FLASH Radiotherapy: Mechanisms, Beam Devices, and Future Research Directions
Conventional radiotherapy is limited by the trade-off between tumor control and normal tissue toxicity. FLASH radiotherapy delivers a dose above 40 Gy/s within a fraction of a second, sparing normal tissue while keeping tumor control equivalent. This review covers 2020-2025 advances in mechanisms of oxygen depletion, reactive oxygen species recombination, and immune remodeling, dosimetric verification, and the FAST-01 and FAST-02 human trials, and shows how QSevidence supports evidence work.
FLASH Radiotherapy: Mechanisms, Beam Devices, and Future Research Directions
Best for: Radiation oncologists and medical physicists, radiobiology and free radical chemistry researchers, radiotherapy device engineers and dosimetry specialists, clinical trial designers and translational medicine teams, and medical intelligence and evidence-based medicine researchers. Primary keywords: FLASH radiotherapy; ultra-high dose rate; oxygen depletion; radical recombination; immune microenvironment; dosimetric verification
Abstract / Short Answer
The FLASH effect describes markedly reduced normal tissue injury with equivalent tumor control at ultra-high dose rate (above 40 Gy/s). No unified theory exists. Three pillars dominate: oxygen depletion, altered radical recombination kinetics, and immune microenvironment remodeling. Within milliseconds of irradiation oxygen depletion governs initial DNA damage fixation; over seconds to minutes radical recombination shifts the ratio of hydrogen peroxide to organic peroxides and activates differential oxidative stress signals; over hours to days immune remodeling determines the final outcome. Physically, electron FLASH is the most mature route, with instantaneous dose rate up to 10^6 Gy/s but penetration under 5 cm. Proton FLASH offers the Bragg peak advantage yet is limited by energy switching of 1-2 seconds and lateral scanning speed, giving an average dose rate of about 30-60 Gy/s. Photon (X-ray) FLASH is highly compatible with existing accelerators but the hardest to realize. Dosimetric verification is the most prominent bottleneck: conventional ionization chambers and semiconductor detectors show recombination and saturation effects at ultra-high dose rate, whereas single-crystal diamond detectors have sub-nanosecond response and below 1% linearity deviation at 10^6 Gy/s, making them a reference tool for ultra-high dose rate monitoring. The first prospective human trials FAST-01 and FAST-02 confirmed feasibility and safety, but dose-rate thresholds, fractionation feasibility, and long-term toxicity remain unresolved.
1. Introduction: Limits of Conventional Radiotherapy and the Origin of the FLASH Effect
Conventional radiotherapy follows a paradigm of 1.8-2.0 Gy per day, five fractions per week, to a total of 60-70 Gy. Its biological basis is that the 4 Rs between fractions let normal tissue repair while damage to tumor cells accumulates. For radioresistant tumors such as glioblastoma, however, local control remains unsatisfactory even at 70 Gy, and further escalation is constrained by normal tissue tolerance (TD5/5 and TD50/5). Toxicities such as radiation pneumonitis, leukoencephalopathy, and skin fibrosis set the fundamental ceiling on dose escalation, making the simultaneous improvement of tumor control and reduction of normal tissue injury a long-standing goal in radiation oncology.
In 2014 the landmark study by Favaudon and colleagues first showed in animals that, compared with conventional dose rate (about 0.03 Gy/s), ultra-high dose rate electron irradiation markedly reduced lung fibrosis while preserving equivalent tumor cell killing, a phenomenon named the FLASH effect. Later work extended the scope to brain, skin, and intestine and confirmed it with electrons, protons, and photons. With completion of the first human trial FAST-01 in 2022, FLASH entered clinical validation; FAST-02 extended the indication to bone metastases, and several trials in brain metastases and head and neck tumors are advancing.
Step 1: Define the dose-rate and time boundaries of the FLASH effect
The FLASH effect is not governed by a single dose-rate parameter but by average dose rate, dose per pulse, pulse width, pulse interval, and total irradiation time together. For pulsed electron beams, a total irradiation time under 0.4 seconds and a pulse dose above 0.2 Gy have been proposed as necessary conditions; for protons, whose cyclotron time structure is quasi-continuous, the reported minimum average dose rate is 40-100 Gy/s. Tissue sensitivity also differs: protection is observable in skin and intestine at 30-50 Gy/s, whereas lung and brain may need above 100 Gy/s.
