Toxicity of the respiratory system is quite common after radiotherapy in thoracic tumours. The quantification of lung tissue response to irradiation is important in designing treatments associated with a minimum of complications and maximum tumor control. This work aims to estimate volumes V13, V20 and V30 as an index of radiation pneumonitis occurrence, to evaluate the predictive strength of the relative seriality, Lyman-Kutcher-Burman(LKB) and parallel normal tissue complication probability (NTCP) models regarding the incidence of radiation pneumonitis in a group of patients following lung cancer radiotherapy when lung perceived as paired and single organ respectively and also software development for the determination of the best estimates of the models’ parameters based on maximum likelihood method. The study was based on 46 patients and for each patient, lung dose-volume histograms (DVHs) and the clinical treatment outcome was available. From the 46 patients treated, 28 of them were scored as having radiation induced pneumonitis, with RTOG criteria grade ≥2. Firstly lungs were evaluated as a paired organ. Analyzing this material we failed to associate volume V13, V20 and V30 with radiation pneumonitis occurrence (χ2-test: probability of agreement between observed and predicted results using the 0.05 significance level). By applying ANOVA of the NTCP models examined in the overall group considering lungs as paired organs the LKB with Martel et al parameter set gave the best results, whereas when lungs perceived as individual organ (unhealthy lung volume-PTV) the best model was appeared to be LKB with Burman et al parameter set. However, in this relatively small group of lung cancer patients NTCP models didn’t show excessive correlation with the clinical outcome. Nevertheless, when total lung volume irradiated and total dose received were taken into account as factors of radiation pneumonitis prediction, correlation was almost duplicated for both perception of lungs. In order to achieve the best fitting of models to the clinical outcome for the specific patient group, maximum likelihood analysis was applied via software development using mle programming language, to find those parameters that maximize the likelihood function. When lungs perceived as single organ, the best fitting of models to the clinical outcome for relative seriality were D50 = 22Gy, γ= 2, s=0.031, LKB model D50 = 23Gy, m=0.18, n=1 and for parallel model, D50 = 20Gy, m=0.2, n=0.6. Maximum likelihood analysis was not applied for paired lung assumption as constraints did not allow us to properly fit the models. / -

- / The treatment techniques applied in the chest area (breast and lung cancer) in radiotherapy, increase the lung toxicity leading to the development of pulmonary complications. The aim of this study is to compare the predictive strength of different radiobiological models in the evaluation of radiation pneumonitis, correlate the absorbed dose with the severity of the clinical outcome and examine biological factors that may affect or induce complications after irradiation. Furthermore, due to the fact that the value of each parameter is accompanied by its confidence interval, every model is represented by a group of dose-response curves that create a range in which each radiobiological model can vary. The range of each model is very important when selecting the values of parameters used, due to the existence of coincendence areas between the models. The study was based on 179 breast cancer patients undergoing radiotherapy. Dose volume histograms and the clnical treatment outcome for every patient were available. 24 patients scored radiation pneumonitis grade 2 and 65 showed milder symptoms of grade 1. Normal tissue complication probabilities were computed for every model for each patient. Moreover, statistical analysis was applied to investigate whether the absorbed dose is the only factor that influeces the development of radiation pneumonitis in breast cancer radiotherapy (x2 test) and the ability of the radiobiological models used to discriminate cases that developed radiation pneumonitis from those that did not (ROC curves). The Relative Seriality model described with greater accuracy the clinical outcome in contrast to the LKB and Parallel model. The statistical results showed that radiation pneumonitis in the case of breast cancer radiotherapy does not depend only on the absorbed dose but on other radiobiological factors that induce the development of complication, such as the intrinsic radiosensitivity of each patient. ROC curves pointed out the weakness of the models to discriminate cases of complication from cases of non-complication. Finally, this study accented the importance to use parameter values extracted from patients groups with similar clinical characteristics as the ones examined in order to avoid the coincidence areas between the models.

The main purpose of this thesis is to encourage the development of bioeffect planning as an experimental tool by which means bioeffect plans may be compared with standard isodose plans. This thesis also addresses the fundamental problems of the derivation of useful biological models for clinical application and the description of tumour and normal tissue parameter values and their variability. Particular emphasis has been placed on comparing the predictive value of the models and parameters against clinical results of fractionated and continuous irradiation either alone or combined.

