Sleep Architecture in Children with Idiopathic Generalized Epilepsy


Address correspondence and reprint requests to Dr. R. Maganti at Marshfield Clinic-Neurology, 2727 Plaza Drive, Wausau, WI 53792, U.S.A.


Summary: Purpose: Children with epilepsy experience sleep disturbances, behavioral and attentional problems at higher rates than their peers. However, the relation between sleep disturbances and the observed behavioral and attentional abnormalities is poorly defined.

Methods: Children with primary generalized epilepsy who were seizure free and between the ages of 5 and 18 years were matched with age- and gender-matched healthy controls and underwent two consecutive nights of nocturnal polysomnography with extended electroencephalography. Connor's Continuous Performance Test (CPT) was administered to assess daytime attentional function. Parents completed the Child Behavior Checklist (CBCL) to assess their emotional-behavioral status. Two sample t tests were used to examine group differences. Spearman correlations were used to examine the relation between sleep variables and behavior and attention variables. Multiple regression analysis was used to identify independent predictors of abnormal behavior and attention among patients.

Results: Eleven children with primary generalized epilepsy and eight age- and sex-matched controls participated in the study. Children with epilepsy had longer stage 1 sleep percentage (7.19 ± 3.2 vs. 4.8 ± 3.5; p = 0.05) and latency to rapid-eye-movement (REM) sleep (123.5 ± 40.1 vs. 101.75 ± 24.3; p = 0.018) compared with controls. Children with epilepsy had worse attention (CPT index, 10.94 ± 6.55 vs. 3.42 ± 4.04; p = 0.004) and exhibited significantly higher CBCL Total Behavior and Internalizing Behavior Problem scales. Whereas regression analysis showed no independent predictors of abnormal behavior and attention, a tendency toward association between CBCL total behavior scale and REM percentage (r= 0.55; p = 0.07), and between CPT overall index and stage 1 sleep percentage (r= 0.40; p = 0.10) was noted.

Conclusions: Sleep architecture is abnormal in children with primary generalized epilepsy. Further studies are needed to determine whether abnormalities in sleep architecture contribute to poor daytime behavior and attention.

Adults with epilepsy frequently experience excessive daytime sleepiness and fatigue (1–3). They also experience deficits in attention and concentration (1–3). The pathogenic mechanism underlying these abnormalities is unclear, although frequency of seizures, antiepileptic medications (AEDs), and etiology of underlying epilepsy have been thought to play a causative role. A suggestion has been made that a previously unrecognized sleep disorder may be a contributing factor (4). Polysomnographic studies (PSGs) in adults with epilepsy demonstrate reduced sleep efficiency, sleep fragmentation with frequent arousals and awakenings, and reduced slow-wave and rapid-eye-movement (REM) sleep; these findings may result in a chronic sleep-deprived state (5–7). In adults, chronic sleep deprivation can lead to cognitive complaints such poor attention and short-term memory (8).

Children with epilepsy have been shown to experience abnormalities in sleep architecture, although their clinical significance remains to be defined. As in adults, chronically sleep-deprived children experience reduced attention spans and behavioral aberrations. Subjective reporting by parents of children with epilepsy includes hyperactivity, decreased alertness, impulsiveness, emotional liability, and mood changes (9–11). Objective examination of the relation between abnormal sleep architecture, daytime behavior, and cognitive problems has not been undertaken. Such data for generalized epilepsy also are lacking, although, given the occurrence of generalized EEG seizures during sleep, theoretical reasons exist to expect a greater effect on quality and efficiency of sleep.

We sought to characterize sleep architecture in patients with idiopathic generalized epilepsy who were seizure free. This study was designed to examine the direct and indirect influence of epilepsy on objective measures of sleep architecture and relate this to quantified measures of daytime behavior and attention. We hypothesized that abnormal sleep architecture in children with idiopathic generalized epilepsy who are seizure free may underlie deficits in attention and behavioral disturbances.


Patients with epilepsy between ages 5 years and 18 years who had a diagnosis of primary idiopathic generalized epilepsy were eligible to participate. Only patients who were seizure free were approached for participation. Seizure freedom was defined as absence of clinical seizures by history. To increase the degree of certainty of seizure freedom, only those patients who had a normalized EEG while receiving treatment were enrolled in the study. Patients with preexisting mental retardation (IQ < 70), known history of any psychiatric diagnoses, known attention deficit disorder, and use of any psychoactive medications including stimulants were excluded from the study. Controls were normal healthy children, between the ages of 5 and 18 years, who were not experiencing psychiatric or behavioral problems and had no history of seizures. Additionally, a detailed sleep history was obtained from each subject and the parents, and those with a history of snoring, a history suggestive of sleep apnea, restless leg syndrome, or other parasomnias/dyssomnias, and those with previously formally diagnosed sleep disorders were excluded from the study.

