Ion Centre Institutional Review Board.

Consent for publication
Not applicable

Availability of data and materials
The datasets used and analysed during the current study are available from the corresponding author on reasonable request. and analysed data. ZXY and WXD analysed data. LJD and KL designed the work and revised the manuscript. All authors read and approved the final manuscript.