Particle Arc Therapy in Cancer Radiotherapy: A Scoping Review of Feasibility and Dosimetric Evidence Authors: Indiana Neumann, Eva Bezak, Shahraam Afshar Vahid, Edward Simpson, Scott Penfold Published: 2026 ABSTRACT Simple SummaryParticle arc therapy is an emerging radiotherapy technique that delivers radiation from many angles around the patient, with the aim of improving tumour targeting while reducing radiation exposure to surrounding healthy tissue. This scoping review evaluated the current evidence on this technique, including its potential benefits, limitations, and stage of clinical implementation. The findings suggest that particle arc therapy can improve dose precision and reduce radiation exposure to critical organs, while also offering potential biological advantages that may enhance tumour control. However, most of the available evidence is based on computational and early-stage studies. Further research, including clinical studies and technological development, is needed before this approach can be widely implemented in routine cancer treatment. 1. Introduction Recently, particle arc therapy (PArc) has emerged as an advanced delivery technique that combines pencil beam scanning (PBS) with a rotating gantry to deliver doses across a full or partial arc. Unlike conventional particle therapy, which delivers PBS beams from a limited number of discrete angles, PArc distributes partial doses across a continuous or quasi-continuous range of beam angles, cumulatively producing conformal target coverage while potentially enhancing normal tissue sparing [1]. Clinically and in the published literature, proton arc therapy (PAT) is more prevalent than carbon-ion arc therapy (CArc), owing to the greater global availability of proton therapy facilities and the more advanced clinical translation of proton-based techniques [2]. Arc-based approaches for particle therapy were first proposed by Deasy et al. [3] using helical proton beam delivery with distal edge tracking (DET), demonstrating that highly conformal and uniform dose distributions could be achieved for nasopharyngeal tumours. Shortly thereafter, Sandison et al. [4] reported that PAT reduced lung irradiation and integral lung dose by approximately half compared with electron arc therapy, while improving target dose uniformity. In 2016, Ding et al. [5] proposed spot-scanning proton arc (SPArc) therapy, a novel optimisation algorithm that integrates control point resampling with energy layer redistribution, filtration, and resampling, enabling robust and delivery-efficient proton arc plans suitable for continuous delivery. Since then, numerous optimisation strategies and delivery techniques have been developed to improve plan quality and delivery efficiency, with particular emphasis on energy layer optimisation (ELO), reflecting the rapid evolution of PArc delivery and optimisation methodologies. More recently, these technological developments have enabled the first reported clinical implementations of step-and-shoot PAT for head and neck cancer (HNC) [6,7]. PArc has been proposed to further improve dose conformity and reduce normal tissue exposure, albeit with a potential increase in the low-dose bath [8]. Additional proposed advantages include a lower integral dose, a reduced risk of treatment-related toxicities and secondary primary malignancies, and improvements in treatment workflow and delivery efficiency [9]. Importantly, the increased degrees of freedom associated with PArc may also facilitate enhanced dose-averaged linear energy transfer (LETd) optimisation and redistribution compared with conventional particle therapy techniques [10,11,12]. Despite these potential advantages, several technical and clinical challenges remain. Preliminary investigations have identified prolonged delivery times associated with the number of energy jumps and the energy layer switching time (ELST) [13,14], as well as concerns related to range uncertainties and interfractional robustness [15,16,17,18]. These factors highlight the need for further technical development in delivery systems and continued advances in optimisation strategies to enable widespread clinical implementation. Over the past decade, research investigating PArc has expanded rapidly, with numerous studies exploring its dosimetric feasibility and potential clinical advantages across a range of tumour sites. However, the literature remains highly heterogeneous and is predominantly composed of in silico treatment planning and proof-of-concept studies, with considerable variation in planning methodologies, delivery strategies, and evaluation metrics. Accordingly, a comprehensive synthesis of the available evidence is required. This scoping review aims to systematically map and synthesise the current literature on PArc, by summarising dosimetric outcomes, biological considerations, delivery feasibility, and technological developments across particle types and tumour sites. 2. Materials and Methods This scoping review was conducted in accordance with the Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) 2018 guideline [19]. The review protocol was not registered with any registry. Eligibility criteria were defined a priori and are detailed in Table A1. Peer-reviewed journal articles or extended peer-reviewed conference proceedings published from 2010 onwards and written in English were eligible for inclusion. Eligible studies evaluated PArc using protons or carbon ions in direct comparison with conventional particle therapy and reported quantitative or qualitative treatment-planning-related outcomes. Studies employing passive-scattering techniques as the particle arc intervention were excluded; however, passive scattering was permitted when used as the comparator for conventional particle therapy. Conference abstracts, review articles, clonogenic assay studies, and grey literature were also excluded. A systematic search strategy was developed in consultation with an academic librarian and was designed to map and synthesise the available evidence on PArc in cancer radiotherapy across all tumour sites. The search strategy, detailed in Table A2, was applied in MEDLINE® and Scopus, with the final search conducted in December 2025. In addition, the first 10 pages of Google Scholar were searched to identify relevant peer-reviewed journal articles that may not have been indexed in MEDLINE® or Scopus. All records retrieved from the database searches were imported into Covidence [20] for reference management and study screening. Duplicate records were identified and removed prior to screening. A total of 282 titles and abstracts were screened independently by two reviewers (I.N. and E.B.) according to predefined eligibility criteria (Table A1), with disagreements resolved by consensus and discussion with additional reviewers where required, resulting in 145 records being excluded. Full-text review of the remaining 137 articles was performed by the primary investigator (I.N.) using the same eligibility criteria (Table A1). Eligibility decisions were confirmed in consultation with the research team, resulting in a further 63 studies being excluded. In total, 74 studies were included in this review following the screening process (Figure 1). Data extraction and charting were performed by the primary reviewer (I.N.) using a predefined extraction table to facilitate structured comparisons and synthesis. Extracted data included study characteristics: authors, year of publication, country of origin, and study design; treatment planning parameters: tumour site(s), sample size, particle type, and arc delivery approach; dosimetric outcomes; plan quality metrics; and qualitative observations derived from visual analysis of dose distributions and dose–volume histograms. Due to heterogeneity in study designs, treatment techniques, and reported outcome metrics, quantitative meta-analysis was not feasible. Instead, the results were synthesised across multiple analytical frameworks—including tumour site, reported dosimetric outcomes, treatment plan quality and deliverability metrics, and particle type—to provide a comprehensive assessment of the feasibility of PArc. 3. Results and Discussion As a relatively novel treatment technique in radiotherapy, PArc has been explored through a range of treatment planning approaches, optimisation frameworks, and delivery strategies. The introduction of arc geometry enables delivery across a continuous or quasi-continuous range of beam angles using a rotating gantry or rotating chair. Arc configurations may be implemented as full, partial, or multiple arcs depending on tumour and organ-at-risk (OAR) location, with partial arcs typically used for peripheral targets to reduce normal tissue dose and full or multiple arcs used for centrally located tumours to improve conformity and coverage. Arc geometries may also be partitioned into sub-arcs and delivered