Using PET with [18-F]-2-deoxyglucose (FDG), we studied the effects of antiepileptic drugs on cerebral glucose metabolism. Serial scans were performed before and after the test drug was added to or removed from the patient's regimen. At least 3 weeks elapsed after achieving steady-state plasma levels when drugs were added, or after plasma levels were undetectable when drugs were tapered, before repeat scans were obtained. Only a single drug was changed between scans. In the phenobarbital (PB) study, the "on-drug" scan was performed first in each case. In this instance, a mean of 14 weeks elapsed between the time blood levels were undetectable and repeat scanning in order to avoid the possibility of withdrawal effects. Scanning in each group was performed 30 min after injection of 5 mCi of FDG, with EEG monitoring to exclude ictal activity. Regional glucose metabolic rates were calculated in 8 to 20 regions of interest. PB reduced LCMRglu in seven of eight regions studied, with a mean reduction over all regions of 37 +/- 3%. Phenytoin (PHT) reduced LCMRglu in only two of 10 regions (mean = 13%). We studied the effect of PHT on cerebellar metabolism in 42 patients using unpaired scans. Cerebellar LCMRglu was lower when patients were taking PHT at the time of scan, as well as in those who were taking PHT for 5 years or more, but the differences were not significant. There was a weak inverse correlation between PHT serum level and cerebellar LCMRglu in patients taking the drug at the time of scan (r = -0.36; 0.05 less than p less than 0.1).(ABSTRACT TRUNCATED AT 250 WORDS)