Prediction of nidal penetration of the glue is difficult in the embolisation of brain arteriovenous malformations (AVM). We evaluated penetration differences of glue-lipiodol mixtures with and without tungsten powder for embolisation of brain AVM. Fifty eight injections during embolisation of brain AVM in 15 patients were evaluated. Glue contained lipiodol only (N=25), or lipiodol and tungsten powder (N=33). Glue concentration ranged from 20 to 80%. Injection was done at the pedicle (N=41) or nidus (N=17). Degree of penetration was rated as good, fair, or poor according to penetrated glue amount in the nidus. Penetration difference was compared in each injection at the pedicle or nidus. When the glue mixtures were injected at the pedicle, the glue-lipiodol mixture (GLM) penetrated into the nidus in eight of 16 injections (50%) and the glue-lipiodol-tungsten mixtures (GLTM) in 23 of 25 (92%). When glue was injected at the nidus, GLM was confined to the nidus in nine of nine injections (100%) and GLTM in seven of eight (88%). Penetration into the vein was noted in two of 25 injections (8%) in GLM and six of 33 (18%) in GLTM. Penetration difference into the nidus between GLM and GLTM was statistically significant (p < 0.05) when the glue mixtures were injected at the pedicle. GLM and GLTM can be used effectively at the nidus. In contrast to GLM, GLTM can be used effectively at the pedicle. However, the risk of venous penetration seemed to be higher with GLTM than in GLM.