using either coplanar or non-coplanar trajectories to enhance geometric flexibility. Treatment plan optimisation is a major focus in PArc, aiming to balance plan quality, robustness, computational efficiency, and delivery time. Two primary paradigms are employed: heuristic and direct optimisation. Heuristic approaches involve pre-selection of energy layers and spot placement, followed by iterative refinement, whereas direct optimisation performs energy layer selection, sequencing, and spot weighting simultaneously within a single optimisation step. A wide range of optimisation algorithms have been explored and applied in PArc, including proximal gradient descent and a fast iterative shrinkage-thresholding algorithm (FISTA) [14,21], coordinate descent and alternating direction method of multipliers (ADMM) [21,22], orthogonal matching pursuit [21,23], and mixed-integer programming (MIP) frameworks [14,24]. A defining aspect of PArc optimisation is the configuration of energy layers, including selection, reduction, sequencing, and filtering. Techniques such as energy layer reduction (ELR) [25], energy layer optimisation for spot-scanning proton arc therapy (ELO-SPAT) [26], early layer and spot assignment (ELSA) [13], and energy layer filtering (ELF) [16] have been proposed to address this challenge. These approaches aim to balance dosimetric quality and robustness with delivery complexity and duration, often quantified by the number of energy jumps and associated ELST. Several optimisation frameworks have been developed to address these challenges, including SPArc for protons and spot-scanning hadron arc (SHArc) for heavier ions, both of which have been widely investigated. SPArc, introduced by Ding et al. [5], is an advanced intensity-modulated proton therapy (IMPT)-based technique that follows a heuristic-iterative optimisation approach, combining robust optimisation with iterative refinement of control points, energy layers, and spot placement and weighting following pre-selection of energy layers and spots. SHArc therapy, developed by Mein et al. [27], extends this framework to light and heavy ions using a heuristic-sequential approach, in which arc parameters, including gantry angle sampling and associated energy layer and spot distributions, are pre-optimised prior to a multi-stage optimisation process. Heavy ions, particularly carbon ions, were of interest owing to their high-LET characteristics and potential to overcome hypoxia-induced radioresistance. Intensity modulation strategies in PArc differ in how Bragg peaks are distributed across target depth and gantry angles to achieve the desired 3D dose distribution. In conventional 3D spot-scanning (Figure 2a), pencil beams are laterally scanned across sequential energy layers extending from the proximal to distal target boundaries, distributing Bragg peaks throughout the target volume. In contrast, DET, introduced by Deasy et al. [3], replaces volumetric Bragg peak placement with pristine Bragg peaks positioned along the distal edge of the target for each gantry angle, requiring multiple beam energies to conform to variations in target depth (Figure 2b). Although DET can reduce the integral dose through sharper distal dose gradients, its reliance on distal-edge placement reduces robustness to proton range uncertainties. Mid-range proton arc therapy (MRPAT) [23] addresses this limitation by shifting the Bragg peaks to the mid-range water-equivalent path length (Figure 2c), improving robustness while maintaining delivery efficiency and potentially enhancing biological effectiveness through favourable LETd distributions. Single or mono-energetic beam strategies have been proposed to reduce delivery time by minimising the number of energy layers and transitions. Proton mono-energetic arc therapy (PMAT) [11] uses multiple single-energy partial arcs to deposit Bragg peaks near the centre of the target, enhancing LETd within the target while reducing the dose to surrounding tissues and improving delivery efficiency. Similar approaches have been investigated using single-energy proton modulated arc therapy [30,31]. Volumetric-modulated proton arc therapy (VPAT) [32] employs a constant beam energy with external modulation across sectors, enabling shorter delivery times while maintaining plan quality and robustness. The SpeleoFilter framework proposed by Smith et al. [33] utilises a patient-specific, 3D-printed beam filtration device to achieve uniform dose distributions from a single incident energy, reducing the number of beam energies and spots. More recently, a constant-energy-per-arc approach has been proposed to further reduce energy switching while maintaining plan quality through the use of multiple arcs [23]. PArc may be delivered using either static (‘step-and-shoot’) or dynamic (‘continuous’) methods (Figure 3). In static delivery, several energy layers are delivered at each gantry angle, whereas dynamic delivery involves continuous gantry rotation simultaneously with spot scanning and potential energy layer transitions, resulting in a spiral-like delivery pattern [1]. Static delivery has recently been translated into early clinical implementation through ELR [6], as well as step-and-shoot SPArc delivery [7] in patients with HNC. In contrast, dynamic delivery remains an area of ongoing technical development, with systems such as DynamicARC™ [34] demonstrating the feasibility of simultaneous spot-scanning PBS delivery and energy switching during continuous gantry rotation on the IBA Proteus®One proton machine. Emerging delivery and planning approaches include the integration of arc delivery with FLASH-based techniques [35,36], which deliver radiation at ultra-high dose rates (≥40 Gy s−1), and lattice radiotherapy [37,38], a form of three-dimensional spatially fractionated radiotherapy. Additional developments include strategies that combine conventional Bragg peak dose deposition with shoot-through proton beam layers [39,40], a technique designed to sharpen the beam penumbra. Dynamically collimated proton arc therapy (DC-PAT) [40,41,42] has also been proposed, integrating energy-specific collimation with arc delivery to enhance plan quality. 3.1. Clinical Site-Specific Findings Across 74 studies, PArc approaches were investigated in a wide range of clinical sites—most commonly brain, head and neck, lung, and prostate cancers, with fewer studies in pancreatic, liver, breast, and other thoracic malignancies. Of the included studies, 17 evaluated multiple tumour sites and reported site-specific dosimetric outcomes, while one additional study [37] combined results across multiple anatomical regions. Brain, head and neck, lung, prostate, and liver were the most frequently represented tumour sites among these multi-site investigations. Overall, studies primarily evaluated target and OAR dosimetry, LETd optimisation and enhancement, and optimisation strategies to improve plan quality, robustness, and delivery efficiency, with normal tissue complication probabilities (NTCPs) and other biological outcomes reported in a subset of studies. Reporting of dosimetric, biological, and delivery metrics was highly variable across studies, and the literature was predominantly focused on PAT, with comparatively fewer studies investigating CArc. 3.1.1. Central Nervous System A total of 30 studies investigated PArc for central nervous system tumours (Table 1), comprising 25 studies of primary brain tumours and eight studies of skull-base tumours, with some studies including both sites. The majority of studies focused on target and OAR dosimetry, beam and energy layer configuration, and LETd enhancement, with a few also reporting predicted risks of secondary cancers. Brain Across the 25 brain cancer studies, PArc generally demonstrated maintained or improved target coverage, frequent improvements in conformity, and variable effects on dose homogeneity relative to conventional proton or carbon-ion therapy techniques. OAR sparing was generally improved across studies, albeit not across all structures consistently. Dose reductions were consistently reported for optic structures [31,35,50,58], cochlea [31,44,56], and healthy brain [22,34,45,46,57], whereas brainstem and hippocampus dose metrics showed variations across and within studies. Integral dose findings were mixed, with Toussaint et al. [51] reporting a 17% reduction with PAT, whereas Wuyckens et al. [58] reported increases in integral dose with ELSA and Flexi-Arc relative to IMPT. Furthermore, Henjum et al. [50] demonstrated that PAT could either increase or decrease the integral dose in ependymoma patients—depending on the level of energy-layer pruning—compared with IMPT. Improvements in predicted toxicity outcomes were observed in biological modelling studies. Chang et al. [46] found reduced NTCP for brain necrosis, and Toussaint et al. [51] reported reductions in the predicted risk of