The main purpose of this thesis is to encourage the development of bioeffect planning as an experimental tool by which means bioeffect plans may be compared with standard isodose plans. This thesis also addresses the fundamental problems of the derivation of useful biological models for clinical application and the description of tumour and normal tissue parameter values and their variability. Particular emphasis has been placed on comparing the predictive value of the models and parameters against clinical results of fractionated and continuous irradiation either alone or combined.

The search for the optimal choice of treatment time, dose and fractionation regimen is one of the major challenges in radiation therapy. Several aspects of the radiation response of tumours and normal tissues give different indications of how the parameters defining a fractionation schedule should be altered relative to each other which often results in contradictory conclusions. For example, the increased sensitivity to fractionation in late-reacting as opposed to early-reacting tissues indicates that a large number of fractions is beneficial, while the issue of accelerated repopulation of tumour cells starting at about three weeks into a radiotherapy treatment would suggest as short overall treatment time as possible. Another tumour-to-normal tissue differential relevant to the sensitivity as well as the fractionation and overall treatment time is the issue of tumour hypoxia and reoxygenation. The tumour oxygenation is one of the most influential factors impacting on the outcome of many types of treatment modalities. Hypoxic cells are up to three times as resistant to radiation as well-oxygenated cells, presenting a significant obstacle to overcome in radiotherapy as solid tumours often contain hypoxic areas as a result of their poorly functioning vasculature. Furthermore, the oxygenation is highly dynamic, with changes being observed both from fraction to  fraction and over a time period of weeks as a result of fast and slow reoxygenation of acute and chronic hypoxia. With an increasing number of patients treated with hypofractionated stereotactic body radiotherapy (SBRT), the clinical implications of a substantially reduced number of fractions and hence also treatment time thus have to be evaluated with respect to the oxygenation status of the tumour. One of the most promising tools available for the type of study aiming at determining the optimal radiotherapy approach with respect to fractionation is radiobiological modelling. With clinically validated in vitro-derived tissue-specific radiobiological parameters and well-established survival models, in silico modelling offers a wide range of opportunities to test various hypotheses with respect to time, dose, fractionation and details of the tumour microenvironment. Any type of radiobiological modelling study intended to provide a realistic representation of a clinical tumour should therefore take into account details of both the spatial and temporal tumour oxygenation. This thesis presents the results of three-dimensional radiobiological modelling of the response of tumours with heterogeneous oxygenation to various fractionation schemes, and oxygenation levels and dynamics using different survival models. The results of this work indicate that hypoxia and its dynamics play a major role in the outcome of radiotherapy, and that neglecting the oxygenation status of tumours treated with e.g. SBRT may compromise the treatment outcome substantially. Furthermore, the possibilities offered by incorporating modelling into the clinical routine are explored and demonstrated by the development of a new calibration function for converting the uptake of the hypoxia-PET tracer 18F-HX4 to oxygen partial pressure, and applying it for calculations of the doses needed to overcome hypoxia-induced radiation resistance. By hence demonstrating how the clinical impact of hypoxia on dose prescription and the choice of fractionation schedule can be investigated, this project will hopefully advance the evolution towards routinely incorporating functional imaging of hypoxia into treatment planning. This is ultimately expected to result in increased levels of local control with more patients being cured from their cancer. / <p>At the time of the doctoral defense, the following papers were unpublished and had a status as follows: Paper 2: Manuscript. Paper 6: Manuscript.</p>

Thulani Nyathi, Student no: 0413256X, MSc thesis, Physics, Faculty of science. 2006. Supervisor: Prof D van der Merwe. / This study aims to calculate tumour control probabilities (TCP) and normal tissue complication probabilities (NTCP) using radiobiological models and correlate these probabilities with clinically observed treatment outcome from follow-up records. These radiobiological calculations were applied retrospectively to thirty-nine paediatric patients who were treated with radiation at Johannesburg Hospital during the period January 1990 to December 2000 and had histologically proven rhabdomyosarcoma. Computer software, BIOPLAN, was used to calculate the TCP and NTCP arising from the dose distribution calculated by the treatment planning system and characterized by dosevolume histograms (DVHs). There was a weak correlation between the calculated TCP and the observed 5-year overall survival status. Furthermore, potential prognostic factors for survival were examined. Statistical analysis was performed using the Cox proportional hazards regression model. The 5-year overall survival rate was 55 %. The findings of this study are a yardstick against which more aggressive radiotherapy fractionation regimes can be compared.

