Oxygenation and Spatial Heterogeneity Shape Radiotherapy Protocol Ranking Through Phenotypic Adaptation Authors: Francesco Albanese, Giulia Chiari, Marcello Edoardo Delitala Published: 2026 ABSTRACT Tumor response to radiotherapy is strongly influenced by oxygen availability and phenotypic heterogeneity, yet their combined impact on the relative performance of fractionation schedules remains unclear. Here, we develop a mathematical model that integrates spatial oxygen dynamics with continuous phenotypic adaptation to hypoxia and radiation, and use it to systematically compare radiotherapy protocols under a common normal-tissue toxicity constraint. Under spatially uniform oxygenation, we find that alternative fractionation schedules provide little improvement over standard-of-care protocols in normoxic conditions. Under moderate hypoxia, however, a distinct class of protracted schedules with longer inter-fraction intervals substantially increases time-to-progression, in some cases by up to twofold. This regime-dependent benefit is consistent with a shift in the balance between reoxygenation and selection for resistant phenotypes. When oxygen delivery is spatially heterogeneous, treatment outcomes depend strongly on the geometric organization of oxygen sources. Even with identical total oxygen supply, different spatial configurations lead to large variability in time-to-progression and can alter the relative ranking of radiotherapy protocols. These results show that radiotherapy effectiveness is not an intrinsic property of a treatment schedule alone, but emerges from its interaction with tumor microenvironmental structure and evolutionary dynamics. Incorporating both spatial heterogeneity and phenotypic adaptation may therefore be important for the consistent evaluation and design of fractionation strategies in heterogeneous tumors.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1007/s11538-026-01758-5. Introduction Solid tumors are widely recognized as heterogeneous and evolving systems, in which malignant cells interact dynamically with their microenvironment (Hanahan and Weinberg 2000). Among microenvironmental factors, oxygen availability plays a central role in shaping both tumor progression and response to therapy. Radiotherapy represents a standard treatment modality (Siegel et al. 2023), whose efficacy is strongly modulated by local oxygen levels. Oxygen enhances radiation-induced DNA damage, an effect quantified through the Oxygen Enhancement Ratio (OER), defined as the dose increase required under hypoxia to match the biological effect observed in normoxia. However, oxygen influences treatment response in more than one way. Beyond its direct physico-chemical role in damage fixation, hypoxia induces transcriptional and metabolic reprogramming, promoting aggressive phenotypes that contribute to radioresistance (Beckers et al. 2024). Importantly, experimental studies indicate that OER is not uniform across intratumoral subpopulations. More aggressive cells are characterized by lower intrinsic radiosensitivity (Lagadec et al. 2012). Moreover, oxygen distribution within the tumor microenvironment (TME) is typically spatially heterogeneous, owing to the irregular arrangement and perfusion of the tumor vasculature. As a consequence, not only the overall level of oxygenation, but also the geometry of oxygen sources and the resulting spatial gradients may contribute to shaping local radiotherapy response (Hormuth et al. 2021). Mathematical models of tumor response to radiotherapy combine population growth dynamics with radiobiological descriptions of radiation-induced cell killing. Tumor kinetics are typically represented through proliferation laws incorporating carrying-capacity effects (Zheng et al. 2025), whereas radiation response is described by the linear–quadratic (LQ) formalism, in which the surviving fraction after a dose d is \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$S(d)=\exp (-\alpha d - \beta d^2)$$\end{document}S(d)=exp(-αd-βd2), with α α and β β denoting radiosensitivity parameters inferred from experimental and clinical studies (van Leeuwen et al. 2018). The LQ formulation also underlies the definition of biologically effective dose (BED), enabling comparisons across fractionation schedules. Experimental and clinical evidence on hypoxia led to extensions of the classical LQ model incorporating oxygen dependence. In particular, radiosensitivity parameters have been rescaled through an OER factor to account for variations in radiation response between normoxic and hypoxic conditions (Wenzl and Wilkens 2011). Alternative approaches have relaxed the assumption of homogeneous radiosensitivity by allowing the parameters α α and β β to vary across tumor cell populations or between patients, thereby representing intrinsic biological heterogeneity (Alfonso and Berk 2019). Trait-structured models have recently been employed to describe tumor response to radiotherapy in the presence of both phenotypic heterogeneity and microenvironmental modulation (Celora et al. 2023; Chiari et al. 2023). In these formulations, spatial dynamics and continuous phenotypic structure are coupled within a unified PDE system, so that tumor response under irradiation can be interpreted as an eco-evolutionary process in which selection emerges from the redistribution of cells across phenotypic states. Several mathematical studies have explored alternative radiotherapy fractionation strategies (Prokopiou et al. 2015; Henares-Molina et al. 2017; Brüningk et al. 2021). To our knowledge, however, the combined impact of spatially heterogeneous oxygenation and continuous phenotypic adaptation on the comparative performance of treatment schedules has not been systematically investigated. In particular, several protocols currently adopted in clinical practice are considered optimal for tumors exhibiting specific cellular and microenvironmental features; however, a mechanistic modeling framework capable of explaining these