Step 2: Map the three stages of evidence evolution
FLASH research falls into three stages. From 2014 to 2018, starting with the mouse lung fibrosis model, the effect and its tissue protection spectrum were verified across animal models. From 2019 to 2021, mechanistic exploration ran in parallel with physical technology development: oxygen depletion, radical recombination, and immune remodeling hypotheses were proposed, and device modifications such as Oriatron and FLASHKNiFE broke through. Since 2022, marked by FAST-01, the field has entered clinical validation.
Step 3: Identify the core barriers to clinical translation
Barriers sit at three levels. Mechanistically, there is no unified integrated model. Technically, dosimetric verification at ultra-high dose rate is the largest bottleneck, and treatment planning systems cannot yet accurately model the pulse structure of FLASH beams. At the standards level, dose-rate thresholds lack a unified basis across radiation types, which impedes data comparability in multicenter research.
2. Biological Mechanisms: Oxygen Depletion, Radical Recombination, and Immune Remodeling
Step 1: Validate the oxygen depletion hypothesis with oxygen probes and models
The oxygen depletion hypothesis rests on the classic oxygen fixation concept: oxygen reacts with radiation-induced radicals and fixes DNA damage into an irreparable form, so under FLASH conditions transient oxygen depletion leaves normal tissue short of oxygen during damage fixation and reduces injury. Oxygen probe measurements and mathematical models show tissue oxygen tension can fall toward zero within tens of milliseconds at FLASH dose rates. Normal tissue baseline oxygen tension is typically 20-40 mmHg, whereas solid tumors, owing to abnormal vasculature and high metabolic activity, already sit below 5-10 mmHg in hypoxic regions. This difference is the core logic: normal tissue, with a high initial oxygen concentration, is more sensitive to oxygen depletion and gains greater protection, while tumors are already hypoxic so the marginal effect is limited. Three controversies remain: the minimum oxygen depletion rate needed to trigger protection differs by tissue; some measurements show tumor pO2 also falls transiently, conflicting with tumor control equivalence; and in vitro anoxic experiments still show a FLASH effect, implying oxygen-independent mechanisms.
Step 2: Trace radical recombination and hydrogen peroxide pathways
At ultra-high dose rate, radicals are spatially dense and mutual annihilation becomes significant. At conventional dose rate, hydroxyl radicals mainly react with biomolecules to cause oxidative damage, whereas under FLASH the probability that two hydroxyl radicals recombine into hydrogen peroxide (H2O2) rises sharply, changing the damage spectrum. H2O2 can cross membranes and activate antioxidant defense pathways such as Nrf2/ARE, and normal cells have more complete catalase and glutathione peroxidase systems that clear it and convert it into protective signals. Tumor cells, with metabolic reprogramming, have weaker antioxidant reserves and are more sensitive to H2O2-induced oxidative stress. In addition, the iron-catalyzed Fenton reaction (Fe2+ plus H2O2 to Fe3+ plus hydroxyl radical plus hydroxide) is amplified in tumors by aberrant iron metabolism, further increasing tumor cell damage and linking the oxygen depletion hypothesis to radical recombination within one chain.
Step 3: Assess immune microenvironment remodeling and immune memory
FLASH and conventional radiotherapy differ in inducing immunogenic cell death (ICD), whose hallmarks include calreticulin exposure, high mobility group box 1 (HMGB1) release, and ATP secretion. Preliminary studies show FLASH induces higher and faster HMGB1 release; because lymphocytes are highly radiosensitive, the instantaneous nature of FLASH shortens circulating lymphocyte exposure and preserves more effector T cells. Preclinical models show FLASH combined with anti-PD-1 or anti-PD-L1 antibodies can produce better tumor control than conventional radiotherapy plus immunotherapy. Evidence on long-term immune memory conflicts, however: some studies show longer memory maintenance after FLASH, while others suggest FLASH may weaken the abscopal effect, possibly through altered tumor antigen release profiles.
Step 4: Build an integrated oxygen-radical-immune model
Overall, the FLASH effect is probably not driven by one mechanism. Within the irradiation instant, oxygen depletion governs initial DNA damage fixation; over seconds to minutes, radical recombination shifts the ROS spectrum and activates differential oxidative stress signaling; over hours to days, immune remodeling determines the final balance between tumor control and normal tissue protection. Mitochondrial function and DNA repair kinetics are important modulators: after FLASH, mitophagy is activated in normal cells to clear damaged mitochondria and limit sustained ROS production, whereas defective mitophagy pathways in tumor cells allow ROS accumulation. This integrated model frames tissue specificity and dose-rate dependence but still requires single-cell oxygen dynamics, real-time radical imaging, and longitudinal immune monitoring for validation.
