Hybrid classical-Bayesian approaches for sample size calculations are used to formally integrate prior beliefs or uncertainty about unknown parameters involved in the traditional power. In this paper, we consider a hybrid procedure that exploits different prior distributions according to the role played by the unknown quantities. The approach is applied to obtain the optimal sample sizes of comparative studies based on Poisson data by using the rate ratio as the measure of interest. A numerical example is provided, allowing the comparison between classical and hybrid methods.