Treatment planning comparison of proton arc therapy and intensity modulated proton therapy for synchronous bilateral breast or chest wall and regional nodal irradiation.

1
Introduction
Synchronous bilateral breast cancer (SBBC) accounts for only 1–2% of all breast cancer diagnoses but may present a significant challenge for radiation treatment planning [1]. When treating patients who require bilateral breast or chest wall and regional nodal irradiation (RNI), radiation target volumes are extensive, encircle the thorax, and lie close to critical normal tissues, including the heart, coronary vessels, and lungs [2]. Achieving adequate target coverage while sparing organs-at-risk (OARs) typically requires a deep inspiration breath-hold technique and more sophisticated photon radiation delivery methods, such volumetric modulated arc therapy (VMAT) or Tomotherapy [3], [4], [5]. Treatment planning is particularly challenging in this setting, especially with modern organ-sparing goals and the increasing use of breast reconstruction, where maintaining skin integrity and cosmetic outcome is essential [6], [7]. Rotational photon therapy provides improved dose distribution over three-dimensional conformal radiation therapy in this setting [8], [9], but exposes thoracic organs to a broad low-dose bath [10]. Therefore, proton therapy, with its superior conformality and elimination of exit dose, has emerged as a promising alternative for reducing heart and lung exposure [11], [12].
Proton therapy has progressed from passive scattering to pencil beam scanning and Intensity Modulated Proton Therapy (IMPT), offering increasingly conformal dose distributions and better sparing of normal tissues [13]. Despite these advances, anterior beam arrangements of protons are not inherently skin-sparing [14]. This is particularly concerning for patients with implant-based reconstruction, where higher skin dose has been associated with moist desquamation and unplanned surgical revisions [15], [16]. While generating and implementing skin-rind structures as optimization parameters helps minimize near-surface dose when using IMPT, next-generation delivery methods such as proton arc therapy aim to further improve performance, especially in complex anatomical settings.
Proton arc therapy (PAT) is a novel technique that combines the precision of pencil beam scanning with the geometric advantages of rotational delivery, conceptually similar to VMAT in photon therapy [17]. By continuously rotating the gantry and dynamically modulating spot delivery, PAT enables more flexible beam angles and potentially more homogeneous dose distributions, while improving normal tissue sparing [17]. Earlier studies in head and neck, unilateral breast, and thoracic cancers suggest improved robustness and favorable dose–volume outcomes, including reduction in dose to organs-at-risk (OARs) and improved conformity, compared to fixed-field IMPT [17], [18], [19], [20]. However, its application in bilateral breast cancer remains underexplored.
This study aimed to evaluate whether PAT provides advantages over IMPT in patients with SBBC requiring bilateral RNI with respect to target coverage, plan robustness, OAR dose, skin sparing effect, and delivery efficiency.

2
Materials and methods
2.1
Patient Selection and contouring
This retrospective study was approved by the Institutional Review Board of Corewell Health (IRB No. 2017-455). Twenty patients were identified with unilateral, locally advanced breast cancer who had previously undergone computed tomography (CT) based simulation and received treatment at our institution. The cohort included ten patients with bilateral intact breast anatomy, five patients who had undergone bilateral mastectomy with immediate reconstruction using either permanent implants or tissue expanders present at the time of simulation, and five patients who had bilateral mastectomy without reconstruction. Based on the presence or absence of natural or reconstructed breast tissue, patients were categorized as having either a “thick chest wall”, defined as intact bilateral breasts or bilateral reconstructions (n = 15), or a “thin chest wall”, defined as bilateral mastectomy without reconstruction (n = 5).
To simulate the clinical scenario of SBBC requiring bilateral, comprehensive regional nodal irradiation, bilateral target volumes were generated on each patient’s existing simulation CT. This included contouring the contralateral breast or chest wall and the associated regional nodal areas, specifically the bilateral axillary levels I–III, supraclavicular, posterior cervical, and internal mammary lymph chain in the first three intercostal spaces. Contouring was performed in accordance with established breast cancer contouring guidelines, including RTOG, ASTRO, and the RADCOMP trial atlas (Supplemental material Fig. S1). Additionally, a near-surface structure referred to as the “skin rind” was contoured. This was defined as the tissue extending from the body surface to a depth of 3 mm overlying the bilateral breast and bound by the chest wall target volumes (Fig. 1).

