Radiotherapy, delivered through any modality, aims to administer a high dose of radiation to a target volume whilst minimising the dose delivered to normal, healthy tissue. The fundamental characteristics of proton beams provide an advantage over photons in this regard but make them more sensitive to various uncertainties.To ensure their safe delivery, proton plans should be evaluated to identify any detriment to the delivered dose distribution that may result from these uncertainties. The effect of potential errors should be mitigated by reducing uncertainties and by generating plans that are robust to any residual uncertainty. However, robustness to uncertainty comes at the cost of plan quality in the absence of error and larger uncertainties require greater compromises. For this reason, it is important that the evaluated uncertainties are not overly conservative and that the trade-off between plan quality and robustness to uncertainty is carefully considered. In this thesis, approaches to the evaluation and mitigation of uncertainties are presented with a focus on the effect of fractionated treatment deliveries on the uncertainty in dose distributions resulting from setup uncertainties. A method of evaluating the robustness of treatment plans to uncertainties, incorporating the effect of fractionation on those uncertainties resulting from random setup errors, was developed. This approach uses the dose distribution calculated in a limited number of âerror scenariosâ to estimate the upper and lower bounds on the dose distribution over the entire treatment course. Validation against a Monte Carlo type simulation of treatment courses for clinical intensity modulated proton therapy (IMPT) plans was performed. Building on this method of plan evaluation, a ârobustness databaseâ of acceptable parameters was developed for head and neck patients treated with IMPT. This approach allows for the evaluation of future treatment plans to identify those plans which lie outside of the range of previous experience. Numerous methods of including uncertainties in the plan optimisation exist. However, they neglect to account for the impact of fractionated deliveries, resulting in unnecessary compromise to plan quality. A novel optimisation method that evaluates the bounds of the dose distribution over a treatment course was developed. This fractionation incorporated robust optimisation was demonstrated using a 2-dimensional optimisation environment in which a comparison with plans robustly optimised without considering the effect of fractionation and conventionally optimised plans was performed. The proposed approach demonstrated a reduced sensitivity to uncertainty compared to conventionally optimised plans and a reduced integral dose compared to robustly optimised plans. As such, fractionation incorporated robust optimisation presents a basis upon which plans could be developed that are safe and reduce both the unnecessary irradiation of normal tissues and the unnecessary compromise of target coverage.