Sleep Medicine 4 (2003) 51–55

www.elsevier.com/locate/sleep
Original article

Effects of antiepileptic drugs on sleep architecture: a pilot study

Benjamin Legros a,b, Carl W. Bazil a,*
a
Comprehensive Epilepsy Center, Columbia-Presbyterian Medical Center, New York, NY, USA
b
Hopital Erasme, Service de Neurologie, Brussels, Belgium
Received 12 March 2002; received in revised form 24 September 2002; accepted 27 September 2002

Abstract
Objectives: The effects of antiepileptic drugs (AEDs) on sleep architecture are not well understood, especially in patients with localiza-
tion-related epilepsy, in whom seizures themselves can disrupt sleep. To clarify the effects of AEDs on sleep architecture, we performed a
prospective study, looking at sleep architecture in patients with epilepsy admitted for video-EEG monitoring.
Methods: Adult patients with localization-related epilepsy treated with a single AED and admitted between 10/1997 and 04/2001 were
included. Control patients on no AEDs were also included. Both groups were withdrawn from other AEDs. Overnight polysomnography was
recorded and was scored according to the standard method. Adult patients with localization-related epilepsy on no medication were also
recorded and served as controls. Patients with no seizure during the recording and no seizure in the 24 h preceding the recording were
analyzed in this paper. Patients with a seizure in the 24 h preceding the recording and patients with a seizure during the recording were
analyzed separately.
Results: A total of 72 nights were recorded in 39 patients, and patients taking each AED were compared to controls. We did not find any
statistically significant effect of carbamazepine (CBZ). Phenytoin (PHT) disrupted sleep by increasing stage 1 sleep (PHT: 13.2 ^ 7.3%;
control: 7.7 ^ 4.8%; P ¼ 0:008), and decreasing slow wave sleep (SWS) (PHT: 7.9 ^ 4.2%; control: 11.3 ^ 4.4%; P ¼ 0:03) and REM
sleep (PHT: 13.9 ^ 6.2; control: 18.8 ^ 5.1; P ¼ 0:01). Valproic acid (VPA) disrupted sleep by increasing stage 1 sleep (VPA: 16.8 ^ 9.8%;
control: 7.7 ^ 4.8%; P ¼ 0:007). Gabapentin (GBP) improved sleep by increasing SWS (GBP: 19.4 ^ 4.2%; control: 11.3 ^ 4.4%;
P ¼ 0:0009). PHT and VPA disrupt sleep in the absence of seizures, while CBZ and lamotrigine have no significant effects. GBP improves
sleep by increasing SWS.
Conclusions: AEDs have differing effects on sleep structure, which can be beneficial or detrimental. Consideration of these potential
effects is important in maintaining optimal sleep in patients with epilepsy.
q 2002 Elsevier Science B.V. All rights reserved.
Keywords: Antiepileptic drugs; Sleep architecture; Pilot study

1. Introduction and often difficult to distinguish from the effects of seizures.

Two of the most prevalent complaints of patients with
epilepsy are disturbed sleep and excessive daytime drowsi-
ness [1]. These symptoms have several possible causes,
including coexisting sleep disorders, effects of seizures,
and effects of anticonvulsant drugs. In studies of temporal
and frontal lobe epilepsy patients, complex partial seizures
have been shown to profoundly disrupt sleep [2–4]. Both
diurnal and nocturnal seizures decrease REM sleep; noctur-
nal seizures increase stage 1 sleep and decrease stages 2 and
4 sleep and sleep efficiency (SE) [2]. Antiepileptic drugs
(AEDs) may also disrupt sleep, but their effects are variable
* Corresponding author. Comprehensive Epilepsy Center, The Neurolo-
gical Institute, 710 West 168th Street, New York, NY 10032, USA. Tel.:
11-212-305-1742; fax: 11-212-305-1450.
E-mail address: cwb11@columbia.edu (C.W. Bazil).
1389-9457/02/$ - see front matter q 2002 Elsevier Science B.V. All rights reserved.
doi:10.1016/S1 389-9457(02)00 217-4
In order to control for the occurrence of seizures, only
patients with no seizure during the recording and no seizure
in the 24 h preceding the recording were analyzed in this
paper. Results of the recordings were compared to record-
ings from patients taking no AEDs.
2. Methods
In order to examine a relatively homogeneous population,
only patients with localization-related epilepsy (based on
history, interictal and ictal recordings) were included.
They were admitted consecutively to the Epilepsy Monitor-
ing Unit at Columbia-Presbyterian Medical Center for diag-
nosis or surgical evaluation between February 1997 and
June 2001. Patients were included if they were taking only
one AED. The minimum dosage for inclusion and the mini-
52 B. Legros, C.W. Bazil / Sleep Medicine 4 (2003) 51–55