A structured interview was performed for all recruited subjects, to review their medical and seizure history including age at onset, duration of epilepsy, and current AEDs. For children with epilepsy, AED levels were obtained at the time of recruitment. This study was reviewed and approved by the University of Wisconsin Institutional Review Board. Consent and assent were obtained from subjects and controls before recruitment.


All subjects underwent two consecutive nights of nocturnal PSGs as inpatients by using Grass Instruments (Grass, Quincy, MA, U.S.A.). The studies were performed at the general clinical research center sleep laboratory of the University of Wisconsin Hospital. The standard PSG recordings were modified and expanded to 30 channels. This consisted of 20 channels of EEG (using the 10-20 system), two lateral canthus eye leads, two submental leads, one electrocardiogram lead, one piezoelectric transducer for thoracic respiratory effort, one nasal thermistor for air flow, one channel for oxygen saturation, and two anterior tibialis electromyography leads to monitor leg movements.

After gathering the usual sleep timings, all subjects were encouraged to go to sleep at their usual time. Precise lights-off and lights-on times were recorded. A minimum of 7 h of sleep was recorded. Patients with epilepsy were allowed to take their AEDs as per their prescribed schedule during the days of study. The first night was used as adaptation night, and these data were not used in the analysis to control for the “first night effect.”

The obtained sleep data were manually scored in 30-s epochs by using the Rechtschaffen and Kales (R&K) scoring criteria (12), by two independent scorers. The following sleep parameters were gathered for all subjects: total time in bed, sleep efficiency, stage 1 sleep percentage total sleep time (TST), stage 2 percentage TST, slow-wave sleep percentage TST (stages 3 and 4), REM sleep percentage TST, latency to sleep onset, REM latency, wakefulness after sleep onset (WASO), respiratory disturbance index (RDI; defined as number of apneas and hypopneas per hour of sleep), periodic limb movements index (PLMSI; defined as number of periodic limb movements per hour of sleep), and arousal index. Apneas, hypopneas, and PLMS were calculated manually based on the recommended criteria by the American Academy of Sleep Medicine Task Force (13,14).

For subjects with epilepsy, spike-and-wave index (defined as total duration of spike-and-wave discharges from sleep onset to sleep offset measured in minutes) also was determined. Thus, the time durations of each generalized spike-and-wave, polyspikes-and-wave, and 3-Hz spike–wave discharges were separately measured to the 200th millisecond, from onset to offset, and totaled in minutes. We hypothesized that a higher density of epileptiform discharges would lead to more sleep fragmentation. The distribution of the spike–wave discharges during various sleep stages also was tallied. EEG seizures, defined as runs of rhythmic generalized spike-and-wave discharges lasting >3 s were noted and used in analysis. In the presence of epileptiform discharges or EEG seizures, sleep-stage scoring was based on R&K criteria. However, the epoch with EEG seizure or spike–wave discharge was carefully examined to differentiate vertex waves and K complexes from epileptiform discharges, and scoring for that stage was done by excluding the duration of the spike–wave discharge and based on what the remainder of the epoch showed. The slow waves following the spikes were excluded from sleep-stage scoring.

Tests of behavior and attention

Parents were asked to complete the Child Behavior Checklist (CBCL) (15). This is a standardized inventory consisting of 113 items that yield information on overall social competence and both broad- and narrow-band problems. Dependent measures selected from the CBCL were the summary indices of Total Behavioral Problems and Total Internalizing and Externalizing Behavior Problems.

Attention function was tested by using the Connor's Continuous Performance Task (CPT), which is computer-based task of sustained visual attention. Dependent measures include indices of correct responses (hits), errors of omission (reflecting passive inattention), errors of commission (reflecting impulsive responses), and an overall summary scale.

Statistical analysis

Differences in means of sleep measures between normal controls and children with epilepsy were determined by using two-sample t tests. Differences in measures of attention measures (CPT overall index, hits, omissions, and commissions) and behavioral measures (CBCL total problem score, internalizing problem score, and externalizing problem score) also were determined by using two-sample t tests.

