Clinical Neurophysiology 123 (2012) 883-891
Contents lists available at SciVerse ScienceDirect
Clinical Neurophysiology
j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / c l i n p h
Effect of levetiracetam monotherapy on background EEG activity and cognition
in drug-naive epilepsy patients
Jounhong Ryan Cho a,b,1, Dae Lim Koo a,1, Eun Yeon Joo a, So Mi Yoon a, Eunji Ju c,
James Lee d, Dae Young Kim a, Seung Bong Hong a,b,
a Department of Neurology, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Republic of Korea
b Samsung Biomedical Research Institute, Seoul, Republic of Korea
c Neuroscience and Behavioral Biology Program, Emory University, Atlanta, GA, USA
d Department of Biochemistry, University of Washington, Seattle, WA, USA
a r t i c l e
i n f o
h i g h l i g h t s
Article history:
We evaluated the effect of LEV on background EEG and cognition in epilepsy patients.
Accepted 10 September 2011
LEV decreases slow frequency power and increases fast frequency power.
Available online 13 October 2011
LEV causes improvement in diverse neuropsychological tests.
Some EEG changes across LEV therapy were correlated with change in cognitive tests.
Keywords:
Antiepileptic drugs
Levetiracetam
a b s t r a c t
EEG
Quantitative analysis
Objective: To investigate the cognitive effect of levetiracetam (LEV) monotherapy with quantitative elec-
Cognitive function
troencephalogram (EEG) analysis and neuropsychological (NP) tests.
Methods: Twenty-two drug-naive epilepsy patients were enrolled. EEG recordings were performed
before and after LEV therapy. Relative power of discrete frequency bands was computed, as well as alpha
peak frequency (APF) at occipital electrodes. Eighteen patients performed a battery of NP tests twice
across LEV treatment.
Results: LEV therapy decreased the power of delta (1-3 Hz, p < 0.01) and theta (3-7 Hz, p < 0.05) bands
and increased that of alpha-2 (10-13 Hz, p < 0.05) and beta-2 (19-24 Hz, p < 0.05) bands. Region-specific
spectral change was observed: delta power change was significant in fronto-polar region, theta in ante-
rior region, alpha-2 in broad region, and beta-2 in left fronto-central region. APF change was not signif-
icant. Improvement in diverse NP tests requiring attention, working memory, language and executive
function was observed. Change in theta, alpha-2, and beta-2 power was correlated with improvement
in several NP tests.
Conclusions: Our data suggest LEV is associated with acceleration of background EEG frequencies and
improved cognitive function. Change in frequency band power could predict improvement in several cog-
nitive domains across LEV therapy.
Significance: Combined study of quantitative EEG analysis and NP tests can be useful in identifying cog-
nitive effect of antiepileptic drugs.
O 2011 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights
reserved.
1. Introduction
and Kimford, 2004). Neuropsychological testing has been the
preferred method of examining the impact of AEDs on cognitive
Antiepileptic drugs (AEDs) remain the primary treatment
functions, but suffers from several limitations, such as test-retest
option in epilepsy. Although they can be effective for seizure
variability and practice effect (Clemens et al., 2006). Inclusion of
control, AEDs can also cause neurotoxicity, adversely affecting
quantitative EEG measures can provide additional insight in
normal functions of the central nervous system (CNS) (Ortinski
investigating the impact of pharmacological intervention on
cognitive function (Gevins et al., 2002; Smith et al., 2006). Quanti-
Corresponding author at: Department of Neurology, Samsung Medical Center,
tative analysis of background EEG frequencies can be a simple and
Sungkyunkwan University School of Medicine, 50 Irwon-Dong, Gangnam-Gu, Seoul
objective method of evaluating the effect of AEDs and disclosing its
135-710, Republic of Korea. Tel.: +82 2 3410 3592; fax: +82 2 3410 0052.
implication on cognitive function, since background EEG activity
E-mail addresses: sbhong@skku.edu, sbhong2010@hotmail.com (S.B. Hong).
1 Two co-first authors contributed equally to this work.
represents the functional state of the brain.
1388-2457/$36.00 O 2011 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights reserved.
doi:10.1016/j.clinph.2011.09.012