Table 1
AED parameters

Drug Inclusion Washout

Minimum dose (mg) Blood level (mg/ml) Half-life (hour) Time from last dose (day)

Carbamazepine 600 4 8 1
Gabapentin 900 2 8 1
Lamotrigine (with inducers) 200 2 16 2
Lamotrigine (if baseline 150 2 24 4
monotherapy)
Lamotrigine (with valproate) 100 2 50 7
Oxcarbazepine 900 NA a 8 1
Phenytoin 200 10 24 3
Phenobarbital 60 15 72 9
Tiagabine 32 NA 8 1
Topiramate 100 NA 18 3
Valproic acid 500 50 12 2
Levetiracetam 1000 NA 8 1
Zonisamide 100 NA 60 7
a
NA, not available.

mum blood level of the AED (when available) are indicated
in Table 1. In order to provoke seizures, medication was
often decreased. If a patient was admitted on polytherapy
but, after having been tapered off his medication, was on
monotherapy, he was also included. The minimum time
between stopping medication and inclusion (time from last
dose) was determined according to the known half-life of
the drug. Washout times for each drug are also indicated in
Table 1. Patients taking no medication or who were finally
completely tapered off their medications were included and
served as control patients. Patients who were taking psycho-
tropic drugs at the time of the recording, had a definite
psychiatric diagnosis, had non-epileptic seizures, or who
were sleep-deprived were excluded. Patients were ques-
tioned regarding sleep disorders and excluded if a history
of sleep apnea or periodic limb movements was present,
although no specific testing was performed. Beverages
containing caffeine were not allowed.
No patient had exclusively diurnal or nocturnal seizures.
Focal onset seizures were verified by video-EEG monitor-
ing, and computerized seizure detection insured that unwit-
nessed seizures were detected. In addition to the usual 10–
20 array of scalp electrodes, patients had bilateral outer
canthus electrodes and chin EMG electrodes plus a subtem-
poral chain of electrodes (a total of 29 scalp electrodes).
Seizures were diagnosed by video-EEG using the full
scalp array. The subjects were not sleep-deprived on the
nights before recording or during recording, and were not
allowed daytime naps. Nurses instructed the patients to
sleep at 23:00 h, and to awaken at 07:00 h, however precise
‘lights off’ times were not recorded. Although the patients
were in shared rooms on a hospital ward, efforts were made
to optimize sleeping conditions. During recordings, the
doors of the room remained closed, the lights off, and the
patients undisturbed unless a seizure occurred. Following
seizures, patients were encouraged to return to sleep. If a
benzodiazepine was given after prolonged or repeated
seizures, the night was not included in the analysis.
Polysomnography was scored in 30 s epochs, according
to the standard technique [5], by reformatting digital EEG to
polysomnographic channels and parameters. Because there
was not always a precise ‘lights off’ time, scoring began at
sleep onset (defined by three consecutive epochs of stage 1
or one epoch of stage 2) and continued until awakening in
the morning by the staff at about 07:00 h. Therefore, sleep
latency could not be determined and total recording time
varied somewhat. SE is generally defined as time asleep as a
percentage of time in bed (‘lights off’ to ‘lights on’). For this
study, therefore, SE was also calculated somewhat differ-
ently from the standard, as percentage asleep from sleep
onset until awakening. This would likely result in higher
numbers, but comparisons between groups should still be
valid. The initial night of recording was not used in the
analysis (in order to control for ‘first night’ effects [6],
however patients adhered to the sleep schedule on that
night).
In order to control for the effects of seizures on sleep
architecture, only patients with no seizure in the 24 h
preceding the recording and no seizure during the recording
were analyzed. Patients with a complex partial or seconda-
rily generalized seizure during the recording or in the 24 h
preceding the recording were analyzed separately. SE,
percentages in each sleep stage, and total sleep time
(TST) were calculated. Stage 3 and stage 4 sleep were
pooled as slow wave sleep (SWS).
All results were compared to control patients, using a t-
test.
3. Results
A total of 72 nights were recorded in 39 patients (one to
 B. Legros, C.W. Bazil / Sleep Medicine 4 (2003) 51–55 53

