Summary:Purpose: Vigabatrin (VGB) is a new antiepileptic drug that increases the human brain ␥-aminobutyric acid (GABA) level by irreversibly inhibiting GABA transaminase. Although some patients respond to VGB with a significant seizure reduction, others do not. The aim of this study was to identify possible responders before or in an early phase of VGB treatment by measuring the GABA and homocarnosine contaminated with macromolecules/creatine and phosphocreatine ratio (GABA+/Cr) signal by means of proton-nuclear magnetic resonance ( 1 H NMR) spectroscopy. Methods: Measurements were performed immediately before and after a titration period of 1 month (2 g/day during the past 2 weeks). A third measurement followed a maintenance period of 3 months (2 or 3 g/day). In 14 patients with drugresistant temporal lobe epilepsy and 3 patients with occipital lobe epilepsy, GABA+/Cr was measured in the ipsilateral (i.e., epileptogenic) hemisphere and contralateral (i.e., nonepileptogenic) hemisphere in a volume of 8 cm 3 .Results: Depending on the therapeutic efficacy of VGB, we defined three groups: (a) full responders (n ס 7), (b) nonresponders (n ס 7), and (c) partial responders (n ס 3). The nonresponders had no significant change in the GABA+/Cr signal during the treatment compared with baseline. The full responders had a significant increase of the GABA+/Cr signal during the whole treatment phase and a lower ipsilateral level at baseline. The partial responders had also a lowered ipsilateral GABA+/Cr signal at baseline and an increase during treatment but a decrease when the seizures started again.Conclusions: Responders to VGB could be identified by a lower ipsilateral baseline GABA+/Cr signal and a steeper increase during VGB treatment. However, it was not possible to predict the duration of the response (full versus partial responder) with these criteria.