---

884
J.R. Cho et al. / Clinical Neurophysiology 123 (2012) 883-891
Levetiracetam (LEV) is one of the ``new'' AEDs that have demon-
Patients were laid in a bed during EEG recording, and an EEG
strated its effectiveness in the treatment of symptomatic and idio-
technician constantly monitored their state of alertness. Patients
pathic epilepsy (Rocamora et al., 2006; Gambardella et al., 2008).
were asked to close their eyes for 10 s and then open for another
LEV appears to be as effective as the older AEDs in seizure control,
10 s. This procedure was repeated 10 times.
but is more tolerable and has more favorable neuropsychological
profile (Berkovic et al., 2007). Neuropsychological examinations
2.3. EEG data processing and analysis
on patients with LEV therapy consistently show no negative effect,
and even improvement in some studies (Piazzini et al., 2006;
For quantitative analysis of background EEG frequencies, one of
Gomer et al., 2007; Meador et al., 2007; Helmstaedter and Witt,
epilepsy specialists (Koo), who was completely blind to the pa-
2008; Zhou et al., 2008). However, the effect of LEV on background
tients' information, reviewed all EEG recordings and selected
EEG activity and neurophysiological relevance of these reported
epochs. Epoch selection criteria included: (1) the presence of
favorable neuropsychological findings are largely unexplored.
physiological alpha activity with maximum amplitude in posterior
The primary objective of this study is to analyze the effect of
regions while eyes closed, (2) the absence of physiological or non-
LEV monotherapy on background EEG activity and to relate quan-
physiological artifacts, (3) the absence of epileptiform discharges
titative EEG findings with neuropsychological outcomes. We have
(i.e. spikes) or other paroxysmal activities, and (4) the absence of
quantified the relative power of discrete frequency bands before
electrophysiologic patterns indicating drowsiness or arousal. For
and after the LEV treatment, to observe whether LEV monotherapy
each patient and condition, 20 of 2 s-long epochs were collected
causes background EEG change globally and/or locally at a certain
and exported to MATLAB (Version 7.6.0, The Mathworks Inc.,
brain region. Alpha peak frequency in occipital electrodes was also
Natick, MA, USA) for further analysis.
calculated. In parallel, standard neuropsychological tests have been
Overview of spectral analysis is given in Supplementary Fig. S1.
administered to evaluate the effect of LEV on cognitive function.
Selected epochs were first multiplied by Hanning window (MAT-
We hypothesized that the cognitive effect of LEV monotherapy
LAB function: hann) to reduce leakage, and then submitted to dis-
on epilepsy patients can be better explored by combined neuro-
crete Fourier transformation (DFT). Hann-windowed epochs were
physiological and neuropsychological measures.
discrete-Fourier-transformed using second-order Goertzel algo-
rithm (MATLAB signal processing toolbox function: goertzel or
2. Methods
spectrogram). The Goertzel algorithm is a computationally efficient
digital signal processing technique for identifying frequency com-
2.1. Patients
ponents of a given signal (Mitra, 1998). While the general fast Fou-
rier transform algorithm computes evenly across the bandwidth of
Twenty-two consecutive drug-naive patients were recruited
the input signal, the Goertzel algorithm can calculate spectral com-
from an outpatient epilepsy clinic at Samsung Medical Center
ponents at specific, predetermined frequencies. In this study, a
(SMC), between August 2007 and January 2010. Patients with new-
range from 1 to 30 Hz with 0.1 Hz step was used. Transformed sig-
ly or recently diagnosed epilepsy were included, who had least two
nals were squared to calculate power spectrum, and 20 of these
unprovoked seizures (>48 h apart) during the last year at the time
were averaged for each electrode. The relative power of seven dis-
of study enrollment. The study was approved by the internal re-
crete bands (delta: 1-3 Hz, theta: 3-7 Hz, alpha 1: 7-10 Hz, alpha
view committee in SMC (IRB file No. 2007-08-063-001) and all pa-
2: 10-13 Hz, beta 1: 13-19 Hz, beta 2: 19-24 Hz, beta 3: 24-30 Hz)
tients gave their written informed consent to participate in this
was computed for 19 monopolar derivations. Relative power of a
study before prescription.
certain frequency band is defined as the ratio of total power within
Each patient received LEV monotherapy for 8-22 months. Be-
a given band to total power of entire frequency range, thus ranging
fore and after the LEV treatment, all patients underwent neurolog-
from 0 to 1. In addition to relative power at each electrode, average
ical examination, routine laboratory tests, and EEG recordings.
of the 19 derivations was calculated to observe ``global'' effect of
Eighteen of them performed full battery of neuropsychological
LEV on different frequency ranges in background EEG.
examinations. LEV was initially given at 500-1000 mg per day
Furthermore, the peak frequency within alpha band (alpha
for first 2 weeks. If seizures remained during the initial evaluation
1 + alpha 2) was computed for occipital electrodes (O1 and O2)
period, LEV was titrated up to a maximum 1500 mg per day. Dose
where posterior alpha activity was maximal. To calculate alpha
level was adjusted individually based on clinical decision to obtain
peak frequency (APF), each of Hann-windowed epochs was band-
appropriate balance between seizure control and tolerability.
passed from 7 to 13 Hz using 3rd-order Butterworth filter
Seizure frequency was checked with a self-recorded seizure diary
(MATLAB filter design toolbox function: butter) in both forward
when the patient visited the outpatient clinic.
and reverse directions to ensure zero phase distortion (MATLAB
filter design toolbox function: filtfilt). And the power spectrum of
band-passed signal was computed as described above, with fre-
2.2. EEG recording
quency range from 7 to 13 Hz with 0.01 Hz step. The frequency
corresponding to maximum power was found, and 20 of these
EEG recordings were performed twice during the study period:
were averaged to obtain mean APF for each patient and condition.
``baseline'' EEG (before LEV treatment) and ``follow-up'' EEG (after
LEV treatment). EEGs were recorded in a silent room at forenoon
hours with a same type of digital EEG machine (Vanguard Instru-
2.4. Neuropsychological assessments
ments Imaging Products, Cleveland, OH, USA). Ag-AgCl electrodes
were placed according to the International 10-20 system plus
Eighteen patients were administered a full battery of neuropsy-
extra electrodes in bilateral ears. Four additional electrodes were
chological (NP) tests. All NP tests were conducted in a silent, tem-
used to monitor eye movement. Impedance was kept below
perature-controlled room by a single certified neuropsychologist
5 kX. EEG was initially recorded against Pz electrode and later
(Yoon) at forenoon hours. Battery items were selected to monitor
re-referenced to linked ear channels (average of A1 and A2
various domains of cognitive function, including memory, atten-
channels) for analysis. EEG signals were high-passed at 0.5 Hz
tion, language, executive function and general intelligence. All of
and low-passed at 70 Hz by a hardware-imposed filter, digitized
NP tests described below are widely used to assess cognitive func-
on a 12-bit A-to-D converter with sampling rate of 200 Hz.
tion in adult epilepsy patients.

