Protonen töten Zellen wirksamer ab als Photonen. Die klinisch verwendete konstante relative biologische Wirksamkeit (RBW) für Protonen vernachlässigt jedoch erste klinische Evidenz einer RBW-Variabilität, die vom linearen Energietransfer (LET) abhängt. Diese Arbeit trägt dazu bei, die RBW-Variabilität in Protonen-Bestrahlungsplänen zu berücksichtigen, um potenzielle Nebenwirkungen zu vermindern. Zuerst wurde ein erhöhtes Risiko für RBW-induzierte Nebenwirkungen bei Hirntumorpatienten festgestellt. Dies konnte jedoch nicht systematisch durch klinische Planungsstrategien reduziert werden. Zweitens ergab eine multizentrische europäische Studie, dass die zentrums-spezifischen, nicht standardisierten LET-Berechnungen erheblich voneinander abweichen. Eine harmonisierte LET-Definition wurde vorgeschlagen und reduzierte die Variabilität zwischen den Zentren auf ein klinisch akzeptables Niveau, was künftig eine einheitliche Dokumentation des Therapieergebnisses ermöglicht. Abschließend wurden vier Strategien zur RBW-Reduktion in der Planoptimierung bei Hirntumorpatienten angewandt, die das Risiko für Nekrose und Erblindung erheblich reduzierten. LET-Optimierung in Hochdosisregionen erscheint besonders geeignet, um die Sicherheit der Patientenbehandlung künftig weiter zu verbessern.:List of Figures vii List of Tables viii List of Acronyms and Abbreviations ix 1 Introduction 1 2 Theoretical background 3 2.1 Proton interactions with matter 4 2.2 Biological effect of radiation 8 2.2.1 Linear-quadratic model 8 2.2.2 Relative biological effectiveness 9 2.3 Proton beam delivery and field formation 13 2.4 Treatment planning 14 2.4.1 Patient modelling and structure definition 15 2.4.2 Treatment plan optimisation 16 2.4.3 Treatment plan evaluation 19 2.5 Proton therapy uncertainties and mitigation strategies 22 2.5.1 Clinical mitigation strategies 23 2.5.2 Optimisation approaches beyond absorbed dose 26 3 Variable biological effectiveness in PBS treatment plans 29 3.1 LET and RBE recalculations of proton treatment plans with RayStation 30 3.1.1 Monte Carlo dose engine 30 3.1.2 Monte Carlo scoring extensions 32 3.1.3 Graphical user interface 33 3.2 LET assessment and the role of range uncertainties 36 3.2.1 Patient cohort and treatment plan creation 37 3.2.2 Simulation of range deviations 38 3.2.3 Treatment plan recalculation settings 39 3.2.4 Resulting impact of range deviations 40 3.3 Patient recalculations in case of side effects 46 3.3.1 Image registration and range prediction 48 3.3.2 Retrospective treatment plan assessment 49 3.4 Benefit of an additional treatment field 50 3.4.1 Patient and treatment plan information 50 3.4.2 Results of variable RBE recalculations 51 3.5 Discussion 51 3.6 Summary 59 4 Status of LET and RBE calculations in European proton therapy 61 4.1 Study design 62 4.1.1 Treatment planning information 64 4.1.2 Data processing and treatment plan evaluation 67 4.2 Treatment plan comparisons in the water phantom 68 4.2.1 Absorbed dose evaluation 69 4.2.2 Centre-specific LET calculations 69 4.2.3 Harmonised LET calculations 71 4.3 Treatment plan comparisons in patient cases 72 4.3.1 Dose-averaged linear energy transfer for protons 73 4.3.2 Centre-specific RBE models and parameters 76 4.4 Discussion 77 4.5 Summary 82 5 Biological treatment plan optimisation 83 5.1 Treatment plan design 84 5.1.1 Clinical goals 86 5.1.2 Novel treatment plan optimisation approaches 87 5.2 Treatment plan quality assessment with a constant RBE 90 5.3 Assessment of NTCP reductions with a variable RBE 90 5.4 Discussion 95 5.5 Conclusion 100 6 Summary 103 7 Zusammenfassung 107 Bibliography 111 Danksagung 137 / Protons are more effective in cell killing than photons. However, the clinically applied constant proton relative biological effectiveness (RBE) neglects emerging clinical evidence for RBE variability driven by the linear energy transfer (LET). This thesis aims to safely account for RBE variability in proton treatment plans to mitigate potential side effects. First, an elevated risk for RBE induced overdosage was found in brain tumour patients. However, this could not be mitigated systematically by clinical planning strategies. Second, a multicentric European study revealed that centre-specific non-standardised LET