The current status of FLASH particle therapy: a systematic review

Particle therapies are becoming increasingly available clinically due to their beneficial energy deposition profile, sparing healthy tissues. This may be further promoted with ultra-high dose rates, termed FLASH. This review comprehensively summarises current knowledge based on studies relevant to proton- and carbon-FLASH therapy. As electron-FLASH literature presents important radiobiological findings that form the basis of proton and carbon-based FLASH studies, a summary of key electron-FLASH papers is also included. Preclinical data suggest three key mechanisms by which proton and carbon-FLASH are able to reduce normal tissue toxicities compared to conventional dose rates, with equipotent, or enhanced, tumour kill efficacy. However, a degree of caution is needed in clinically translating these findings as: most studies use transmission and do not conform the Bragg peak to tumour volume; mechanistic understanding is still in its infancy; stringent verification of dosimetry is rarely provided; biological assays are prone to limitations which need greater acknowledgement.


Introduction
It is recommended that X-ray radiation therapy (RT) be included in the treatment of 50% of all cancer patients in developed countries worldwide, making it one of the most common modes of treatment available currently [1]. Accounting for only 5% of total cancer therapy costs, as well as being a non-invasive technique, it is appealing in both its low economic burden and practicality [2]. Although there are significant advantages to utilising conventional X-ray RT, this modality has the potential to cause detrimental off-target effects, inherent in the physics involved. Due will allow for a lower incidence of negative side effects in patients, and even allow for greater radiation doses to be delivered to improve the probability of tumour control [8].
Particle therapy is one such modality that is continually being developed and currently being employed clinically for this reason. It employs the use of high Linear Energy Transfer (LET) particles to irradiate treatment volumes with minimal exit dosage compared to X-ray photons (low LET), reducing irradiation of surrounding healthy tissues. X-rays deliver their maximum dose at the point of electronic equilibrium and then attenuate exponentially as they pass through the remaining tissue, prior to and upon exiting the tumour. Conversely high LET particles, including protons or carbon ions, have low energy deposition upon entry and maximum deposition at the Bragg Peak towards the end of their track. Particle therapy centres are now employing intensity modulated and pencil beam scanning techniques, which target tumours in layers of irradiation spots at differing doses, depths, and positions [9]. The benefits of such treatment include better dose conformation, minimisation of damage to healthy tissues and an increased total dose that can be administered [10]. Use of high LET particles, particularly carbon ions, allows for efficient treatment of radioresistant tumours localised within critical organs where X-ray or proton beam therapies would be ineffective.
The resurgence of studies involving irradiation of tissue at ultra-high dose rates, also known as FLASH radiation therapy (FLASH-RT), alongside development of proton and carbon ion therapies, are steps towards further improving the therapeutic efficacy of RT. FLASH-RT typically delivers absorbed dose rates greater than 40 Gy/s to irradiate tumours in very small timeframes (< 1 s), as opposed to conventional dose rates (CONV, < 0.1 Gy/s) which are normally fractionated and delivered over the course of minutes [11]. There is some evidence that the utilisation of FLASH technique, even with photon or electron beams, results in a radioprotective effect in normal tissues, which reduces toxicity compared to CONV. Multiple mechanisms have been hypothesised to explain this phenomenon, including rapid intracellular oxygen depletion preventing indirect DNA damage via reactive oxygen species (ROS) [12], alterations to the nature of DNA damage and DNA repair pathways, and immune response modulation [13]. However, the radiobiological processes by which these mechanisms occur during FLASH-RT are not well understood. All these factors are thought to contribute towards the unique normal tissue sparing effect of FLASH-RT, which has been previously demonstrated in vivo for photons, electrons, and heavy ions.
Incorporating the differential tissue response of the FLASH concept with the normal tissue sparing of particle therapy may prove to be beneficial when utilised in combination. Taking advantage of proton or carbon ion beam's characteristics -particularly their minimal exit dose -alongside the normal tissue sparing properties afforded by FLASH, there is the potential to synergistically minimise healthy tissue toxicities. Although there are multiple studies outlining the biological outcomes of the FLASH effect with electrons and X-ray photons [14][15][16], few studies have explored the implications of using proton or carbon ions at ultra-high dose rates in vitro and in vivo. This systematic review summarises available literature detailing FLASH-RT's discovery, its implementation in electron, proton and carbon ion-based studies, as well as discussion of conflicting data concerning the various hypotheses that form the basis of these studies. Overviews of theoretical, in vitro, and in vivo particle beam studies relevant to ultra-high dose rates, and their effect upon normal and cancer cell biology, are the primary focus with the aim of understanding current state of knowledge on mechanistic benefits that could further improve particle therapy outcomes in the future. From literature captured by the systematic search strategy, this review provides: a brief history of electron FLASH experiments; its recent resurgence in the literature in terms of in vitro, in vivo, and even clinical application; and the current state of research into FLASH delivery of protons and carbon ions.

Systematic review process
To produce an accurate overview of the biological mechanisms of the FLASH effect, a systematic review of all literature pertaining specifically to particle-FLASH (excluding electrons) has been compiled. Since electron-FLASH publications were amongst the most prevalent throughout initial FLASH searches, the key electron-FLASH studies were included in a non-systematic manner in this manuscript to provide background and context to the particle-FLASH studies identified throughout the systematic process. The progression of the electron-FLASH field has been essential in providing preliminary data that will aid in translation into the particle-FLASH field.
The inclusion criteria for literature within review include: • Scientific articles. Conference abstracts and reviews were excluded. • Findings only pertaining to the radiobiological effect of ultra-high dose irradiation on normal and cancerous cellular function and response. Technical developments pertaining to proton and carbon-ion delivery were excluded. • In silico, in vitro, in vivo, and clinical studies.
• Studies utilising ultra-high dose rates, defined as an average dose rate of greater or equal to 40 Gy/s. • Articles published between the years 2009-2022. • Language restricted to English (translations also accepted).
In completing this process systematically, multiple search criteria were utilised with increasing degrees of complexity (additional search terms, higher search specificity). Table 1 below outlines the set of search criteria used to identify proton and carbon-FLASH literature performed on the 8th December, 2022. Studies from search #4 of Table 1 were added to the studies found in the initial literature searches. In addition to utilising Scopus and Medline to compile resources, other articles were found via selecting relevant citations from systematically identified literature (pearling). Finally, additional sources were identified via grey literature searching with Google Scholar. This resulted in the culmination of 110 publications, which were then screened by IK to validate their relevance, followed by EB to resolve any disagreement in relevance between IK and JA. The final number of systematically identified proton and carbon therapy papers was 34. A diagrammatic summary of the PRISMA search strategy utilised for this systematic review is displayed in Fig. 1. PRISMA guidelines were followed in the literature identification and exclusion process [17]. Relevant electron-FLASH papers identified in a non-systematic manner were summarised in table format (Table 2), with systematically identified particle-FLASH literature being summarised in the same manner (Table 3), listing year of publication, particle type, irradiation source and mode, model (e.g., in vitro, in vivo), particle energy (MeV), doses (Gy), delivery time, pulse/burst count and rate, FLASH average and instantaneous dose rate per fraction (Gy/s), comparative low or CONV dose rate, (Gy/s) and key experimental conclusions. Throughout the text, average dose rates are discussed unless specified.