radiation-induced secondary cancer. A larger low-dose bath was one of the most consistent trade-offs of arc-based delivery reported across studies [35,45,47,49,53]. LETd findings were predominantly favourable, with studies commonly reporting improved LETd distributions such that high LETd was redistributed from OARs to the target volume, consequently enhancing the biological dose to the tumour whilst reducing hotspots to surrounding OARs. Delivery efficiency findings were mixed and highly patient- and technique-specific, with some studies reporting reductions in total beam delivery time [24,32,34,44,58] and others reporting increases [42,46,48]. Skull Base Of the seven studies that investigated skull base tumours, the majority reported comparable target coverage with improved dose conformity but slightly degraded dose homogeneity. OAR sparing, including the brainstem, spinal cord, chiasm, cochleas, and optic nerves, was often found to improve with PArc compared with conventional particle therapy. Despite an occasional increase in the low-dose bath with PArc [39,41,48], the integral dose was reduced. Smith et al. [41] reported reductions in integral dose from 6.47 GyL with IMPT to 6.40 and 5.43 GyL with uncollimated PAT and DC-PAT, respectively, while Cong et al. [55] found a reduction from 41.43 GyL with IMPT to 32.77 GyL with SPArc. LETd investigations consistently reported reductions in OAR LETd; however, target LETd metrics varied across studies, demonstrating greater dependence on the specific arc technique. Beam delivery time increased with PArc [48,55]; however, Gu et al. [26] demonstrated a reduction in the ELST using an integrated energy layer selection and sequencing optimisation (ELO-SPAT) approach relative to SPArc, showing potential for improved delivery efficiency within PArc. 3.1.2. Head and Neck Across 27 studies (Table 2), PArc was investigated in a range of head and neck sites, including oropharyngeal, nasopharyngeal, and sinonasal cancers. Most of these studies used simultaneous integrated boost prescriptions with high- and low-risk target volumes. Optimisation strategies were commonly evaluated to improve plan quality, robustness, LETd distributions, and delivery efficiency, while NTCP modelling was frequently incorporated, particularly for xerostomia and dysphagia. Overall, PArc maintained target coverage relative to conventional particle therapy, while often improving conformity and, in many cases, homogeneity. Comparable target coverage was reported across multiple studies, along with reduced target overdose in a few [21,23,50,59,60,66,68]. Reductions in selected coverage metrics were reported in a minority of studies, particularly for low-risk volumes and CArc approaches [5,50,55,63,66]. Several studies reported improved conformity with PArc and, occasionally, improved homogeneity; however, these outcomes were strongly dependent on factors such as target volume size (high- vs. low-risk) and treatment technique. OAR sparing generally favoured PArc, particularly for serial OARs and salivary and swallowing structures. Consistent reductions in spinal cord and brainstem Dmax were reported [6,7,18,43,52,55,59,60,64,65,68], alongside reductions in Dmean to parotids, submandibular glands, the oral cavity, and pharyngeal structures [18,59,60,61,62,65]. These translated into consistent reductions in NTCP for xerostomia and dysphagia across multiple studies [16,43,60,61,65], with additional improvements reported for other toxicity endpoints [6,17,18,62]. Meanwhile, de Jong et al. [43] demonstrated consistent reductions in grade 2+ and grade 3+ xerostomia and dysphagia using various optimisation algorithms—including ELF, ELSA, and SPArc—compared with IMPT for oropharyngeal cancer. Furthermore, Tattenberg et al. [70] evaluated quality-adjusted life expectancy (QALE) in oropharyngeal cancer and reported increases of up to 0.8 quality-adjusted life years (equivalent to 9.6 months) with PAT compared with IMPT. The integral dose was frequently reduced, although increases were observed in specific cases [50], and several studies reported an increased low-dose bath despite improved intermediate- and high-dose sparing [17,59,63,67,68]. Although LETd-related findings were more heterogeneous, they generally suggested that PArc can reshape LETd distributions in potentially favourable ways while also introducing some biological trade-offs. In oropharyngeal cancer, de Jong et al. [60] reported maintained mean LETd within the high-risk target volume and reduced LETd in the parotid and submandibular glands with PAT. However, LET1% increased in the spinal cord and brainstem despite reductions in both physical and relative biological effectiveness (RBE)-weighted dose to these structures—a finding also reported by Reiners et al. [69]. CArc studies generally reported improved LET metrics, with LET1% and LET50% increasing within the target and decreasing within the brainstem [63,66]. Delivery time was highly dependent on arc implementation, beam and energy layer configuration, and machine assumptions. Several studies reported longer delivery times for PArc under conventional spot-scanning delivery methods [5,15,55,60], while others demonstrated comparable or reduced delivery times through reductions in ELST [5,59] or optimisation strategies [6,7,16,23,26,40,43,60,64]. 3.1.3. Thoracic A total of 23 studies (Table 3) investigated PArc for thoracic malignancies. Most of these studies focused on lung cancer, with a smaller number evaluating oesophageal cancer [58,71], breast cancer [52,72], and other thoracic malignancies, including mediastinal Hodgkin’s lymphoma and thoracic spine tumours [52]. These studies commonly investigated energy-layer and robust optimisation strategies, with a focus on plan quality, deliverability, and the impact of breathing motion, as well as setup and range uncertainties. Lungs Most studies investigating lung cancer found that tumour coverage was generally maintained, with occasional reductions in target overdose. Target dose conformity was consistently improved. Target dose homogeneity varied across the studies and only occasionally improved, as reported by Sanchez-Parcerisa et al. [31], where the homogeneity index (HI) was reduced with PAT relative to IMPT but increased with mono- and bi-energetic modulated PAT (ME-PAT, and BE-PAT). PArc showed consistent OAR sparing for lung cancer across healthy lung tissue, as well as the oesophagus, spinal cord, and heart. Only one lung study reported LETd outcomes, finding that the target LETmean increased from 3.16 keV μm−1 with IMPT to 3.21 and 3.80 keV μm−1 with PAT and ME-PAT, although it decreased to 2.92 keV μm−1 with BE-PAT [31]. Delivery time outcomes across the lung studies were dependent on the optimisation algorithm implemented in the treatment planning. For instance, the majority of the studies using SPArc often found an increase in the total delivery time [24,55,78,79], unless the ELST was reduced to less than 1 s, in which case delivery times were comparable or reduced [76]. Various studies that developed or implemented ELO algorithms found reduced ELST [26] or total delivery time [23,32,38,58,64,77]. Robustness was generally reported to be maintained or improved with PArc compared with conventional particle therapy. Breast For breast malignancies, PArc maintained target coverage with modest improvements in conformity and homogeneity relative to PBS-based techniques. Chang et al. [72] demonstrated improved conformity index (CI) (0.78 vs. 0.77, p = 0.044) and HI (1.05 vs. 1.08, p = 0.005) with SPArc, alongside significant reductions in doses to the heart, left anterior descending artery (LAD), and ipsilateral lung (p = 0.001). These dosimetric improvements translated to reduced predicted risks of major coronary events (p = 0.003), pneumonitis (p = 0.005), and acute skin toxicity (p = 0.007). Delivery time increased at higher ELST but was comparable when ELST ≤ 1 s. Similarly, Johnson et al. [52] reported improved conformity and dose gradients with PArc, with reductions in ipsilateral lung and LAD doses; however, secondary cancer risk modelling showed structure-dependent variation, with reduced risk to ipsilateral OARs but increased risk to contralateral structures. Robustness was evaluated only by Chang et al. [72], who reported comparable target coverage under range and setup uncertainties together with improved robustness of several OAR dose metrics using SPArc. The same study also evaluated respiratory motion interplay using 4D dynamic dose calculations and reported reduced dosimetric impact of interplay with SPArc relative to vertical IMPT. However, these findings were limited to breast cancer patients with relatively small respiratory motion (<2 mm), and more recent evidence suggests that interplay effects may already be minimal in conventional fixed-field proton therapy for breast cancer [81]. Oesophageal