calculations differed substantially. A harmonised LET definition was proposed which reduced the inter-centre variability to a clinically acceptable level and allows for future consistent outcome reporting. Finally, four strategies to include RBE variability in treatment plan optimisation were applied to brain tumour patients, which considerably reduced the estimated risk for necrosis and blindness. Of these, LET optimisation in high dose regions may be suited for clinical practice to further enhance patient safety in view of a variable RBE.:List of Figures vii List of Tables viii List of Acronyms and Abbreviations ix 1 Introduction 1 2 Theoretical background 3 2.1 Proton interactions with matter 4 2.2 Biological effect of radiation 8 2.2.1 Linear-quadratic model 8 2.2.2 Relative biological effectiveness 9 2.3 Proton beam delivery and field formation 13 2.4 Treatment planning 14 2.4.1 Patient modelling and structure definition 15 2.4.2 Treatment plan optimisation 16 2.4.3 Treatment plan evaluation 19 2.5 Proton therapy uncertainties and mitigation strategies 22 2.5.1 Clinical mitigation strategies 23 2.5.2 Optimisation approaches beyond absorbed dose 26 3 Variable biological effectiveness in PBS treatment plans 29 3.1 LET and RBE recalculations of proton treatment plans with RayStation 30 3.1.1 Monte Carlo dose engine 30 3.1.2 Monte Carlo scoring extensions 32 3.1.3 Graphical user interface 33 3.2 LET assessment and the role of range uncertainties 36 3.2.1 Patient cohort and treatment plan creation 37 3.2.2 Simulation of range deviations 38 3.2.3 Treatment plan recalculation settings 39 3.2.4 Resulting impact of range deviations 40 3.3 Patient recalculations in case of side effects 46 3.3.1 Image registration and range prediction 48 3.3.2 Retrospective treatment plan assessment 49 3.4 Benefit of an additional treatment field 50 3.4.1 Patient and treatment plan information 50 3.4.2 Results of variable RBE recalculations 51 3.5 Discussion 51 3.6 Summary 59 4 Status of LET and RBE calculations in European proton therapy 61 4.1 Study design 62 4.1.1 Treatment planning information 64 4.1.2 Data processing and treatment plan evaluation 67 4.2 Treatment plan comparisons in the water phantom 68 4.2.1 Absorbed dose evaluation 69 4.2.2 Centre-specific LET calculations 69 4.2.3 Harmonised LET calculations 71 4.3 Treatment plan comparisons in patient cases 72 4.3.1 Dose-averaged linear energy transfer for protons 73 4.3.2 Centre-specific RBE models and parameters 76 4.4 Discussion 77 4.5 Summary 82 5 Biological treatment plan optimisation 83 5.1 Treatment plan design 84 5.1.1 Clinical goals 86 5.1.2 Novel treatment plan optimisation approaches 87 5.2 Treatment plan quality assessment with a constant RBE 90 5.3 Assessment of NTCP reductions with a variable RBE 90 5.4 Discussion 95 5.5 Conclusion 100 6 Summary 103 7 Zusammenfassung 107 Bibliography 111 Danksagung 137

Monte-Carlo calculations are used to simulate the stochastic effects of fast neutron-induced chemical changes in the radiolysis of the ferrous sulfate (Fricke) dosimeter. To study the dependence of the yield of ferric ions, G(Fe[superscript 3+]), on fast neutron energy, we have simulated, at 25 [degree centigrade], the oxidation of ferrous ions in aerated aqueous 0.4 M H[subscript 2]SO[subscript 4] (pH 0.46) solutions when subjected to ~0.5-10 MeV incident neutrons, as a function of time up to ~50 s. The radiation effects due to fast neutrons are estimated on the basis of track segment (or"escape") yields calculated for the first four recoil protons with appropriate weighting according to the energy deposited by each of these protons. For example, a 0.8-MeV neutron generates recoil protons of 0.505, 0.186, 0.069, and 0.025 MeV, with linear energy transfer (LET) values of ~41, 69, 82, and 62 keV/[micro]m, respectively. In doing so, we consider that further recoils make only a negligible contribution to radiation processes. Our results show that the radiolysis of dilute aqueous solutions by fast neutrons produces smaller radical yields and larger molecular yields (relative to the corresponding yields for the radiolysis of water by [superscript 60]Co [gamma]-rays or fast electrons) due to the high LET associated to fast neutrons. The effect of recoil