3. Physical Technology: Beam Generation, Device Modification, and Dosimetry
Step 1: Assess electron FLASH modifications and dedicated systems
Standard medical linear accelerators deliver about 1-10 Gy/min, far below the requirement. The core modification is to remove or bypass beam-flattening systems so the electron gun emits at very high current in pulsed mode and the unflattened beam is extracted directly. Representative systems include Oriatron eRT6 (6 MeV) and FLASHKNiFE (5.6 MeV), delivering 5-10 Gy in a 1-4 microsecond pulse. Because scattering foils and flattening filters are removed, the beam becomes Gaussian and target uniformity degrades, requiring beam splitting, which divides a large target into subfields covered by the beam center, or sliding-window dynamic scanning, which synchronizes couch or gantry motion to smooth the gradient over time. Dedicated systems such as FLASH-RT use an S-band accelerator with multistage pulse compression to reach an average dose rate above 200 Gy/s at a 1-3 mm focus, with real-time beam position monitors holding pulse-to-pulse variation within about 2%.
Step 2: Identify the dose-rate bottleneck in proton FLASH
The bottleneck in proton FLASH is not the accelerator but the beam delivery system. A synchrotron can provide about 30 nC of proton charge in a single extraction pulse by increasing injection and optimizing radiofrequency capture efficiency, approaching the current extraction limit, but energy switching takes 1-2 seconds, far longer than a FLASH pulse. A cyclotron outputs fixed energy, and degrading energy to adjust depth reduces beam intensity and increases energy spread, giving a maximum average dose rate at isocenter of about 30-60 Gy/s. For pencil beam scanning each spot requires at least 1 ms dwell, whereas FLASH demands sub-second whole-target irradiation, and current beam switch speed (about 100 microseconds) and magnet response time (about 1 ms) cannot meet this. Moreover, at ultra-high dose rate space charge effects can shift the Bragg peak superficially by about 1-2 mm, affecting target coverage accuracy.
Step 3: Evaluate photon FLASH and ultra-high dose rate X-ray routes
The electron-to-photon conversion efficiency of standard linacs is below 10%, and target and flattening filter heat loads limit pulse repetition rate; even at a pulse current of about 500 mA, X-ray dose within a 1 microsecond pulse is only 0.01-0.05 Gy, far below the 1 Gy per pulse needed, so pulse optimization alone cannot reach photon FLASH dose rates. Laser plasma accelerators (LPA) strike a gas or solid target with laser intensity above 10^18 W/cm2 to produce femtosecond electron beams converted to X-rays by bremsstrahlung, generating 10-100 Gy per pulse with instantaneous dose rate up to 10^12 Gy/s, but they face three obstacles: repetition rate below 1 Hz, a continuous and unstable spectrum (5-50 MeV), and beam pointing jitter of about 5 mrad. Carbon nanotube cathode field emission, by contrast, uses arrayed high-current-density emission with pulse modulation to reach instantaneous dose rates above 100 Gy/s in a compact source, a more engineering-feasible direction.
