The aim of this study was to evaluate the usefulness of measuring the standardized uptake value (SUV) in primary brain tumors on fluorine-18 fluorodeoxyglucose (FDG) positron emission tomography (PET) scans. Two groups of patients were studied. Whole-brain glucose cerebral metabolic rates (wCMRs) and SUVs were obtained in 20 normal subjects. Twenty-seven patients with histology-proven malignant primary CNS tumors (high-grade gliomas n=22, primitive neuroectodermal tumors n=3, ependymomas n=2) were also studied. The degree of FDG uptake was assessed by visual inspection and thereafter regions of interest were placed over the lesion, the contralateral cortex and white matter and the whole brain. Average (avg) and maximum (max) pixel values were determined in each site. Based on these measurements, SUV, tumor to cortex (T/C) and tumor to white matter (T/WM) activity ratios were calculated. There was no correlation between wCMRs (4.55+/-0.36 mg min(-1) 100 g(-1)) and wSUVs (5.41+/-0.43) in the normal subjects (r=0.18, P=0.45). In the second group, 17 lesions were described as definitely and seven as probably malignant. However, SUVs in these tumors and in the contralateral cortex were not significantly different. Although the SUVs were generally higher in the tumor than in the contralateral white matter, there was a significant overlap between the values. The range of the SUVs was wide: 2.54-11.8 for the tumors, 2.98-9.96 for the cortex and 1.87-6.76 for the white matter. SUVs in the normal cortex were negatively correlated with blood glucose level at the time of the injection. SUVs in the whole brain and in the cortex were lower in patients previously treated by irradiation, even months after completion of the treatment. No correlation was detectable between any of the SUVs and the age of the patients, tumor type, time post injection, use of dexamethasone, patient weight, dose injected and visual score. With cutoff levels of 1.5 for T max/WM and 0.6 for T max/C, the sensitivity of the activity ratios was 74% and 96% respectively. In conclusion, SUVs do not correlate with CMRs across subjects and appear to be of limited value in characterizing brain tumors. Visual assessment and measurement of the activity ratios currently remain the most reliable methods of analysis.