ions of oxygen, which is also taken into account in the calculations, is shown to decrease G(Fe[superscript 3+]) by about 10%. Our calculated values of G(Fe[superscript 3+]) are found to increase slightly with increasing neutron energy over the energy range covered in this study, in good agreement with available experimental data. We have also simulated the effect of temperature on the G(Fe[superscript 3+]) values in the fast neutron radiolysis of the Fricke dosimeter from 25 to 300 [degree centigrade]. Our results show an increase of G(Fe[superscript 3+]) with increasing temperature, which is readily explained by an increase in the yields of free radicals and a decrease in those of molecular products. For 0.8-MeV incident neutrons (the only case for which experimental data are available in the literature), there is a ~23% increase in G(Fe[superscript 3+]) on going from 25 to 300 [degree centigrade]. Although these results are in reasonable agreement with experiment, more experimental data, in particular for different incident neutron energies, would be needed to test more rigorously our Fe[superscript 3+] ion yield results at elevated temperatures.

Résumé : Le contrôle de la chimie de l'eau dans un réacteur nucléaire refroidi à l'eau nécessite une compréhension détaillée des effets de la radiolysede l'eau afin de limiter la corrosion et la dégradation des matériaux par oxydation générée par les produits de cette radiolyse. Toutefois, la mesure directe de la chimie dans le cœur des réacteurs est extrêmement difficile, sinon impossible, en raison des conditions extrêmes de haute température et haute pression, et les champs d’irradiation mixtes neutrons/γ, qui ne sont pas compatibles avec l'instrumentation chimique normale. Pour ces raisons,des modèles théoriques et des simulations sur ordinateur sont essentielles pour la prédiction de la chimie sous rayonnement de l'eau de refroidissement dans le cœur et son impact sur les matériaux. Dans ce travail, des simulations Monte Carlo ont été utilisées pour calculer les rendements des principales espèces (e[indice supérieur -][indice inférieur aq], H[indice supérieur •], H[indice inférieur 2], [indice supérieur •]OH et H[indice inférieur 2]O[indice inférieur 2]) formées lors de la radiolyse de l’eau liquide neutre par des neutrons mono-énergétiques de 2 MeV à des températures entre 25 et 350 °C. Le choix des neutrons de 2 MeV comme énergie d'intérêt est représentatif du flux de neutrons rapides dans un réacteur. Pour l'eau légère, la contribution la plus significative à la radiolyse vient des quatre premières collisions des neutrons qui produisent, dans la majorité des cas, des protons avec des énergies de recul de ~1.264, 0.465, 0.171 et 0.063 MeV et des transferts d’énergie linéique (TEL) moyens respectivement de ~22, 43, 69et 76 keV/[micro]m. Par ailleurs, nous avons négligé les effets des radiations dus aux ions de recul de l’oxygène. Les rendements moyens finaux peuvent alors être estimés comme étant la somme des rendements résultant de l’action de ces protons après pondérations en fonction de l’énergie déposée. Les rendements ont été calculés à 10[indice supérieur -7], 10[indice supérieur -6] et 10[indice supérieur -5] s. Les valeurs obtenues sont en accord avec les données expérimentales disponibles. En comparant nos résultats avec les données obtenues pour les rayonnements à faible TEL (rayons γ de [indice supérieur 60]Co ou électrons rapides), nos rendements calculés pour les neutrons rapides ont montré une dépendance en température essentiellement similaire, mais avec des valeurs plus faibles pour les rendements en radicaux libres et des valeurs plus élevées pour les rendements moléculaires. Nous avons également utilisé les simulations Monte Carlo pour étudier l'existence de la chute rapide de la constante de vitesse de réaction de l'électron hydraté (e[indice supérieur -][indice inférieur aq]) sur lui-même – l’une des principales sources de formation de H[indice inférieur 2] – au-dessus de 150 °C. Cette dépendance en température a été observée expérimentalement en milieu alcalin par divers auteurs, mais jamais en milieu neutre. Lorsque cette baisse de la constante de vitesse d’auto-réaction de e[indice supérieur -][indice inférieur aq] est