---

J.R. Cho et al. / Clinical Neurophysiology 123 (2012) 883-891
885
To evaluate verbal memory, the Korean California verbal learn-
distribution using Kolmogorov-Smirnov test (MATLAB statistical
ing test (K-CVLT) was administered (Kim and Kang, 1997). K-CVLT
toolbox function: kstest). Since all of null hypotheses were rejected
involves five learning steps of 16 words. After learning, patients
at the 5% significance level, we used non-parametric Wilcoxon
were asked to perform immediate recall, immediate hint recall,
sign-rank test (MATLAB statistical toolbox function: signrank) to
and delayed free recall, delayed hint recall, and recognition of
find any statistical significance between two conditions. For all
memorized words. Visual memory was assessed by the Rey com-
measures, a p-value of <0.05 was considered significant.
plex figure test (RCFT), which is composed of an immediate recall,
To examine correlation between changes in quantitative EEG
a delayed recall, and recognition after copying a complex figure
measures (average of 19 electrodes) and performance change in
(Meyers and Meyers, 1995). Delays for both tests were 20 min.
NP tests, Spearman's rank correlation coefficient (MATLAB statisti-
The digit span tests were administered, both forwards and
cal toolbox function: corr with spearman option), denoted by q,
backwards, to assess verbal attention and working memory
was calculated between two given variables, one from quantitative
(Wechlser, 1987). The Corsi Block tapping tests, both forwards
EEG measures (for example, change in relative power of theta
and backwards, were used to examine visual attention and work-
band) and another from NP tests (for example, change in K-CVLT
ing memory (Corsi, 1972). In addition, the trail making tests
test). To test the statistical significance of calculated q value, we
(TMT-A and B) and the digit symbol tests were administered to
conducted a permutation test with re-sampling, which accounts
monitor visual attention and working memory (Reitan, 1992).
for multiple comparisons problem, under the null hypothesis of
To examine executive function of frontal lobe, we administered
no significant correlation between two variables. We randomly
the Wisconsin card sorting test (WCST, assessing conceptual infor-
shuffled the order of quantitative EEG measures and calculated q
mation and cognitive set-shifting ability according to external de-
value with NP scores in order to destroy any possible inherent cor-
mands) (Heaton et al., 1993), the Stroop test (assessing inhibitory
relation. We repeated the re-sampling process 104 times to esti-
ability) (Lee et al., 2000), the controlled oral word association test
mate the sampling distribution under the condition that the null
(COWAT, assessing phonemic and semantic verbal fluency ability)
hypothesis is correct. We subsequently asked what percentage in
(Kang and Na, 2003), and Raven's colored progressive matrices
the permutation distribution the observed q value corresponds
(RCPM, assessing logical and visual reasoning abilities) (Raven
to. Two-sided test with 1% significance level suggests that the ob-
et al., 2003). To evaluate language function, the Korean Boston
served q value should be larger than 99.5 percentile and smaller
naming test (K-BNT) were used, assessing confrontational naming
than 0.5 percentile to reject the null hypothesis. Empirical p-value
ability (Kim and Na, 1997). To assess general intelligence, Korean
was also calculated from permutation distribution.
Wechsler intelligence scale (KWIS) was administered (Jun et al.,
1963).
3. Results
The higher the scores of tests, the better the neuropsychological
performances in all tests, except for the RCFT copy, TMT tests, per-
3.1. Patients
severative responses and errors in WCST, where lower scores mean
better results.
Patient information and clinical information are summarized in
Table 1. Out of 22 patients enrolled in this study, 13 were female.
2.5. Statistics and correlation analysis
The mean age of the patients was 32.2 13.0 years (range 16-
58 years). The mean duration of the evaluation period (i.e. the
All of quantitative EEG parameters and scores from neuropsy-
interval between baseline and follow-up EEG recordings and neu-
chological examinations were compared between ``baseline'' and
ropsychological tests) was 12.4 3.6 months (range 8-22 months).
``follow-up'' conditions. We first tested the null hypothesis that
The mean seizure frequency before first prescription was 3.4 8.8
quantitative EEG and NP parameters follow standard normal
per month. The mean dose level of LEV at the end of evaluation
Table 1
Clinical characteristic of the patients, N = 22.
Pt No.
Sex
Age (years)
Follow-up interval
Epilepsy type
Dosage of
Seizure frequency
Seizure free
MRI findings
(month)
LEV (mg/day)
before LEV (n/month)
after LEV?
1
M
45
20
IGE
1500
1
Y
Normal
2
M
22
8
PE
1000
0.1
Y
Normal
3
M
52
10
PE
1000
0.05
N (0.01/mo)
Normal
4
F
25
11
IGE
1500
30
Y
Normal
5
F
44
12
PE
1000
0.01
Y
Normal
6
F
20
11
IGE
1000
7
Y
Normal
7
F
20
9
PE
1500
0.5
Y
Normal
8
M
16
12
IGE
1000
30
N (1/mo)
Normal
9
F
40
11
PE
500
0.1
Y
Normal
10
F
31
12
PE
1000
0.1
Y
Normal
11
F
29
11
PE
500
0.1
Y
Pineal Cyst
12
F
19
18
IGE
1000
0.5
Y
Normal
13
F
29
12
PE
625
0.08
Y
NS CA
14
F
47
22
PE
1000
0.0005
Y
B HS
15
F
49
11
PE
1000
1
Y
NS CA
16
F
45
12
PE
1000
3
Y
Normal
17
M
20
17
PE
1000
0.1
Y
Normal
18
M
58
11
IGE
1000
0.001
Y
Normal
19
M
33
10
PE
1000
0.0001
Y
Normal
20
M
20
12
PE
1000
0.1
Y
HS, B
21
F
27
9
PE
1000
0.1
Y
Normal
22
M
19
12
PE
1000
0.01
Y
Normal
Abbreviations: Pt, Patient; IGE, Idiopathic Generalized Epilepsy; PE, Partial Epilepsy; NS CA, Non-specific Cortical Atrophy; B HS, Bilateral Hippocampal Sclerosis.