<p>The present thesis is focused on the development and application of dose response models for radiation therapy. Radiobiological models of tissue response to radiation are an integral part of the radiotherapeutic process and a powerful tool to optimize tumor control and minimize damage to healthy tissues for use in clinical trials. Ideally, the models could work as a historical control arm of a clinical trial eliminating the need to randomize patents to suboptimal therapies. In the thesis overview part, some of the basic properties of the dose response relation are reviewed and the most common radiobiological dose-response models are compared with regard to their ability to describe experimental dose response data for rat spinal cord using the maximum likelihood method. For vascular damage the relative seriality model was clearly superior to the other models, whereas for white matter necrosis all models were quite good except possibly the inverse tumor and critical element models. The radiation sensitivity, seriality and steepness of the dose-response relation of the spinal cord is found to vary considerably along its length. The cervical region is more radiation sensitive, more parallel, expressing much steeper dose-response relation and more volume dependent probability of inducing radiation myelitis than the thoracic part. The higher number of functional subunits (FSUs) consistent with a higher amount of white matter close to the brain may be responsible for these phenomena. With strongly heterogeneous dose delivery and due to the random location of FSUs, the effective size of the FSU and the mean dose deposited in it are of key importance and the radiation sensitivity distribution of the FSU may be an even better descriptor for the response of the organ. An individual optimization of a radiation treatment has the potential to increase the therapeutic window and improve cure for a subgroup of patients.</p>

The present thesis is focused on the development and application of dose response models for radiation therapy. Radiobiological models of tissue response to radiation are an integral part of the radiotherapeutic process and a powerful tool to optimize tumor control and minimize damage to healthy tissues for use in clinical trials. Ideally, the models could work as a historical control arm of a clinical trial eliminating the need to randomize patents to suboptimal therapies. In the thesis overview part, some of the basic properties of the dose response relation are reviewed and the most common radiobiological dose-response models are compared with regard to their ability to describe experimental dose response data for rat spinal cord using the maximum likelihood method. For vascular damage the relative seriality model was clearly superior to the other models, whereas for white matter necrosis all models were quite good except possibly the inverse tumor and critical element models. The radiation sensitivity, seriality and steepness of the dose-response relation of the spinal cord is found to vary considerably along its length. The cervical region is more radiation sensitive, more parallel, expressing much steeper dose-response relation and more volume dependent probability of inducing radiation myelitis than the thoracic part. The higher number of functional subunits (FSUs) consistent with a higher amount of white matter close to the brain may be responsible for these phenomena. With strongly heterogeneous dose delivery and due to the random location of FSUs, the effective size of the FSU and the mean dose deposited in it are of key importance and the radiation sensitivity distribution of the FSU may be an even better descriptor for the response of the organ. An individual optimization of a radiation treatment has the potential to increase the therapeutic window and improve cure for a subgroup of patients.