observed treatment preferences as a consequence of such features is still lacking. In this context, this study is motivated by the need to understand how the interplay between oxygen spatial structure, evolutionary selection of resistant phenotypes, and normal-tissue toxicity constraints may shape the landscape of admissible fractionation protocols. Are there dose–interval combinations, beyond those commonly used in clinical practice, that may lead to improved treatment outcomes? How does spatial heterogeneity in oxygen delivery influence the relative performance of radiotherapy schedules? Are protocols identified under spatially uniform conditions sufficiently robust, or is detailed characterization of spatial oxygen heterogeneity required for reliable treatment selection? To address these questions, we develop a phenotypically structured partial differential equation (PS-PDE) model (Lorenzi et al. 2025). Within this framework, we systematically explore the dose–interval plane under biologically effective dose constraints, enabling comparison of fractionation protocols across different oxygenation regimes and tissue tolerance conditions. This modeling strategy aligns with current efforts toward biologically informed radiotherapy design, including multi-omics–guided treatment adaptation (Boldrini et al. 2024) and spatially targeted dose painting (Bentzen and Grégoire 2011). The remainder of the paper is organized as follows. Section 2 introduces the mathematical model, including tumor spatio–phenotypic dynamics, oxygen dynamics, and the oxygen- and phenotype-dependent formulation of radiotherapy response. Section 3 investigates how oxygenation affects the performance of fractionation schedules under a common toxicity constraint, first under spatially uniform oxygen supply and then under heterogeneous oxygen delivery. Particular attention is given to the role of phenotypic adaptation and to how the spatial organization of oxygen supply can alter time-to-progression and protocol ranking. Conclusions and perspectives are discussed in Sect. 4. Methods We introduce a spatio–phenotypic model coupling tumor-cell dynamics, oxygen diffusion–consumption, and oxygen- and phenotype-dependent radiotherapy response. Model parameters are listed in Table 1. Simulation details and indicator definitions are reported in the Supporting Information S1 and S2.Table 1Model parameters. Background colors distinguish parameter groups associated with cell dynamics (blue), radiotherapy response (red), and oxygen kinetics (green). The final column reports the origin of each parameter value, indicating whether it is taken from the literature, derived in a specific section of the paper, or estimated within the present framework as a model estimate (M.E.). Entries labeled with “S” refer to the corresponding section of the Supporting Information Table 1Model parameters. Background colors distinguish parameter groups associated with cell dynamics (blue), radiotherapy response (red), and oxygen kinetics (green). The final column reports the origin of each parameter value, indicating whether it is taken from the literature, derived in a specific section of the paper, or estimated within the present framework as a model estimate (M.E.). Entries labeled with “S” refer to the corresponding section of the Supporting Information Model parameters. Background colors distinguish parameter groups associated with cell dynamics (blue), radiotherapy response (red), and oxygen kinetics (green). The final column reports the origin of each parameter value, indicating whether it is taken from the literature, derived in a specific section of the paper, or estimated within the present framework as a model estimate (M.E.). Entries labeled with “S” refer to the corresponding section of the Supporting Information Cell Dynamics We consider a portion of biological tissue where the early growth of a tumor mass occurs. The spatio–temporal evolution of the active cancer cell population is described by a density function n(t, x, u), where tt denotes time, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$x \in \Omega _s \subset \mathbb {R}$$\end{document}x∈Ωs⊂R represents the spatial position, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$u \in \Omega _p = [0,1]$$\end{document}u∈Ωp=[0,1] is a continuous phenotypic trait. Although spatial dynamics are modeled along a single coordinate, n(t, x, u) is interpreted as a volumetric cell density. Following the framework introduced in Chiari et al. (2023), we model resistance as a continuous variable ranging from fully hypoxia- and radiation-sensitive cells (u=0u=0) to fully resistant ones (u=1u=1), with higher values of uu corresponding to increased resistance to both stresses. We emphasize that uu represents an intrinsic phenotypic adaptation state, shaped by selection and non-genetic phenotypic changes, rather than a marker of a cell’s current, instantaneous oxygen exposure. A cell with high uu therefore represents a phenotype adapted to hypoxic stress and intrinsically more resistant to radiation, consistently with the well-established association between hypoxia-induced cellular adaptation and radioresistance (Beckers et al. 2024). The spatio–phenotypic avascular evolution of the tumor cell density is modeled through a Phenotypically Structured Partial Differential Equation (PS-PDE):1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \partial _t n(t,x,u) = D_p\,\partial _u^2 n(t,x,u) + D_s\,\partial _x^2 n(t,x,u) + \bigl [R(t,x,u)-T(t,x,u)\bigr ]\,n(t,x,u) \end{aligned}$$\end{document}∂tn(t,x,u)=Dp∂u2n(t,x,u)+Ds∂x2n(t,x,u)+[R(t,x,u)-T(t,x,u)]n(t,x,u)where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$D_p$$\end{document}Dp represents the coefficient of phenotypic diffusion, accounting for stochastic epigenetic changes. The term \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$D_s$$\end{document}Ds corresponds to the linear spatial diffusion coefficient, which describes the random dispersal of cells within the