Patients suffering from Diabetes Mellitus illness need to control their levels of sugar by a restricted diet, a healthy life and in the cases of those patients that do not produce insulin (or with a severe defect on the action of the insulin they produce), by injecting synthetic insulin before and after the meals. The amount of insulin, namely bolus, to be injected is usually estimated based on the experience of the doctor and of the own patient. During the last years, several computational tools have been designed to suggest the boluses for each patient. Some of the successful approaches to solve this problem are based on obtaining a model of the glucose levels which is then applied to estimate the most appropriate dose of insulin. In this paper we describe some advances in the application of evolutionary computation to obtain those models. In particular, we extend some previous works with Grammatical Evolution, a branch of Genetic Programming. We present results for ten real patients on the prediction on several time horizons. We obtain reliable and individualized predictive models of the glucose regulatory system, eliminating restrictions such as linearity or limitation on the input parameters.