Table 2
Demographic data a

AED monotherapy Number of recordings Number of patients Age (years) (mean (SD)) Duration of epilepsy (years) (mean (SD))

NO 23 13 37.7 (10.5) 14.7 (11.7)

CBZ 15 10 42.7 (14.8) 19.4 (12.6)
PHT 15 7 49.6 (12.7)* 23.6 (7.6)
VPA 4 2 42.5 (10.6) 16.5 (19.1)
PB 1 1 60 11
GBP 5 3 38 (16.3) 12.7 (9)
LTG 5 4 42 (19.1) 4 (2.9)
PRI 1 1 31 31
FBM 1 1 38 27
ZNS 1 1 45 29
TG 1 1 20 13
a
NO, no AED, control; CBZ, carbamazepine; PHT, phenytoin; VPA, valproic acid; PB, phenobarbital; GBP, gabapentin; LTG, lamotrigine; PRI, primidone;
FBM, felbamate; ZNS, zonisamide; TG, tiagabine. *Statistical difference from the control group (P , 0:05).

six nights per patient). Thirty-two patients had temporal
lobe epilepsy. One patient had a left hemispheric seizure
onset and in six patients seizure onset was unclear. Demo-
graphic data are presented in Table 2. There were no signif-
icant demographic differences between the groups except
that patients under phenytoin (PHT) were older than control
patients (PHT: 49.6 ^ 12.7; control: 37.7 ^ 10.5;
P ¼ 0:04).
Sleep architecture is presented in Fig. 1 (only for mono-
therapy subgroups including more than two recordings).
PHT increased stage 1 sleep (PHT: 13.2 ^ 7.3%; control:
7.7 ^ 4.8%; P ¼ 0:008), decreased SWS (PHT:
7.9 ^ 4.2%; control: 11.3 ^ 4.4%; P ¼ 0:03) and decreased
REM sleep (PHT: 13.9 ^ 6.2; control: 18.8 ^ 5.1;
P ¼ 0:01). All these results were statistically significant.
Valproic acid (VPA) significantly increased stage 1 sleep
(VPA: 16.8 ^ 9.8%; control: 7.7 ^ 4.8%; P ¼ 0:007).
Gabapentin (GBP) increased SWS (GBP: 19.4 ^ 4.2%;
Fig. 1. Sleep architecture. Stage 1 sleep is significantly increased in patients
taking phenytoin and valproic acid. Slow wave sleep is significantly
decreased in patients taking phenytoin and is increased in patients taking
gabapentin. REM sleep is significantly decreased in patients taking pheny-
toin. % of TST, percentage of total sleep time; NO, no antiepileptic drug,
control; CBZ, carbamazepine; PHT, phenytoin; VPA, valproic acid; GBP,
gabapentin; LTG, lamotrigine; SWS, slow wave sleep; REM, rapid eye
movements. *Statistical difference from the control group (P , 0:05).
control: 11.3 ^ 4.4%; P ¼ 0:0009) significantly. There
was no statistical difference between control patients and
patients under monotherapy for TST and SE.
Sleep architecture for patients with a seizure in the 24 h
preceding the recording or a seizure during the recording
contained too few patients to make reliable conclusions by
drug; however, no definite differences were seen between
agents. In the presence of seizures in the 24 h preceding the
recording, PHT tended to further decrease REM compared
with no drug (PHT: 11.7 ^ 3.4%; control: 16.7 ^ 7.4%) and
GBP to increase SWS (GBP: 19.3 ^ 18.3%; control:
12.7 ^ 10.1%).
Due to the limited number of recordings, we do not have
any data on felbamate, zonisamide, and tiagabine.
4. Comments
In the absence of seizures, we found no significant change
in sleep architecture in carbamazepine (CBZ) treated
patients, although a non-significant decrease in REM was
seen. This drug has been reported to have multiple effects on
sleep architecture, including increased TST, increased
SWS, decreased REM density with unchanged REM latency
and percentage [7]. Acutely, CBZ has been reported to
reduce REM sleep, but in chronically treated patients this
effect did not persist [8]. These authors excluded patients
with seizures within 48 h of the study, although monitoring
was not performed. In another study including both control
and epilepsy patients, decreased REM was seen acutely in
both groups; with chronic treatment (epilepsy patients only)
