Golden-angle radial sparse parallell (GRASP) MRI uses temporal regularization in the reconstruction, which risks distorting temporal profiles and reducing DCE-MRI parameter accuracy and precision. The aim of this study is to investigate this issue for liver DCE-MRI by measuring kinetic parameters on data reconstructed at variable temporal resolution. The results depend on temporal resolution according to well-known patterns also observed in simulations and fully sampled data. A systematic error remains at the highest temporal resolution, but this is more likely due to well-known issues of signal saturation. Image reconstruction at lower temporal resolution risks degrading diagnostic image quality due to the mixing of images with different contrast. We conclude that: (1) temporal regularization in GRASP is unlikely to induce significant error in kinetic parameters; (2) images should be reconstructed at high temporal resolution around 2-4s.