Across the two oesophageal cancer studies, PAT achieved comparable target coverage to IMPT, with improved CI and comparable HI [58,71]. OAR sparing was generally improved, with consistently reduced spinal cord D0.05cc, lung V20, and heart V25 and V40. The integral dose varied between studies, decreasing with PArc in the study by Chocan Vera et al. [71] by 1.8% (p = 0.1) compared with IMPT, but increasing in the study by Wuyckens et al. [58] (Flexi-Arc: 160.9; ELSA: 147.2 vs. IMPT: 136.9 GyL). Predicted toxicity outcomes showed small but significant reductions in pericardial effusion (0.5%, p < 0.001) and pneumonitis (0.1%, p = 0.001). Delivery time was generally longer with PAT than with IMPT; however, Wuyckens et al. [58] reported reduced beam delivery time with Flexi-Arc compared with ELSA, bringing it closer to that of IMPT. Both studies also demonstrated that PAT maintained robustness against setup, range, and breathing motion uncertainties while preserving target coverage and OAR sparing. Other Thoracic Sites In mediastinal lymphomas and thoracic spine tumours, PAT maintained target conformity relative to IMPT [52,75]. Lung V20Gy, oesophageal Dmean, LAD Dmax, and spinal cord Dmax were generally reduced, although trade-offs were observed, including increased low-dose breast exposure or heart dose in selected cases. Organ equivalent dose-based modelling suggested overall reductions in secondary cancer risk for most thoracic OARs, including the oesophagus, lungs, and thyroid, with occasional increases observed for the breast. Liu et al. [75] demonstrated that SPArc delivery times were comparable to IMPT when ELST was ≤1 s. 3.1.4. Abdominal Across abdominal sites, 12 studies investigated PArc (Table 4), including five liver, four pancreatic, and three general abdominal studies. These studies predominantly focused on LETd enhancement and delivery efficiency using a range of optimisation algorithms. Liver For liver malignancies, PArc generally maintained target coverage, while the majority of studies reported improved conformity relative to IMPT. In hepatocellular carcinoma, Liu et al. [82] reported improvements in target coverage metrics, including increased V67.5 and reduced Dmax, alongside improved CI (0.72 vs. 0.53, p = 0.02), and similar HI with PArc relative to IMPT. Reductions in dose to several OARs were reported across studies, including the spinal cord, healthy liver tissue, and kidneys. Model-based analyses suggested potential clinical benefits, with Liu et al. [82] reporting an 8.24% reduction in the predicted probability of radiation-induced liver disease with PArc relative to IMPT (p = 0.002). LET-focused analyses also demonstrated favourable biological dose redistribution, with increased target LETmean (2.33 vs. 2.47 keV μm−1, p = 0.020) and reduced LETmean in adjacent OARs such as the kidneys and healthy liver [82]. Delivery efficiency varied across studies. Beam delivery time was generally longer with PArc compared with IMPT [24,55], although some optimisation strategies were able to mitigate this effect [23]. PArc plans maintained robustness under setup and range uncertainty scenarios. For example, Ma et al. [23] reported improved robustness variance (RV) metrics, with reduced CTV RV110 from 23.48% with IMPT to 19.64% with BOMP and 9.94% with PAT, while RV40 for healthy liver remained comparable across plans (IMPT: 1.64 vs. PAT: 1.84, BOMP: 1.49%). Pancreas For pancreatic cancer, all identified studies evaluated carbon-ion SHArc approaches, with no proton-based PArc studies reported. Across these investigations, SHArc generally maintained or improved target coverage relative to conventional carbon-ion radiotherapy (CIRT) while enhancing dose conformity. Target and OAR dose metrics were reported only by Baltazar et al. [83], who found statistically significant increases in D95 and V95 with SHArc using monitor unit MU-based energy-layer selection compared with CIRT (44.00 vs. 41.57 Gy and 90.77 vs. 83.03%, respectively; both p < 0.05). The maximum spinal cord dose was also reduced; however, body V4.8GyRBE increased from 9.86% with CIRT to 20.98% and 21.47% with SHArc using central and MU-based energy-layer selection strategies, respectively (p < 0.05), indicating an increased low-dose bath. Despite this, no significant differences were observed in NTCP endpoints for gastrointestinal haemorrhage, obstruction/perforation, or diarrhoea. cancers-18-02443-t004_Table 4Table 4Summary of key findings for particle arc therapy in the treatment of abdominal malignancies.Study IDStudy Cohort (n)/ PrescriptionIntervention/ ComparatorStudy Focus/ Planning ContextTarget DosimetryOAR DosimetryLETd OutcomesDelivery TimeMein et al. [47]Pancreatic cancer (n = 1); 3 GyRBE × 1 fxSHArc-C vs. IMCTLETd evaluationTarget dose conformity improved based on isodose distributions.–High LETd was centralised within the CTV for SHArc-C (70–140 keV μm−1), whereas IMCT displaced high LETd into normal tissue (LETmax ≈ 100 keV μm−1), with a reduced LETmean in the CTV (≈50 keV μm−1).–Wuyckens et al. [14]Computational abdominal phantom (n = 1); 60 GySPArc; MIP; FISTA vs. IMPTOptimisation algorithm comparisonD98% increased with SPArc and MIP (57.7 and 58.5 Gy) and decreased with FISTA (55.0 Gy) relative to IMPT (56.8 Gy). CI improved with SPArc (1.3), MIP (1.1), and FISTA (1.2) vs. IMPT (1.4). HI reduced with SPArc (2.9) and MIP (2.2), and marginally with FISTA (4.0), vs. IMPT (4.2).––ELST reduced with SPArc and MIP (18.8 and 4.8 s) but increased with FISTA (28 s) vs. IMPT (24.2 s).Wuyckens et al. [24]Liver cancer (n = 1); 67.5 GyRBE × 15 fxMIP-PAT [Dt, dt, dT] †, SPArc vs. IMPTBi-criteria Pareto optimisationD98% comparable for SPArc and IMPT (66.17 vs. 66.22 Gy) and varied across MIP optimisations (66.37/66.27/60.12 Gy). CIRTOG improved with MIP (1.14/1.00/0.57) but increased with SPArc (1.59) relative to IMPT (1.52). HI generally improved with MIP (1.04/1.05/1.25) but increased with SPArc (1.09) relative to IMPT (1.08).Spinal canal D5% decreased with MIP (4.53/4.77/3.16 Gy) and SPArc (6.45 Gy) relative to IMPT (7.9 Gy). External D5% generally improved with MIP (1.25/0.72/0.10 Gy) but increased with SPArc (2.66 Gy) relative to IMPT (1.10 Gy).–Delivery time varied by optimisation objective for MIP (102.30/40.10/11.20 s) and increased for SPArc (65.80 s) relative to IMPT (54.50 s).Liu et al. [35]Liver cancer (n = 1); 40 GyRBE × 5 fxSPLASH vs. IMPTVoxel-level FLASH dose-rate optimisationCI improved for SPLASH relative to IMPT (0.65 vs. 0.55).A slight increase in low-dose bath was observed with SPLASH.––Liu et al. [82]Hepatocellular carcinoma (n = 10); 67.5 GyRBE × 15 fxPAT vs. IMPTLETd evaluationV67.5 GyRBE improved (97.84 vs. 96.30%). CI improved (0.72 vs. 0.53, p = 0.006). HI increased (1.06 vs. 1.04, p = 0.02).Liver Dmean reduced (10.62 vs. 14.70 GyRBE, p = 0.002). Spinal cord Dmax increased (0.19 vs. 0.03, p = 0.08). Kidneys V14Gy reduced (2.35 vs. 6.88%, p = 0.063). The median predicted probability of RILD reduced by 8.24% (p = 0.002).CTV LETmean improved (2.47 vs. 2.33 keV μm−1, p = 0.020). Liver LETmean reduced (0.81 vs. 0.99 keV μm−1, p = 0.137). Kidneys LETmean reduced (0.24 vs. 0.26 keV μm−1, p = 0.063).–Mairani [53]Pancreatic cancer (n = 1); 38 GyRBE × 12 fxSHArc-C vs. IMCT; LET-optimised IMCT (LET-IMCT)LETd evaluationComparable target coverage. Dose conformity and homogeneity improved based on qualitative analyses of isodose distribution and DHV, respectively.Low-dose bath to normal tissues increased, as shown in DVH.GTV LETmin improved with SHArc-C relative to IMCT and was comparable to LET-IMCT (47.2 vs. 38.4 and 48.6 keV μm−1). GI-tract LETmax reduced with SHArc-C relative to IMCT and LET-IMCT (47.5 vs. 88.0 and 83.0 keV μm−1).–Zhu et al. [38]Abdominal case (n = 1); 20 GyRBE to target volume.PAT; PAT with ELO vs. IMPTLATTICE; EL and LETd optimisationCIvalley improved for PAT and ELO relative to IMPT (0.77 and 0.76 vs. 0.62). CIpeak improved for PAT and ELO relative to IMPT (0.51 and 0.53 vs. 0.43).Stomach Dmean reduced with PAT and ELO relative to IMPT (0.51 and 0.59 vs. 0.69 Gy).–NEJ and ELST were reduced with ELO relative to PAT (53 vs. 71; 379.9 vs. 874.0 s), resulting in a reduced estimated delivery time (564.9 vs. 1058 s) comparable to IMPT.Baltazar et al. [84]Anthropomorphic pancreas phantom (n = 1); 4 Gy to PTVSHArc-C vs. CIRTLETd optimisationTarget dose conformity improved, as indicated by isodose distributions.