---

886
J.R. Cho et al. / Clinical Neurophysiology 123 (2012) 883-891
Table 2
Results for the quantitative EEG parameters (averaged from 19 electrodes) in fixed frequency bands.
Parameters
Baseline
Follow-up
Significance
Mean
Median
S.D.
Mean
Median
S.D.
Delta
0.1165
0.1098
0.0530
0.0913
0.0920
0.0291
p = 0.0099
Theta
0.1837
0.1528
0.0858
0.1419
0.1331
0.0499
p = 0.0168
Alpha-1
0.3378
0.2705
0.1870
0.3348
0.2817
0.1935
N.S.
Alpha-2
0.2170
0.1824
0.1195
0.2788
0.2701
0.1497
p = 0.0148
Beta-1
0.0993
0.0905
0.0478
0.1024
0.0888
0.0408
N.S.
Beta-2
0.0299
0.0277
0.0206
0.0382
0.0376
0.0297
p = 0.0312
Beta-3
0.0218
0.0153
0.0240
0.0223
0.0170
0.0222
N.S.
APF (Hz)*
10.0109
10.0400
0.7198
10.1832
10.2500
0.7645
N.S.
Abbreviations: S.D., Standard deviation; APF, Alpha Peak Frequency; N.S., Not Significant.
* APF was calculated from O1 to O2 electrodes only.
decreased while that of high frequencies was slightly increased
after LEV treatment in frequency-dependent manner (Fig. 1). In
addition, the peak frequency within alpha band is slightly right-
shifted towards higher frequency.
Change in frequency band power across LEV treatment is illus-
trated with box-plot diagrams. Spectral analysis within pre-de-
fined frequency bands reveals band-specific alterations induced
by LEV treatment (Fig. 2). The relative power of low frequencies,
both delta (Fig. 2A) and theta (Fig. 2B) bands, was significantly de-
creased after LEV treatment (p = 0.0099 for delta and p = 0.0168 for
theta). On the other hand, the relative power of faster frequencies,
specifically at alpha-2 and beta-2 bands, was significantly in-
creased after LEV intake (p = 0.0148 for alpha-2 and p = 0.0312
for beta-2). Mean values of alpha-1 (Fig. 2C), beta-1 (Fig. 2E) and
beta-3 (Fig. 2G) power were slightly increased, but the difference
was not significant.
To identify which specific brain regions were influenced by LEV,
relative power of significantly changed frequency bands was com-
pared at each electrode location (Fig. 3). In delta band (Fig. 3A),
Fig. 1. Grande average of power spectra across all channels and all patients. Mean
mean and median values of relative power were decreased in all
power spectrum before LEV treatment (``baseline'' condition) is illustrated with
19 electrodes, but the difference was prominent at three prefrontal
solid line, and mean power spectrum after LEV treatment (``follow-up'' condition)
electrodes (Fp1, Fp2 and Fz) and a right central electrode (C4). In
with dashed line. Power is in decibel scale. Prominent peak at alpha frequency is
theta band (Fig. 3B), mean and median values of relative power
visible due to our epoch selection criteria.
were reduced in all 19 electrodes, similar to delta band. The result
was statistically significant in anterior regions, specifically at bilat-
eral frontal (Fp1, Fp1, F3, F4, F7, F8 and Fz), right temporal (T8), and
period was 1005.7 257.3 mg per day (range 500-1500 mg/day).
central (C3, C4 and Cz) electrodes. Power of alpha-2 band (Fig. 3C)
These amounts were effective in terms of seizure control and toler-
was significantly increased in broad regions, except T7, P8 and Pz
ability. All but two patients became seizure free: even these two
electrodes. Even in these three electrodes, mean and median power
patients experienced significant seizure reduction after LEV treat-
was increased after LEV intake. In beta-2 band (Fig. 3D), relative
ment. Six patients were classified as idiopathic generalized epi-
power was increased in all 19 derivatives, but significantly in left
lepsy (IGE) and sixteen as partial epilepsy (PE). No significant
fronto-central (Fp1, F3, F7 and C3) and parietal (P7) electrodes.
difference was observed in clinical features, including age, gender,
Result for alpha peak frequency is shown in Table 2. There was
duration of epilepsy, dose of LEV between IGE and PE groups. MRI
no significant change in alpha peak frequency after LEV mono-
findings are summarized in Table 1: bilateral hippocampal sclero-
therapy (p = 0.4074). Mean and median values of alpha peak fre-
sis was found in two patients, non-specific cortical atrophy in two
quency were slightly increased at follow-up stage, but did not
patients, and cystic lesion in pineal area was found in one patient.
reach statistical significance.
The remaining 17 patients showed no abnormalities in MRI scans.
At final evaluation, no patient showed neurological symptoms
3.3. Neuropsychological tests
indicating overt neurotoxicity induced by LEV.
Statistical results for NP tests are summarized in Table 3. The ta-
3.2. Quantitative EEG analysis
ble includes baseline and follow-up values for each NP test admin-
istered, and two-condition comparisons with Wilcoxon rank sum
Statistical results for quantitative EEG parameters (across all
test. Performance in several verbal and visual memory tests was
patients and all channels) are presented in Table 2. Prolonged
significantly improved after prolonged LEV monotherapy (p =
LEV monotherapy induced change in power of background fre-
0.0112 for K-CVLT total, p = 0.0321 for K-CVLT long delay free re-
quencies (Fig. 1). Prominent peaks at alpha band in both conditions
call, p = 0.0084 for RCFT copy, p = 0.0274 for RCFT immediate recall,
are present due to our epoch selection criteria (data segments with
and p = 0.0073 for RCFT delayed recall). Improvement in attention
posterior alpha activity). Visual comparison between averaged
and executive function were also observed in some relevant tests
power spectrums shows that power of low frequencies was
(p = 0.0196 for TMT-A, p = 0.0310 for TMT-B, p = 0.0060 for Digit

---

J.R. Cho et al. / Clinical Neurophysiology 123 (2012) 883-891
887
Fig. 2. Graphical presentation (box-plot diagrams) of change in relative power within each frequency band. The lower and upper borders of rectangular box correspond to
25% and 75% percentile of the data with median indicated by the red line. The whiskers extend over the entire range of the data. (A-G) Averaged from all patients and all
channels. Asterisks mean statistically significant results. The lower and upper borders of the box-plots, 25% and 75% percentile of the data, contain the data median indicated
by the red line. The whiskers extend over the entire range of the data. (A) Relative delta power was significantly reduced after LEV monotherapy (p < 0.01). (B) Relative theta
power was significantly reduced after LEV monotherapy (p < 0.05). (C) No statistical significant change was found in alpha-1 band. Almost no difference between conditions
was observed. (d) Relative alpha-2 power was significantly augmented after LEV monotherapy (p < 0.05). (E) Relative beta-1 band power was unchanged after LEV
monotherapy. (F) Relative beta-2 band power was significantly increased after LEV treatment. (G) No significant change was found in beta-3 band. (For interpretation of the
references to colour in this figure legend, the reader is referred to the web version of this article.)
symbol test, p = 0.0313 for WCST category, p = 0.0390 for WCST
4. Discussion
perseverative error, p = 0.0010 for COWAT phonemic word fluency
and p = 0.0362 for COWAT semantic word fluency). Language func-
To our knowledge, this is the first study to characterize changes
tion, as assessed by K-BNT, was also improved after LEV intake
in background EEG activity and NP tests after LEV monotherapy in
(p = 0.0205). No deterioration was observed in any of the tests
drug-naive epilepsy patients. Our study demonstrates that pro-
administered.
longed LEV treatment induced an acceleration of background EEG
frequencies: a decrease in the relative power of delta and theta
bands and an increase in that of fast frequency bands, specifically
3.4. Correlation between EEG and NP measures
at alpha-2 and beta-2 (Figs. 2 and 3). These effects were region-
specific: we found decreased slow frequency power mostly in ante-
Finally, we investigated the possible correlation between
rior regions, increased alpha-2 in broad regions, and increased
changes in background EEG activity and neuropsychological out-
beta-2 in left fronto-temporal region. There was no statistically sig-
comes to find out whether global power change in specific fre-
nificant change in APF. Neuropsychological tests revealed improve-
quency bands correlates with improvement in certain NP tests
ment in broad spectrum of cognitive function after LEV therapy.
using two-sided permutation test. We have run statistical tests
It is interesting that the pattern of spectral change following
on all possible combinations. Six pairs were found to have signifi-
LEV monotherapy is exactly opposite to the data from traditional
cant correlation between changes in quantitative EEG measures
AEDs, which negatively affects patients' cognition. Treatment with
and neuropsychological profile, as summarized in Table 4. Change
``classical'' AEDs such as phenytoin, carbamazepine (CBZ) or phe-
in relative theta power had negative correlation with digit symbol
nobarbitol increases the power of slow frequencies while reducing
test (q = A0.7528, p < 0.01, Fig. 4A) and K-BNT (q = A0.6123,
that of fast frequencies in background EEG (Duncan et al., 1989;
p < 0.01). Therefore, decrease in slow theta power was associated
Besser et al., 1992; Meador et al., 1993). Prolonged intake of car-
with improved performance in selective NP tests requiring visual
bamazepine and gabapentin in healthy volunteers induced EEG
attention, working memory and language function. Change in
slowing and decrease of the peak frequency of the posterior alpha
power of alpha-2 band also correlated with improvement in digit
rhythm, both associated with greater subjective neurotoxicity
symbol test (q = 0.6881, p < 0.01, Fig. 4B), COWAT phonemic word
(Salinsky et al., 2002). Topiramate (TPM) also causes EEG slowing
fluency (q = 0.6030, p < 0.01, Fig. 4C). Change in beta-2 power had
and alters cognitive event-related potential (Neufeld et al., 1999;
positive correlation with change in digit symbol test (q = 0.6249,
Mecarelli et al., 2001; Jung et al., 2010). Since EEG is sensitive to
p < 0.01) and K-BNT tests (q = 0.7089, p < 0.01, Fig. 4D).
disturbances in biochemical environment in the brain, any