a decrease was seen although this was not significant [9].
Our results therefore support the idea that chronically admi-
nistered CBZ does not significantly affect sleep, although it
is possible that a subtle effect would be seen with larger
numbers.
Data on PHT from other studies have also been variable.
Decreases, increases or no changes in each sleep stage have
been reported [7,10]. Our data show that PHT disrupts
54 B. Legros, C.W. Bazil / Sleep Medicine 4 (2003) 51–55
sleep by increasing stage 1 sleep and by decreasing SWS
and REM sleep. With seizures before the recording, PHT
tended to further decrease REM sleep, but this is was not
statistically significant. It may be that the effects of
seizures and PHT are additive, but our numbers were not
large enough to support this. Patients on PHT were signif-
icantly older than controls by about 12 years, on average.
Arousals and stage 1 sleep increase with age as does the
incidence of sleep apnea, another reason to interpret our
findings cautiously. On the other hand, percentages in stage
1 sleep, SWS and REM sleep are not very different
between patients in the range 30–39 and those in the
range 40–49 years of age (age 30–39: % stage 1:
16.0 ^ 6; % SWS: 10.0 ^ 8; % REM: 20.0 ^ 4; age 40–
49: % stage 1: 14.0 ^ 9; % SWS: 8.0 ^ 6; % REM:
22.0 ^ 5) [11]. Therefore, it is unlikely that the difference
of age between the two groups can completely explain the
observed changes in sleep architecture.
VPA has been reported to have little or no effect on sleep
architecture [7]. Our findings suggest that VPA does disrupt
sleep by significantly increasing stage 1 sleep. Furthermore,
as shown in Fig. 1, there is a non-significant decrease in
stage 2, REM and SWS sleep. Although non-significant,
these results might be relevant and could explain sleepiness
in some VPA treated patients. While increased stage 1 is
non-specific, it could reflect mild sleep disruption and this
might further be clarified by studies of drowsiness
performed the following day.
Placidi et al. found that, in epilepsy patients taking other
drugs, GBP increases REM sleep and SWS and reduces
stage 1 sleep after 4 months of treatment [12]. Our mono-
therapy study confirms that GBP improves sleep quality by
increasing SWS significantly. We also found a non-signifi-
cant decrease in stage 1 and REM sleep (Fig. 1). A decrease
in light sleep, in addition to an increase in SWS, could
contribute to sleep quality improvement in GBP treated
patients.
Lamotrigine (LTG) has no significant effect on sleep
architecture. In another add-on study, Placidi et al. reported
that LTG increased REM sleep and decreased SWS after 3
months of treatment, although these differences were not
statistically significant [13].
A major limitation of this study was the inclusion of
patients whose AEDs were tapered off in order to provoke
seizures. An effect of acute withdrawal of AEDs on sleep
architecture can not be excluded, although with the agents
tested this has not previously been described. Ideally,
patients would have been compared to themselves under
control and drug conditions, however this would not have
been possible in most cases without extending hospital stay.
(From previous papers, we know that the presence of
seizures is probably one of the more important parameters
of sleep architecture [2].) Another limitation is that, because
of technological limitations in a video-monitoring unit, we
did not perform any test for sleep apnea and periodic limb
movements. Patients with a history of these sleep disorders
were excluded. Finally, the small sample sizes, especially
for VPA, GBP, and LTG, are a major limitation of this study