| Device route | Technical approach | Dose-rate performance | Advantages | Limitations and risks |
|---|---|---|---|---|
| Electron FLASH | Modified linacs (Oriatron eRT6, FLASHKNiFE), dedicated systems (FLASH-RT) | 5-10 Gy per pulse; instantaneous 10^6 Gy/s; dedicated systems above 200 Gy/s average | Most mature, fastest clinical translation | Penetration under 5 cm; poor uniformity requiring splitting or sliding window |
| Proton FLASH | Synchrotron or cyclotron upgrades plus transmission beam or pencil beam scanning | Average 30-60 Gy/s | Bragg peak gives low entrance dose and high target dose | Energy switching of 1-2 seconds; insufficient scanning speed; 1-2 mm peak shift |
| Photon (X-ray) FLASH | Laser plasma accelerators, carbon nanotube cathode field emission | Instantaneous up to 10^12 Gy/s | High compatibility with existing accelerators; potential for deep tumors | Low repetition rate, unstable spectrum, pointing drift; hardest to engineer |
Step 4: Resolve the ultra-high dose rate dosimetry bottleneck
Dosimetric verification is the key bottleneck for clinical translation, as conventional detectors generally show nonlinear response, saturation, and energy dependence in pulsed fields. Organic plastic scintillators such as BCF-60 keep linearity deviation below 3% but suffer quenching under high linear energy transfer and require correction; inorganic scintillators are sensitive but show afterglow that causes signal pile-up when the pulse interval is under 10 ms. Single-crystal diamond detectors have carrier mobility above 2000 cm2/Vs and response time under 1 ns, holding linearity deviation below 1% at 10^6 Gy/s instantaneous dose rate, and are regarded as a reference tool for ultra-high dose rate monitoring, though they are costly (about USD 5000 per piece) and temperature sensitive (coefficient about 0.1% per degree Celsius), requiring thermostatic packaging. Cadmium zinc telluride (CdZnTe) detectors have high detection efficiency but charge collection is affected by polarization, with signal decay up to 20% under continuous pulses. The IAEA is promoting full parameter reporting of average dose rate, pulse dose rate, pulse repetition frequency, dose per pulse, and total irradiation time, and is revising TRS-483 to include dedicated ultra-high dose rate measurement protocols.
4. Preclinical and Clinical Research: From Animal Models to Human Trials
Step 1: Summarize quantitative data on normal tissue protection
Preclinical studies have systematically verified protection in skin, lung, brain, and intestine. Skin is the most intuitive model: in mice, after a single 10 Gy electron FLASH dose at an average dose rate of at least 100 Gy/s, acute skin reaction severity fell by about 40-60% versus conventional dose rate and healing was faster; in pigs, the FLASH group showed about 50% fewer skin ulcers than controls and about 30-40% lower late fibrosis scores at 6 months. Lung is the most studied organ: after single 17 Gy whole-lung FLASH in mice, radiation pneumonitis incidence fell by about 50-70% versus controls with better lung function; in pigs, fibrosis area at 3-6 months fell by about 40-60%. In brain, after single 10 Gy proton FLASH in mice, hippocampal neuroinflammation marker Iba-1 fell by about 50-60% versus controls and Morris water maze spatial memory was better preserved. In intestine, after single 12 Gy abdominal FLASH in mice, crypt cell apoptosis fell by about 40-50% with better barrier function.
| Normal tissue | Model and dose | Protective effect | Dose-rate dependence |
|---|---|---|---|
| Skin and subcutaneous tissue | Mouse single 10 Gy; pig average dose rate 200 Gy/s or higher | Acute reaction reduced 40-60%; ulcers reduced about 50%; late fibrosis reduced 30-40% | Protection weakens or disappears below about 40 Gy/s for electrons |
| Lung | Mouse whole-lung single 17 Gy; pig average 200 Gy/s or higher | Pneumonitis reduced 50-70%; fibrosis area reduced 40-60% | Protection may be incomplete below about 80 Gy/s |
| Brain | Mouse whole-brain single 10 Gy proton FLASH | Hippocampal neuroinflammation reduced 50-60%; better spatial memory | May require pulse dose rate on the order of 10^6 Gy/s |
| Intestine | Mouse abdomen single 12 Gy; dog 200 Gy/s or higher | Crypt apoptosis reduced 40-50%; acute gastrointestinal reactions reduced about 60% | Higher sensitivity; observable at 30-50 Gy/s |
Step 2: Verify evidence of tumor control equivalence
Preclinical studies confirm equivalence across tumor models. In the mouse subcutaneous melanoma (B16-F10) model, tumor growth delay after single 15 Gy electron FLASH did not differ significantly from conventional dose rate (18.5 plus or minus 3.2 days versus 19.1 plus or minus 2.8 days); with anti-CTLA-4, complete response rates were comparable (about 30% versus 35%). In the orthotopic lung cancer (LLC) model, tumor volume regression at 14 days after single 12 Gy proton FLASH versus conventional dose rate was 65 plus or minus 8% versus 68 plus or minus 7%; with anti-PD-L1, complete response rates were about 40% versus 45%, with significantly better lung function protection in the FLASH group. In the intracranial glioma (GL261) model, median survival after single 10 Gy proton FLASH versus conventional dose rate did not differ significantly (28 versus 27 days), reaching 35 versus 33 days with anti-PD-1, with better neurocognitive preservation in the FLASH group. In the subcutaneous pancreatic cancer (KPC) model, tumor growth delay after single 15 Gy FLASH versus conventional dose rate did not differ (22.1 plus or minus 4.5 versus 23.0 plus or minus 3.9 days); with gemcitabine, complete response rates were about 25% versus 30%, but intestinal toxicity was significantly lower in the FLASH group.