–GTV LET98% and LET2% increased with SHArc-C relative to IMPT (76.3 vs. 63.5 keV μm−1 and 101.3 vs. 73.0 keV μm−1).Delivery time increased with SHArc-C (20 vs. 5 min).Baltazar et al. [83]Pancreatic cancer (n = 7); 48 GyRBE × 12 fxSHArc-C (Central EL; MU-based EL) vs. CIRTEL selection; LETd optimisationD95% improved (CEN: 43.0; MU: 44.0 vs. 41.57 GyRBE, p < 0.05 for MU only). V95% increased for the MU approach (90.77 vs. 83.03%, p < 0.05). HI reduced (CEN: 0.16; MU: 0.14 vs. 0.2).Spinal cord Dmax reduced (CEN: 17.49; MU: 19.25 vs. 32.48 GyRBE, p < 0.05). Body V4.8 GyRBE increased (CEN: 21.47; MU: 20.98 vs. 9.86%, p < 0.05), indicating a larger low-dose bath.GTV LET98% improved (CEN: 54.8; MU: 57.2 vs. 42.6 keV μm−1, p < 0.05). GI-tract LET1% reduced (CEN: 50.2; MU: 53.6 vs. 85.5 keV μm−1, p < 0.05).Delivery time increased for SHArc-C relative to CIRT (25–28 vs. 7–10 min).Cong et al. [55]Liver cancer (n = 1); 67.5 Gy × 15 fxSPArc vs. IMPTSynchrotron-based PBSD99% decreased (66.02 vs. 66.64 Gy). CI improved (2.11 vs. 2.38).Spinal cord Dmax reduced (8.46 vs. 25.85 Gy). Oesophagus Dmean increased (7.73 vs. 7.18 Gy). ID reduced (80.18 vs. 97.67 GyL).–Dynamic delivery time increased for SPArc (416.86 vs. 277.61 s).Ma et al. [23]Liver cancer (n = 1)PAT vs. IMPTConstant-energy-per-arc multi-arc (BOMP optimisation)Dmax reduced (BOMP: 110.8; PAT: 116.3 vs. 116.4%). CI improved (BOMP: 0.84; PAT: 0.83 vs. 0.79).Dmean reduced across liver (BOMP: 11.18; PAT: 12.50 vs. 14.31 Gy), spinal cord (BOMP: 1.69; PAT: 4.10 vs. 10.49 Gy), and skin (BOMP: 1.35; PAT: 1.62 vs. 1.64 Gy).–Total plan delivery time reduced with BOMP (259.9 s) but increased with PAT (2184.9 s) relative to IMPT (354.4 s).Rothwell et al. [36]Anthropomorphic abdominal phantom (n = 1); 60 Gy × 4 fx (per-fraction dose-rate evaluation)PFAT vs. IMPTProton FLASH-arc therapy (PFAT) feasibility; PMAT optimisation [11]Dmax and Dmin increased (17.2 vs. 16.9 Gy; 11.4 vs. 10.5 Gy). HI increased (1.20 vs. 1.16).Spinal cord Dmax increased (0.1 vs. 0.0 Gy). External-GTV Dmax and V40Gy/s increased (16.4 vs. 15.8 Gy and 80.1 vs. 12.3%).High LETd centralised in the GTV core for PFAT, whereas IMPT placed high LETd predominantly outside of the target volume within surrounding healthy tissue.–Dosimetric and radiobiological metric definitions and abbreviations are provided in Table A3. CIRT, carbon-ion radiotherapy; FISTA, fast iterative shrinkage-thresholding algorithm; GI, gastrointestinal; RILD, radiation-induced liver disease; NEJ, number of energy jumps; † MIP-PAT results correspond to three optimisation objectives (D_t, d_t, d_T), representing different Pareto-optimal trade-offs. D_t denotes a MIP-PAT solution with high weighting on the dose-fidelity objective relative to the BDT objective; d_T denotes a solution with high weighting on the BDT objective relative to dose fidelity; d_t represents a balanced weighting of both objectives. A trade-off was observed between target coverage and OAR sparing, with D_t optimisation favouring target dose metrics and d_T optimisation favouring OAR sparing and delivery efficiency. Summary of key findings for particle arc therapy in the treatment of abdominal malignancies. Dosimetric and radiobiological metric definitions and abbreviations are provided in Table A3. CIRT, carbon-ion radiotherapy; FISTA, fast iterative shrinkage-thresholding algorithm; GI, gastrointestinal; RILD, radiation-induced liver disease; NEJ, number of energy jumps; † MIP-PAT results correspond to three optimisation objectives (D_t, d_t, d_T), representing different Pareto-optimal trade-offs. D_t denotes a MIP-PAT solution with high weighting on the dose-fidelity objective relative to the BDT objective; d_T denotes a solution with high weighting on the BDT objective relative to dose fidelity; d_t represents a balanced weighting of both objectives. A trade-off was observed between target coverage and OAR sparing, with D_t optimisation favouring target dose metrics and d_T optimisation favouring OAR sparing and delivery efficiency. All studies included LET-focused analyses and demonstrated favourable LETd redistribution with SHArc. Carbon-ion SHArc studies reported centralisation of elevated LETd within the target relative to CIRT, with increased target LETd, as demonstrated by Baltazar et al. [84], where LET2% increased from 73.0 keV μm−1 with CIRT to 101.3 keV μm−1 with SHArc. Correspondingly, significant reductions in gastrointestinal-tract LETmax were reported by Mairani [53] (47.5 vs. 88.0 keV μm−1) and Baltazar et al. [83] (50.2 and 53.6 vs. 85.5 keV μm−1, for central and MU-based energy-layer selection, respectively; p < 0.05). Delivery time was substantially increased with SHArc relative to CIRT, with reported beam delivery time increases of up to approximately 200% [83,84]. Additionally, SHArc exhibited greater sensitivity to uncertainties and interfractional anatomical changes, resulting in reduced robustness compared with CIRT [23,83,84]. General Abdominal Similar to the other abdominal sites, PArc demonstrated improvements in conformity and target dose distribution in both general abdominal phantom studies [14,36] and case-based studies [38]. OAR sparing varied across studies, with Zhu et al. [38] reporting reductions in mean stomach dose with standard PAT and PAT with ELO compared with IMPT (0.51 and 0.59 vs. 0.69 Gy), while Rothwell et al. [36] reported modest increases in maximum spinal cord and external dose. However, Rothwell et al. [36] observed redistribution of high LETd toward the GTV core, as well as an increase in V40Gy/s with proton FLASH-arc therapy compared with IMPT (80.1 vs. 12.3%). Together, these changes may contribute to reduced healthy tissue toxicity. Delivery efficiency was dependent on the optimisation strategy, with Wuyckens et al. [14] reporting reduced ELST with SPArc and MIP but increased ELST with FISTA relative to IMPT, while Zhu et al. [38] showed that ELO reduced the estimated beam delivery time from 1058 s to 564.9 s, comparable to IMPT. 3.1.5. Pelvic PArc was evaluated for the treatment of prostate cancer across 11 studies (Table 5), with one study also including analysis of a female pelvic cancer case [32]. The studies mostly focused on target and OAR dosimetry using various delivery and optimisation techniques. Target coverage was generally maintained relative to conventional particle therapy, with comparable or improved dose conformity and homogeneity [22,35,85,86]. Improved OAR sparing with PArc was commonly reported across studies, with reductions in Dmean [52,55,85] and high-dose volumes [13,32,85] to the rectum, bladder, and femoral heads. Although some studies found an increase in low-dose bath with PArc [13,35,74], others reported overall decreases in integral dose. Johnson et al. [52] estimated secondary cancer risk using the organ equivalent dose ratio (OEDIMPT/OEDPAT) within a mechanistic model, reporting reduced risk in the rectum (1.8) but a slightly increased risk in the bladder (0.8). A few studies investigated LETd with SPArc [74] and SHArc [47,48], demonstrating that PArc could spatially redistribute elevated LETd toward the target, concentrating high LETd within the prostate while minimising high LETd within surrounding normal tissues compared with IMPT. Most prostate studies reported increased delivery time with arc-based techniques, with delivery typically around 30–50% longer for SPArc/SHArc [48,55,74], but substantially longer (up to 200%) when ELSA optimisation was implemented [13]. Notably, optimisation of ELST (<0.1 s) substantially improved efficiency, demonstrating comparable delivery times [85], while strategies employing single-energy beams demonstrated reduced delivery times [32]. Robustness was generally comparable across studies, except for Mein et al. [48], who reported that carbon-ion SHArc therapy with robust optimisation was less robust than intensity-modulated carbon-ion therapy. 3.2. Dosimetric Outcomes 3.2.1. Target Coverage and Dose Conformity Across the included studies, PArc generally maintained target coverage relative to conventional particle therapy, consistently achieving clinical dose coverage criteria. In a subset of studies, target overdose was reduced with specific optimisation algorithms across various treatment sites [21,22,23]. These findings indicate that arc-based delivery does not compromise fundamental dose coverage requirements, even in anatomically complex treatment sites. Improvements in conformity were frequently observed, with several studies reporting values closer to the ideal (CI → 1) with PArc compared with conventional particle therapy. This trend was particularly evident in sites with irregular or concave target geometries, where the increased angular sampling of arc delivery enables improved dose sculpting. However, this trend was not universal, with some studies reporting comparable or reduced conformity depending on optimisation strategy and arc implementation, highlighting the technique- and patient-specific nature of these outcomes. Figure 4 illustrates this trend by plotting the difference in the absolute deviation from the ideal CI value of 1 between conventional particle therapy and PArc. This approach standardised comparisons across studies employing different CI formulations. Across 37 studies, the median relative improvement in conformity index was 6.3% (IQR 1.7–15.5%), with 