Epithelial-mesenchymal transition (EMT) plays a critical role in the early stages of dissemination of carcinoma leading to metastatic tumors, which are responsible for over 90% of all cancer-related deaths. Current therapeutic regimens, however, have been ineffective in the cure of metastatic cancer, thus an urgent need exists to revisit existing protocols and to improve the efficacy of newly developed therapeutics. Strategies based on preventing EMT could potentially contribute to improving the outcome of advanced stage cancers. To achieve this goal new assays are needed to identify targeted drugs capable of interfering with EMT or to revert the mesenchymal-like phenotype of carcinoma to an epithelial-like state. Current assays are limited to examining the dispersion of carcinoma cells in isolation in conventional 2-dimensional (2D) microwell systems, an approach that fails to account for the 3-dimensional (3D) environment of the tumor or the essential interactions that occur with other nearby cell types in the tumor microenvironment. Here we present a microfluidic system that integrates tumor cell spheroids in a 3D hydrogel scaffold, in close co-culture with an endothelial monolayer. Drug candidates inhibiting receptor activation or signal transduction pathways implicated in EMT have been tested using dispersion of A549 lung adenocarcinoma cell spheroids as a metric of effectiveness. We demonstrate significant differences in response to drugs between 2D and 3D, and between monoculture and co-culture.