We tested the performance of six diffusion models for detecting effects from treatment in pancreatic ductal adenocarcinoma. All models, including the mono-exponential fit, were able to detect parameter changes after treatment in individual patients for five out of nine patients. For multi-parametric models treatment affected different parameters for different patients, allowing potentially discriminating between types of effects. The pseudo diffusion coefficient values from the intravoxel incoherent motion model were in line with instant dephasing due to bulk motion rather than capillary perfusion, limiting its clinical interpretation. Using two mono-exponential fits, one at low and one at high b-values, was preferred.