---

888
J.R. Cho et al. / Clinical Neurophysiology 123 (2012) 883-891
Fig. 3. Topographical arrangement of box-plot diagrams according to electrode locations. The lower and upper borders of rectangular box correspond to 25% and 75%
percentile of the data with median indicated by the red line. The whiskers extend over the entire range of the data. Blue and red arrows indicate channels with significant
decrease and increase, respectively. (A) Relative delta (1.5-3.5 Hz) power was significantly reduced in three frontal electrodes (FP1, FP2 and Fz) and one right central
electrode (C4). (B) Relative theta (3.5-7.5 Hz) power was decreased in all of bilateral frontal and central electrodes (FP1, FP2, F3, F4, F7, F8, Fz, C3, C4 and Cz). Additional
significant change was observed in right temporal electrode (T8). (C) Significant increase in alpha-2 (10-12.5 Hz) power was observed in broad regions, except for three
electrodes (T7, P8 and Pz). (D) Significantly increased beta-2 (20-24 Hz) power was detected in left fronto-central (FP1, F3, F7 and C3) and a left parietal electrode (P7). (For
interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
dysfunctional change relevant to cerebral blood flow or cerebral
Other studies have conducted quantitative analysis on EEG to
oxygen/glucose uptake may be manifested in EEG as slowing of
investigate the cognitive effect of LEV. LEV add-on therapy was
background rhythms (Ingvar et al., 1976). Therefore, EEG slowing
not associated with EEG background slowing and decreased alpha
is an indicator of cerebral dysfunction and a common feature in
peak frequency in occipital electrodes (Veauthier et al., 2009). LEV
intoxication of CNS-active drug. Therapy with conventional AEDs
adjunctive therapy increased EEG background activity in the beta
is associated with significant EEG slowing, which may connect to
band (22-30 Hz) in the prefrontal cortex and left hippocampus
deterioration in cognitive measures and subjective neurotoxicity.
(Park and Kwon, 2009). Short-term LEV treatment did not induce
On the other hand, our results suggest that LEV monotherapy
any change in quantitative EEG parameters in healthy volunteers
causes acceleration of EEG background rhythms. Our finding is
(Mecarelli et al., 2004). These studies, however, do not solely re-
comparable to reports of lamotrigine therapy. Lamotrigine did
flect the effect of LEV on epilepsy patients, since the examined pa-
not cause significant change in relative power of delta, theta and
tients were already taking other AEDs that may have different
alpha bands, but increased the percentage of beta band (Marciani
effects on background EEG. Moreover, the duration of taking LEV
et al., 1998). Absence of EEG slowing and increase of fast EEG activ-
was relatively short, compared to the present study.
ity may explain a positive role of lamotrigine in attention and
LEV is associated with good tolerability and favorable NP profile
alertness, as supported by neuropsychological findings that it did
compared to other AEDs (Berkovic et al., 2007; Meador et al., 2007;
not induce cognitive dysfunction (Marciani et al., 1999; Joo et al.,
Gambardella et al., 2008; Helmstaedter and Witt, 2008). From
2006a,b; Seo et al., 2007). Our results are similar, but unique in that
basic level, amygdala-kindled rats consistently performed well in
we observed a significant decrease in relative power of slow fre-
Morris water maze along LEV therapy, unlike valproate-treated
quency bands in addition to an increase in fast frequency power.
epileptic rats showing impairment in spatial memory tasks