A set of mathematical models, known as radiobiological Dose-Response models, have been developed, to model the biological effects and complications that arise following irradiation. The overall objective is to be able to apply these in clinical practice with confidence, and ensure more successful treatments are given to patients. This investigation serves to assess these models and their predictive power of NTCP following irradiation of the lung. Clinical data, from patients treated for inoperable stage III non-small cell lung cancer is obtained and the consequent biological effect (severity of pneumonitis) observed as a result of this radiation treatment is assessed by the models. By gaining more knowledge about the 3D dose-distribution and the incidence of radiation pneumonitis through the evaluation of the models, the main treatment goal, which is to maximise TCP and minimise NTCP can be achieved. Post treatment data is obtained regarding the clinical outcome or clinical endpoint for each patient, considered to be Radiation Pneumonitis. The clinical endpoint is a specific biological effect that may or may not have occurred,after a certain period, following irradiation. The models are assessed on their ability to predict a NTCP value that corresponds to the resulting clinical endpoint following treatment. Furthermore a software tool for the calculation of NTCP’s by the models is developed, in an attempt to provide an important tool for optimization of radiotherapy treatment planning. With our findings from this study, our aim is to further strengthen, support and challenge already existing literature on dose-response modelling. / -

The aim of this study is the clinical derivation of the dose-response relations of bladder and prostate regarding PSA progression and urinary complications using patients treated for prostate cancer with both photon and proton beams. Such data are necessary for a prospective estimation of the clinical effectiveness of radiation therapy using combinations of different radiation modalities. Material During the period from 2002 until 2006, at the Academic Hospital in Uppsala, Sweden 189 patients underwent radiotherapy for prostate cancer, which combined photon and proton beams therapy. Of these patients, 100 have been included in this study and have been analysed for the prostate. The analyses for urinary complications were made for 72 patients who didn’t have final clinical urinary outcome equal to one . The dose distribution delivered to the prostate, the two regions of the bladder and the clinical treatment outcome, were available for each patient. The patients were given a proton boost of 20 Gy in 4 fractions of 5 Gy in addition to a conventional photon beam treatment, which was prescribed to a dose of 50 Gy in 25 fractions of 2 Gy. In this analysis, the delineated regions of interest were the prostate, the whole urinary bladder and the lower 3 cm part of the bladder. It is known that most urinary complications come from the lower 3cm part of bladder due to its anatomical position near to urethra and prostate. The photon and proton doses were calculated using the BED (biologically effective dose) concept. Furthermore, for the calculation of the proton dose an RBE value of 1.1 was considered. Finally, the combined effective dose was chosen to be the sum of the maximum dose of protons and the mean dose of photons for the whole bladder and the bladder-3cm, while for the prostate, the effective dose was considered as the sum of the mean dose for photons and the minimum dose for protons. The radiobiological parameter acquisition was performed for the Poisson Binomial and Probit models using the Maximum Likelihood method. Results Of the 100 patients, 94 had tumor control (94 %), whereas 6 patients had treatment failure (6 %). Of the 72 patients, 15 (21%) showed urinary complications, whereas 57 (79%) were