History of FLASH radiotherapy
Although the first publications that cite the use of radiation at ultra-high dose rates did not initially quote the term 'FLASH' irradiation, coined by Favaudon et al. in 2014 [18], interest in the field was motivated earlier by exploring the relationship between radiation dose rate and biological response. In 1958, Kirby-Smith and Dolphin published one of the first papers that suggests a dose rate dependent effect on DNA lesion formation, showing that in a Tradescantia (spiderwort plant) model, the total number of chromosome aberrations was reduced when electrons were administered at higher dose rates of up to 4 × 10 6 Gy/s [19]. The effect of cellular oxygenation on cellular radiosensitivity was also topical, and in 1959 Dewey and Boag presented a radiobiological relationship between dose rate, cellular oxygenation, and cellular survival. They aimed to observe the effect of dose rate on the surviving fraction of Serratia marcescens bacteria, whilst modifying both total irradiation dose and the concentration of dissolved oxygen. Bacteria were irradiated in the presence of both CONV X-rays (0.1 Gy/s) and ultra-high dose rate electrons (5 × 10 7 -10 × 10 8 Gy/s) with total absorbed doses ranging between 0 and 200 Gy, and also varying oxygen concentrations (100% O 2 , 1% O 2 in N 2 and 100% N 2 ). The proportion of surviving bacteria was higher at the ultra-high dose rate irradiation, which was comparable to the radioprotection afforded under anaerobic conditions at the conventional dose rate [20]. This was theorised to be a result of the initial absorbed dose of radiation 'consuming' the oxygen dissolved within the irradiated bacterium, thus preventing the production of harmful reactive oxygen species that would usually result in cytotoxic DNA damage and apoptosis. Also lending to the theory, the authors suggested that the irradiation timeframe was so minute, that even if dissolved oxygen was not completely removed from the extracellular environment, it would not be able to reenter the cell rapidly enough for the radiation pulse to dissociate it. This theory, termed the 'oxygen effect', would consequently form part of the hypothesis behind the healthy tissue sparing effects of FLASH irradiation. Similar publications around this time held interest in the relationship Table 1 Search strategy for proton-and carbon-FLASH articles, utilising an expanded set of FLASH, proton, and carbon therapy search terms. A total of 62 experimental studies were yielded from Scopus and Medline searches after duplicates were removed Search # Search Terms Scopus Medline 1 "FLASH-RT" OR "FLASH radiation therapy" OR "FLASH radiotherapy*" OR "FLASH effect*" OR "FLASH irradiation" OR "Ultra-high dose rate*" 662 396 2 "Proton therapy" OR "Proton beam therapy" OR "Proton radiation therapy" OR "Proton radiotherapy" OR "Proton beam irradiation" 12,562 8,447 3 "Heavy ion radiotherapy" OR "Heavy ion radiation therapy" OR "Carbon ion therapy" OR "Carbon ion irradiation" OR "Carbon ion radiotherapy" OR "Carbon ion radiation therapy" OR "Carbon beam therapy"

Zebrafish studies
Montay Gruel et al. (2019) displayed a FLASH sparing effect in zebrafish embryos [36]. In this study, alterations to the length of FLASH irradiated zebrafish embryos was significantly less than CONV irradiated embryos 5 days postfertilization, where irradiations occurred at 4 h postfertilization. When zebrafish were preincubated with an antioxidant, FLASH exhibited no further sparing effect, whilst the embryos were spared from CONV induced radiation damage [36]. Another zebrafish embryo model was used by Pawelke et al. (2021) to validate whether the pulse dose rate and oxygen levels used during irradiations masked a potential FLASH effect. 26 Gy of 20 MeV electrons at a continuous, conventional dose rate of 0.1116 Gy/s were compared to FLASH with a mean dose rate of 1 × 10 5 Gy/s, and zebrafish embryos were irradiated within vessels at differing pO 2 levels. A sparing effect was observed in this study at pO 2 levels below atmospheric oxygen levels (< 148 mmHg) with a greater degree of protection afforded at < 5 mmHg (hypoxic conditions). Whilst FLASH-RT appeared to have a mild protective effect over CONV-RT at both pO 2 levels, at high pO 2 levels, FLASH and CONV-RT zebrafish morphologies exhibited minimal differences, whereas the effect was more substantial at low pO 2 [48]. This included preventing the reduction of embryo length and eye diameter by 4%, as well as a 20% reduced rate of PE and SC in comparison to low dose rate, quasi-continuous irradiation.
between dose rate and ROS concentration [21], and if there were analogous FLASH effects in E. coli [22]. Further work looking into the dose rate effect in the survival of mammalian cells was conducted by C.D. Town in 1967. Using 3.5 × 10 7 Gy/s electrons, experiments were conducted comparing HeLa cell survival as a function of radiation dose, between single and double pulses of radiation. When cells were exposed to doses higher than 10 Gy, there was an apparent sparing of tissues; cells treated with doses administered in a single burst of radiation had a higher survival fraction compared to irradiation over two pulses. To validate whether this tissue sparing was related to the oxygen effect, an additional test, wherein cell suspensions were exposed to either nitrogen or air, was performed. Single pulse irradiations under aerated conditions resulted in similar cell survival curves to that of irradiation under anaerobic conditions (at doses exceeding 10 Gy) [23]. These findings, similar to Dewey and Boag's S. marcescens experiments, supported that the 'oxygen effect' is not limited to bacterial cells; indeed, there was sufficient evidence alluding to its role in mammalian cell protection. In the present day, the definition of the 'FLASH effect' phenomenon has formed as the sparing effect ultra-high dose rates have upon healthy, but not cancerous, tissues in vivo, whereas this was not the primary aim of these initial studies. Nevertheless, they highlight the discovery of dose-rate dependent radiobiological responses, and paved the way for FLASH's re-emergence as a potential cutting-edge treatment modality decades later.

