Study and development of physical models to evaluate biological effects of ion therapy: the study of local control of prostate cancer

External beam radiotherapy (EBRT) is a therapy technique aiming at treating locoregional tumors with high efficiency. However, many tumors remain uncontrolled. Newest EBRT techniques always aim at increasing the dose to the tumor while sparing the surrounding healthy tissues. Carbon-ion beam therapy is one of these promising techniques. The number of clinical centres offering carbon-ion beam radiotherapy has been increasing over the world for the last decade. This keen interest spread after very promising results from pilot projects at Berkeley (USA), Chiba (Japan) and Darmstadt (Germany). The theoretical advantages of carbon-ions are better spatial selectivity in dose deposition and better efficiency in cell killing. They have thus the potential to increase the control of tumors, particularly for unresectable radioresistant tumors. In high linear-energy-transfer (LET) radiations, such as carbon-ion beams, biological effects vary along the ion track, hence, to quantify them, specific radiobiological models are needed. There exist several radiobiological models based on very different theoretical approaches and approximations. They were created and improved in each of the pilot institutions. At the current state of knowledge, no convergence between the model results seems to be possible in the very near future. Clinically employed radiobiological models are the Local Effect Model (LEM) developed in Germany and implemented in CE-certified treatment planning systems, the National Institute of Radiological Science (NIRS) model employed in Japanese centres with passive beam delivery systems and the microdosimetric kinetic model (MKM) in Japanese centres with active scanning beam delivery systems. Mathematical models can be used to predict, for instance, Tumor Control Probability (TCP) and then evaluate treatment outcomes. For a given radiation dose delivered in a defined schedule, TCP models calculate the level of tumor control. In the first step, this work studies the influence of the LEM (first version) input parameters on the TCP predictions in the specific case of prostate cancer. Several published input parameters and their combination were tested. Their influence on the dose distribution calculated in a water phantom and in a patient geometry was evaluated. Changing input parameters induced clinically significant modifications of mean dose, spatial dose distribution, and TCP predictions. TCP predictions were found more sensitive to the parameter threshold dose (D_t) than to the biological parameters α and β. In the second step, methods for evaluation, comparison and optimization of clinical protocols are presented. For defined boundary conditions, 'optimized' D_t are proposed for the treatment of prostate cancer. The theoretical methodologies developed in this thesis can only be validated when they will be challenged with clinical results from on-going trials. More generally, the improvement of radiobiological models and the comparison of clinical results from different institutions will be only achieved when more patient outcome data with well-defined patient groups, fractionation schemes, well-defined end-points and harmonized reporting will be available. (author)