incluse dans nos codes de simulation, tant pour des rayonnements de bas TEL (grappes isolés) que de haut TEL (trajectoires cylindriques), g(H[indice inférieur 2]) montre une discontinuité marquée à la baisse à ~150°C, ce qui n'est pas observée expérimentalement. Les conséquences de la présence de cette discontinuité dans le rendement en H[indice inférieur 2] pour les rayonnements à bas et haut TEL sont discutées. Enfin, nous avons tenté d’expliquer l'augmentation – considérée comme anormale – du rendement en H[indice inférieur 2] en fonction de la température au-dessus de 200 °C par l’intervention de la réaction des atomes H[indice supérieur •] avec l'eau, préalablement proposée par Swiatła-Wojcik et Buxton en 2005. La constante de vitesse de cette réaction est toujours controversée. // Abstract : Controlling the water chemistry in a water-cooled nuclear power reactor requires understanding and mitigating the effects of water radiolysis to limit the corrosion and degradation of materials by oxidizing radiolysis products. However, direct measurement of the chemistry in reactor cores is extremely difficult due to the extreme conditions of high temperature, pressure, and mixed neutron/γ-radiation fields, which are not compatible with normal chemical instrumentation. For these reasons, theoretical models and computer simulations are essential for predicting the detailed radiation chemistry of the cooling water in the core and the impact on materials. Monte Carlo simulations were used to calculate the yields for the primary species (e[superscript -][subscript aq], H[superscript •], H[subscript 2], [superscript •]OH, and H[subscript 2]O[subscript 2]) formed from the radiolysis of neutral liquid water by mono-energetic 2-MeV neutrons and the mechanisms involved at temperatures between 25 and 350 °C. In this work, we chose 2-MeV neutron as our energy of interest since it is known as representative of a fast neutron flux in a nuclear reactor. For light water, for that chosen energy, the most significant contribution to the radiolysis comes from the first four neutron collisions that generate mostly ejected protons with energies of ~1.264, 0.465, 0.171, and 0.063 MeV, which had, at 25 °C, mean linear energy transfers (LETs) of ~22, 43, 69, and 76 keV/[micro]m, respectively. In this work, we simply neglected the radiation effects due to oxygen ion recoils. The average final fast neutron yields could be estimated as the sum of the yields for these protons after allowance was made for the appropriate weightings (by using the Eq (2) in Chapter 4) according to their deposited energy. Yields were calculated at 10[superscript -7], 10[superscript -6] and 10[superscript -5] s. Our computed yield agreed reasonably well with the available experimental data. By comparing our results with data obtained for low-LET radiation ([superscript 60]Co γ-rays or fast electrons), our computed yields for fast neutron radiation showed essentially similar temperature dependences over the range of temperature studied, but with lower values for yields of free radicals and higher values for molecular yields. In this work, we also used our Monte Carlo simulation to investigate the existence of drop of hydrated electron (e[superscript -][subscript aq]) self-reaction rate constant at 150 °C. One of the main sources of H[subscript 2] formation is the self-reaction of hydrated electrons. The temperature dependence of the rate constant of this reaction (k[subscript 1]), measured under alkaline conditions, reveals that the rate constant drops abruptly above ~150 °C. However, when this drop in the e[superscript -][subscript aq] self-reaction rate constant is included in our code for low (isolated spurs) and high (cylindrical tracks) linear energy transfer (LET), g(H[subscript 2]) shows a marked downward discontinuity at ~150 °C which is not observed experimentally. The consequences of the presence of this discontinuity in H[subscript 2] yield for both low and high LET radiation are discussed. Another reaction that might explain the anomalous increasing of H[subscript 2] yield with temperature is the reaction of H[superscript •] atoms with water previously proposed by Swiatla-Wojcik and Buxton (2005) whose rate constant is still in controversial.