tissue. The function R(t,x,u)R(t,x,u) defines the net proliferation rate of the tumor cell population, whereas T(t,x,u)T(t,x,u) quantifies the effective rate of cell death induced by radiotherapy. Both terms depend on time, space, and phenotype, reflecting the heterogeneous nature of the tumor response to oxygen availability and radiation exposure. Reaction Term The proliferation rate R(t,x,u)R(t,x,u) models cell growth as influenced by oxygen availability, phenotypic adaptation, and crowding effects:2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} R\bigl (t,x,u\bigr ) = r_0 + \gamma _O\,M\bigl (O(t,x)\bigr )\,(1-u^2) - \gamma _H\bigl (1-M\bigl (O(t,x)\bigr )\bigr )\,(1-u)^2 - \kappa \,\rho (t,x) \end{aligned}$$\end{document}R(t,x,u)=r0+γOM(O(t,x))(1-u2)-γH(1-M(O(t,x)))(1-u)2-κρ(t,x)where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r_0$$\end{document}r0, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\gamma _O$$\end{document}γO, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\gamma _H$$\end{document}γH, and κ κ are model parameters, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$O(t,x):[0,t_{end}]\times \Omega _s\rightarrow \mathbb {R}_{+}$$\end{document}O(t,x):[0,tend]×Ωs→R+ denotes the local oxygen concentration. The first term \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r_0$$\end{document}r0 represents the baseline proliferation rate of fully resistant cells (u=1u=1), which sustain a minimal growth rate even under adverse environmental conditions. The second term provides an additional proliferative contribution, weighted by two effects: (i) a Hill-type function enhancing proliferation under normoxic conditions,3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} M\bigl (O(t,x)\bigr ) = \frac{O(t,x)^4}{O(t,x)^4+(\alpha _O)^4} \end{aligned}$$\end{document}M(O(t,x))=O(t,x)4O(t,x)4+(αO)4and (ii) a quadratic dependence on uu, expressing that phenotypes less resistant to radiotherapy and less adapted to hypoxia (u → 0u→0) exhibit higher proliferative potential. This formulation reflects an energetic trade-off, whereby cells investing metabolic resources in resistance to hypoxia or radiation damage proliferate less efficiently. The Hill exponent is set to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n_H=4$$\end{document}nH=4 as a modeling choice, providing a sharp but smooth transition and an effectively saturated oxygen-dependent proliferative contribution within the normoxic range. This value is not intended to represent a specific molecular cooperativity mechanism. The robustness of the results presented below with respect to this choice is assessed in Supporting Information S10 by repeating the analysis for different Hill coefficients. The third term penalizes poorly adapted phenotypes under low-oxygen conditions, thereby acting as a selective pressure toward higher-resistance traits. The parameter \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\gamma _H$$\end{document}γH is set to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\gamma _H = r_0+\gamma _O$$\end{document}γH=r0+γO, so that under extreme hypoxia the penalty experienced by poorly adapted phenotypes offsets the largest intrinsic growth rate attained under favorable oxygenation. The quadratic functional form in both the second and third terms is chosen so as to induce a concave fitness landscape: for fixed tt and xx, and depending on the oxygen concentration, the reaction term admits a unique global maximum with respect to the phenotypic variable uu, as discussed in Supporting Information S3. This modeling assumption is consistent with analogous choices adopted in the literature on PS-PDEs (Lorenzi et al. 2025). Finally, the crowding term enforces population saturation due to spatial limitations. Here, ρ (t,x)ρ(t,x) denotes the total local cell density,4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \rho (t,x) = \int _{\Omega _p} n(t,x,u)\,du. \end{aligned}$$\end{document}ρ(t,x)=∫Ωpn(t,x,u)du.For fixed (t, x, u), the net proliferation rate R(t, x, u) is decreasing with respect to the local density ρ ρ, thereby limiting local tumor growth as cell density increases. The crowding coefficient κ κ is calibrated in Supporting Information S4 by prescribing KK as an upper-bound carrying capacity under the most favorable proliferative conditions. Overall, the reaction term defines a dynamic fitness landscape that evolves in time with both the oxygen distribution and the phenotypic composition of the tumor population. Radiotherapy Term Radiation-induced cell death is modeled through a phenotype- and oxygen-dependent extension of the linear–quadratic (LQ) formulation. We write5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} T(t,x,u) = \bigl [\alpha (O(t,x),u)\, d + \beta (O(t,x),u)\, d^2 \bigr ]\, \delta _{\mathcal {T}}(t), \end{aligned}$$\end{document}T(t,x,u)=[α(O(t,x),u)d+β(O(t,x),u)d2]δT(t),where d denotes the dose delivered per fraction (Gy), and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\delta _{\mathcal {T}}(t)$$\end{document}δT(t) is a sum of Dirac delta functions that localize irradiation at prescribed treatment times, as detailed below. The coefficients α (O,u)α(O,u) and β (O,u)β(O,u) are continuous counterparts of the classical constant LQ parameters and determine the survival fraction \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$S(d)=\exp \bigl (-T(t,x,u)\big )$$\end{document}S(d)=exp(-T(t,x,u)) in the standard formulation (McMahon 2018). In our model they depend on both oxygen concentration and phenotype:6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \begin{aligned} \alpha (O,u)&= \dfrac{\tilde{\alpha } + \Delta _\alpha (1-u)^2}{\textrm{OER}(O)} \\ \beta (O,u)&= \dfrac{\tilde{\beta } + \Delta _\beta (1-u)^2}{\textrm{OER}^2(O)} \end{aligned} \end{aligned}$$\end{document}α(O,u)=α~+Δα(1-u)2OER(O)β(O,u)=β~+Δβ(1-u)2OER2(O)The quadratic dependence on (1-u)(1-u) introduces phenotype-specific modulation of radiosensitivity, consistently with patient-specific modeling