Resurgence of FLASH
Studies concerning ultra-high dose rate irradiation faded into obscurity for some 30 years until its sudden resurgence in the 2010s. Although in previous publications in vitro experiments predominated, more recent work aimed to identify how ultra-high dose rates benefited normal tissues in vivo. Radiation-induced pulmonary fibrosis is a side effect [24] with the potential for treatment-related death following external beam irradiation [25]. As such, multiple in vivo studies have been performed to determine whether utilising ultra-high dose rates may elicit a greater differential response between normal and tumour tissue damage compared to current clinical dose rates. A summary of the experimental results of these electron-FLASH papers is presented in Table 2    0.09 FLASH reduced apoptosis in neurogenic brain regions 1-day post irradiation. Vascular dilation and downregulation of tight junction proteins occurred 1-4 weeks after CONV treatment but was minimal in FLASH mice, thereby preserving vasculature integrity in the brain.

cells,
BRγc-/-mice  0.1 Antitumour efficacy of irradiations was not affected by use of hypofractionated regimens; FLASH and CONV irradiations were equipotent in the delay of GBM tumour growth in mice. The discrimination index of mice exposed to FLASH irradiations in the 10 Gy, 2 × 7 Gy, and 3 × 10 Gy groups showed no significant difference to unirradiated controls.

Table 2
(continued) with regards to increasing total dosages (thereby improving tumour control) whilst also mitigating side effects.
Recently, a similar in vivo study by Kondradsson et al. (2021) attempted to identify potential adverse effects and treatment procedures required to administer electron-FLASH in canines with microscopic residual disease and spontaneous superficial solid tumours. Ten patients were prescribed doses ranging from 15 to 35 Gy, with average dose rates ranging between 400 and 500 Gy/s. 11 out of 13 irradiated tumours presented with either a complete response, partial response, or stable disease. Out of the 10 patients, only one experienced a grade 3 adverse skin event after a 35 Gy dose to the nasal planum, characterised by moist desquamation. Other adverse events appeared to be mild during follow up examinations 3-6 months posttreatment, including alopecia, dry desquamation, and erythema [50]. Akin to the feline-based study conducted by Vozenin et al., the key limitation to this preclinical trial was no comparative conventional dose rate group to compare adverse effects and disease response efficacy. It would be greatly beneficial for future in vivo work with canine and feline models to include conventional dose rate groups so that more accurate comparisons may be drawn. However, the feasibility in completing comparative experiments such as these may be difficult due to ethics considerations.

First-In-Human case study
The first-in-human case study of FLASH irradiation was reported in 2019 on a 75-year-old male with a multi-resistant cutaneous lymphoma, which had metastasised to multiple sites across his skin's surface [34]. The patient had experienced unfavourable side effects from the previous treatment of approximately 110 tumour sites with either MV/kV X-rays or low energy electrons. Regardless of the relatively low radiation doses and fractionation regimens employed (20 Gy in 10 fractions or 21 Gy in 6 fractions), skin continued to respond poorly, requiring up to 4 months to heal damage at irradiation sites 3-4 cm in diameter. FLASH was therefore administered due to its previous implications in the sparing of healthy tissues whilst maintaining tumour control efficacy comparable to that of conventional dose rates. In this trial study, 5.6 MeV electrons at a dose rate of 167 Gy/s and dose of 15 Gy was administered to one of the patient's most progressive ulcero-infiltrating tumours 3.5 cm in diameter. A 5 mm bolus was utilised, resulting in a 90% isodose coverage depth of 1.3 cm. Grade 1 epithelitis presented at 3 weeks post-FLASH, however 5 months after therapy almost all trace of negative skin reactions from treatment had receded (Figs. 2, [34]).
This study presents some valuable qualitative results depicting the temporal evolution of the treated area to

Mice brain studies
With the prospect of FLASH minimising healthy tissue damage, its effectiveness in the treatment of brain tumours is of extreme relevance clinically, particularly in paediatrics. In this respect, in vivo whole brain irradiation electron-FLASH studies are prevalent. Alaghband et al. (2020) displayed an apparent radioprotective effect in brain tissue of mice following FLASH irradiation. After exposure to 8 Gy,6 MeV electron irradiation at 4.4 × 10 6 Gy/s (FLASH), neurocognitive test results were indistinguishable from the control group, whereas conventional irradiation at 0.077 Gy/s with 6 MeV electrons caused substantial cognitive detriment. The benefit of FLASH, when utilised in the brain, was reduced neurocognitive impairment, attributed to the preservation of neurogenetic niche and neurogenesis. Mice irradiated at conventional dose rates observed a lower proportion of mature and immature neurons 4 months after irradiation. Additionally, a two-fold reduction in plasma growth hormone expression was also observed a week post treatment in comparison to controls, suggesting conventional irradiation also impairs pituitary gland function. These side effects were not observed in FLASH-irradiated mice [38]. Other whole brain irradiation studies displayed similar findings at and exceeding 100 Gy/s [32], including but not limited to: 200 Gy/s, 300 Gy/s [37], 2.5 × 10 3 and 5.6 × 10 6 Gy/s [39,41].