2.2
Treatment planning
For each patient, two types of proton plans were generated on Raystation version 2023B (RayResearch, Sweden): a conventional IMPT plan and a Spot-Scanning Proton Arc (SPArc®) plan. Single-field optimization (SFO) with dual-isocenter and 2 beams was utilized in IMPT plans. Each beam was restricted to one side and was not allowed to treat the contralateral side. SPArc® is an in-house developed proton arc delivery technique that enables dynamic spot scanning during continuous gantry rotation. The SPArc® plans were generated through an in-house developed optimization algorithm and implemented through scripting [17]. A half arc with a 2.5-degree sampling frequency was applied for SPArc® [21]. The detail of beam arrangement could be found in supplemental material Table S1. Worst-case scenario robust optimization was applied in both plans to account for ± 5 mm setup uncertainties and ± 3.5% range uncertainties for both IMPT and PAT plans [22]. All the plans were based on the Monte Carlo algorithm and calculated with a dose grid of 0.3 × 0.3 × 0.3 cm3 (calculation engine version 5.5). Value 1.1 was used for the calculation of generic relative biological effectiveness (RBE)2
[23]. Both plans were prescribed a total dose of 50 Gy (RBE) in 25 fractions to the bilateral breast or chest wall and regional nodal target volumes. The planning requirement for both techniques was that at least 95% of the total clinical target volume received a minimum of 95% of the prescribed dose (D95 ≥ 95%). The same criterion applied to each individual subsite clinical target volume, including axillary levels I/II/III, supraclavicular, and posterior cervical regions. The only exception was the internal mammary node (IMN), where a minimum of 90% of the volume was required to receive at least 95% of the prescribed dose (D90 ≥ 95%). The OAR dose constraints applied were as follows: mean heart dose ≤ 1.7 Gy (RBE), heart V5 ≤ 10%, heart V25 < 2%; left anterior descending artery (LAD) mean dose ≤ 4 Gy (RBE) and maximum dose ≤ 15 Gy (RBE); mean lung dose ≤ 15 Gy (RBE), with lung V5, V10, V20, and V30 constrained to ≤ 65%, 50%, 30%, and 20%, respectively. Additional constraints included esophagus Dmax ≤ 30 Gy (RBE), trachea Dmax ≤ 40 Gy (RBE), and spinal cord Dmax ≤ 28 Gy (RBE). To evaluate and control dose to the near-surface region, a dose constraint of 42 Gy (RBE) mean dose to the skin rind was incorporated into the optimization process for this structure in both planning techniques to minimize excessive dose to the superficial layer. For each patient, dose-volume histograms were generated for clinical target volumes and OARs. Plans were assessed to ensure comparable target coverage before evaluating differences in normal tissue exposure between techniques. Key endpoints included mean lung and heart dose, LAD dose, and skin rind dose.

2.3
Delivery time
The delivery time for IMPT was calculated using the cyclotron (IBA Proteus®Plus model) and accounted for the spot spilling time (SSPT), spot switch time (SSWT), energy layer switch time (ELST), and beam switch time [24]. Additionally, 5 min were included for gantry mechanical rotation and couch iso-shift movement [19]. For PAT, the treatment delivery time was calculated using Controller-DynamicARC®. Specifically, Controller-DynamicARC® integrates SSPT, SSWT, and ELST with the gantry’s mechanical rotation, which includes a maximum rotation speed of 6°/s and acceleration/deceleration limits of 0.6°/s2
[25].

2.4
Statistical analysis
Normality was assessed for each dose-volume parameter. Paired comparisons for each dose-volume parameter were performed using paired t-tests when the data were normally distributed; otherwise, the Wilcoxon signed-rank test was applied. Subgroup analyses were also performed to compare dose-volume outcomes between patients with thick and thin chest walls. Multiple comparisons were corrected using the Benjamini–Hochberg false discovery rate (FDR) procedure. Results were considered statistically significant at FDR q value < 0.05.