studies showing that higher proliferation rates are associated with increased radiosensitivity (Rockne et al. 2010). The OER formulation effectively rescales the delivered dose, so that \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$d/\textrm{OER}(O)$$\end{document}d/OER(O) represents an oxygen-dependent effective dose, with the quadratic term scaling accordingly as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$d^2/\textrm{OER}^2(O)$$\end{document}d2/OER2(O). The values of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tilde{\alpha }$$\end{document}α~, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tilde{\beta }$$\end{document}β~, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta _{\alpha }$$\end{document}Δα, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta _{\beta }$$\end{document}Δβ are reported in Table 1. Further details on their derivation are provided in Supporting Information S5, where we also explore in depth the rationale underlying Eq. (6) and the theoretical implications of this formulation, which are used throughout the results presented in the following sections. In addition to this intrinsic phenotypic contribution, radiosensitivity is further modulated by the local oxygen concentration: increasing oxygen availability enhances radiation-induced killing through an OER-based rescaling of the LQ parameters α (O,u)α(O,u) and β (O,u)β(O,u). We adopt the following functional form7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \textrm{OER}(O) = 1 + \frac{(\textrm{OER}_{\max }-1)\, K_O}{\textrm{OER}_{\max }\,O + K_O} \end{aligned}$$\end{document}OER(O)=1+(OERmax-1)KOOERmaxO+KOin line with the saturation law proposed by Howard-Flanders and Alper (1957). Since the OER rescales the effective delivered dose as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$d/\textrm{OER}(O)$$\end{document}d/OER(O), the reciprocal factor \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$1/\textrm{OER}(O)$$\end{document}1/OER(O) is the quantity that increases with oxygenation in the radiation-response term. This factor ranges from \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$1/\textrm{OER}_{\max }$$\end{document}1/OERmax under anoxic conditions (O→ 0)(O→0) to 11 under well-oxygenated conditions \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(O\rightarrow +\infty )$$\end{document}(O→+∞). The parameter \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$K_O$$\end{document}KO corresponds to the oxygen level at which \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$1/\textrm{OER}(O)$$\end{document}1/OER(O) reaches the midpoint between its anoxic limit \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$1/\textrm{OER}_{\max }$$\end{document}1/OERmax and its well-oxygenated limit 11. Radiotherapy protocols are characterized by equally spaced isodoses and are denoted throughout by the triple \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathcal {P}=(d,\Delta ,N_f)$$\end{document}P=(d,Δ,Nf), where d is the dose per fraction, Δ Δ the inter-fraction interval, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$N_f$$\end{document}Nf the total number of fractions. If treatment starts at time \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$t^{RT}_0$$\end{document}t0RT, the set of irradiation times is \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathcal {T}=\{t^{RT}_0+j\Delta \;|\; j=0,\dots ,N_f-1\}$$\end{document}T={t0RT+jΔ|j=0,⋯,Nf-1}. The formulation in Eq. (5) is equivalent to imposing an instantaneous update of the cell density at each irradiation time \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$t^{RT}_i$$\end{document}tiRT:8\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} n(t^{RT,+}_i,x,u) = S(d)\, n(t^{RT,-}_i,x,u), \end{aligned}$$\end{document}n(tiRT,+,x,u)=S(d)n(tiRT,-,x,u),where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$t^{RT,-}_i$$\end{document}tiRT,-, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$t^{RT,+}_i$$\end{document}tiRT,+ denote instants immediately before and after dose delivery. Cells lethally damaged by radiation are assumed to be removed instantaneously. Delayed effects (e.g., senescence, necrotic accumulation, and radiation-induced microenvironmental remodeling) as well as sublethal damage repair are not explicitly modeled; accordingly, corrections such as the Lea-Catcheside protraction factor (Brenner 2008) are not included. Oxygen Dynamics The spatio–temporal evolution of the oxygen concentration O(t, x) is described by the following diffusion–consumption equation:9\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \partial _t O(t,x) = D_O\,\partial _{x}^2 O(t,x) - \lambda _O\,O(t,x) - \zeta _O\,\rho (t,x)\,O(t,x) + V_O(x) \end{aligned}$$\end{document}∂tO(t,x)=DO∂x2O(t,x)-λOO(t,x)-ζOρ(t,x)O(t,x)+VO(x)Here, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$D_O$$\end{document}DO denotes the oxygen diffusion coefficient in the tissue, and a Fickian diffusion process is assumed for simplicity. The term \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-\lambda _O O$$\end{document}-λOO accounts for physiological oxygen consumption by healthy tissue at reference density. Oxygen consumption by tumor cells is modeled through the nonlinear term \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-\zeta _O \rho O$$\end{document}-ζOρO. The consumption rate is assumed phenotype-independent for simplicity, as the present framework does not explicitly resolve alternative metabolic pathways or phenotype-dependent oxygen-consumption mechanisms. Since the previous term already accounts for oxygen consumption by healthy tissue at physiological density, the tumor contribution is interpreted as an effective excess uptake associated with the replacement of healthy cells by tumor cells. The oxygen variable O(t,x)O(t,x) is expressed throughout the paper in oxygen partial-pressure units (mmHg). Parameter values for \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\lambda _O$$\end{document}λO