Boron-10 (10B) concentrations were measured in 107 surgical samples from 15 patients with glioblastoma multiforme who were infused with 95 atom% 10B-enriched p-boronophenylalanine (BPA) intravenously for 2 h just prior to surgery at doses ranging from 98 to 290 mg BPA/kg body weight. The blood 10B concentration reached a maximum at the end of the infusion (ranging from 9.3 to 26.0 microg 10B/g) and was proportional to the amount of BPA infused. The boron concentrations in excised tumor samples ranged from 2.7 to 41.3 microg 10B/g over the range of administered BPA doses and varied considerably among multiple samples from individual patients and among patients at the same BPA dose. A morphometric index of the density of viable-appearing tumor cells in histological sections obtained from samples adjacent to, and macroscopically similar to, the tumor samples used for boron analysis correlated linearly with the boron concentrations. From that correlation it is estimated that 10B concentrations in glioblastoma tumor cells were over four times greater than concurrent blood 10B concentrations. Thus, in the dose range of 98 to 290 mg BPA/kg, the accumulation of boron in tumor cells is a linear function of BPA dose and the variations observed in boron concentrations of tumor specimens obtained surgically are largely due to differences in the proportion of nontumor tissue (i.e. necrotic tissue, normal brain) present in the samples submitted for boron analysis. The tumor:blood 10B concentration ratio derived from this analysis provides a rationale for estimating the fraction of the radiation dose to viable tumor cells resulting from the boron neutron capture reaction based on measured boron concentrations in the blood at the time of BNCT without the need for analysis of tumor samples from individual patients.