3
Results
PAT and IMPT achieved comparable target coverage, with a mean total clinical target volume coverage of 96.9% for both techniques (p = 0.94). All plans met predefined robustness criteria. Under worst-case uncertainty conditions, target coverage remained acceptable, with at least 95% of the volume receiving 90% of the prescription dose. The worst-scenario D90 was 97.4% ± 1.3% for PAT and 97.8% ± 1.1% for IMPT (Table S2 and Fig. S2), with no statistically significant difference observed between the two techniques (p = 0.42).
When analyzing individual clinical tumor volume (CTV) subregions, PAT maintained comparable coverage in the bilateral breast and chest wall (96.6% for the left side and 96.0% for the right side, identical between techniques, Table S3). Statistically significant increases in coverage were observed in selected nodal subregions, including left axillary level II (99.8% vs 97.8%, p < 0.001) and the left supraclavicular region (98.6% vs 96.8%, p < 0.001), as detailed in Table S3.
Regarding the lung dose, compared to IMPT, PAT lowered the mean dose to the left lung by nearly 4 Gy (RBE) (8.7 vs. 4.8 Gy (RBE), p < 0.001) and to the right lung by nearly 3 Gy (RBE) (8.0 vs. 5.1 Gy (RBE), p < 0.001). Corresponding reductions in V5, V10, and V20 were all statistically significant (p < 0.001), as shown in Fig. 2 and Table S4. In the thick chest wall group, PAT reduced left and right lung mean doses by approximately 4.4 Gy (RBE) and 3.1 Gy (RBE), respectively (p < 0.001) (Supplemental material Table S5). In the thin chest wall group, the reductions were smaller but remained significant for most metrics, including right lung mean dose (p = 0.012), right lung V10 (p = 0.002), left lung V10 (p = 0.005), right lung V20 (p = 0.009) and left lung V10 (p = 0.013) (Supplemental material Table S6).
PAT also demonstrated advantages in cardiac sparing, particularly for low- and intermediate-dose regions. Across all patients, heart V5 and V25 were significantly lower with PAT (V5: 2.9% vs. 5.2%, p = 0.003; V25: 0% vs. 0.3%, p = 0.004). While the mean heart dose was comparable with both techniques (p = 0.410), PAT achieved a statistically significant reduction in LAD and RCA maximum doses (LAD: p = 0.044; RCA: p = 0.040). In patients with thick chest walls, PAT also reduced heart V5 (p = 0.005) and V25 (p = 0.021) (Supplemental material Table S5).
In terms of skin-sparing, PAT substantially reduced dose to the superficial skin rind. For the full cohort, the average mean dose to the skin rind was 40.8 Gy (RBE) with PAT compared to 45.0 Gy (RBE) with IMPT (p < 0.001). Subgroup analysis revealed this skin-sparing advantage was greater in patients with thick chest walls, where the skin rind mean dose was reduced by 4.7 Gy (RBE) (p < 0.001). Among patients with thin chest walls, PAT achieved a smaller yet statistically significant reduction in skin dose of 2.3 Gy (RBE) (p = 0.020).
Regarding delivery efficiency, as shown in Supplemental material Table S1, PAT consistently demonstrated shorter treatment delivery times than IMPT across the 20 cases. The mean delivery time for PAT was 195 ± 24 s, compared with 497 ± 23 s for IMPT (p < 0.001).