and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\zeta _O$$\end{document}ζO are reported in Table 1, and further details on their calibration are provided in Supporting Information S6. These parameters are chosen to ensure that oxygen dynamics rapidly reach quasi-steady conditions relative to the slower timescale of cell population dynamics. Finally, the source term \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$V_O(x)$$\end{document}VO(x) represents oxygen supply from the vasculature. Allowing \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$V_O$$\end{document}VO to depend on space enables the modeling of heterogeneous vascular structures, ranging from simplified and controlled configurations (e.g., in vitro conditions) to more complex and spatially irregular oxygen inputs characteristic of in vivo tumor environments. In this work, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$V_O$$\end{document}VO is assumed to be time-independent for the sake of model simplicity, since vascular dynamics are not explicitly modeled and lie outside the scope of the present study. Results We use the model to compare radiotherapy schedules across oxygenation regimes, using time-to-progression as the primary outcome. We first analyze spatially uniform oxygen supply to characterize the dose–interval response landscape and its mechanistic drivers, and then examine how toxicity constraints, intermediate oxygenation levels, and spatial oxygen-source heterogeneity reshape protocol ranking. Schedule Ranking under Uniform Oxygenation A natural starting point for applying the model is to consider tumor evolution under a spatially uniform oxygen supply. In the absence of tumor cells (ρ =0ρ=0) and under spatially uniform conditions, the steady state for the oxygen equation (9) is given by \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$O^* = V_O/\lambda _O$$\end{document}O∗=VO/λO. In the following, we fix the tumor-free steady state to a reference oxygenation level \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I_0$$\end{document}I0, with \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I_0 \in \{O_M, O_m, O_h\}$$\end{document}I0∈{OM,Om,Oh} (see Table 1), which determines the oxygen supply as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$V_O(x) = I_0 \lambda _O$$\end{document}VO(x)=I0λO for all \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$x \in \Omega _s$$\end{document}x∈Ωs, yielding a homogeneous and constant oxygen input. From a biological perspective, this setting can be interpreted as tumor evolution occurring near a vascular structure, such as a single vessel or a cluster of vessels, resulting in an approximately uniform oxygen supply along the spatial direction under consideration and constant over time. Under this configuration, oxygen rapidly becomes nearly uniform within the tumor core, with only small boundary-layer deviations, owing to its fast diffusion relative to the timescale of cell dynamics. For this reason, this scenario serves as a reference configuration before considering more complex spatial oxygen distributions. Normal tissues differ in their sensitivity to fraction size, which in the LQ framework is characterized by the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(\alpha /\beta )_H$$\end{document}(α/β)H ratio. Protocol comparisons are therefore performed under a fixed constraint on the biologically effective dose (BED), defined in Supporting Information S2, by requiring that the BED of each admissible protocol does not exceed that of a conventional standard-of-care (SoC) schedule \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathcal {P}_{SoC}=(d_{SoC}=1.8~\textrm{Gy}, \Delta _{SoC}=1~\textrm{day}, N_{f,SoC}=30)$$\end{document}PSoC=(dSoC=1.8Gy,ΔSoC=1day,Nf,SoC=30), namely10\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \textrm{BED}(\mathcal {P}) \le \textrm{BED}(\mathcal {P}_{SoC}) . \end{aligned}$$\end{document}BED(P)≤BED(PSoC).For each prescribed dose per fraction d, the number of fractions is chosen as the largest integer compatible with this constraint,11\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} N_f= \left\lfloor \frac{\textrm{BED}(\mathcal {P}_{SoC})}{d\left( 1+d/(\alpha /\beta )_H\right) } \right\rfloor . \end{aligned}$$\end{document}Nf=BED(PSoC)d1+d/(α/β)H.Equivalently, among all integer numbers of fractions satisfying the BED constraint, we select the one that yields the largest BED not exceeding \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\textrm{BED}(\mathcal {P}_{SoC})$$\end{document}BED(PSoC). Thus, for each dose–interval pair (d,Δ )(d,Δ), the protocol \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathcal {P}=(d,\Delta ,N_f)$$\end{document}P=(d,Δ,Nf) is uniquely specified. This construction enforces comparable levels of healthy-tissue toxicity across treatment schedules. By preventing an artificial increase in the effective dose, it allows the model to test the effects of fractionation and temporal scheduling separately from trivial dose escalation. Following the setup proposed by Henares-Molina et al. (2017), we explore alternative radiotherapy schedules by varying the dose per fraction d and the inter-fraction interval Δ Δ. Calendar effects associated with clinical delivery schedules, such as weekend breaks, are not explicitly considered in the present model. The explored dose range is \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$d \in (0,5]\,\textrm{Gy}$$\end{document}d∈(0,5]Gy, consistent with the validity limits of the LQ model, which is known to lose accuracy beyond this interval (Kirkpatrick et al. 2008; Cui et al. 2022). Time intervals are taken as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta \in (0,100]$$\end{document}Δ∈(0,100] days, again following Henares-Molina et al. (2017). This extended interval range is introduced to reveal the structural dependence of treatment outcome on the dose–interval trade-off. Although many of the explored intervals exceed typical clinical