Abstract
Abstract
Objective. FLASH radiotherapy spares normal tissue at ultra-high dose rates, but the dependence of this protection on linear energy transfer (LET) and tissue oxygenation is not mechanistically understood. We developed a model coupling Geant4-DNA track-structure G-values to the Labarbe peroxyl-radical (ROO•) kinetic framework to predict FLASH normal-tissue sparing as a joint function of LET and oxygenation.
Approach. G-values for •OH, e⁻aq, H• and H₂O₂ were computed with Geant4-DNA chem6 for electrons (0.16–8.70 keV µm⁻¹) and protons (0.69–9.5 keV µm⁻¹), with 500 primary events per LET point, and taken from published Monte Carlo data for carbon ions (10–200 keV µm⁻¹). These were supplied to the Labarbe ordinary differential equation system using the published rate constants without adjustment, and integrated across dose rates of 0.03–10⁵ Gy s⁻¹, doses of 5–30 Gy and oxygen concentrations of 2–200 µM. FLASH sparing was quantified as the fractional reduction in time-integrated ROO• exposure. The mechanism was tested by removing individual reactions from the kinetic system, and predictions were compared with three independent preclinical datasets.
Main results. Using the NTCP parameters published by Labarbe et al (2020), with no quantity fitted in the present work, the model gives a dose-modifying factor of 1.42; with the parameters refitted here it gives 1.44, against experimentally reported values of 1.30–1.40. Sparing decreases monotonically with LET, from 32.3% at 0.16 keV µm⁻¹ to 11.2% at 200 keV µm⁻¹. At matched LET, proton and electron yields converge below 2 keV µm⁻¹ but diverge by up to 19% in G(e⁻aq) at 8.7 keV µm⁻¹, setting a validity limit on LET-only parameterisation. Removing the ROO•–ROO• recombination term collapses sparing from 32.3% to 2.4%, identifying second-order peroxyl recombination as the operative mechanism. The oxygen dependence is dose- and dose-rate-gated, with the sparing difference across 2–200 µM ranging from below 1 to approximately 40 percentage points; at 30 Gy and 100 Gy s⁻¹ the model predicts 8.2 percentage points less sparing in bone osteocytes (15 µM) than in normoxic tissue, consistent with the absence of FLASH bone sparing reported in canine mandible.
Significance. Peroxyl-radical bimolecular recombination, driven by LET-dependent track-structure yields, provides a parameter-free chemical basis for the modality dependence of the FLASH effect. The predicted dose-, dose-rate- and oxygen-gating of sparing gives testable criteria for where FLASH protection should and should not occur.