Higher order animal models
Demonstration of the electron FLASH effect is not limited to in vivo mice experiments, with implications in both mini-pig test animals and cat patients. Mini-pig irradiation at 300 Gy/s compared to 0.083 Gy/s markedly reduced damage to skin tissues, avoiding signs of acute toxicity including inflammation, ulcer formation and hair follicle destruction [33]. Additionally, late skin fibronecrosis was limited to CONV irradiated pigs, with FLASH experiencing no tissue complications at 36 weeks after 28-34 Gy irradiations. Six cat patients diagnosed with squamous cell carcinoma of the nasal planum were also FLASH treated, resulting in a progression free survival (PFS) of 84% at 16 months [33]. Side effects were limited to acute mucositis in half of the patients, and depilation across all cat patients. Although the sample size of this study is limited, it is some of the first evidence displaying no notable toxicity of FLASH in higher mammal models. One key limitation is that there are no experimental groups within this study irradiated at conventional dose rates; only retrospective data are referred to in drawing conclusions in this aspect. PFS ranged from 50 to 80% within other conventional dose rate trials [49]. This is promising in its potential applications in human patients, utilise the high LET Bragg peak region for proton-FLASH irradiations.
Buonanno et al. (2019) presented one of the first studies to analyse the long-term radiobiological effects of FLASH proton irradiation in vitro, particularly in non-cancerous human cells. At three different instantaneous dose rates: 0.05 Gy/s, 100 Gy/s or 1000 Gy/s, IMR90 lung fibroblast cells were irradiated with 0-10 Gy of 4.5 MeV protons and a colony forming assay conducted. Altering dose rate did not appear to have a dramatic impact upon cell survival, with no significant difference between the fraction of surviving cells across all dose rates and total doses administered. This aligns with previous studies observing the effect of proton dose rate on cellular survival, with dose rates exceeding 10 9 Gy/s having no effect on colony formation compared to conventional dose rates for Human umbilical vein endothelial cells (HUVEC) [58], V-79 Chinese Hamster cells [59], or human-hamster hybrid cells (A L ) [60]. Interestingly, this effect carries over into cancer cell clonogenicity, with human derived cervical cancer cell line HeLa [61,62] exhibiting no dose rate dependence upon cellular survival post irradiation.
In quantifying DNA damage, exposure to 20 Gy absorbed proton dose delivered at a FLASH instantaneous dose rate of 1000 Gy/s displayed a statistically significant reduction in γH2AX fluorescence 30-minutes post-irradiation compared to both 0.05 Gy/s and 100 Gy/s protons [63]. Although this finding initially appears promising, it is limited by the immunolabelling techniques used to acquire data. Specifically, the resolution at which microscopy was performed (40x) and an apparent plateau of γH2AX fluorescence for doses ≥ 10 Gy suggests that accurate distinction of foci could not be obtained due to saturation above this threshold. At a dose of 20 Gy, approximately 700 doublestrand breaks (DSBs) could be expected considering the formation roughly 35 DSBs/1 Gy [64]. One may assume that such a high number of γH2AX foci would prevent definitive resolution of the true number of DSBs formed, and thus lead to the saturation observed ≥ 10 Gy. This is a factor that needs to be taken into serious consideration in future work, as this will determine the doses at which the FLASH effect maximally prevents normal tissue DNA damage. Additionally, these results seem to contradict other studies analysing dose rate effects on DSB production via γH2AX [65][66][67] and 53BP1 [66,68] immunofluorescence, which show a non-significant induction of these foci between FLASH and CONV dose rates.
Further tests to quantify markers of cellular senescence and inflammation (at 20 Gy, 1000 Gy/s) also showed reduced proportions of β-galactosidase positive cells 1-month post-irradiation (40% fewer vs. CONV) and TGFβ1 induction 24 h and 1-month (4.4-fold and 4.3-fold provide comparisons between healthy and irradiated tissue. The fact that this is the first clinical study using FLASH dose rates is also substantial. However, it is limited in that they exclusively analysed the short-term effect of FLASH on cancerous skin tissue alone; the effect that electron FLASH would have on other organs of different tissue depths and morphologies based on this data is unknown. Regardless, this result bodes well for further clinical translation, as limiting inflammation is still greatly beneficial for patients experiencing radiation induced skin complications. Further translation clearly requires additional indications to display the benefits of this treatment modality, (e.g., irradiation of brain, lung, or liver tumours) would represent more complicated anatomy where improvement in outcomes is needed. Additionally, there were no conventional dose rate irradiations conducted within this study to directly compare the FLASH treatment to; all irradiations were FLASH only [34].

In vitro studies
Although proton-FLASH studies are not as widespread as those conducted with electrons or X-ray photons, many of these models also reinforce the relationship between ultrahigh dose rates and normal tissue protection. It is of note that save for proof-of-concept dosimetry planning by Wei et al. [51] and 6 experimental studies [52] [53][54][55][56][57], herein to our knowledge, that no other papers either explicitly state or However, an apparent differential effect was observed for 16 Gy irradiations, with 100% of FLASH mice surviving versus ~ 40% at the CONV dose rate. In addition, FLASH mice exhibited better weight retention 9 days post-irradiation. Hematoxylin and eosin stains were also performed to observe late effects 90 days post-irradiation. Both dose rates showed mild signs of inflammation with inflammatory cells infiltrating intestinal villi. Remarkably, FLASH irradiated intestines appeared to exhibit reduced signs of inflammation compared to CONV protons, evidenced by a lower proportion of infiltrating inflammatory cells and thinner layer of hyperplastic submucosa and muscularis [73]. In syngeneic mice models, there is evidence that proton-FLASH produces an improved tumorigenic response due to increased recruitment and infiltration of CD3 + T cells within the tumour microenvironment [70]. These findings may be reflective of the ability of FLASH to minimise interference with, or perhaps aid in stimulating, the immune response.
Similar to the previously described electron-FLASH experiment conducted by Pawelke et al. [48], Beyreuther et al. [69] also performed zebrafish embryo irradiations, but with 224 MeV protons at a CONV dose rate of 0.08 Gy/s and FLASH at 100 Gy/s. The endpoints of this study were relative embryonic survival and the rate at which morphological changes occurred, including pericardial edema (PE) and spinal curvature (SC). Only 23 Gy irradiation appeared to exhibit statistically significant differences between CONV and FLASH embryos with PE at 3-and 4-days postirradiation; dose rate did not appear to have a substantial effect upon any other endpoints or at other doses [69]. Akin to in vitro studies, the Bragg peak is not reported to conform to targeted tissues throughout these experiments.

First-In-Human clinical trial
In December of 2020, recruitment began for the first proton-FLASH clinical investigation, conducted at Cincinnati Children's Proton Therapy Center [83]. The focus of this study was to assess the workflow of a palliative FLASH treatment for bone metastases in the extremities of 10 patients aged 18 years or older. FLASH treatments were delivered at a dose rate of 51-61 Gy/s, with single dose regimen of 8 Gy being utilised on a total of 12 metastatic sites across all patients. This was followed by assessment of radiation related normal tissue toxicities and adverse side effects in addition to pain response, use of pain relief and pain flare [84]. Key results from the study include 6 out of the 12 treated sites experiencing a complete pain relief response, 2 of 12 sites reporting partial pain relief, and 4 of 12 sites experiencing pain flare following treatment. Twelve adverse events were reported within this patient group, with eleven of these being classed as grade 1 including skin hyperpigmentation, decrease respectively) post-irradiation [63]. At 100 Gy/s, the only significant difference was ~ 20% fewer β-gal positive cells compared to CONV, which may suggest that higher dose rates more effectively mitigate long-term cellular senescence.
These data scratch the surface of which factors modulate the proton-FLASH effect, illustrating that tissue protection may be more heavily reliant upon modifying gene expression in response to radiation exposure, especially concerning inflammatory and senescent cellular pathways. A summary table outlining all systematically identified proton and carbon-FLASH publications is presented in Table 3.

