Cancer spheroids are a valuable model for screening anticancer strategies. However, studies are published using various numbers of spheroids and sections per spheroid. Here, we establish the sample size requirements for valid screening strategies to treat glioma: how many spheroids per experimental group and how many sections per spheroid are required to detect one-third reduction in an endpoint measurement after treatment? From two glioblastoma patients, 32 untreated organotypic spheroids were cultured and sectioned entirely (14-100 sections per spheroid). The viable fraction was determined as endpoint by automated image analysis in sections and used to establish the minimally-detectable difference between a treatment and reference group. Variance was considerable with a coefficient of variance of 21%. The biological variation in viability in sections of spheroids produced 97% of variance when sample size was large. Variance increased when numbers of spheroids but not numbers of sections per spheroid were reduced. A minimum of 12 spheroids per group and one section per spheroid was required for a valid comparison of a treatment group and a control group. When 10 treatment groups and one control group were compared, 16 spheroids per group were required. Thus, the statistical power depended almost entirely on the number of organotypic glioma spheroids and hardly on the number of sections per spheroid. The organotypic glioma spheroid model does not appear to be suitable for high-throughput screening of anticancer strategies, because of the relatively large number of spheroids required. It is the model of choice for low-throughput screening, because this model is far more representative for the parental tumor than any other more efficient glioma model.