31 of 37 studies reporting improved conformity compared with conventional particle therapy. Overall, PArc improved or maintained target dose conformity across a range of tumour sites and conformity index formulations. However, the most commonly used CI metric—the Radiation Therapy Oncology Group (RTOG) conformity index—also exhibited the greatest variability, with some studies favouring conventional particle therapy and no consistent trend observed across specific tumour regions. This variability may reflect known limitations of the RTOG formulation, which has been reported to be less robust than alternative indices such as van’t Riet, Lomax, and Saint-Anne, Lariboisière, and Tenon (SALT) indicies [87]. Homogeneity results were more variable. While some studies demonstrated improved or comparable homogeneity, others reported modest reductions with PArc relative to conventional particle therapy. This variability likely reflects the trade-off between conformity and homogeneity inherent in inverse planning, particularly when additional optimisation objectives (e.g., LETd or OAR sparing) are introduced. Notably, reduced target dose homogeneity may not necessarily be detrimental, as biological dose robustness does not account for factors such as hypoxia, and hotspots or increased LETd within the tumour core may provide a compensatory biological advantage [27,88]. Moreover, Lucia et al. [89] demonstrated that in patients with brain metastases treated with fractionated stereotactic radiotherapy, inhomogeneous dose distributions were associated with improved local tumour control (93% vs. 78%) and lower rates of radionecrosis (0% vs. 7.5%) than homogeneous dose distributions. Overall, these findings suggest that PArc can achieve clinically acceptable target dose distributions, with potential advantages in conformity under appropriate optimisation conditions. 3.2.2. Organ-at-Risk Sparing OAR dose reduction was one of the most consistently reported advantages of PArc. Across multiple tumour sites, reductions in mean and intermediate-to-high-dose volumes were observed, reflecting the ability of arc delivery to redistribute the beam entrance dose over a larger volume of normal tissue. This resulted in reduced dose to many critical structures, particularly those adjacent to or overlapping with the target; however, this was often accompanied by an increase in the low-dose bath. Although this may appear counterintuitive, the integral dose reflects the total energy deposited across all irradiated tissues and is influenced by both the dose level and the volume irradiated. Consequently, the increase in low-dose exposure was outweighed by larger reductions in intermediate-to-high-dose regions, resulting in a net reduction in integral dose. As a result, the increase in low-dose exposure was generally considered clinically acceptable. As shown in Figure 5, the median relative reduction in integral dose across studies was 10.5% (IQR 3.3–17.0%). Overall, 22 of 25 studies demonstrated reduced integral dose with PArc compared with conventional particle therapy, although increases were observed in a small number of studies, primarily in central nervous system and thoracic malignancies. A skin-sparing effect with PAT was also consistently reported across studies, irrespective of tumour site. Reported improvements included reductions in mean skin dose [23,72], maximum skin dose [5,41,44,85], high-dose skin volumes [52], and predicted skin toxicity [72]. These findings suggest that distributing the beam entrance dose over multiple beam angles may reduce skin dose while maintaining target coverage. Several studies further explored the clinical implications of these dosimetric changes through NTCP modelling and secondary cancer risk estimation. Reductions in predicted NTCP were consistently reported across multiple organs and endpoints, particularly for xerostomia and dysphagia in patients with HNC, reflecting improved sparing of salivary glands and surrounding critical structures. Similar trends were observed in other anatomical regions, where decreased OAR dose was associated with lower predicted risks of radiation-induced toxicities [46,71,72,76,77,78,82,86]. Decreases in estimates of secondary cancer risk were also associated with PArc [51,52,70]. Collectively, these findings suggest that, despite an increased low-dose bath, the overall dose redistribution achieved with PArc may result in a net reduction in both acute toxicity and long-term malignancy risk. 3.2.3. LET Enhancement A key advantage of PArc identified across the literature is its ability to modulate and enhance LETd distributions. These effects were most frequently reported in central nervous system (brain and skull base) and abdominal tumour sites, with studies involving carbon-ion therapy more commonly evaluating LETd and often demonstrating a greater magnitude of enhancement. Increased LETd within the target volume, particularly in the tumour core, was frequently achieved alongside reductions in LETd to surrounding OARs. This spatial redistribution is enabled by the increased flexibility in beam angle selection and energy layer configuration afforded by arc delivery. By steering high-LET regions away from critical structures and concentrating them within the tumour, PArc may enhance biological effectiveness without increasing the physical dose. These effects have important implications for dose escalation strategies, particularly in radioresistant tumours such as pancreatic cancer [47], where improved tumour control remains a key challenge. Increased LETd within the target may enhance tumour cell killing and help overcome radioresistance, while maintaining or reducing normal tissue toxicity. However, the extent of LETd enhancement was highly dependent on whether specific LETd optimisation was implemented, in addition to the optimisation strategy, delivery constraints, and anatomical context. As such, the development of robust and clinically feasible LETd-guided optimisation frameworks remain an important area for future research. 3.3. Plan Quality and Deliverability 3.3.1. Robustness Across the included studies (n = 74), a majority implemented both robust optimisation and robust analysis (n = 41), with smaller subsets employing only robust optimisation or evaluation, or no robustness framework at all. Robust optimisation approaches were broadly consistent, with most studies adopting setup uncertainties of 2–5 mm and range uncertainties of approximately 2–3.5%, typically evaluated using 21 worst-case scenarios. More advanced implementations were observed in recent studies, including voxel-wise optimisation strategies, incorporation of robustness objectives directly on target volumes or OARs, and 4D robust optimisation to account for organ motion. Overall, robustness was most commonly improved or maintained; however, reduced or mixed robustness was also frequently reported, indicating strong dependence on tumour site, delivery strategy, and evaluation method. Robustness to setup and range uncertainties was generally achievable under standard uncertainty settings, although challenges were noted for CArc techniques [83,84], particularly with single-energy beams or control points [48,54,63]. In these cases, trade-offs among robustness, target dose conformity and homogeneity, delivery efficiency, and LETd enhancement were observed. More recently, a novel treatment strategy proposed by Li et al. [66]—LEt bOost by heavy Particle Arc RaDiation (LEOPARD)—was shown to mitigate robustness limitations associated with SHArc, while maintaining dose conformity and LETd benefits. Interfractional robustness emerged as a key limitation, particularly in sites with substantial anatomical variation, where the increased conformity of arc plans may reduce tolerance to anatomical changes and necessitate adaptive planning [6]. This was most evident in HNCs, which are particularly susceptible to anatomical changes throughout treatment [7,15,18,23,43,63,68], although several studies also reported comparable or improved robustness in this setting [5,6,16,40,59,60,61,62,66]. The use of range shifters may improve robustness in these cases and reduce the need for replanning [15,18,43]. In contrast, intrafractional robustness, including interplay effects, was often improved for thoracic sites, particularly for lung treatments, where motion management and arc delivery reduced sensitivity to respiratory motion [72,74,76,79,80]. Collectively, these findings indicate that while PArc can improve or maintain robustness, outcomes remain highly dependent on the treatment context. 