In vivo studies
One of the first papers to demonstrate in vivo FLASH radioprotection using protons was published in 2020 by Diffenderfer et al. [72]. A 230 MeV proton beamline with dose rates of 78 Gy/s for FLASH and 0.9 Gy/s for conventional irradiation were utilised in 15 Gy whole abdominal irradiations of C57BL/6J mice. Akin to studies performed with electron irradiation, FLASH irradiated mice exhibited greatly reduced levels of acute intestinal damage, defined by an increased number of EdU positive cells within abdominal crypts and increased crypt regeneration compared to the conventional dose rate. Moreover, where symptoms of fibrosis were severe in conventionally irradiated mice, FLASH irradiated tissues displayed intestinal morphology similar to that of non-irradiated tissues 8-weeks post irradiation. In a follow up experiment testing for tumour growth and control probability in a flank injection, murine pancreatic cancer model, no difference was observed between either dose rate used and doses of 12 and 18 Gy [72]. This aligns with previous claims that proton-FLASH has comparable tumour kill efficiency to conventional dose rates. Yet, this data is exemplary in highlighting the favourability of ultrahigh dose rates to both mitigate the inflammatory response and maintain cellular proliferation after exposure. A murine model conducted by Cunningham et al. also supports this, wherein equivalent tumour control was maintained in a squamous cell carcinoma FLASH irradiation compared to CONV [74]. Additionally, when FLASH irradiating normal hind leg tissue, there were decreased levels of TGFβ1 24and 96-hours post irradiation, as well as a higher G-CSF to GM-CSF ratio, both indicative of reduced inflammation compared to CONV [74].

0.01
Laser and conventionally accelerated protons induce γH2AX and 53BP1 foci with similar efficacy. Induction of 3-nitrotyrosine, a marker for nitroxidative stress, was 2-3-fold lower in laser accelerated protons compared to conventionally accelerated protons, with similar induction compared to X-rays at 1 Gy.
[58] 2017 Proton, LAP:         There was no substantial difference in zebrafish viability or fish length between FLASH and CONV irradiation modes, and both irradiation modes sparing zebrafish morphology compared to photon and electron irradiation at a dose of 10 Gy.
[82] 2022 Proton, LAP:  findings include comparable tumour control under either the FLASH (100 Gy/s) or CONV (0.3 Gy/s) dose rates, however there was a substantial reduction in the proportion of mice with lung metastases (~ 10% of mice under FLASH versus ~ 30% under CONV). In normal tissues, morphology was greatly spared under FLASH conditions compared to CONV, providing further evidence in support of a FLASH effect when applying carbon ions at ultra-high dose rates in vivo [79].
Additional simulation work explores the interaction of multiple interacting carbon ion tracks, instantaneous irradiations, and the effect these have upon radiolytic oxygen consumption and peroxide ion formation. At the highest dose rate utilised of approximately 10 10 Gy/s, 300 MeV/ nucleon (~ 11.6 keV/µm) carbon ions consumed 90% of oxygen present in solution, suggesting that carbon ions at FLASH dose rates are capable of inducing transient intracellular hypoxia [86]. Intriguingly, they also show that peroxide ion formation increases with increasing dose rate, and draw comparison to previous work conducted by Montay-Gruel et al. [36] where lower concentrations of H 2 O 2 were produced at ~ 1000 Gy/s compared to ~ 0.1 Gy/s. However, as the dose rates and particles utilised (carbon ions vs. electrons) between these studies differ so greatly, it is very difficult for accurate comparisons to be drawn.

Mechanisms of the FLASH effect
Data from studies reviewed thus far collectively indicate three potential mechanisms for enhancing tumour control probability against the normal tissue complication probability during FLASH irradiation with particles. Specifically, normal tissue sparing effects due to rapid oxygen depletion; increased complexity in the nature of DNA damage that is more easily repaired by normal cells; and induction of antitumour immune response. Thus, the FLASH effect edema, erythema, fatigue, and puritus. One patient also experienced grade 2 extremity pain 1 month post treatment. FLASH was deemed clinically feasible in the palliative treatment of bone metastases, with the efficacy of treatment and the profile of adverse effects being analogous to conventional dose rate radiotherapies regimes [85].

Carbon-FLASH
The representation of experimental carbon-FLASH literature, in a radiobiological context, is sparse, with only six papers fitting the systematic criteria of this review. Two of these papers, authored by Zakaria et al. [12,86], outline the potential anti-tumour benefits of carbon FLASH in silico. Monte Carlo simulations produced 3-dimensional track segments of low and high LET carbon ions to draw comparisons between their energy deposition profiles. High LET, 4.1 MeV/nucleon carbon ions (~ 330 keV/µm) exhibited a significant production of radiolytic oxygen after movement of the ions through water compared to lower LET carbon ions. As molecular oxygen can act as a radiosensitiser, they suggest that this radiolytic formation of oxygen at high LET (i.e., Bragg peak) may offer enhanced tumour control. Oxygenation remains relatively unchanged within the low LET, 'normal tissue' region in these simulations, also suggesting that the sparing capacity of carbon-FLASH would be unchanged [12]. In the first in vitro study by Tinganelli et al., CHO-K1 cells exposed to 7.5 Gy of 70 Gy/s carbon-ions had greater surviving fractions at 0.5% and 4% oxygenation (hypoxia) post-FLASH compared to CONV, but not at 21% O 2 (physoxia). These irradiations were performed within the plateau region of the dose-depth distribution (~ 13 keV/ µm), with minimal sparing even at lower oxygen concentrations [76]. Following this, Tinganelli et al. published the first in vivo carbon FLASH study, analysing normal and tumour response in a C3H/He mouse model. Notable   Fig. 3 Representation of the oxygen depletion hypothesis, depicting relative radiosensitivity of normal versus tumour tissues as a function of oxygen tension. Normal tissues see a more significant change in intracellular O 2 levels, affording a brief period of hypoxia which protects normal tissues from ROS related damage during FLASH irradiation. On average, cancerous cells are already relatively hypoxic, so radiation resistance is unchanged by O 2 consumption via FLASH. Permitted reprinting from Wilson et al. 2020 [87] tissue sparing, and that there are multiple other factors that contribute towards this radiobiological outcome [12,92].