Correlation between sleep variables, and epilepsy variables was computed for the patient group by using the Spearman correlation coefficient. The relation between sleep variables and dependent variables; and between epilepsy variables and dependent variables, also was examined by using Spearman Correlations. Correlation between sleep variables and behavioral/attentional variables also was computed in the control group. Multiple regression analysis was used to identify independent predictors of attention and behavior among the patient group. A value of p > 0.05 was set for statistical significance.



A total of 11 patients (five boys and six girls) with primary generalized epilepsy and eight age- and sex-matched controls (four boys and four girls) completed the study. Mean age of patients with epilepsy was 13.36 years (range, 9–17 years), and the mean age of controls was 14.25 years (range, 13–17 years). Mean duration of epilepsy for patients with epilepsy was 5.6 years (range, 2–12 years). Mean age at onset of epilepsy was 8.72 years (range, 3–14 years). Five of the 11 had a diagnosis of childhood absence epilepsy, four had juvenile absence epilepsy, and two had juvenile myoclonic epilepsy (JME). All patients were treated with AEDs; seven were receiving monotherapy with valproate (VPA), one patient was treated with ethosuximide (ESM), two were treated with a combination of VPA and ESM, and one patient was using a combination of VPA and topiramate (TPM).

Behavioral measures

Children with epilepsy exhibited significantly higher (worse) scores on the CBCL Total Behavioral Problem scale and Total Internalizing but not Total Externalizing Behavioral Problem scales (Table 2). With regard to attentional function, patients with epilepsy performed significantly worse than controls on the overall CPT index, as well as the number of correct hits (fewer than controls) and errors of omission (more than controls). No group differences were seen in the number of errors of commission (see Table 1).

Table 2. Differences in sleep measures among patients and controls

Sleep variable
(Mean ± SD)
(Mean ± SD)

p Value
  1. WASO, Wakefulness after sleep onset; PLMs, periodic limb movements of sleep; PLM, periodic limb movement; RDI, respiratory disturbance index.

Total recording time505.56 ± 57.97490.12 ± 37.38 0.35
Sleep efficiency 90.86 ± 5.15 90.66 ± 3.72 0.41
Stage 1 sleep (%)  7.19 ± 3.27  4.87 ± 2.58 0.05*
Stage 2 sleep (%) 49.49 ± 8.90 47.46 ± 8.02 0.30
Slow-wave sleep (%) 17.33 ± 5.45 17.31 ± 5.77 0.44
Rapid-eye-movement (REM) sleep (%) 17.91 ± 3.84 22.03 ± 7.93 0.10
Sleep latency 15.06 ± 11.55 18.34 ± 16.43 0.25
REM latency123.54 ± 40.13101.75 ± 24.35 0.018*
WASO 32.45 ± 24.65 26.87 ± 21.37 0.30
Total PLMs 52.40 ± 30.12 36.37 ± 22.72 0.18
PLM Index  5.08 ± 3.61  4.23 ± 2.40 0.57
RDI  2.24 ± 2.7  1.45 ± 1.39 0.26
Average oxygen saturation 93.63 ± 1.64 94.62 ± 1.31 0.49
Lowest oxygen saturation 89.90 ± 1.06 89.18 ± 1.10 0.50
Arousal index   3.2 ± 2.3  3.55 ± 2.0 0.44
PLM arousal index  0.84 ± 1.05  0.52 ± 0.74 0.46
Table 1. Behavioral measures among patients and controls
(mean ± SD)
(Mean ± SD)

p Value
  1. CPT, Continuous performance task; CBCL, Child Behavior Checklist.

  2. *Statistically significant.

CPT Overall Index10.94 ± 6.55 3.42 ± 4.040.004*
CPT Hits  303 ± 15.91  320 ± 3.330.0074*
CPT Omissions20.63 ± 15.37  3.5 ± 3.330.0068*
CPT Commissions   56 ± 11.1815.12 ± 5.790.39
CBCL Total problem score   56 ± 11.1842.87 ± 14.140.02*
CBCL–Internalizing scale56.45 ± 11.6844.25 ± 14.540.05*
CBCL–Externalizing scale50.36 ± 9.4845.14 ± 8.340.14

Sleep measures

All recruited subjects and controls completed 2 nights of nocturnal PSG. Sleep data are presented in Table 1. The mean total recording time was not significantly different between the subjects and controls [505 min for children with epilepsy (range, 417–577 min), and 490 min (range, 422–535 min) for controls (p = 0.25)]. The mean total sleep time was also not statistically different between the two groups (449.09 ± 53.97 min vs. 446.28 ± 29.84 min; p = 0.85). All subjects reported that they felt well rested in the morning-after questionnaire after the second night of study.