3.3.2. Treatment Delivery Efficacy Delivery efficiency findings were mixed and highly dependent on technical and planning parameters. This variability was consistently linked to key factors, including ELST, optimisation algorithms, beam configuration, and delivery strategy. The treatment planning system, delivery system, and beam configuration parameters for each included study are summarised in Supplementary Table S1. While several studies demonstrated reduced beam delivery time with PArc, others reported increased delivery times relative to conventional particle therapy, most commonly for SPArc [12,46,55,78] and CArc techniques [48,63,66,83,84]. However, when ELST was reduced to <1 s, SPArc delivery times were often comparable to or lower than IMPT [5,44,59,72,74,75,85]. Several optimisation frameworks were shown to reduce ELST or total beam delivery time relative to conventional implementations, including ELO-SPAT [26], ELF [43], filtered PAT [64], and MIP approaches [14,24], although their effectiveness remained technique- and site-dependent. Delivery efficiency was also improved through several planning strategies, including the implementation of autobeam sequencing [60], subplan partitioning [16], reduction in the number of beams [6], and the use of uncollimated shoot-through techniques, all of which were shown to reduce beam delivery time while maintaining plan quality. In addition, delivery techniques such as the SpeleoFilter [33] and VPAT [32] reduced dependence on ELST by enabling single-energy delivery through external energy modulation, while delivery systems such as proton dynamic arc delivery [34] facilitated continuous gantry rotation. Overall, these findings indicate that delivery time in PAT is not inherently increased or reduced but is instead dependent on optimisation strategy, machine characteristics (particularly ELST), and delivery configuration. 3.4. Particle Type At present, PAT has progressed further toward clinical implementation than CArc, primarily due to the greater availability of proton therapy facilities and the relative compatibility with existing delivery systems. Proton therapy is substantially more accessible worldwide, with approximately 129 proton facilities and only 17 carbon-ion facilities [2]. This imbalance reflects the greater cost and technical demands associated with heavy-ion therapy. In particular, the large and complex gantry systems required for carbon ions present significant challenges for arc delivery, although alternative approaches such as upright patient positioning using a rotating chair have been proposed [27,47,48,54]. Despite these limitations, carbon ions offer several physical and biological advantages, including a sharper Bragg peak, reduced lateral scattering, and higher LETd, resulting in increased RBE. These properties reduce dependence on tissue oxygenation and cell-cycle effects, making CArc particularly well suited for the treatment of hypoxic and radioresistant tumours. However, variable RBE models and the characteristic fragmentation tail of carbon ions—which increases the exit dose beyond the distal fall-off of the Bragg peak—introduce additional uncertainties and may reduce robustness to setup and range variations compared with protons [27,57]. Across the 10 studies investigating CArc, research predominantly focused on radioresistant tumour sites, including pancreatic cancer [47,53,83,84], glioblastoma [47,48,53,54], HNC [63,66], prostate cancer [47,48], multiple brain metastases [57], and skull base chordomas [48]. Compared with PAT, CArc demonstrated greater biological dose enhancement. All studies evaluated LETd enhancement and consistently demonstrated increased tumour LETd, particularly within the tumour core, together with reductions in high-LETd regions of surrounding OARs. These biological advantages may improve tumour control by overcoming hypoxia-induced radioresistance, with SHArc-based approaches demonstrating enhanced tumour cell killing under hypoxic conditions in both in vitro [54] and computational studies [27]. While target coverage and OAR sparing were generally maintained or improved, an increased low-dose bath was frequently observed, consistent with PAT [27,48,53,54,83], although this did not appear to impact plan quality in a clinically meaningful way. Furthermore, several studies suggested that the biological benefits of enhanced LETd and improved management of tumour radioresistance may outweigh modest reductions in target dose homogeneity and increased target hotspots [27,63,84]. Similar to PAT, robustness outcomes with CArc were variable. Several studies reported reduced robustness, particularly in the absence of robust optimisation [53,54,84]; however, studies incorporating robust optimisation occasionally demonstrated comparable or improved robustness relative to conventional carbon-ion approaches [27,66]. Overall, CArc highlights a fundamental trade-off among LETd enhancement, target dose homogeneity, and robustness, commonly referred to as the “LET trilemma” [57,63,84,90]. Emerging strategies, such as LEOPARD [66], aim to address this trade-off by combining IMPT with SHArc to enhance LETd in hypoxic tumour regions while simultaneously improving dose conformity, robustness, and delivery efficiency, thereby providing a more balanced optimisation across competing planning objectives. 3.5. Early Clinical Implementation Early clinical implementation of PAT has only recently begun, with the first treatments reported in late 2024–2025. At the Trento Proton Therapy Centre (Italy), PAT has been delivered to HNC patients, with initial reports demonstrating high delivery accuracy, reflected by a mean patient-specific QA (PSQA) gamma passing rate (3%, 3 mm) of 99.6%, and feasible treatment times of approximately 12–24 min depending on arc configuration [6]. Since then, clinical implementation has expanded to other anatomical sites, with more than 50 patients treated or currently undergoing PAT to date [91]. Similarly, step-and-shoot SPArc has been implemented clinically for HNC at Corewell Health (USA), demonstrating delivery feasibility with PSQA gamma passing rates (3%, 3 mm) exceeding clinical thresholds (95%). However, recorded step-and-shoot SPArc treatment times were longer than simulated IMPT delivery (16.7 min vs. 6.32 min), whereas dynamic SPArc was estimated to reduce treatment time to approximately 4.31 minutes, comparable to conventional delivery [7]. These early studies represent important milestones towards routine clinical implementation of PAT, although dynamic arc delivery remains under active development. 3.6. Summary of Key Findings Overall, PArc consistently maintained target coverage and improved dose conformity, although effects on target dose homogeneity were more variable. While some studies reported reduced homogeneity, this may be beneficial in hypoxic tumours, where heterogeneous dose distributions could enhance tumour control. Improvements in OAR sparing were consistently observed, alongside reduced integral dose; however, these benefits were sometimes accompanied by an increased low-dose bath, reflecting a trade-off between low-dose and intermediate–high-dose regions. Notably, increases in low-dose volumes were often offset by reductions in intermediate-to-high-dose volumes, which may be more clinically relevant, as reflected in improved biological outcomes such as reductions in NTCPs and secondary cancer risk. In addition, PAT demonstrated consistent improvements in skin sparing across tumour sites, suggesting that redistribution of the entrance dose with arc delivery may reduce the skin dose and, consequently, the risk of radiation-induced skin toxicity. Furthermore, PArc demonstrated improved LETd distributions by redistributing high-LETd regions from OARs to the target volume, which may be particularly beneficial for hypoxic tumours, especially within the brain and abdomen. Robustness and delivery efficiency were less consistent, with reduced robustness often observed in thoracic and abdominal sites owing to motion and anatomical variability, and delivery time largely dependent on machine capabilities and optimisation strategies. These findings are analogous to the well-established comparison between IMRT and VMAT, where arc-based delivery increases the low-dose bath while improving conformity and reducing intermediate-to-high-dose volumes. However, unlike VMAT, reduced treatment times are not yet consistently observed with PAT compared with IMPT using current technology. Overall, the findings support the feasibility and potential clinical benefits of PArc, although outcomes remain dependent on tumour site, arc technique, and optimisation strategy. 