We evaluate the reproducibility of ktrans values derived from dynamic contrast MRI images in patients with newly diagnosed glioblastoma. Particular focus is put on the reproducibility across choices of T1 values and arterial input functions (AIFs) to the Tofts-Kermode model. Reproducibility is assessed across multiple pre-therapy baseline visits in a 45 patient cohort. Our model based on static population AIFs and static global T1 values had excellent reproducibility compared to other models including unique individual AIFs and T1 maps. There is relatively little concordance between all tested models, but individual AIFs led to higher mean ktrans values.