practice, they are included to characterize the underlying response behavior. As shown below, slices of the treatment landscape along either the dose or interval axis retain a similar qualitative structure, with systematic shifts in response, indicating a robust correlation between fraction spacing and dose rather than isolated optima. Figure 1 displays the predicted time-to-progression (TTP) across the (d,Δ )(d,Δ) parameter space under spatially uniform oxygen supply, in normoxic (left panel) and hypoxic (right panel) conditions. Here, TTP denotes the time elapsed from the initiation of radiotherapy to tumor relapse, defined as the time at which the total tumor burden returns to the prescribed detection threshold; its formal definition is reported in Supporting Information S2. Each dose–interval map was obtained through a systematic scan of the protocol space based on 25002500 independent model runs. Each run corresponds to a single dose–interval pair (d,Δ )(d,Δ), for which the number of fractions is determined by the BED constraint and the resulting TTP is recorded. Under high oxygenation, the TTP landscape is relatively flat, with a broad region of dose–interval combinations yielding comparable outcomes. In this regime, the predicted time-to-progression remains close to the value obtained under the SoC schedule, typically around two years, and deviations from this reference produce only marginal variations in relapse timing. In contrast, under hypoxic conditions the structure of the map changes qualitatively. A distinct region of the parameter space emerges in which TTP values increase substantially, reaching almost four years along a well-defined ridge in the (d,Δ )(d,Δ) plane. This structure delineates a therapeutic efficiency frontier, separating protocols with limited benefit from schedules capable of nearly doubling the time-to-progression relative to standard fractionation. Notably, the frontier is located in the region of the map characterized by longer inter-fraction intervals, indicating that more protracted or metronomic-like fractionation schedules can achieve the largest TTP gains under hypoxic conditions. The linear-like geometry of this high-TTP region further indicates the presence of a persistent trade-off between dose per fraction and inter-fraction spacing: rather than a single isolated optimum, comparable outcomes can be achieved along a continuous band of dose–interval combinations in the (d,Δ )(d,Δ) plane.Fig. 1Time-to-progression (TTP), measured in years, as a function of single-fraction dose and inter-fraction interval under a spatially uniform oxygen supply. Left: normoxic conditions (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I_0=O_M$$\end{document}I0=OM). Right: hypoxic conditions (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I_0=O_m$$\end{document}I0=Om). Protocol comparisons are performed under the constraint \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\textrm{BED}(\mathcal {P}) \le \textrm{BED}(\mathcal {P}_{\textrm{SoC}}) \approx 64\,\textrm{Gy}$$\end{document}BED(P)≤BED(PSoC)≈64Gy, where the standard-of-care protocol \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathcal {P}_{\textrm{SoC}}$$\end{document}PSoC consists of 3030 fractions of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$1.8\,\textrm{Gy}$$\end{document}1.8Gy delivered at 11-day intervals. Healthy-tissue fractionation sensitivity is characterized by \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(\alpha /\beta )_H = 10\,\textrm{Gy}$$\end{document}(α/β)H=10Gy, consistent with values reported in Henares-Molina et al. (2017). Oxygenation qualitatively reshapes the dose–interval response landscape: under normoxia, protocol performance is relatively insensitive to fractionation choice, whereas under hypoxia, protracted schedules with longer inter-fraction intervals achieve substantially greater TTP than the standard-of-care schedule Time-to-progression (TTP), measured in years, as a function of single-fraction dose and inter-fraction interval under a spatially uniform oxygen supply. Left: normoxic conditions (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I_0=O_M$$\end{document}I0=OM). Right: hypoxic conditions (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I_0=O_m$$\end{document}I0=Om). Protocol comparisons are performed under the constraint \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\textrm{BED}(\mathcal {P}) \le \textrm{BED}(\mathcal {P}_{\textrm{SoC}}) \approx 64\,\textrm{Gy}$$\end{document}BED(P)≤BED(PSoC)≈64Gy, where the standard-of-care protocol \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathcal {P}_{\textrm{SoC}}$$\end{document}PSoC consists of 3030 fractions of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$1.8\,\textrm{Gy}$$\end{document}1.8Gy delivered at 11-day intervals. Healthy-tissue fractionation sensitivity is characterized by \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(\alpha /\beta )_H = 10\,\textrm{Gy}$$\end{document}(α/β)H=10Gy, consistent with values reported in Henares-Molina et al. (2017). Oxygenation qualitatively reshapes the dose–interval response landscape: under normoxia, protocol performance is relatively insensitive to fractionation choice, whereas under hypoxia, protracted schedules with longer inter-fraction intervals achieve substantially greater TTP than the standard-of-care schedule Interplay between Reoxygenation and Phenotypic Selection To qualitatively elucidate the mechanisms underlying the observed structure of the dose–interval map, we compare tumor dynamics generated by two representative protocols located in distinct regions of Fig. 1: the SoC schedule and a protocol attaining maximal TTP within the same map. These cases serve as prototypical examples of low- and high-performance regimes and illustrate the dynamical mechanisms shaping treatment response. The resulting insights are consistent with the broader analysis presented in Sect. 3.4, where we systematically explore a range of