complication-free.. The estimated values of the parameters for tumour are D50= 49.4 Gy (68% CI = 47.90-52.80 Gy) and γ = 2.25 (68% CI = 1.95-2.80) for the Poisson , D50= 49.55Gy (68% CI= 47.56-51.45Gy) and γ = 2.25 (68% CI = 1.95-2.80) for Binomial, whereas for the Probit model the values of D50 and γ50 are 47,27Gy (68% CI = 45.25-50.01 Gy) and 1,33 (68% CI = 1.20-1.37), respectively. The estimated values of the parameters for the whole bladder are D50= 104 Gy (68% CI = 103.12-105.01 Gy) and γ = 0.7 (68% CI = 0.67-0.72) for the Poisson , D50= 108 Gy (68% CI= 106-108.8 Gy) and γ = 0.6 (68% CI = 0.58-0.70) for Binomial, whereas for the Probit model the values of D50 and γ50 are 97 Gy (68% CI = 95.30-97.56 Gy) and 1 (68% CI = 0.94-1.12), respectively. Finally, the estimated values of the parameters for bladder 3cm are D50= 88.4 Gy (68% CI = 85.4-89.5 Gy) and γ = 1.30 (68% CI = 1.18-1.45) for the Poisson , D50= 88.58 Gy (68% CI= 86.21-89.85 Gy) and γ = 1.28 (68% CI = 1.12-1.51) for Binomial, whereas for the Probit model the values of D50 and γ50 are 85.58 Gy (68% CI = 83.23-89.21Gy) and 1.78 (68% CI = 1.56-1.83), respectively. From the derived mean DVHs of the prostate it is concluded that photon and proton therapies contribute the same in the toxicity of the patients and both proton and photon therapy provide the patients with the prescribed dose in the target – prostate gland. From the derived mean DVHs of the bladder and the bladder 3cm, it is observed that photon therapy provides the patients with more dose than the proton therapy, thus it can be assumed that the urinary complications were mainly due to the photon treatment. Bladder 3cm receives more dose both during photon and proton therapy in comparison with the whole bladder for patients with and without complications. Thus, the lower 3cm part of the bladder contributes more in the possibility of urinary complications. In ROC analysis, for the prostate the area under the ROC curve is 0.71. This indicates that the model distinguishes quite well the group of the patients with and without PSA progression. For the whole bladder and the lower 3cm of the bladder, the results are 0.61 and 0.65 respectively; the model does not separate well the two groups (with and without the complications). These results suggest that other factors may also be important for urinary toxicity. Conclusions The dose-response relations of bladder and prostate appear to be described well by the estimated parameters of the Poisson, Poisson and Probit models. Future studies incorporating more radiobiological models and more detailed factors describing the combined treatment and endpoint registration will be needed until an accurate prospective estimation of the expected urinary complications is reached. / Ο σκοπός αυτής της μελέτης είναι ο προσδιορισμός των παραμέτρων δόσης απόκρισης αναφορικά με το κλινικό αποτέλεσμα της εξέλιξης της PSA και των ουροποιητικών επιπλοκών μετά από ακτινοθεραπεία προστάτη με φωτόνια και πρωτόνια. Αυτά τα δεδομμένα είναι πού χρήσιμα στην κλινική πράξη για την εκτίμηση και σύγκριση των πλάνων ακτινοθεραπείας. Υλικά και μέθοδοι Κατά την περίοδο 2002 με 2006, στο Ακαδημαϊκό Νοσοκομείο της Ουψάλα, Σουηδίας, 189 ασθενείς υποβλήθηκαν σε ακτινοθεραπεία φωτονίων και πρωτονίων για προστάτη. Από το σύνολο των ασθενών, 100 μελετήθηκαν και συμπεριλήφθηκαν στην παρούσα εργασία. Ειδικότερα για τις ουροποιητικές επιπλοκές, η ανάλυση πραγματοποιήθηκε για 72 ασθενείς χωρίς κλινικό αποτέλεσμα ίσο με 1. Το κλινικό αποτέλεσμα και οι κατανομές δόσεις της ουροδόχου κύστης και του προστάτη ήταν διαθέσιμα για κάθε ασθενή. Οι ασθενείς δέχθηκαν θεραπεία πρωτονίων των 20 Gy σε 4 συνεδρίες των 5 Gy καθώς και θεραπεία φωτονίων των 50 Gy σε 25 συνεδρίες των 2 Gy. Οι περιοχές ενδιαφέροντος που απεικονίστηκαν είναι ο προστάτης, ολόκληρη η ουροδόχος κύστη και τα χαμηλότερα 3 cm από την βάση της ουροδόχου