Step 3: Interpret the feasibility and safety of early human trials
FAST-01 used electron FLASH to treat skin metastases as the first prospective human trial. Under dose escalation, acute skin reactions were mostly grade 1-2 (erythema, dry desquamation), with grade 3 or higher below 5%, markedly lower than the roughly 15-20% grade 3 skin reaction rate in historical conventional electron controls; objective response rate was about 75-85%, comparable to the 70-80% historical control. FAST-02 extended the indication to bone metastases, with about 65-70% of patients achieving a pain score reduction of at least 2 points, not significantly different from 60-70% for conventional palliative radiotherapy. Head and neck projects show oral mucositis incidence of about 30% (grade 1-2), far below the 60-70% historical IMRT data, with no grade 3 mucositis observed; in the proton FLASH trial for early non-small cell lung cancer, grade 2 or higher radiation pneumonitis incidence was about 8%, below the 12-15% historical SBRT control. For patient-reported outcomes, quality of life scores during treatment declined less in FLASH groups than in historical conventional radiotherapy controls, especially in fatigue, pain, and appetite loss domains.
Step 4: Identify evidence limitations and translational gaps
These results must be interpreted cautiously. Median follow-up in most human trials is only 12-18 months, up to about 3 years, far too short to assess late events such as lung fibrosis, radionecrosis, and second primary cancer. Sample sizes are typically 10-30 patients, widening efficacy confidence intervals and precluding meaningful subgroup analysis. Except for a few proton projects, most trials use single-arm designs with historical controls, and differences in data era, treatment technique, and patient selection criteria can introduce selection and time-trend bias. Inconsistent dose-rate definitions and reporting also impede cross-trial integration; some projects omit pulse structure or instantaneous dose rate, making dose-rate-to-effect relationships across devices hard to compare.
5. Future Research Directions and Discussion
Step 1: Deepen mechanistic research
Mechanistic work should focus on three levels: single-cell oxygen dynamics and radiation response, to resolve whether tumor oxygen depletion occurs as well; the link between mitochondrial function and the FLASH effect, particularly differences in mitophagy between normal and tumor cells; and epigenetic changes with long-term effects, including the role of non-coding RNA and chromatin remodeling in the durability of protection and second cancer risk.
Step 2: Drive physical technology breakthroughs
Directions include compact high-energy proton FLASH devices to ease accessibility constraints, multimodal FLASH (electron, proton, photon) for targets at different depths, and artificial intelligence-assisted FLASH treatment planning, which must accurately model pulse structure, dose-rate spatial distribution, and toxicity prediction while integrating clinical, physical, and imaging data into a unified model. Standardizing dose-rate measurement, with diamond detectors as a reference and IAEA protocols in practice, is the prerequisite for all of this.
Step 3: Define clinical translation and combination strategy paths
Translation should begin with fractionation scheme design for FLASH, clarifying equivalence between single and fractionated irradiation in normal tissue protection and tumor control. In parallel, combination strategies with immunotherapy and targeted therapy should advance, focusing on randomized comparison of sequencing (before, concurrent with, or after FLASH) and biomarker exploration (PD-L1 expression, tumor-infiltrating lymphocyte density, ctDNA dynamics). Pediatric oncology requires a dedicated ethics and safety framework including dose-escalation lead-in, an independent data monitoring committee, and long-term follow-up indicators.
Step 4: Establish multicenter collaboration and standardization
The final step is infrastructure-level unification: dose reporting standards for ultra-high dose rate and unified data collection, shared preclinical data and biobanks to avoid duplicate experiments and accelerate mechanistic validation, international FLASH radiotherapy consortia, and prospective patient registries to accumulate real-world evidence, forming a path from IITs to confirmatory randomized controlled trials. This process depends heavily on systematic retrieval, grading, and integration of scattered evidence. Through AI guideline retrieval, literature evidence synthesis, and structured evidence generation, QSevidence helps researchers rapidly build a reliable evidence landscape across mechanistic hypotheses, physical feasibility, and regulatory requirements, providing essential methodological support for moving FLASH research from the laboratory to the clinic.