oxygenation levels and evaluate the distribution of gains across near-optimal protocols. This analysis reveals a non-monotone dependence of treatment benefit on oxygen availability, with intermediate hypoxia emerging as the regime in which fractionation and timing have the largest impact. Figure 2 shows the time evolution of tumor burden, oxygen concentration, and mean phenotype (see Supporting Information S2 for its definition) for both protocols under normoxic and physiological hypoxic conditions. During the initial stage, before treatment delivery, the time evolution of the three reported quantities coincides for the two protocols, since tumor growth and the associated microenvironmental pressure are identical. Differences emerge only after radiation delivery, when the trajectories associated with the two protocols start to diverge. In the hypoxic case (panels (b,d,f) of Fig. 2), the visible change in slope of the pre-treatment tumor-burden curve around t≈ 1000t≈1000 days reflects a progressive slowing of tumor growth as the population expands. Increasing local cell density strengthens the crowding contribution in Eq. (2), while the accompanying decrease in oxygen availability further limits proliferation and promotes selection toward more hypoxia-adapted, less proliferative phenotypes. These combined effects reduce the net growth rate and explain the longer time required to reach the detection threshold \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Gamma _{\textrm{RT}}$$\end{document}ΓRT compared with the normoxic case. As shown in panels (a,c,e) of Fig. 2, under normoxic conditions, the maximal-TTP protocol operates within a mildly deoxygenated TME and induces only a limited phenotypic shift. In contrast, the SoC protocol promotes tumor reoxygenation while delivering more densely spaced radiation fractions, thereby redistributing the population toward higher-resistance trait values. As a result, microenvironmental effects and therapy-induced selection partially offset one another, leading to comparable TTP values despite markedly different underlying dynamics. Under hypoxic conditions (panels (b,d,f) of Fig. 2), the maximal-TTP protocol again maintains low oxygenation levels with only minor changes in the phenotypic distribution. The SoC protocol still increases oxygen availability; however, unlike the normoxic case, this does not induce a substantial redistribution of phenotypes. In this regime, the TME has already selected for highly resistant phenotypes, leaving little room for further therapy-driven phenotypic switching. Consequently, the balance observed under normoxia disappears: resistance is already established, resulting in earlier relapse of the SoC protocol after reoxygenation.Fig. 2Time evolution of tumor burden (top), oxygen level (middle), and mean phenotypic trait (bottom) under the standard-of-care protocol (blue solid line) and the maximal-TTP protocol (red solid line). Panels (a,b) show tumor burden, panels (c,d) the oxygen level, and panels (e,f) the mean phenotypic trait. Tumor burden Γ (t)Γ(t) is expressed as the total number of cells, while oxygen levels and thresholds are expressed in \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\textrm{mmHg}$$\end{document}mmHg. The standard-of-care protocol consists of 3030 fractions of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$1.8\,\textrm{Gy}$$\end{document}1.8Gy delivered at 11-day intervals. The detection threshold \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Gamma _{\textrm{RT}}$$\end{document}ΓRT and the oxygenation thresholds \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$O_M$$\end{document}OM, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$O_m$$\end{document}Om, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$O_h$$\end{document}Oh are reported in Table 1. Panels (a,c,e) correspond to the normoxic case \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(I_0 = O_M)$$\end{document}(I0=OM): the maximal-TTP protocol delivers approximately \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$2.3\,\textrm{Gy}$$\end{document}2.3Gy every 1515 days for 2222 fractions (total dose \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sim 51\,\textrm{Gy}$$\end{document}∼51Gy), yielding a TTP gain of approximately 113113 days. Panels (b,d,f) correspond to the hypoxic case \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(I_0 = O_m)$$\end{document}(I0=Om): the maximal-TTP protocol delivers approximately \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$1.3\,\textrm{Gy}$$\end{document}1.3Gy every 3535 days for 4242 fractions (total dose \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sim 55\,\textrm{Gy}$$\end{document}∼55Gy), yielding a TTP gain of approximately 22 years. Treatment-induced reoxygenation interacts differently with phenotypic selection across oxygenation regimes: under normoxia, SoC-induced reoxygenation is partially offset by redistribution toward resistant phenotypes, whereas under hypoxia resistance is already established, leading to earlier relapse under SoC than under the protracted maximal-TTP schedule (Color figure online) Time evolution of tumor burden (top), oxygen level (middle), and mean phenotypic trait (bottom) under the standard-of-care protocol (blue solid line) and the maximal-TTP protocol (red solid line). Panels (a,b) show tumor burden, panels (c,d) the oxygen level, and panels (e,f) the mean phenotypic trait. Tumor burden Γ (t)Γ(t) is expressed as the total number of cells, while oxygen levels and thresholds are expressed in \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\textrm{mmHg}$$\end{document}mmHg. The standard-of-care protocol consists of 3030 fractions of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$1.8\,\textrm{Gy}$$\end{document}1.8Gy delivered at 11-day intervals. The detection threshold \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Gamma _{\textrm{RT}}$$\end{document}ΓRT and the oxygenation thresholds \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$O_M$$\end{document}OM, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$O_m$$\end{document}Om, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$O_h$$\end{document}Oh are reported in Table 1. Panels (a,c,e) correspond to the normoxic case \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(I_0 = O_M)$$\end{document}(I0=OM): the maximal-TTP protocol delivers approximately \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$2.3\,\textrm{Gy}$$\end{document}2.3Gy every 1515 days for 2222 fractions (total dose \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sim 51\,\textrm{Gy}$$\end{document}∼51Gy), yielding a TTP gain of approximately 113113 days. Panels (b,d,f) correspond to the hypoxic case \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(I_0 = O_m)$$\end{document}(I0=Om): the maximal-TTP protocol delivers approximately \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$1.3\,\textrm{Gy}$$\end{document}1.3Gy every 3535 days for 4242 fractions (total dose \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sim 55\,\textrm{Gy}$$\end{document}∼55Gy), yielding a TTP gain of approximately 22 years. Treatment-induced reoxygenation interacts differently with phenotypic selection across oxygenation regimes: under normoxia, SoC-induced reoxygenation is partially offset by redistribution toward resistant phenotypes, whereas under hypoxia resistance is already established, leading to earlier relapse under SoC than under the protracted maximal-TTP schedule (Color figure online) Impact of Toxicity Constraints So far, protocol comparisons have been performed under the constraint that the equivalent toxicity does not exceed that of a SoC schedule. This assumption, however, is not universally appropriate. In some clinical scenarios, dose escalation may be considered to improve tumor control (Brower et al. 2016), whereas in others stricter toxicity limits may require a reduction of the admissible dose (Timmerman et al. 2006). From a clinical standpoint, treatment planning is inherently multi-objective, involving a trade-off between tumor control probability (TCP) (Gong et al. 2013) and normal-tissue toxicity, commonly quantified through the Normal Tissue Complication Probability (NTCP) (Gaito et al. 2022). Additional variability arises from patient- and site-specific factors, including intrinsic radiosensitivity (Nuijens et al. 2025) and the geometric configuration of organs at risk (OARs) (Eber et al. 2025). These considerations suggest that differences in normal-tissue responsiveness and tolerance to fraction size modulate the feasible treatment space and should therefore be incorporated into protocol comparison. The aim of the following analysis is to characterize how protocol efficiency depends on oxygen availability, normal-tissue radiosensitivity, and admissible toxicity levels. To this end, we extend the exploration of simulated radiotherapy schedules by considering multiple combinations of BED, normal-tissue radiosensitivity \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(\alpha /\beta )_H$$\end{document}(α/β)H, and oxygen availability. TTP is again adopted as the efficacy metric. The therapeutic efficiency frontier – formally defined in Supporting Information S7 – is identified as the subset of dose–interval combinations attaining at least 80%80% of the maximum predicted TTP. For each parameter configuration, we compute a TTP-weighted barycenter of the frontier (see Supporting Information S7 for details). This barycenter is not intended to identify an optimal schedule; rather, it provides a compact descriptor of the region in the dose–interval plane associated with highly effective regimens, thereby enabling systematic comparison across oxygenation and toxicity scenarios. Nine parameter configurations are analyzed by combining three oxygenation states—normoxic \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(I_0 = O_M)$$\end{document}(I0=OM), hypoxic \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(I_0 = O_m)$$\end{document}(I0=Om), and severely hypoxic \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(I_0 = O_h)$$\end{document}(I0=Oh)—with three paired toxicity settings: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(\textrm{BED},(\alpha /\beta )_H)=(90\,\textrm{Gy},10\,\textrm{Gy})$$\end{document}(BED,(α/β)H)=(90Gy,10Gy), \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(60\,\textrm{Gy},6\,\textrm{Gy})$$\end{document}(60Gy,6Gy), and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(30\,\textrm{Gy},2\,\textrm{Gy})$$\end{document}(30Gy,2Gy).Fig. 3Therapeutic efficiency frontiers in the dose–interval plane under spatially uniform oxygen supply. Each column corresponds to a fixed normal-tissue sensitivity \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(\alpha /\beta )_H$$\end{document}(α/β)H and BED constraint, common to all oxygenation regimes shown in that column. Rows correspond to increasing hypoxia: (a) normoxic conditions \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(I_0=O_M)$$\end{document}(I0=OM), (b) moderately hypoxic conditions \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(I_0=O_m)$$\end{document}(I0=Om), and (c) severely hypoxic conditions \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$(I_0=O_h)$$\end{document}(I0=Oh). For each parameter configuration, the therapeutic efficiency frontier contains the protocol combinations achieving at least 80%80% of the maximum TTP. Star markers indicate the dose–interval pair achieving maximal TTP, while circular markers indicate the corresponding TTP-weighted barycenters. The dose–interval plane is partitioned into three regions to distinguish schedules typically classified as hyperfractionated (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$d \le 1.5\,\textrm{Gy}$$\end{document}d≤1.5Gy), standard-to-moderately hypofractionated (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$1.5\,\textrm{Gy}< d < 3.5\,\textrm{Gy}$$\end{document}1.5Gy