DNA damage: repair and response
The frequency of DSBs present within irradiated cells is one of the primary indicators of radiation induced DNA damage. Understanding the differences between how normal and tumour cells accumulate and respond to DSBs is therefore pivotal in defining the radiobiological characteristics of FLASH. Reports have indicated that ultra-high dose rates may reduce the number of γH2AX foci in both normal human fibroblasts and mice crypt base columnar cells, after administration of 20 Gy of 1000 Gy/s 4.5 MeV protons [63] or 14 Gy of 216 Gy/s 16 MeV electrons [43] respectively. This effect has also been illustrated in mouse Lewis lung carcinoma, observing a significantly higher number of γH2AX foci after 0.06 Gy/s versus 352 Gy/s 16 MeV electron irradiation [93]. However, clonogenic assays presented non-significant differences in cellular survival after 3 Gy irradiation. This appears contradictory to the equipotent antitumour effect ultra-high dose rates have in comparison to low dose rates; with present data, the implications of these findings are unclear.
Analysis and interpretation of γH2AX foci are fraught with challenges. One potential explanation for the differences between foci measurements and clonogenic survival is that although particle-FLASH reduces the total number of DSB foci observed in irradiated cells, the complexity of these DNA lesions increases with increasing dose rate [13,94]. The pattern of DNA damage that FLASH irradiation produces compared to CONV has not yet been thoroughly analysed, and as such, there is difficulty in making conclusions concerning the degree of DNA damage that FLASH causes or prevents based upon clonogenic and γH2AX immunofluorescence assays alone. If under conventional irradiation conditions, DNA lesions are more broadly dispersed amongst foci, this would allow for resolving additional DNA DSBs during microscopic examination. Conversely, FLASH irradiation may produce clustered DSBs, where breakages occur within proximity to one another and are not discretely resolved. The initial conclusion for these observations would be that DNA damage appears to be of higher severity for CONV. Furthermore, γH2AX foci are indicators of the initiation of the DSB repair process, not of DSBs explicitly. If the nature of DNA damage between low and ultra-high dose rates differs, then these foci measurements are not directly comparable. This raises questions as to whether γH2AX foci assays can be presented as reliable predictors of cell death in this context.
Clustered DSBs are significantly more difficult for cells to repair than isolated lesions [95], and cancerous cells may both reduce normal tissue toxicity as well as enhance tumour kill efficiency. While these mechanisms are the most frequently proposed and investigated [13,87], data are not outright conclusive and, as is typical for radiobiology, likely to be highly convoluted and complex. Research captured by the systematic review specifically contributing to the discussion of these mechanisms are reviewed in the following sub-sections.

Oxygen depletion and reactive oxygen species
The cell killing capability of radiation therapy relies upon causing irreversible DNA damage to cancerous cells. When any cell is exposed to ionizing radiation (including protons) energy is transferred to intracellular chemical species upon physical interaction along its track. One of the sources of damage for particles is via direct collision and consequent ionization of DNA. Indirect DNA damage can also occur via the interaction of ionizing radiation with the intracellular species surrounding it, including oxygen and water [88]. Energetic charged particles can dissociate water and other molecules to form reactive oxygen species (ROS), DNAdamaging molecules which can disrupt DNA nucleotide sequences or the sugar-phosphate backbone [89]. It is suggested that the production of ROS is limited during FLASH irradiation, the hypothesis being that local oxygen is rapidly depleted at ultra-high dose rates, quicker than reoxygenation can occur. This means that any extra dose within the short timeframe does not contribute to further ROS production, thereby avoiding further damage to DNA due to this mechanism. With respect to normal tissues this leads to a state of radio-resistance. Within a hypoxic tumour environment however, then the change in radiosensitivity is less pronounced.
Solid tumours are often deficient in cellular oxygen due to poor vascularisation, meaning fewer ROS can be produced. As a result, hypoxic tumours are 2-3 times more radioresistant than under normoxic conditions [90]. By the oxygen-depletion theory, then the difference in intracellular O 2 is not as substantial for tumorigenic tissues (hypoxic à hypoxic) compared to normal tissues (physoxic à hypoxic) after FLASH exposure, meaning this protective effect is exacerbated in healthy tissue compared to cancerous tissue (Fig. 3).
Although this theory holds credence for in vitro models of oxygen depletion, there is debate in the radiobiology community as to the translation of this phenomena in vivo [91]. If a region of cancerous tissue has comparable levels of oxygenation to normal tissues, then would one not expect a tumour sparing effect as well? The controversy surrounding this theory suggests that transient oxygen depletion during FLASH is not the prime mechanism responsible for healthy 0.2 Gy/s (CONV) protons to in vitro lung fibroblasts, Buonanno et al. observed a 4.7-fold reduced TGFβ1 induction during FLASH compared to CONV, suggesting that higher dose rates may be able to substantially reduce the degree of chronic inflammation in normal tissues [63]. Other factors, including modulation of the tumour microenvironment [70,93], differential cytokine expression [74], and sparing of stem cell niches [44,103] are also thought to contribute towards the FLASH effect, with further preclinical studies required to substantiate these observations.