Stage 1 sleep percentage and REM latency are increased compared with age- and sex-matched controls (Table 2). No significant differences were found in stage II sleep percentage, slow-wave sleep percentage, REM sleep percentage, wakefulness after sleep onset, and arousal index between subjects and controls. However, significant differences were present in stage 1 sleep percentage and latency to REM sleep (Table 2, Figs. 1 and 2). PLMSI and the apnea/hypopnea index were similar for both groups (Table 2).

Epileptiform discharges

Three of the 11 patients with epilepsy had epileptiform discharges consisting of generalized spike–wave, generalized polyspike and wave, and 3-Hz spike-and-wave discharges during sleep. For these patients, quantification of epileptiform discharge duration demonstrated a total duration of 0.058, 1.75, and 1.66 min for the study period. Two of these three patients had occasional EEG seizures; the longest duration of these discharges was 8 s. Epileptiform discharges were noted mainly during stages 1 and 2 sleep, whereas they were scantily present in slow-wave sleep. No epileptiform discharges were seen during REM sleep.


When sleep variables were correlated with attention and behavior measures, a trend was observed between the CBCL Total Behavioral Problem scale and REM percentage (r= 0.55; p = 0.07); and measure of attention (CPT overall index) and stage 1 percentage (r= 0.40; p = 0.10). CBCL subscales (Total Internalizing and Externalizing Behavioral problem scales) and other measures of attention (e.g., hits, omissions, and commissions) did not correlate significantly with sleep variables among children with epilepsy. When epilepsy variables were correlated to sleep variables among children with epilepsy, a significant relation was noted between SWS percentage TST and epilepsy duration (r= 0.588; p = 0.05). Additionally, children with epileptiform discharges on PSG had lower SWS percentage TST compared with those whose sleep EEGs were normal, although this did not reach statistical significance (p = 0.08). Epilepsy variables did not correlate well with attentional measures as well as CBCL behavioral scales (p = 0.42). Finally, regression analysis did not identify any independent predictors of poor attention and behavior among children with epilepsy.


This pilot study demonstrated that sleep architecture and certain measures of attention and behavior were all worse in children with idiopathic generalized epilepsy compared with age- and sex-matched controls. Abnormalities in sleep architecture included (a) increases in stage 1 sleep percentage, and (b) latency to REM sleep compared with age- and sex-matched controls. Additionally, a trend toward lower REM percentage of total sleep time was observed (p = 0.10). Although these abnormalities were seen in the patients with epilepsy as a group, subgroup analysis showed no differences in sleep architecture between those with JME and those absence epilepsy, but children with active epileptiform discharges on EEG had relatively poor slow-wave sleep percentage compared with those that did not (12.76 ± 5.65 vs. 19.05 ± 4.59; p = 0.08). Interestingly, reduction in the duration of slow-wave sleep was seen despite the total density of epileptiform discharges occupying ≤1.75 min of the entire sleep duration. This raises the possibility that epileptiform discharges may themselves disrupt the quality of sleep and result in a chronically sleep-deprived state. The epileptiform discharges were primarily distributed in stages 1 and 2 of sleep, whereas they were rarely present during wakefulness and slow-wave sleep, and scoring in slow-wave sleep was technically difficult. Sleep deprivation resulting from such disruption in either the short term or the long term can facilitate seizures and epileptiform discharges.

Interestingly, these seizure-free children with primary generalized epilepsies nonetheless exhibited significantly poorer attentional performance and were rated as exhibiting more behavioral problems compared with healthy controls. Attentional performance was characterized predominantly by passive inattention (increased errors of omission) and not impulsive responding. Behaviorally, children with epilepsy exhibited more overall behavioral problems relative to controls, characterized by internalizing behavioral traits (e.g., withdrawal, somatic complaints, anxiety, depression) than externalizing behavioral problems (e.g., aggression). Accordingly, these findings suggest that even children with well-controlled primary generalized epilepsies should be carefully screened for cognitive and behavioral difficulties. Attempts at multiple regression analysis to identify independent predictors of poor behavior and attention were unsuccessful because of the small sample size.

Prior descriptive studies reported abnormalities in sleep architecture in children with epilepsy, although methodologically, these studies did not use controls. Furthermore, the absence of EEG data limited the interpretation of the reported findings. Baldy-Moulinier (16), studying childhood absence epilepsy, described several changes in sleep architecture including increased sleep latency, REM latency, REM percentage, and increased awakenings. Similarly, in JME, Gigli et al. (17) described increases in the ratio between total cyclic alternating pattern (CAP) duration and non-REM sleep. In these studies, a control population was not used, and it also was unclear whether these abnormalities were the result of active seizures at the time of the study.