3.7. Limitations and Bias 3.7.1. In the Literature PArc remains a relatively novel research area, with the current literature largely limited to feasibility, in silico, and proof-of-concept studies, alongside investigations focused on optimisation algorithm development. Patient cohorts were typically small (often n≤10, with several case studies n=1), limiting statistical power and the ability to account for inter-patient variability or derive site-specific clinical recommendations. Additionally, studies were concentrated on a limited number of disease sites—predominantly brain, head and neck, lung, and prostate cancers—with several phantom-based investigations further limiting clinical applicability. Importantly, because most studies were computational in nature, implementation of patient-specific quality assurance was limited, and comparisons between planned and measured dose distributions were infrequent, restricting validation of delivery accuracy. Additionally, beam delivery time was frequently estimated rather than experimentally validated, introducing uncertainty due to reliance on machine-based assumptions. Clinical implementation remains in its early stages, having only commenced in mid-2024 [6,7]; consequently, clinical outcome data are limited. Further limitations arise from considerable diversity in treatment planning approaches, evaluation methodologies, and modelling frameworks, making direct comparison between studies challenging. A wide range of treatment planning systems (TPSs), including numerous in-house platforms, were used, with no currently available standardised guidelines for PArc planning. Considerable variation existed in arc delivery techniques, including static versus dynamic delivery methods, the number and angular range of arcs, and beam configuration parameters, such as the number of beams/control points, sampling frequency, and the use of mono-energetic beams, with these choices often influenced by system capabilities and computational constraints. This methodological heterogeneity is illustrated in Supplementary Table S1, which summarises the treatment planning systems, delivery systems, and beam configuration characteristics reported across the included studies. Optimisation strategies varied substantially across studies, with a wide range of dose and energy-layer optimisation algorithms, alongside differences in whether LETd and robust optimisation were incorporated. For dose optimisation, algorithms also differed depending on whether they were developed in-house or implemented within commercial TPSs. This may introduce variability in optimisation performance and limit standardisation, reproducibility, and comparability across studies, thereby affecting the validity and reliability of the reported outcomes. Robust optimisation is recommended for particle therapy to ensure that target coverage and OAR constraints are maintained under range and setup uncertainties, while also accounting for inter- and intrafractional variations, including day-to-day anatomical changes and organ motion [8,31]. However, approximately one-third of studies did not implement robust optimisation, and this, together with a lack of established criteria for robustness evaluation, remains a limitation. Modelling and evaluation of biological endpoints are subject to several limitations. Tumour control probability (TCP) and NTCP models are often derived from photon therapy data [78,82], reducing their accuracy when applied to particle therapy. For secondary cancer risk, both linear and mechanistic models are associated with inherent uncertainties in dose–response relationships and underlying model parameters [52,70], thereby limiting the reliability of conclusions regarding biological effects. A lack of standardised reporting frameworks across studies led to heterogeneity in reported planning parameters and outcome data, ultimately precluding meta-analysis. Variation in the definition and calculation of key metrics, such as CI (Section 3.2.1), further limits the comparability and synthesis of findings. Finally, inconsistencies within individual studies—including discrepancies between reported values across tables, figures, and text—limit the interpretation of the reported outcomes and, consequently, reduce confidence in the reported conclusions. 3.7.2. In This Review A structured systematic search across multiple databases (MEDLINE and Scopus) and Google Scholar was conducted to minimise bias. However, restriction to English-language studies and exclusion of conference abstracts, review articles, and non-peer-reviewed sources (e.g., theses/dissertations) may have resulted in relevant studies being missed. A formal meta-analysis was not feasible due to substantial heterogeneity in study designs, outcome definitions, and reported metrics. Moreover, the inclusion of qualitative data alongside quantitative metrics meant that some findings relied on author interpretation, potentially introducing bias and reducing the reliability of the review conclusions. Finally, the findings reflect only the evidence available up to the final search conducted in December 2025. 3.8. Research Gaps and Future Directions Significant gaps remain in the evaluation and clinical translation of PArc, primarily due to the lack of treatment planning guidelines, reporting protocols, and robust clinical evidence. Current studies are largely limited to small cohorts, with limited statistical analysis and minimal Monte Carlo-based validation. Additionally, research is predominantly concentrated in brain and HNCs, with limited evidence in abdominal and thoracic sites. Future work should therefore prioritise the development of standardised evaluation frameworks and larger, more diverse patient cohorts to identify disease sites most likely to benefit from PArc. In particular, for tumours with poor clinical outcomes, further investigation is needed to determine whether arc-based delivery can enable improved tumour control through dose escalation strategies while maintaining acceptable normal tissue dose constraints. Despite increasing interest in biological optimisation, its integration remains limited, with relatively few studies incorporating LETd-guided optimisation. Furthermore, by consensus, a constant RBE of 1.1 is currently assigned to protons, although this is known to underestimate RBE in the Bragg peak [48]. Future work should therefore focus on refining LET-based optimisation strategies and evaluation frameworks—particularly in the context of tumour radioresistance and hypoxia—and assessing the implementation of variable proton RBE models to properly incorporate biological robustness into treatment planning. In addition, further investigation into the impact of arc-based techniques on radiation-induced lymphopenia is warranted. While reductions in integral dose and dose to secondary lymphoid organs may be beneficial, the increased low-dose bath associated with PArc may adversely affect circulating lymphocytes due to their high radiosensitivity. The overall effect therefore remains uncertain and therefore requires further modelling, given its association with survival outcomes [92,93]. Finally, addressing uncertainties related to motion, anatomical variation, and range errors through improved motion management and robustness evaluation will be critical for the safe clinical implementation of PArc. From a technical perspective, further research is required to develop clinically viable optimisation algorithms that balance plan quality, delivery efficiency, and computational feasibility. A key barrier to clinical translation is the limited integration of PArc techniques into commercial treatment planning and delivery systems, with only a small number of commercial implementations currently available. Technological advancements in delivery systems, energy modulation, and gantry design are also essential for widespread clinical implementation, with additional challenges for CArc related to limited facility availability, high infrastructure costs, RBE modelling, and delivery feasibility, particularly for continuous arc delivery. Collectively, these gaps highlight that while PArc demonstrates significant potential to improve treatment outcomes, substantial methodological, biological, and technological challenges must be addressed before it can be routinely implemented in clinical practice. 4. Conclusions PArc demonstrates significant potential to improve radiotherapy outcomes through enhanced dose conformity, improved OAR sparing, and favourable LETd redistribution. Despite these promising dosimetric and biological advantages, PArc remains in an early stage of development, with current evidence largely limited to in silico and feasibility studies. Key challenges include optimisation complexity, motion and range uncertainties, limited integration of biological optimisation, and the need for technological advancements to support efficient delivery and implementation of PArc techniques into commercial treatment planning and delivery systems. Future research should focus on establishing standardised methodologies, integrating biological optimisation strategies, improving robustness and delivery efficiency, and validating outcomes through prospective clinical studies. Addressing these challenges will be essential to fully realise the clinical potential of PArc and enable its routine clinical implementation.