This paper deals with the estimation of regions of attraction (RoAs) for a cancer dynamical model. The estimation of this type of sets is very interesting in the field of control for cancer dynamics, since it provides the set of possible initial health indicators, for which there exists a treatment protocol allowing to heal the patient. Therefore, we propose a methodology to estimate the region of attraction of a nonlinear dynamical system describing the interaction between a tumor, the immune system and combined therapies of cancer. Then, we use the characterization of the RoA for a given model parameters vector, in order to derive an estimation of an outer approximation of the robust RoA under parametric uncertainties.