κύστης (κύστη 3εκ.). Είναι γνωστό ότι οι περισσότερες επιπλοκές προέρχονται από το τμήμα που περιλαμβάνει τα χαμηλότερα 3 cm από την βάση της ουροδόχου κύστης εξαιτίας της ανατομικής του θέσης κοντά στην ουρήθρα και στον προστάτη. Οι δόσεις φωτονίων και πρωτονίων υπολογίστηκαν χρησιμοποιώντας την BED. Επίσης, στον υπολογισμό της δόσης πρωτονίων χρησιμοποιήθηκε η RBE ίση με 1.1. Τέλος, η ισοδύναμη δόση για την κύστη και τα 3 εκ. της κύστης είναι το άθροισμα της μέγιστης δόσης πρωτονίων και της μέσης δόσης φωτονίων. Ενώ για τον προστάτη, είναι το άθροισμα της μέσης δόσης φωτονίων και της ελάχιστης δόσης των πρωτονίων. Τα δεδομένα χρησιμοποιήθηκαν σε μια διαδικασία προσαρμογής μέγιστης πιθανοφάνειας (maximum likelihood fitting) ώστε να υπολογιστούν οι βέλτιστες τιμές των παραμέτρων που χρησιμοποιούνται από τα μοντέλα Poisson, Binomial και Probit. Αποτελέσματα Από τους 100 ασθενείς , 94 ήταν ασυμπτωματικοί ασθενείς και 6 εμφάνισαν επιπλοκή όσο αφορά την εξέλιξης της PSA . Από τους 72 ασθενείς, 15 (21%) παρουσίασαν ουροποιητικές επιπλοκές ενώ, 57 (79%) δεν παρουσίασαν. Οι βέλτιστες εκτιμήσεις των παραμέτρων δόσης απόκρισης για τον όγκο είναι are D50= 49.4 Gy (68% CI = 47.90-52.80 Gy) και γ = 2.25 (68% CI = 1.95-2.80)για το Poisson, για το Binomial μοντέλο D50= 49.55Gy (68% CI= 47.56-51.45Gy) and γ = 2.25 (68% CI = 1.95-2.80), ενώ για το Probit μοντέλο οι τιμές για τα D50 και γ50 είναι 47,27Gy (68% CI = 45.25-50.01 Gy) και 1,33 (68% CI = 1.20-1.37) αντίστοιχα. . Οι βέλτιστες εκτιμήσεις των παραμέτρων δόσης απόκρισης για την κύστη είναι D50= 104 Gy (68% CI = 103.12-105.01 Gy) και γ = 0.7 (68% CI = 0.67-0.72) για το Poisson , D50= 108 Gy (68% CI= 106-108.8 Gy) και γ = 0.6 (68% CI = 0.58-0.70) για το Binomial, ενώ για το Probit μοντέλο οι τιμές για τα D50 και γ50 είναι 97 Gy (68% CI = 95.30-97.56 Gy) και 1 (68% CI = 0.94-1.12), αντίστοιχα. Τέλος, οι βέλτιστες εκτιμήσεις των παραμέτρων δόσης απόκρισης για την κύστη 3cm είναι D50= 88.4 Gy (68% CI = 85.4-89.5 Gy) και γ = 1.30 (68% CI = 1.18-1.45) για το Poisson, D50= 88.58 Gy (68% CI= 86.21-89.85 Gy) and γ = 1.28 (68% CI = 1.12-1.51) για το Binomial μοντέλο, , ενώ για το Probit μοντέλο οι τιμές είναι 85.58 Gy (68% CI = 83.23-89.21Gy) and 1.78 (68% CI = 1.56-1.83). Από τα μέσα αθροιστικά διαγράμματα του προστάτη προκύπτει ότι και η θεραπεία φωτονίων και πρωτονίων συνεισφέρουν το ίδιο στην τοξικότητα των ασθενών, αλλά και ότι και οι δυο θεραπείες δίνουν στον στόχο την καθορισμένη δόση. Από τα μέσα αθροιστικά διαγράμματα της κύστης και της κύστης 3cm, παρατηρείται ότι η θεραπεία φωτονίων προσφέρει στους ασθενείς με επιπλοκές μεγαλύτερη δόση από την θεραπεία πρωτονίων, με αποτέλεσμα να μπορεί να υποτεθεί ότι οι ουροποιητικές επιπλοκές οφείλονται κυρίως στην θεραπεία φωτονίων. Επίσης, παρατηρείται ότι η κύστη 3cm λαμβάνει περισσότερη δόση και στην θεραπεία φωτονίων και στην θεραπεία πρωτονίων συγκριτικά με ολόκληρη την κύστη για όλους τους ασθενείς με ή χωρίς επιπλοκές. Έτσι, συμπεραίνεται ότι αυτό το μέρος της κύστης συμβάλλει περισσότερο στην εμφάνιση επιπλοκών. Επιπλέον, στην ROC ανάλυση, για τον προστάτη η περιοχή κάτω από τη ROC καμπύλη είναι 0.71, δηλαδή τα μοντέλα φαίνονται να διαφοροποιούν καλά τις ομάδες ασθενών με και χωρίς επιπλοκή της PSA. Για την ουροδόχο κύστη και την κύστη 3cm , τα αποτελέσματα είναι 0.61 και 0.65 αντίστοιχα. Τα μοντέλα δεν διαφοροποιούν καλά τις ομάδες με ή χωρίς. Συμπεραίνεται ότι υπάρχουν και άλλοι παράγοντες που επηρεάζουν την ουροποιητική τοξικότητα. Συμπεράσματα Οι σχέσεις δόσης απόκρισης για την κύστη και τον προστάτη φαίνεται να περιγράφονται αρκετά καλά από τις εκτιμημένες παραμέτρους δόσης απόκρισης για τα Poisson, Poisson και Probit μοντέλα. Μελλοντικές μελέτες με περισσότερα ακτινοβιολογικά μοντέλα και λεπτομερέστερους παράγοντες, που θα περιγράφουν την συνδυασμένη θεραπεία και το κλινικό αποτέλεσμα είναι απαραίτητες για μια ακριβή πρόβλεψη των επιπλοκών.