The relationship between platelet GABA-transaminase (GABA-T) activity and either epilepsy or its treatment has been studied in 281 epileptic children: 55 were newly diagnosed untreated patients and 226 were chronically receiving anticonvulsants (154 in monotherapy and 72 in polytherapy). Results were compared with those from 48 control children. Untreated children had a GABA-T activity of 9.1 +/- 3.7 pmol/min/mg protein, lower than the control group (10.6 +/- 3.8 pmol/min/mg, P < 0.05), whereas treated epileptic children had higher values (11.9 +/- 6.3 pmol/min/mg) than those untreated (P < 0.01). In untreated children, the seven with absences and the nine with simple partial seizures had a GABA-T activity of 6.9 +/- 3.3 and 7.8 +/- 3.2 pmol/min/mg, respectively, lower than the control group (P < 0.05). In treated patients, those receiving valproate (VPA) in monotherapy had a GABA-T activity of 15.3 +/- 7.5 pmol/min/mg, higher than both the control group and the untreated children (P < 0.001). All patients receiving VPA in mono- or polytherapy had a higher activity than those receiving other anticonvulsants (16.4 +/- 8.4 vs. 9.9 +/- 3.9 pmol/min/mg, P < 0.001), the activity in Lennox syndrome and myoclonic epilepsies being significantly higher than in those with absences and partial epilepsy. GABA-T activity did not correlate with doses or trough steady-state serum levels of VPA. Platelet GABA-T could be useful as a peripheral marker of GABAergic alterations and GABAergic effects of antiepileptic drugs in epileptic patients.