4
Discussion
This study evaluated whether PAT offers advantages over IMPT in patients with SBBC requiring RNI. We explicitly incorporated near-surface skin dose constraints and assessed the feasibility of a skin-sparing approach in rotational proton planning for bilateral breast cancer. We found that, while maintaining comparable target coverage, PAT significantly reduced dose to the lungs, heart, and near-surface skin rind.
Our planning approach was more stringent than typical clinical practice. We required D95 ≥ 95% in both total clinical tumor volume (CTV) and each individual subsite (e.g., level I/II/III Axilla, supraclavicular fossa, and posterior cervical triangle) with the only exception being CTV IMN, where we accepted a slightly relaxed constraint of 90% coverage ((D90 ≥ 95%) in order to prioritize heart and LAD sparing in anatomically challenging cases. This approach remains more conservative than commonly used clinical thresholds in prior reported studies [12], [26]. PAT consistently achieved more rigorous goals of target coverage, suggesting improved robustness and potential reliability in real-world clinical settings.
It is suggested that proton therapy for breast cancer has been associated with increased rates of acute skin toxicity, including moist desquamation, as well as long-term effects such as telangiectasia and skin hypo- or hyper-pigmentation [26], [27], [28]. In patients undergoing post-mastectomy irradiation with implants, some studies have reported higher rates of reconstruction failure with proton therapy, though the data remain mixed [16], [29], [30], [31], [32], [33]. Previous research has established a correlation between elevated near-surface dose and adverse skin outcomes, including moist desquamation and unplanned reconstructive surgery [15]. In response, our institution developed and prospectively implemented a near-surface dose avoidance technique, applying a mean skin rind dose constraint of ≤ 80–85% of the prescription dose in selected patients [34]. This constraint was considered a secondary goal, subordinate to target coverage and other organ-at-risk constraints. In the current study, PAT further reduced mean skin rind dose by 4.1 Gy (RBE) compared to IMPT without compromising target coverage.
Subgroup analysis showed greater skin rind dose reduction with PAT in patients with thicker chest walls, including those with intact breasts or reconstruction. This likely reflects increased separation between skin and target, allowing deeper dose placement. Even in cases with thin chest walls, PAT still reduced skin rind dose compared to IMPT. We hypothesize that the in PAT, the additional degree of freedom from limited beam angle to arc trajectory provides the optimizer of treatment planning system with greater flexibility to balance target coverage and organ-at-risk constraints, making it easier to meet skin dose limits without compromising plan quality. This effect is not driven by an increase in the number of spots (Table S1). Plan robustness was not compromised by these dose–volume improvements. Clinically acceptable target coverage was maintained across all evaluated uncertainty conditions, consistent with European Society for Radiotherapy and Oncology consensus recommendations [35]. Delivery time was also reduced by eliminating isocenter shifts and manual gantry rotation, potentially decreasing intra-fraction motion, further enhancing the robustness of treatment delivery [36]. A single-isocenter approach may also simplify setup compared to two-isocenter IMPT techniques. However, we acknowledge that the feasibility of a single-isocenter technique depends on the treatment machine characteristics, such as the maximum field size, which may vary across proton centers.
Long-term cardiac and lung toxicity remain critical concerns in breast radiation. The risk of major coronary events increases by approximately 7.4% per Gy of mean heart dose without an apparent threshold, with risk emerging within 5 years and persisting for decades [37]. Cardiac substructure doses, including LAD and ventricular low-dose exposure, are associated with cardiotoxicity, supporting constraints on heart V5, heart V25, and LAD [38], [39]. Lung dose similarly correlates with secondary cancer risk, with excess relative risk rising about 8.5% per Gy [40]. Consistent with the principle of keeping doses as low as reasonably achievable (ALARA), contemporary techniques yield predictable dose ranges. Contemporary photon techniques typically yield mean heart doses of 5–7 Gy and lung V20 of 30–40%, reduced to mean heart dose of 2–3 Gy and lung V20 around 25% with deep inspiration breath hold [41], [42], [43], while proton therapy can achieve mean heart doses near 1 Gy and Lung V20 of 5–15% in selected patients [42], [43], [44], [45], [46], [47]. In our cohort, PAT further reduced heart V5, heart V25, and maximal LAD dose compared to IMPT. Differences in mean heart and LAD mean dose were not statistically significant, likely due to already low baseline IMPT values. PAT also achieved significant reductions across all lung dose-volume indices, including an approximate 40% relative reduction in mean lung dose and more than a 50% relative reduction in V5–V20, highlighting the potential for incremental cardiopulmonary sparing with advanced proton delivery techniques.
This study has several limitations. As a retrospective planning analysis, it lacks clinical outcome data, and observed reductions in skin, cardiac, and pulmonary dose require prospective validation. Individual factors such as surgical technique, body habitus, and bolus use were not incorporated. Robustness to anatomical changes was not directly evaluated, as repeat CT or synthetic CT data were unavailable. Potential dose variations due to inter-fractional anatomical changes remain to be investigated. Future work will focus on incorporating synthetic CT-based evaluations and adaptive proton therapy strategies to better characterize PAT’s robustness under realistic anatomical variations, which has been increasingly emphasized in recent studies on adaptive planning and inter-fraction motion management in proton therapy [48], [49], [50]. Prospective trials assessing toxicity, cosmesis, and cardiopulmonary outcomes are warranted. Technical considerations such as arc quality assurance and workflow optimization are under active investigation.
In conclusion, PAT demonstrated significant advantages over IMPT in the treatment of SBBC requiring bilateral RNI, particularly in reducing skin, lungs, and heart exposure while maintaining robust target coverage. These findings support the feasibility of incorporating near-surface skin dose constraints in proton planning and highlight PAT as a promising technique for anatomically challenging cases. Further clinical validation is warranted to confirm these benefits and facilitate broader implementation.

CRediT authorship contribution statement
Xi Cao: Writing – review & editing, Writing – original draft, Visualization, Validation, Resources, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Peilin Liu: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Software, Resources, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Lewei Zhao: Writing – review & editing, Writing – original draft, Visualization, Validation, Software, Resources, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Kamran Salari: Writing – review & editing, Writing – original draft, Visualization, Validation, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Xiaoda Cong: Writing – review & editing, Writing – original draft, Visualization, Validation, Software, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Xiangkun Xu: Writing – review & editing, Writing – original draft, Visualization, Validation, Software, Resources, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Xuanfeng Ding: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. Joshua T. Dilworth: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Resources, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.