RPE patients were recruited from those attending our epilepsy clinics. Enrollment was limited to adult patients (aged 19-60 years) whose partial-onset seizures (simple or complex partial with or without secondary generalization, according to the International League Against Epilepsy classification) were poorly controlled by AEDs. Our definition of RPE was partial epilepsy that is uncontrolled by two or more AEDs and having a frequency of at least once per month over 6 months before entry into the study. We included only right-handed persons to exclude the effects of handedness on the cerebral current-source distribution. Patients with progressive neurological disorders, head injury, mental retardation (IQ<70), alcohol or drug abuse, ongoing use of any centrally acting medications, severe psychiatric problems, or other severe medical disorders were excluded. We acquired demographic and epilepsy-related characteristics through an epilepsy questionnaire and the patients' diaries.

A cohort of 31 RPE patients was enrolled in a 24-week, open-label, prospective study. Before prescribing LEV, all patients gave their written informed consent to participate in the study. We added LEV to each patient's current AED regimen. The study included a baseline phase of 12 weeks and then a 4-week interval of titration of LEV (250 mg twice daily in the first 2 weeks, 500 mg twice daily in the 3rd and 4th weeks); thereafter, the dose regimens were adjusted individually based on the investigator's clinical judgment according to the patient's clinical response and tolerability, to obtain the best seizure control and tolerability. The LEV dosage was titrated up to 3,000 mg/day if the attacks had not subsided. Seizure frequency and scalp EEG were recorded, and neuropsychological tests (NPTs) were conducted before the LEV trial and after 24 weeks of therapy. Mood tests were conducted to clarify the effect of mood on cognitive performance. EEG recordings and NPTs were conducted on the same day.

We observed LEV-induced CSD changes and determined whether there was any association between these changes and the presence of freedom from seizure or dosage of LEV. For the comparisons, we subdivided patients into the following two groups: 1) patients with seizure freedom versus those who were not seizure-free, and 2) patients taking at least 2,000 mg/day LEV versus those taking less than 2,000 mg/day LEV. We also compared the results of NPTs conducted at 24 weeks with those conducted at baseline.

EEGs were recorded for at least 30 min in each patient using a 32-channel digital EEG machine (Telefactor Aurora® EEG machine, Grass-Telefactor, Melbourne, Australia). All EEGs were recorded at 200 Hz using 19 channels based on the international 10-20 system (Fp1, Fp2, F3, F4, C3, C4, P3, P4, O1, O2, F7, F8, T3, T4, T5, T6, Fz, Cz, and Pz).

After recomputing to the average referential montage, five 3-s, artifact-free segments were selected from the recording of each EEG (filtered from 1.6 to 70 Hz) from segments of the waking state with no epileptiform discharge. We used the data review and processing module in the Brain Electrical Source Analysis software (v. 5.1, MEGIS, Grafelfing, Germany) to obtain the segments. Frequency-domain analysis in the 22-30 Hz frequency band was applied to the selected 3-s artifact-free segments (600 sample points). LORETA-KEY (KEY Institute for Brain-Mind Research, Switzerland) was used to calculate the intracerebral current density in the frequency-domain analysis.

Our version of LORETA used a three-shell spherical head model, registered to the Talairach human brain atlas, and the data available as digitized magnetic resonance images from the Brain Imaging Center of the Montreal Neurologic Institute.23 The registration between the spherical and Talairach head geometries used the realistic EEG electrode coordinates reported by Towle et al.24

The LORETA solution space was restricted to the cortical gray matter and hippocampus in the Talairach atlas, as defined by the corresponding digitized probability atlas. A total of 2,394 voxels was produced at 7-mm spatial resolution under this neuroanatomical constraint.22,25

All NPTs were conducted in a sound-attenuated, temperature-controlled room by a single examiner. According to the literature and our own clinical experience, several cognitive measures were selected as being particularly sensitive in epilepsy patients. The following parameters were assessed:

In addition, we evaluated mood states using the Beck Depression Inventory30 and Beck Anxiety Inventory31 to examine the effect of mood on cognitive performance. Testing sessions lasted about 40 min. If a seizure occurred during a NPT, that test was suspended and the data were not evaluated.

Five segments of 3-s epochs were selected from a test EEG and a retest EEG from each patient. Thus, a total of 120 EEG data pairs (24 patients×5 epochs×2 conditions) were obtained. Paired sample t-tests were computed for the log-transformed LORETA power at each voxel in the 22-30 Hz frequency band to evaluate differences between the two conditions. In all analyses, the level of statistical significance was set at p<0.05. These voxel-by-voxel t values were displayed as statistical nonparametric maps (SnPMs).

For the NPTs, higher scores indicated better performance in all cognitive tests, except the TMTA and TMTB, for which higher scores indicated worse performance because the time required was defined as the dependent measure for these tests. Data for continuous variables are expressed as mean±SD values, and data for categorized variables are expressed as frequencies (percentages). The paired t-test was used to evaluate the test-retest changes in neuropsychological scores induced by LEV adjunctive therapy. The statistical analyses were conducted with SPSS 12.0 (SPSS, Chicago, IL, USA), and results were deemed to be statistically significant at p<0.05.

Four of the 31 enrolled patients dropped out during the titration period due to central-nervous-system-related side effects or rash. One patient stopped using LEV due to aggravation of seizures. Two patients were excluded due to unsuitable EEG data. A total of 24 patients completed follow-up EEGs and NPTs. Their demographic and clinical characteristics are summarized in Table 1. After 24 weeks of treatment, the frequency of seizures was significantly decreased (p<0.05). The response rate was 54%, and seizure freedom occurred in 33% of the patients at a mean LEV dosage of 2,260 mg/day. The incidence of adverse events in the last 4 weeks was 29%. The most common adverse events were somnolence and asthenia. The findings of brain magnetic resonance imaging (MRI) were normal in 17 of 24 patients, with the remaining 7 patients showing the following abnormal MRI findings: right frontal schizencephaly, left frontal traumatic contusion, left hippocampal sclerosis, left hippocampal sclerosis with temporal arachnoid cyst, both temporal encephalitis, left temporoparietal arachnoid cyst, and left occipital hypoxia.

The CSD in the 22-30 Hz frequency band was increased in the bilateral anterior cingulate gyri, left parahippocampal gyrus, and a small area of the right anterior parahippocampal gyrus after the LEV trial (p<0.05)(Fig. 1). The CSD in a small area of the bilateral superior parietal gyri was higher in patients who were free of seizures (n=8) than in those who were not seizure-free (n=16, p<0.05)(Fig. 2A). The CSD in a small area of the bilateral precunei was lower in patients taking at least 2,000 mg/day LEV (n=15) than in those taking less than 2,000 mg/day LEV (n=9, p<0.05)(Fig. 2B).

The result of NPTs and mood tests before and after the LEV trial are summarized in Table 2. Performances in verbal memory tests, such as LL, IWR, and DWR, were significantly improved after the 24-week LEV trial (p=0.007, 0.008, and 0.023, respectively). Executive function, such as TMTA, was also improved at the follow-up test (p=0.008). There was no deterioration on any of the tests. Mood states were not significantly changed at 24 weeks.