Pancreatic ductal adenocarcinoma (PDAC) contains a subset of exclusively tumorigenic cancer stem cells (CSCs) which have been shown to drive tumor initiation, metastasis and resistance to radio- and chemotherapy. Here we describe a specific methodology for culturing primary human pancreatic CSCs as tumor spheres in anchorage-independent conditions. Cells are grown in serum-free, non-adherent conditions in order to enrich for CSCs while their more differentiated progenies do not survive and proliferate during the initial phase following seeding of single cells. This assay can be used to estimate the percentage of CSCs present in a population of tumor cells. Both size (which can range from 35 to 250 micrometers) and number of tumor spheres formed represents CSC activity harbored in either bulk populations of cultured cancer cells or freshly harvested and digested tumors. Using this assay, we recently found that metformin selectively ablates pancreatic CSCs; a finding that was subsequently further corroborated by demonstrating diminished expression of pluripotency-associated genes/surface markers and reduced in vivo tumorigenicity of metformin-treated cells. As the final step for preclinical development we treated mice bearing established tumors with metformin and found significantly prolonged survival. Clinical studies testing the use of metformin in patients with PDAC are currently underway (e.g., NCT01210911, NCT01167738, and NCT01488552). Mechanistically, we found that metformin induces a fatal energy crisis in CSCs by enhancing reactive oxygen species (ROS) production and reducing mitochondrial transmembrane potential. In contrast, non-CSCs were not eliminated by metformin treatment, but rather underwent reversible cell cycle arrest. Therefore, our study serves as a successful example for the potential of in vitro sphere formation as a screening tool to identify compounds that potentially target CSCs, but this technique will require further in vitro and in vivo validation to eliminate false discoveries.