PURPOSE: To observe the pattern of glue cast formation when glue was injected into the different fow layersof a flow phantom, and to demonstrate the effect in an animal experiment. MATERIALS AND METHODS: A flow phantom model consisted of a circulating pump (Simon varistaltic pump, Manostat Corporation, U.S.A.) connected via a tygon tube with a specially designed glass tube. Flow of about 4ml/sec was maintained by saline. The glass tube consisted of (1) a small-caliber (2.4-3.2mm inner diameter) straight inflow portion (feeder), (2) a round, larg-caliber portion (nidus), and (3) a medium-caliber (6.7mm inner diameter) straight outflow portion (vein). Glue mixed with Lipiodol (glue: Lipiodol=1:1-3:1) was loaded into a 1cc syringe and slowly injected through a microcatheter ; cast formation along the glass tube was observed. The feeder and nidus portions were injected twelve times and the shape of the cast was compared. In an animal experiment, glue at various concentrations was used to stop rapid flow through a surgically-constructed arteriovenous fistula in porcine carotid artery. RESULTS: When mixture was injected into the straight inflow portion (1), it formed a cast along the wall of the glass tube, with a large cast in the nidus portion. Further injection caused enlargement of the cast in the nidus, but no change in the wall of the inflow portion. When mixture was injected into the nidus portion (2) of the glass tube, a cast formed continuously within the nidus. In the animal experiment, rapid injection of glue at a high conentration staunched flow through the arteriovenous fistula. CONCLUSION: Glue-Lipiodol mixture can be injected slowly and continuously under the surface of the polymerized glue cast which forms along the wall or slow flow portion (nidus) of the glass tube in an experimental flow phantom. However, the effect did not appear during rapid flow in the animal arteriovenous fistular model in which only rapid injection of glue which was either pure or at a glue concentration of more than 75% was effective.