Threein vitroclonogenic assays were used to determine the sensitivity of an established human glioblastoma cell line (U251-MG) to five chemotherapeutic agents. The colony-forming efficiency of untreated culture was 0.695±0.170 in a monolayer assay with irradiated feeder cells, 0.018±0.006 in a low-O2agar assay, and 0.049±0.021 in a two-layer agar system with nutrient-enriched medium (p<0.001). Comparison of the slope of the regression line for the dose-response curve and the interpolated ID90for each drug showed that U251-MG was equally sensitive to aziridinylbenzoquinone and dianhydrogalactitol in all three assays. The sensitivity of this cell line to 1,3-bis-(2-chloroethyl)-1-nitrosourea (BCNU), cis-dichlorodiammineplatinum (II) (CDDP) and 9-hydroxy-2-N-methylellipticine (HME), however, varied depending on the assay used. In no instance did U251 -MG show greater sensitivity (lower ID9o or steeper slope) in the loW-O2agar assay than in the other assays. BCNU and CDDP were least active in the monolayer assay, whereas HME showed both the lowest ID90and steepest slope using this technique. We conclude that differentin vitrotumour clonogenic assays show different colony-forming efficiencies for the same cell line and may show different responses to certain drugs. Identification of accurate predictive models of drug sensitivity will require correlativein vivoandin vitrostudies.