Résumé : Les ions hydronium (H3O + ) sont formés, à temps courts, dans les grappes ou le long des trajectoires de la radiolyse de l'eau par des rayonnements ionisants à faible transfert d’énergie linéaire (TEL) ou à TEL élevé. Cette formation in situ de H3O + rend la région des grappes/trajectoires du rayonnement temporairement plus acide que le milieu environnant. Bien que des preuves expérimentales de l’acidité d’une grappe aient déjà été signalées, il n'y a que des informations fragmentaires quant à son ampleur et sa dépendance en temps. Dans ce travail, nous déterminons les concentrations en H3O + et les valeurs de pH correspondantes en fonction du temps à partir des rendements de H3O + calculés à l’aide de simulations Monte Carlo de la chimie intervenant dans les trajectoires. Quatre ions incidents de différents TEL ont été sélectionnés et deux modèles de grappe/trajectoire ont été utilisés : 1) un modèle de grappe isolée "sphérique" (faible TEL) et 2) un modèle de trajectoire "cylindrique" (TEL élevé). Dans tous les cas étudiés, un effet de pH acide brusque transitoire, que nous appelons un effet de "pic acide", est observé immédiatement après l’irradiation. Cet effet ne semble pas avoir été exploré dans l'eau ou un milieu cellulaire soumis à un rayonnement ionisant, en particulier à haut TEL. À cet égard, ce travail soulève des questions sur les implications possibles de cet effet en radiobiologie, dont certaines sont évoquées brièvement. Nos calculs ont ensuite été étendus à l’étude de l'influence de la température, de 25 à 350 °C, sur la formation in situ d’ions H3O + et l’effet de pic acide qui intervient à temps courts lors de la radiolyse de l’eau à faible TEL. Les résultats montrent une augmentation marquée de la réponse de pic acide à hautes températures. Comme de nombreux processus intervenant dans le cœur d’un réacteur nucléaire refroidi à l'eau dépendent de façon critique du pH, la question ici est de savoir si ces fortes variations d’acidité, même si elles sont hautement localisées et transitoires, contribuent à la corrosion et l’endommagement des matériaux. / Abstract : Hydronium ions (H3O+) are formed within spurs or tracks of the low or high linear energy transfer (LET) radiolysis of pure, deaerated water at early times. The in situ radiolytic formation of H3O+ renders the spur and track regions temporarily more acid than the surrounding medium. Although experimental evidence for an acidic spur has already been reported, there is only fragmentary information on its magnitude and time dependence. In this work, spur or track H3O+ concentrations and the corresponding pH values are obtained from our calculated yields of H3O+ as a function of time, using Monte Carlo track chemistry simulations. We selected four impacting ions and we used two different spur and track models: 1) an isolated “spherical” spur model characteristic of low-LET radiation and 2) an axially homogeneous “cylindrical” track model for high-LET radiation. Very good agreement was found between our calculated time evolution of G(H3O+) in the radiolysis of pure, deaerated water by 300-MeV incident protons (which mimic 60Co gamma/fast electron irradiation) and the available experimental data at 25 °C. For all cases studied, an abrupt transient acid pH effect, which we call an “acid spike”, is observed during and shortly after the initial energy release. This acid-spike effect is virtually unexplored in water or in a cellular environment subject to the action of ionizing radiation, especially high-LET radiation. In this regard, this work raises a number of questions about the potential implications of this effect for radiobiology, some of which are briefly evoked. Our calculations were then extended to examine the effect of temperature from 25 to 350 °C on the yield of H3O+ ions that are formed in spurs of the low-LET radiolysis of water. The results showed an increasingly acidic spike response at higher temperatures. As many in-core processes in a water-cooled nuclear reactor critically depend on pH, the question here is whether these variations in acidity, even highly localized and transitory, contribute to material corrosion and damage.

At low doses, relevant to nuclear incidents and accidental releases of radioactivity, the detriment of radiation extends beyond direct effects. This thesis investigates genomic instability, a subclass of non-targeted effects where damage and lethality is transmitted vertically and expressed in the progeny of cells many generations after initial radiation exposure. Through a series of experiments using clonogenic assay of human and fish cell culture, studies described in this thesis describe lethal mutations, hyper radiosensitivity and increased radioresistance – processes involving repair mechanisms that dictate survival in cells exposed to low doses. Further study investigates the difference in the relative biological effect of alpha particle radiation compared to what is expected at high doses. Results demonstrate increased radioresistance in a human cell line while also revealing increased lethality in a fish cell line confirming the need for consideration of dose-dependence as well as variance in behaviors of different cell lines and species. It is hoped the conclusions of this thesis will inspire the creation of protocols with greater attention to the indirect consequences of exposure to radiation at doses relevant to nuclear incidents and accidents. / Thesis / Master of Science (MSc)