The importance of nuclear interactions for ion therapy arises from the influence of the particle spectrum on, first, radiobiology and therefore also on treatment planning, second, the accuracy of measuring dose and, third, the delivered dose distribution. This study tries to determine the qualitative as well as the quantitative influence of the modeling of inelastic nuclear interactions on ion therapy. Thereby, three key disciplines are investigated, namely dose delivery, dose assessment and radiobiology. In order to perform a quantitative analysis, a relative comparison between six different descriptions of nuclear interactions is carried out for carbon ions. The particle transport is simulated with the Monte Carlo code SHIELD-HIT10A while dose planning and radiobiology are covered by the analytic treatment planning program for particles TRiP, which determines the relative biological effectiveness (RBE) with the local effect model. The obtained results show that the physical dose distribution can in principle be significantly influenced by the modeling of fragmentation (about 10% for a 20% change in all inelastic nuclear cross sections for a target volume ranging from 15 to 25 cm). While the impact of nuclear fragmentation on stopping power ratios can be neglected, the fluence correction factor may be influenced by the applied nuclear models. In contrast to the results for the physical dose, the variation of the RBE is only small (about 1% for a 20% change in all inelastic nuclear cross sections) suggesting a relatively weak dependence of radiobiology on the detailed composition of the particle energy spectrum of the mixed radiation field. Also, no significant change (about 0.2 mm) of the lateral penumbra of the RBE-weighted dose is observed.