and our results need to be interpreted with caution. Our
findings will require confirmation in a more definitive
study using a larger sample of patients. This follow-up
study, which would need to be performed in newly identi-
fied, untreated patients, would ideally compare the same
patients prior to treatment and on treatment with anticon-
vulsant medications.
In conclusion, our results suggest that CBZ and LTG have
no significant effect on sleep architecture. PHT disrupts
sleep by increasing stage 1 sleep, and decreasing SWS
and REM sleep. VPA disrupts sleep by increasing stage 1
sleep. GBP improves sleep by increasing SWS. It may be
that more subtle changes would be seen if a larger patient
population had been/were to be examined. Our data suggest,
however, that treatment with PHT or VPA may contribute to
drowsiness in some patients. In addition, it has been
suggested that REM sleep is important in the consolidation
of certain types of memory [14], so that drugs which
decrease REM (such as PHT) could contribute to memory
dysfunction in patients with epilepsy. This interesting
hypothesis will need to await further study.
Acknowledgements
Benjamin Legros received grants from the Belgian Amer-
ican Educational Foundation and the Commission for
Educational Exchange Between Belgium, Luxembourg
and America (Fulbright grant).
References
[1] Hoeppner JB, Garron DC, Cartwright RD. Self-reported sleep disor-
der symptoms in epilepsy. Epilepsia 1984;25(4):434–437.
[2] Bazil CW, Castro LHM, Walczak TS. Diurnal and nocturnal seizures
reduce REM in patients with temporal lobe epilepsy. Arch Neurol
2000;57:363–368.
[3] Crespel A, Baldy-Moulinier M, Coubes P. The relationship between
sleep and epilepsy in frontal and temporal lobe epilepsies: practical
and physiopathologic considerations. Epilepsia 1998;39(2):150–157.
[4] Tachibana N, Shinde A, Ikeda A, et al. Supplementary motor area
seizure resembling sleep disorder. Sleep 1996;19(10):811–816.
[5] Rechtschaffen A, Kales A, editors. A manual of standardized termi-
nology, techniques, and scoring system for sleep stages of human
subjects Los Angeles, CA: UCLA Brain Information Service/Brain
Research Institute, 1968.
[6] Agnew HW, Webb WS, Williams RL. The first night effect: an EEG
study of sleep. Psychophysiology 1966;2:263–266.
[7] Sammaritano M, Sherwin A. Effect of anticonvulsant on sleep.
Neurology 2000;54(Suppl 1):S16–S24.
[8] Gigli GL, Baldinetti F, Placidi F, et al. Nocturnal sleep and daytime
somnolence in untreated patients with temporal lobe epilepsy. Change
after treatment with controlled-release carbamazepine. Epilepsia
1997;38:696–701.
[9] Placidi F, Diomedi M, Scalise A, et al. Effect of anticonvulsants on
nocturnal sleep in epilepsy. Neurology 2000;54(Suppl 1):S25–S32.
[10] Wolfe P, Roder-Wanner UU, Brede M, et al. Influence of antiepileptic
drugs on sleep. In: Martins da Silva A, Binnie CD, Meinardi H, et al.,
 B. Legros, C.W. Bazil / Sleep Medicine 4 (2003) 51–55 55

editors. Biorhythms and epilepsy. New York: Raven Press, 1985. pp.
137–153.
[11] Roth T, Roehrs T. Sleep organization and regulation. Neurology
2000;54(Suppl 1):S2–S7.
[12] Placidi F, Mattia D, Romigi A, et al. Gabapentin-induced modulation
of interictal epileptiform activity related to different vigilance levels.
Clin Neurophysiol 2000;111:1637–1642.
[13] Placidi F, Marciani MG, Diomedi M, et al. Effects of lamotrigine on
nocturnal sleep, daytime somnolence and cognitive functions in focal
epilepsy. Acta Neurol Scand 2000;102:81–86.
[14] Stickgold R, Whidbee D, Schirmer B, et al. Visual discrimination task
improvement: a multi-step process occurring during sleep. J Cognit
Neurosci 2000;12:246–254.