Dynamic contrast enhanced (DCE) MRI is increasingly used for tumor imaging, both for clinical use and research. However, existing DCE-based methods have limitations with regard to standardization and reproducibility, limiting the possi- bility to compare results from different studies, or even between examinations in the same patient. In this work, different parameters regarding the acquisition of the DCE data, and the subsequent analysis of the data have been investigated in order to gain some insight into what is important to consider when doing such analysis. Synthetic time curves have been produced, simulating different situations such as different sampling frequencies, dispersion of the arterial input function (AIF), the effect of incorrect estimation of the bolus arrival time (BAT), and the effect of noise. Finally, the results from the simulation study were compared to sample clinical DCE data. Three established DCE based pharmacokinetic models were specifically in- vestigated; the so-called Tofts, and extended Tofts models, as well as the two compartment exchange model. The latter model was defined as the "ground truth" model throughout, providing accurate a-priori knowledge of both perfusion (flow, Fp), plasma, and extracellular, extravascular volume fractions (vp, ve), and capillary permeability (Ktrans). The research suggests a method, using the transfer functions of the kinetic models, to analytically determine if the sampling frequency that is used is adequate for the kinetic parameters reported by the model used to describe the data. The results show good correspondence between the simulated results and the analytical estimates. Specifically, confident estimation of plasma flow (Fp) required a sampling rate closely related to the mean transit time (MTT) of the bolus (vp/Fp). In relation, the sampling rate requirement for confident estimation of Ktrans is given by ve/Ktrans. In brain DCE, the fraction (vp/Fp)/(ve/Ktrans) turns out to be about in the order of 10-100, meaning the sampling interval needs to be about 10-100 times faster for accurate measurement of Fp compared to Ktrans, depending on the physiology. A frequency analysis on the AIF bolus dispersion (BD) was made, suggesting that low degree of BD (sharp AIF) is preferable for flow estimation. For accurate estimation of Ktrans, it was shown that the sampling frequency requirement is somewhat lower for a large degree of BD. Incorrect estimation of the BAT was shown to lead to inaccurate results, and that there is not a one-to-one relationship between goodness of fit, and correctness of fit, the latter being accurate parameter estimation, and the former being the fit which best matches the input curve.