This article addresses the modeling issue of phenotypic heterogeneity in tumors treated by radiotherapy. Cell population is described as a superposition of independent dynamics. The behavior of each cell is defined by a finite collection of states expressing the possible evolution of some phenotypes after treatment. The population dynamics is given by an age-dependent multi-type branching process. From this representation, we use classical probabilistic tools to derive equations of the mean population size and its variance. Those equations have been implemented into Scilab© to perform numerical simulations. The latter allows biologists to test different assumptions and more particularly to estimate the effect of each phenotypic state on the treatment response.