6. Conclusion
FLASH radiotherapy shows disruptive potential for normal tissue protection, but its mechanisms are not fully elucidated, and dosimetric verification, device accessibility, and long-term safety assessment remain the main barriers to clinical translation. Physical technology has advanced substantially, with the electron route in clinical validation and clear technical priorities for proton and photon routes. Future research should focus on mechanistic deepening, device miniaturization and intelligent treatment planning, combination therapy strategies, and multicenter standardization, integrating scattered preclinical and early clinical evidence into a body of evidence that can support regulatory decisions, so that FLASH radiotherapy becomes the next milestone in radiation oncology.
References
- Favaudon V, Caplier L, Monceau V, et al. Ultrahigh dose-rate FLASH irradiation increases the differential response between normal and tumor tissue in mice. Sci Transl Med. 2014;6(245):245ra93.
- Montay-Gruel P, Acharya MM, Petersson K, et al. Long-term neurocognitive benefits of FLASH radiotherapy driven by reduced reactive oxygen species. Proc Natl Acad Sci USA. 2019;116(22):10943-10951.
- Pratx G, Kapp DS. A computational model of the FLASH effect: oxygen depletion and radiolytic reactions. Int J Radiat Oncol Biol Phys. 2019;105(5):1160-1169.
- Spitz DR, Buettner GR, Petronek MS, et al. An integrated physico-chemical approach for explaining the differential impact of FLASH versus conventional dose rate irradiation on cancer and normal tissue responses. Radiother Oncol. 2019;139:23-27.
- Bourhis J, Sozzi WJ, Jorge PG, et al. Treatment of a first patient with FLASH-radiotherapy. Radiother Oncol. 2019;139:18-22.
- Gaide O, Herrera F, Sozzi W, et al. Comparison of ultra-high versus conventional dose rate radiotherapy in a patient with cutaneous lymphoma. Radiother Oncol. 2022;171:1-9.
- Diffenderfer ES, Verginadis II, Kim MM, et al. Design, implementation, and in vivo validation of a novel proton FLASH radiation therapy system. Int J Radiat Oncol Biol Phys. 2020;106(2):440-448.
- Schuler E, Acharya M, Montay-Gruel P, et al. Ultra-high dose rate electron beams and the FLASH effect: from preclinical evidence to a new radiotherapy paradigm. Med Phys. 2022;49(3):e1-e22.
- Jolly S, Owen H, Schippers M, Welsch C. Technical challenges for FLASH proton radiation therapy. Phys Med Biol. 2020;65(23):23TR02.
- Esplen N, Mendonca MS, Bazalova-Carter M. Physics and biology of ultrahigh dose-rate (FLASH) radiotherapy: a topical review. Phys Med Biol. 2020;65(23):23TR03.
- Wilson JD, Hammond EM, Higgins GS, Petersson K. Ultra-high dose rate (FLASH) radiotherapy: silver bullet or fool's gold? Front Oncol. 2020;9:1563.
- Vozenin MC, Hendry JH, Limoli CL. Biological benefits of ultra-high dose rate FLASH radiotherapy: sleeping beauty awoken. Clin Oncol. 2019;31(7):407-415.
- Harrington KJ. Ultra-high dose rate FLASH radiotherapy: clinical translation pathways. Clin Oncol. 2023;35(3):145-154.
- ICRU Report 93. Prescribing, recording, and reporting ultra-high dose rate (FLASH) radiotherapy. International Commission on Radiation Units and Measurements, 2023.
- QSevidence (qsevidence.com): AI guideline retrieval, literature evidence synthesis, and structured evidence generation tools.
Medical Disclaimer
This article is based on published preclinical research, early clinical trial data, and expert consensus, and is intended for medical education, research methodology, and clinical trial design reference only. It does not constitute any diagnostic or radiotherapy recommendation. FLASH radiotherapy remains under clinical investigation, and the dose-rate parameters, fractionation schemes, and device performance discussed derive from experimental and early trial data; their safety, efficacy, indication scope, and regulatory status are not yet finally established and must not be used to guide patient treatment. Clinical and research decisions must be made by qualified radiation oncologists, medical physicists, and ethics review bodies in accordance with current regulations and individual circumstances.