Discussion
Whilst current literature appears to be explicit in biological methodology and outcomes of particle-FLASH experiments, the reporting of dosimetry and beam parameters are often not as thorough. To draw more accurate comparisons and conclusions between different FLASH studies, there is a necessity for both more stringent quality control and detailed descriptions of the irradiation characteristics used in these experiments. Particle-and electron-FLASH papers cite usage of substantially high dose rates, with average and/ or instantaneous dose rates on the order of 1 × 10 6 Gy/s or higher [38, 41, 59-61, 65, 66, 68, 104]. This is potentially a considerable source of error in delivered dose as delivery of a conventional clinical dose (e.g., 50 Gy) in a matter of microseconds could have a significant degree of variability if FLASH pulses are not time gated with extreme precision, this being exacerbated at lower doses used experimentally.
As this often appears to be neglected, there may be data which incorrectly correlates radiobiological effects to a predicted 'dose' for which there is substantial degrees of error. More thorough documentation of the methodology utilised to measure dosimetry needs to be provided, as it is intrinsically linked to the radiobiological data presented.
In considering the literature on potential therapeutic benefit of particle-FLASH, the data reviewed here illustrate differences in radiobiological responses compared to conventional irradiation dose rates. Although the effectiveness of particle-FLASH to eradicate tumours in vivo appears to be comparable to current clinical strategies, the outstanding result to date appears to be that the differential effect between normal and healthy tissues is increased considerably. Diffenderfer et al. [72] address these points effectively in their study, providing good justifications from past proton and electron-FLASH literature to form the basis of their experiments. One of the main limitations is that whilst protons are used for tumour irradiations instead of photons, their energy deposition profile is not taken advantage of, i.e. protons are transmitted through target volumes and rather than stopping as would be conducted clinically. Of these often contain deleterious mutations within DNA damage repair (DDR) pathways. Particularly, defects in the nonhomologous end joining (NHEJ) and homologous recombination (HR) repair pathways, responsible for DSB repair, result in increased radiosensitivity [96], an effect which is exacerbated for proton irradiation [97]. Initial observations show that the key radiobiological difference in response to particle-FLASH is that normal cells can better withstand ultra-high dose rates compared to cancerous cells. It is possible that tumour kill efficacy is maintained regardless of a reduction in γH2AX foci formation due to the inherent differences in DDR factor expression between normal and cancerous tissues. Normal cells may have a favourable predisposition for processing the pattern of DNA damage FLASH induces due to being more genetically stable, maintaining the NHEJ and HR repair mechanisms responsible for overcoming otherwise lethal DNA damage [13,98]. Due to the inherent genomic instability of cancerous cells, dysfunction of these mechanisms would prevent cell survival after exposure, as clustered DSBs would be accumulated almost instantaneously at multiple foci under particle-FLASH conditions. Therefore, the differential effect within normal and tumorigenic tissues may lie within their respective abilities to resolve DNA damage of increasing complexity. Characterising the activation of the DNA repair factors expressed in each cell type post-FLASH would aid in determining which pathways drive FLASH's normal tissue sparing capabilities and provide explanations for FLASH's unchanged tumour kill efficacy.

Immune modulation: oncolytic and antiinflammatory
Pathways involved in immune response are also thought to play a major role in both the enhancement of tumour control and reduction of inflammation during FLASH. It has been previously shown that ionizing radiation stimulates anti-tumour responses via expression of damage-associated molecular pattern (DAMP) molecules, consequently recruiting dendritic and cytotoxic T cells to neutralise tumour bodies [99,100]. Indeed, the abscopal effect would be a prime example of radiation induced immune response, hypothesised to act via anti-tumour specific T-cells [101]. Yet, not much is understood concerning the implications of how FLASH interferes with the immune response in normal cells. Computational models predict that ultra-high dose rates may spare circulating immune cells, showing a reduction in immune cell death from 90 to 100% at conventional dose rates to 5-10% for dose rates greater than 40 Gy/s [102]. There is also evidence that TGFβ1, a proinflammatory cytokine, may be modulated during FLASH. After administration of 20 Gy at 1000 Gy/s (FLASH) or to or higher than intracellular O 2 , and as there is a much shorter timeframe between radiation pulses, the chance of radical-radical combination increases. With H 2 O 2 , H 2 O and H 2 being the major products of these interactions, it is hypothesised that cellular O 2 is consumed throughout the process, hence explaining why the FLASH effect induces radiolytic oxygen depletion. This also explains why normal tissues are protected during FLASH radiotherapy, as a lack of reactive oxygen species prevents DNA damage in irradiated cells. The duration over which radiation is delivered may also correlate with this. Zlobinskaya presented FLASH instantaneous dose rates of 2 × 10 10 , however irradiation pulses were delivered over 35 min, resulting in a low average dose rate of 0.009 Gy/s [54]. Pulses at ultra-high dose rates, however at a lower frequency and as a result, a longer total irradiation time, may help elucidate mechanisms such as reoxygenation which could occur between pulses if the duration is long enough. Alternative FLASH mechanisms could dominate depending on the total duration of irradiation and instantaneous dose rates used, and not be entirely dependent upon the average dose rate administered.
These predictions outlined by Koch [107] and Spitz et al. [108] lay the foundation for Abolfath et al. [109] and their simulation-based study to test this hypothesis of oxygen depletion. Via analysis of sub-picosecond timescale interactions with Geant4-DNA Monte Carlo modelling, they aimed to perform a molecular dynamics simulation to help explain tissue damage mitigation observed during FLASH irradiation. They simulated damage to a confined portion of DNA irradiated with protons and measured how the molecules surrounding it (including protein, H 2 O and O 2 molecules) interacted. Conclusions included an increase in ROS with increasing radiation dose, a decrease in ROS with an increase in dose rate, and that under physoxia conditions (4-5% O 2 ) the radioprotection of FLASH is maximised. In this instance, Abolfath et al. present that ROS interact through a vast series of hydrogen bonds, preventing ROS diffusion around DNA and reducing the chance of DNA damage. Modelling conducted by Zhu et al. [106] and Zakaria et al. [12] also support these findings. On the contrary, Labarbe et al. [110] present that transient oxygen depletion is not the prime mechanism for the FLASH effect; their model suggesting that a reduction of peroxyl radicals during FLASH is the prime mechanism by which normoxic tissues are protected.
Whilst the majority of simulation work appears to be in support of the oxygen effect and its influence upon particle-FLASH sparing, the translation and application of these results in experimental work is mixed. Beyreuther et al. [69] and their study analysing the effect of 224 MeV proton-FLASH irradiations upon zebrafish morphology depicted minimal differences in pericardial edema and spinal 34 particle-FLASH papers, only 6 utilise the Bragg peak for irradiations, the remaining employing the initial plateau region. Modulation of the Bragg peak and analysis of the damage that might be spared if FLASH dosage was conformed to the tumour volume is lacking in the proton studies mentioned, which is not reflective of clinical treatments using particle therapy that do employ this. For this technique to translate accurately from the preclinical phase to clinical use, it is vitally important that experiments mimic current clinical treatment methods as closely as possible. Ideally, this would include performing 3-dimensional dosimetry and conformation of the proton beam to a xenografted, orthotopic or spontaneous tumour volume. Nevertheless, this is not always feasible due to the limitations of the equipment used for irradiation or the scope of the experiments being performed (in vivo mice models for instance). This does not negate the value of transmission particle-FLASH experiments such as these, as they are invaluable in identifying how particle-FLASH affects healthy tissue proliferation and survival, as well as providing a general overview of which biological pathways respond to irradiation and in what manner. However, it should still be kept in mind that LET of the protons used in these studies and clinical practice differ substantially, and as such conclusions drawn from preclinical data should be taken with caution when attempting to translate them into clinical trials.
As mentioned previously, there is debate concerning which mechanisms contribute towards the particle-FLASH effect. Specifically, clinical scientists and some academics are uncertain whether the oxygen depletion hypothesis is valid, and if its contribution to normal tissue sparing in vivo is overstated compared to other hypothesised mechanisms. Alongside a multitude of studies conducted in vitro and in vivo, Monte Carlo computational modelling has been applied in attempts to validate this theory, aiming to identify the chemical interactions that occur along a single particle's trajectory during FLASH to determine the reaction kinetics and redox chemistry during radiolysis [92,105,106]. During conventional fractionated therapy at low dose rates, interactions between incoming ionised particles and intracellular species may occur, but the products of these reactions are unlikely to react with one another due to existing up to minutes apart. Therefore, the chance of ROS produced from each fraction interacting with each other is negligible. Under FLASH conditions, higher dose rates are used that are many orders of magnitude higher than conventional dose rates (> 40 Gy/s vs. 0.03 Gy/s), with ionizing radiation assumed to be evenly distributed almost instantaneously over the entire tumour volume. As such, spurs and tracks interact and overlap at a higher frequency, thereby increasing the initial concentration of e − eq , H * and * OH radicals after radiation absorption. These radicals are at a concentration equal