---

J.R. Cho et al. / Clinical Neurophysiology 123 (2012) 883-891
889
Table 3
Results for neuropsychological examinations, N = 18.
NP battery item
Baseline
Follow-up
Significance
Mean
Median
S.D.
Mean
Median
S.D.
Memory
K-CVLT total
51.39
46.50
10.12
56.28
57.00
10.60
p = 0.0112
K-CVLT short delay free recall
10.44
10.50
2.87
11.44
11.00
2.68
N.S.
K-CVLT long delay free recall
11.39
11.00
2.30
12.78
13.00
2.84
p = 0.0321
K-CVLT recognition
14.50
15.00
2.46
15.28
16.00
0.95
N.S.
RCFT copya
35.39
35.50
0.69
33.75
34.00
2.40
p = 0.0084
RCFT immediate recall
20.75
22.25
6.77
23.81
25.25
5.75
p = 0.0274
RCFT delayed recall
18.64
20.25
7.64
22.97
23.75
5.34
p = 0.0073
RCFT recognition
20.44
21.00
1.46
20.67
21.00
2.09
N.S.
Attention and executive function
Digit span - forward
8.89
10.00
1.78
9.56
10.00
1.76
N.S.
Digit span - backward
7.28
7.00
1.32
7.61
7.00
2.22
N.S.
Corsi block - forward
7.78
8.00
1.80
8.67
9.50
2.35
N.S.
Corsi block - backward
7.39
7.50
1.79
7.50
8.00
1.76
N.S.
TMT - Aa
33.44
33.50
8.28
28.38
26.50
9.29
p = 0.0196
TMT - Ba
73.39
72.50
25.28
61.33
58.50
17.64
p = 0.0310
Digit symbol test
67.17
69.00
13.04
73.28
73.00
8.05
p = 0.0060
WCST category
4.56
6.00
2.12
5.89
6.00
0.47
p = 0.0313
WCST correct response
66.11
66.50
15.24
69.28
66.00
8.84
N.S.
WCST perseverative responsea
14.89
14.00
11.54
12.33
11.00
7.28
N.S.
WCST perseverative errora
13.11
11.50
9.82
8.61
7.00
6.17
p = 0.0390
Stroop - word correct response
111.78
112.00
0.55
112.00
112.00
0
N.S.
Stroop - color correct response
107.28
112.00
11.41
108.78
112.00
6.42
N.S.
COWAT phonemic word fluency
32.44
31.00
12.83
42.78
43.00
9.65
p = 0.0010
COWAT semantic word fluency
33.24
30.50
10.25
38.77
37.50
9.48
p = 0.0362
RCPM
33.44
34.00
3.15
32.50
34.50
7.76
N.S.
Language
K-BNT
52.83
53.00
4.22
54.11
54.50
3.91
p = 0.0205
General intelligence
KWIS
113.61
113.50
18.09
114.56
115.00
16.14
N.S.
Abbreviations: K-CVLT, Korean California verbal learning test; RCFT, Rey complex figure test; TMT, Trail making test; WCST, Wisconsin card sorting test; COWAT, Controlled
oral word association test; RCPM, Raven's colored progressive matrices; K-BNT, Korean-Boston naming test; KWIS, Korean Wechsler intelligence scale; N.S., Not significant.
a Variables for which lower scores mean better performance.
Table 4
memory performance. From comparative EEG studies, large power
Combination of quantitative EEG measures and neuropsychological tests that showed
in the upper alpha band and small power in the theta band at
significant correlation.
resting state predict good cognitive performance, and alpha peak
Change in
Change in
Spearman's
p-
frequency is positively correlated to cognitive performance
quantitative EEG
neuropsychological
correlation
Value
(Klimesch, 1999; Angelakis et al., 2004). From these findings, our
measures
tests
coefficient (q)
results of increased upper alpha activity and decreased theta
Theta
Digit symbol test
A0.7528
0.0008
activity may explain positive NP findings. Functional interpretation
Theta
K-BNT
A0.6123
0.0058
of increased activity in beta-2 band is not as clear as theta and
Alpha-2
Digit symbol test
0.6881
0.0021
upper alpha, but beta rhythms are linked to attention and complex
Alpha-2
COWAT phonemic
0.6030
0.0078
cognitive processes (Ray and Cole, 1985; Marciani et al., 1998).
word fluency
Beta-2
Digit symbol test
0.6249
0.0064
Therefore, increased activity of fast frequencies in background
Beta-2
K-BNT
0.7089
0.0012
EEG might be indicative of a positive effect of LEV.
Alpha peak frequency (APF) is another parameter in background
Abbreviations: K-BNT, Korean Boston naming test; COWAT, Controlled oral word
association test.
EEG closely related to cognitive function. APF is an EEG measure of
cognitive preparedness, and can be a predictive measure that mon-
itors ``individual response'' of the patient to a given AED (Angelakis
et al., 2004; Clemens et al., 2006). Drug-related APF slowing more
(Lamberty et al., 2000). NP studies of LEV reported mixed results:
than 0.5 Hz is closely associated with significant decline in cogni-
in one study, LEV showed worse effects than non-drug condition,
tion and intelligence (Frost et al., 1995; Salinsky et al., 2002). In
although better than CBZ group (Meador et al., 2007). Other studies
our study, APF change following LEV administration was not signif-
have reported that LEV administration had no negative impact or
icant, although mean and median values were slightly increased.
even improvement in some of cognitive measures, such as atten-
Quantitative EEG studies to investigate the effect of AEDs often
tion, verbal fluency and memory functions (Piazzini et al., 2006;
report their results either averaged from all electrodes (Clemens
Zhou et al., 2008; Park and Kwon, 2009). In the present study of
et al., 2004) or using occipital electrodes only (Veauthier et al.,
small populations with recently diagnosed epilepsy, LEV induced
2009). In addition to global change in frequency band power, we
improvement in wide spectrum of NP tests.
conducted statistical analysis on neurophysiological measures at
The relationship between EEG and cognitive processes is a com-
each electrode contact to identify which brain regions are
plex problem. Besides the relationship between EEG slowing and
specifically affected by LEV. Our approach could demonstrate
cognitive dysfunction, activity of specific frequency bands such
region-specific changes in background EEG rhythms. Decrease in
as theta and alpha oscillations may be relevant to cognitive and
delta power was prominent in most anterior electrodes, consistent

