The present study aims at evaluating the ability of a gastro-intestinal in vitro lipolysis model to predict the performance of two lipid formulations and a conventional tablet containing a poorly soluble drug, cinnarizine, in dogs, both in the fasted and fed state. A self-nano-emulsifying drug delivery system (SNEDDS) was either dosed in a hard gelatin capsule (SNEDDS-C) or loaded onto a porous tablet core (SNEDDS-T) and compared to a marketed conventional tablet (Conv) in an in vitro lipolysis model. The model simulates the digestion in the stomach and intestine during either the fasted or the fed state. Whole fat milk (3.5 %) was used in the fed state model to mimic the dynamic lipolysis events after ingestion of food. The results were compared to a dog study in this issue. In the fasted state in vitro lipolysis model, the amount of solubilized cinnarizine decreased in the order SNEDDS-C>SNEDDS-T>Conv, which correlated well with the in vivo bioavailability. In the fed state in vitro lipolysis model, cinnarizine was solubilized to the same degree for all formulations. Compared to the fasted state model, only the performance of the conventional tablet was improved, indicating food effect. This correlated with the in vivo study, where the tablet was the only formulation with a significant food effect. The fasted state model correlated well with the in vivo results and although the fed state model did not accurately predict the fed state in vivo results, it could predict which formulation that would exhibit a food effect.