Conclusion
Particle therapy delivered in a particle-FLASH context appears to hold much potential in discovery of beneficial radiobiological response in both normal tissues and tumours. However, there are several areas which must be understood before this treatment modality can be translated safely in clinical settings. Many current data appear to present experimental artefacts and limitations. A major omission in many publications is a solid reporting on dosimetry and quality control. For ultra-high dose rates, the errors in dose are much more prone to significantly deviate from quoted doses compared to low dose rates. The impacts due to irradiation within the Bragg peak compared to the plateau region also remain to be identified. Key limitations of currently published literature stem from a lack of resolving the precise biological processes and mechanisms responsible for the normal tissue sparing effect, leaving much to be addressed in future preclinical work. Whilst the oxygen depletion hypothesis appears to be a compelling explanation to these underlying questions, there should be more emphasis on experiments which accurately represent in vivo conditions during oxygen depletion that are truly reflective of the tumour microenvironment. Along with this, whilst analysing DSB formation under FLASH irradiation conditions is valuable, more rigorous reporting of beam characteristics, doses administered, and DSB quantification techniques are also required to draw more accurate comparisons of the biological outcomes between experiments. Producing stricter, better-defined characteristics of "FLASH" or "continuous" beams of set dose rates is also of relevance, for instance there is evidence that quasi-continuous beams with a relatively low instantaneous dose rate (5 × 10 4 Gy/s) may contribute to not observing a FLASH effect [69]. Factors such as pulse structure, time between pulses and instantaneous versus average dose rate further complicate the biological impact of FLASH. Quantifying gene and protein expression under FLASH is also necessary to determine the differential action of DNA repair pathways between cancer and normal cells, as well as determining differences in cell cycle regulation, immune system modulation, and inflammation. Future work focusing on which dose rates, irradiation times, reactive species, and biochemical conditions optimally promote a beneficial particle-FLASH effect will contribute greatly towards progression of this phenomenon into clinical application.

Funding Open Access funding enabled and organized by CAUL and its Member Institutions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the curvature in comparison to CONV treatments. A similar study was conducted by Pawelke et al. [48], instead using 20 MeV electrons, observed reduced incidence of pericardial edema and spinal curvature under low pO 2 conditions after FLASH irradiations compared to high pO 2 conditions. This suggests that local oxygen concentration may have an impact upon the effectiveness of FLASH, where high pO 2 irradiations masked a potential FLASH effect and zebrafish morphology exhibited minimal differences compared to CONV. This also aligns with whole brain irradiation studies in mice, where Montay-Gruel et al. [36] displayed that the neurocognitive benefits of electron-FLASH were lost when brain tissue pO 2 was artificially doubled via carbogen breathing.
Water phantom studies, such as those conducted by Jansen et al. [111], seem to oppose these conclusions. Contrary to previously mentioned modelling studies, they state that for 10 Gy FLASH irradiations, oxygen consumption decreases with increasing dose rate nonlinearly due to lower steady state values of e − aq radicals, contradicting the apparent sparing effect of FLASH at low pO 2 . Bovine serum albumin irradiations with electron-FLASH conducted by Cao et al.
[46] also support these findings (Table 2), although in vivo measurements failed to achieve oxygen depletion under CONV conditions due to reoxygenation from vasculature. However, additional analogous experiments performed in vivo are required to further link oxygen depletion to the FLASH effect and to identify alternative biological explanations. Interestingly, the increased normal tissue sparing effect of FLASH at low pO 2 levels also appears contradictory to previous claims hypothesising that this effect would be inconsequential in comparable treatments of hypoxic tumourigenic tissues. An in vitro study by Adrian et al. [40] aligns with this, showing a FLASH tissue sparing effect in hypoxic prostate cancer cells. Spitz et al. [108] postulate that normal tissues can limit Fenton type reactions that yield ROS, thereby providing a potential explanation for this discrepancy. Regardless, translation of this sparing effect on in vivo, hypoxic tumourigenic models is yet to be thoroughly explored.
From these experimental data, the consensus appears to be that oxygen depletion does contribute towards healthy tissue sparing, this effect being greater at low pO 2 levels.
Modelling of oxygen kinetics [105] and experiments utilising multicellular spheroids [112] suggest that FLASH depletion has negligible impact upon antitumour efficacy in already highly hypoxic tumour cores. Regardless, these data present multiple avenues for future work in addressing key questions concerning the feasibility of FLASH, limited not only to the oxygen effect but also other unknown biochemical mechanisms of the FLASH effect. source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons. org/licenses/by/4.0/.