Type 2 Diabetes affects over 415 million people worldwide. The condition is associated with an increased risk of blindness, kidney failure, heart attacks and premature death. Improved glucose control is shown to substantially reduce these risks. However, most guildelines for Type 2 Diabetes management are reactive and can be seen to be more of a trial and error system until a potential cure is found. There are major uncertainties regarding the likely effectiveness of different treatments at an individual level. Personalised Medicine aims to look at the individual instead of the whole population to estimate how they will respond on each treatment to determine the most effective therapy for them personally. This thesis develops different models for starting drugs for individuals with newly diagnosed Type 2 Diabetes. Using data from the UK Prospective Diabetes Study (UKPDS) two models were created. The first uses a Bayesian mixture regression model to predict glycated haemoglobin (HbA1c), an overall measure of glycaemic exposure, after the first year of therapy. Then increments of HbA1c were modelled via a Wiener process with baseline covariate effects and an individual fraility parameter. This second model predicts the next HbA1c level as well as an estimate for how long an individual can stay on a particular therapy before they reach a critical HbA1c boundary. These two models could assist diabetes management by indicating a patients most effect therapy for them, as well as estimating the likely time before an additional or replacement therapy would be required.