---

890
J.R. Cho et al. / Clinical Neurophysiology 123 (2012) 883-891
Fig. 4. Representative examples that show significant correlation between change in frequency band power and improvement in neuropsychological tests. (A) Change in
relative power of theta band was negatively correlated with score change in digit symbol test (r = A0.7528, p < 0.001). (B) Power change in alpha-2 band was positively
correlated with performance change in digit symbol test (r = 0.6881, p < 0.01). In this test, increased score means improvement. Therefore, alpha-2 power increase predicted
better performance in this test. (C) Change in alpha-2 power was positively correlated with improvement in COWAT phonemic word fluency test (r = 0.6030, p < 0.01). (D)
Power change in beta-2 band was positively correlated with better performance in K-BNT task (r = 0.7089, p < 0.01).
with previous report that maximum source of delta activity lies in
from that of LEV itself. From this finding, we can speculate that LEV
the deepest and most anterior brain regions (Michel et al., 1992).
may ``normalize'' the altered power spectrum and cognitive func-
Reduced theta power was significant in anterior regions. Frontal
tion towards healthy state, but EEG data and NP test results from
theta rhythm at resting state is often interpreted as a pathological
patient with remaining seizures on LEV in addition to neurologi-
EEG finding in relation to various neurological diseases (Palmer
cally intact subjects are necessary to confirm this hypothesis.
et al., 1976). Increased power of upper alpha was observed in broad
regions. In contrast, increased beta-2 power was prominent in left
Disclosure
fronto-temporal region.
Our results from background EEG and NP tests shows that
The authors report no conflicts of interest.
change in neurophysiological parameters can be connected with
improvement in cognitive function. We asked if power change in
Acknowledgments
certain frequency bands is correlated to changes in score of NP
tests. Interestingly, functionally relevant frequency bands that
This work was supported by a grant (2010K000817) from Brain
significantly changed after LEV therapy could predict the improve-
Research Center of the 21st Century Frontier Research Program
ment in several NP tests (Fig. 4 and Table 4). Decrease in theta
funded by the Ministry of Science and by a Grant of the Korean
power and increase in alpha-2 power correlated with improve-
Health Technology R&D Project, Ministry for Health, Welfare &
ment in several NP tests requiring attention, executive function,
Family Affairs, Republic of Korea (No. A110097).
and language function. These results further support that activity
of theta and upper alpha band is involved in attention and memory
Appendix A. Supplementary data
performance (Klimesch, 1999). Increase in beta-2 power was pos-
itively correlated with improvement in executive and language
Supplementary data associated with this article can be found, in
function. In summary, decrease in slow frequencies and increase
the online version, at doi:10.1016/j.clinph.2011.09.012.
in fast frequencies could predict positive change in patient's cogni-
tive function.
The present study suffers from several limitations. The sample
References
of patients was relatively small. There may be group difference be-
Angelakis E, Lubar JF, Stathopoulou S, Kounios J. Peak alpha frequency: an
tween IGE and PE patients, but our sample size was too small to ad-
electroencephalographic
measure
of
cognitive
preparedness.
Clin
dress this issue. Moreover, we did not have control group.
Neurophysiol 2004;115:887-97.
Comparison of quantitative EEG parameters with neurologically
Berkovic SF, Knowlton RC, Leroy RF, Shiemann J, Falter U. Placebo-controlled study
of
levetiracetam
in
idiopathic
generalized
epilepsy.
Neurology
normal subjects may give additional insight on the effect of epi-
2007;669:1751-60.
lepsy and LEV on background EEG rhythms. It was reported that
Besser R, Hornung K, Theisohn M, Rothacher G, Kramer G. EEG changes in patients
epilepsy itself may cause EEG slowing without any AED therapy
during
the
introduction
of
carbamazepine.
Electroencephalogr
Clin
Neurophysiol 1992;83:19-23.
(Clemens et al., 2006). Inclusion of patients on LEV but with
Clemens B, Menes A, Nagy Z. Objective assessment of neurotoxicity while shifting
remaining seizures may differentiate the effect of seizure freedom
from carbamazepine to oxcarbazepine. Acta Neurol Scand 2004;109:324-9.

---

J.R. Cho et al. / Clinical Neurophysiology 123 (2012) 883-891
891
Clemens B, Menes A, Piros P, Bessenyei M, Altmann A, Jerney J, et al. Quantitative
Meador KJ, Gevins A, Loring DW, McEvoy LK, Ray PG, Smith ME, et al.
EEG effects of carbamazepine, oxcarbazepine, valproate, lamotrigine, and
Neuropsychological and neurophysiologic effects of carbamazepine and
possible clinical relevance of the findings. Epilepsy Res 2006;70:190-9.
levetiracetam. Neurology 2007;69:2076-84.
Corsi PM. Human memory and the medial temporal region of the brain. Dissertation
Mecarelli O, Piacenti A, Pultino P, Vicenzini E, Rizzo C, Rinalduzzi S, et al. Clinical
Abstr Int 1972;34:891B (University Microfilms No. AAI05-77717).
and electroencephalographic effects of topiramate in patients with epilepsy and
Duncan JS, Smith SJ, Forster A. Effects of the removal of phenytoin, carbamazepine
healthy volunteers. Clin Neuropharmacol 2001;24:284-9.
and valproate on cognitive function. Epilepsia 1989;30:590-6.
Mecarelli O, Vicenzini E, Pulitano P, Vanacore N, Romolo FS, Di Piero V, et al. Clinical,
Frost JD, Hrachovy RA, Glaze DG, Rettig GM. Alpha rhythm slowing during initiation
cognitive, and neurophysiologic correlates of short-term treatment with
of carbamazepine therapy: implications for future cognitive performance. J Clin
carbamazepine, oxcarbazepine, and levetiracetam in healthy volunteers. Ann
Neurophysiol 1995;12:57-63.
Pharmacothr 2004;38:1816-22.
Gambardella A, Labate A, Colosimo E, Ambrosio R, Quattrone A. Monotherapy for
Meyers JE, Meyers KR. Rey complex figure test and recognition trial. Lutz,
partial epilepsy: focus on levetiracetam. Neuropsychitr Dis Treat 2008;4:33-8.
FL: Psychological assessment resources, Inc.; 1995.
Gevins A, Smith ME, McEvoy LK. Tracking the cognitive pharmacodynamics of
Michel CM, Lehmann D, Henggeler B, Brandeis D. Localization of the sources of EEG
psychoactive
substances
with
combinations
of
behavioral
and
delta, theta, alpha and beta frequency bands using the FFT dipole
neurophysiological measures. Neuropsychopharmacology 2002;26:27-39.
approximation. Electroencephalogr Clin Neurophysiol 1992;82:38-44.
Gomer B, Wagner K, Frings L, Saar J, Carius A, Harle M, et al. The influence of
Mitra
SK.
Digital
signal
processing:
a
computer-based
approach.
New
antiepileptic drugs on cognition: A comparison of levetiracetam with
York: McGraw-Hill; 1998.
topiramate. Epilepsy Behav 2007;10:486-94.
Neufeld MY, Kogan E, Chistik V, Korczyn AD. Comparison of the effects of vigabatrin,
Heaton RK, Chelune GJ, Talley JL, Kay GG, Curtiss G. Wisconsin card sorting test
lamotrigine and topiramate on quantitative EEG in patients with epilepsy. Clin
manual revised and expanded. Lutz, FL: Psychological Assessment Resources,
Neuropharmacol 1999;22:80-6.
Inc.; 1993.
Ortinski P, Kimford JM. Cognitive side effects of antiepileptic drugs. Epilepsy Behav
Helmstaedter C, Witt JA. The effects of levetiracetam on cognition: a non-
2004;5(Suppl. 1):60-5.
interventional surveillance study. Epilepsy Behav 2008;13:642-9.
Palmer FB, Yarworth S, Niedermeyer E. Frontal midline theta rhythm. Clin
Ingvar E, Sjolund B, Ardo A. Correlation between dominant EEG frequency, cerebral
Electroencephalogr 1976;7:131-8.
oxygen uptake, and blood flow. Electroencephalogr Clin Neurophysiol
Park SP, Kwon OY. Increased EEG current-source density in the high beta frequency
1976;41:268-76.
band induced by levetiracetam adjunctive therapy in refractory partial epilepsy.
Joo EY, Tae WS, Hong SB. Regional effects of lamotrigine on cerebral glucose
J Clin Neurol 2009;5:178-85.
metabolism in idiopathic generalized epilepsy. Arch Neurol 2006a;63:1282-6.
Piazzini A, Chifari R, Canevini MP, Turner K, Fontana SP, Canger R. Levetiracetam: an
Joo EY, Hong SB, Tae WS, Han SJ, Seo DW, Lee KH, et al. Effect of lamotrigine on
improvement of attention and of oral fluency in patients with partial epilepsy.
cerebral blood flow in patients with idiopathic generalised epilepsy. Eur J Nucl
Epilepsy Res 2006;68:181-8.
Med Mol Imaging 2006b;33:724-9.
Raven J, Raven JC, Court JH. Manual for Raven's progressive matrices and vocabulary
Jun YS, Suh BY, Lee CW. Instructions for KWIS. Seoul: Jung-Ang Institute of
scales. San Antonio, TX: Harcourt Assessment; 2003.
Education; 1963.
Ray WJ, Cole HW. EEG alpha activity reflects attentional demands, and beta activity
Jung KY, Cho JW, Joo EY, Kim SH, Choi KM, Chin J, et al. Cognitive effects of
reflects emotional and cognitive processes. Science 1985;228:750-2.
topiramate revealed by standardised low-resolution brain electromagnetic
Reitan RM. Trail making test: manual for administration and scoring. Tucson
tomography
(sLORETA)
of
event-related
potentials.
Clin
Neurophysiol
AZ: Neuropsychology Press; 1992.
2010;121:1494-501.
Rocamora R, Wagner K, Schulze-Bonhage A. Levetiracetam reduces frequency and
Kang YW, Na DL. Seoul neuropsychological screening battery. Seoul: Human Brain
duration of epileptic activity in patients with refractory primary generalized
Research & Consulting; 2003.
epilepsy. Seizure 2006;15:428-33.
Kim J, Kang Y. Korean-California verbal learning test. Seoul: Special Education; 1997.
Salinsky MC, Binder LM, Oken BS, Storzbach D, Aron CR, Dodrill CB. Effects of
Kim HH, Na DL. Korean-Boston naming test (K-BNT). Seoul: Hakjisa; 1997.
gabapentine and carbamazepine on the EEG and cognition in healthy
Klimesch W. EEG alpha and theta oscillations reflect cognitive and memory
volunteers. Epilepsia 2002;43:482-90.
performance. A review and analysis. Brain Res Rev 1999;29:169-95.
Seo JG, Lee DI, Hwang YH, Lee HW, Jung DK, Suh CK, et al. Comparison of cognitive
Lamberty Y, Margineanu DG, Klitgaard H. Absence of negative impact of
effects of lamotrigine and oxcarbazepine in epilepsy patients. J Clin Neurol
levetiracetam on cognitive function and memory in normal and amygdala-
2007;3:31-7.
kindled rats. Epilepsy Behav 2000;1:333-42.
Smith ME, Gevins A, McEvoy LK, Meador KJ, Ray PG, Gilliam F. Distinct cognitive
Lee JH, Kang YW, Na DL. Efficiencies of the Stroop interference indexes in healthy
neurophysiologic
profiles
for
lamotrigine
and
topiramate.
Epilepsia
older adults and dementia patients. Korean J Clin Psychol 2000;19:807-18.
2006;47:695-703.
Marciani MG, Stanzione P, Mattia D, Spanedda F, Bassetti MA, Maschio M, et al.
Veauthier J, Haettig H, Meencke HJ. Impact of levetiracetam add-on therapy on
Lamotrigine add-on therapy in focal epilepsy: electroencephalographic and
different EEG occipital frequencies in epileptic patients. Seizure 2009;18:392-5.
neuropsychological evaluation. Clin Neuropharmacol 1998;21:41-7.
Wechlser D. Wechsler memory scale - Revised manual. San Antonio, TX: The
Marciani MG, Spanedda F, Mattia D. Neurophysiologic and neuropsychologic
Psychological Corporation; 1987.
profiles of lamotrigine in epilepsy. Clin Neuropharmacol 1999;22:159-63.
Zhou B, Zhang Q, Tian L, Xiao J, Stefan H, Zhou D. Effects of levetiracetam as an add-
Meador KJ, Loring DW, Abney OL, Allen ME, Moore EE, Zamrini EY, et al. Effects of
on therapy on cognitive function and quality of life in patients with refractory
carbamazepine and phenytoin on EEG and memory in healthy adults. Epilepsia
partial seizures. Epilepsy Behav 2008;12:305-10.
1993;34:153-7.

---

Document Outline

• Effect of levetiracetam monotherapy on background EEG activity and cognition in drug-nave epilepsy patients
  • Introduction
  • Methods
    • Patients
    • EEG recording
    • EEG data processing and analysis
    • Neuropsychological assessments
    • Statistics and correlation analysis
  • Results
    • Patients
    • Quantitative EEG analysis
    • Neuropsychological tests
    • Correlation between EEG and NP measures
  • Discussion
  • Disclosure
  • Acknowledgments
  • Supplementary data
  • References

---