In adults with primary or secondarily generalized epilepsy, prior PSG studies reported a number of sleep architecture changes including reduced total sleep time, slow-wave sleep, and increased upward stage shifts toward awakenings. These changes were, however, more significant when patients had nocturnal seizures compared with those who had no seizures (18). PSG studies of adults with focal epilepsy showed mixed results. In temporal lobe epilepsy, for example, reduced sleep efficiency and REM sleep and increased wakefulness after sleep onset and lighter stages of sleep were found, especially with diurnal and nocturnal seizures (19,20). In frontal lobe epilepsy, reduced sleep efficiency and increased ratio of cyclic alternating pattern duration to non-REM sleep was noted (21,22). Others have shown that sleep architecture can be abnormal in patients with focal epilepsy, even in the absence of clinical seizures, because interictal discharges can themselves cause repeated arousals (5,23–25). In our population, although we believe that the epilepsy syndrome itself caused the observed abnormalities in sleep architecture, it also is possible that the respiratory events, although minimal, may have contributed to the findings.

Questionnaire-based studies showed that sleep complaints can be common in children with epilepsy, and sleep deprivation can manifest as behavioral abnormalities such as hyperactivity, irritability, and or mood abnormalities (11,26). Previous studies also showed that children with primary generalized epilepsy demonstrate more behavioral psychopathology on various CBCL problem scales as well on CPT measures compared with the control population (27,28).

The results of the current study suggest that sleep disturbances in children with primary generalized epilepsy may contribute to the observed reductions in attention and behavior. Although none of these children had preexisting psychiatric diagnoses, significant differences were noted in CBCL Total Problem Scale and Internalizing Scale, with the latter indicating more depressive types of symptoms.

The goal of this pilot study was to determine whether sleep abnormalities underlie poor behavior and attention. Although the results were not statistically significant, a weak trend toward correlation between sleep measures and behavioral and attention measures in children with epilepsy was observed. This might indicate that abnormalities in sleep architecture could be clinically significant, and further studies are needed to determine whether such abnormalities could lead to poor behavior and attention in patients with idiopathic generalized epilepsy. These associations may not have been detectable in our study because of the small number (11) of patients with epilepsy in the current study. For example, to have 90% (80%) power to detect a correlation of 0.6, roughly equal to that observed between CBCL Total Problem Score and REM percentage, a sample size of 24 is needed, twice that in this study. For 90% (80%) power to detect a smaller but realistic correlation of 0.4, similar to that for CPCT Overall Index and stage 1 percentage, 62 patients with epilepsy are needed. Our pilot study indicates that a larger sample size is necessary to support our hypothesis.

The second limitation of our study is that we were unable to control for medication effects. Therefore all attempts were made to recruit patients on a uniform medication regimen, either VPA or ESM, or both. Although abnormalities in sleep architecture might be caused by AEDs, the results of studies in general are mixed. PHT, CBZ, gabapentin (GBP), and lamotrigine (LTG) have all been shown to affect sleep architecture (29), although others have shown that AEDs improve sleep quality by reducing the interictal discharges and EEG arousals (30). Importantly, excessive daytime sleepiness was documented in de novo patients with epilepsy before AEDs were initiated and after sustained discontinuation (31). Data on effects of VPA on sleep structure are not consistent. VPA is known to have some hypnotic properties; it has been shown not to alter the sleep architecture itself by some (29), whereas other studies have shown that it may reduce REM sleep (32) or increase stage 1 sleep (33). Thus sleep architecture can be abnormal in children with epilepsy, an abnormality that may be independent of AED effect. It also is possible that the behavioral abnormalities may have been due to AEDs as well. Several studies have shown that behavioral effects may be caused by a variety of AEDs, but literature in this area is confounded by inconsistent and contradictory results. Therefore no consistent conclusions can be made in this regard (34). A study of children with newly diagnosed idiopathic generalized epilepsy before and after treatment may answer this question, but the study would be confounded by behavioral and attentional abnormalities caused by untreated epilepsy, and therefore methodologically difficult. In our population of patients it is possible that AEDs may have contributed to the changes seen in the sleep architecture. The third limitation of our study is that we measured respiratory events with only three channels (nasal thermistor, piezoelectric transducer, and oxygen saturation), unlike in conventional PSG. Therefore we may have underestimated the number and severity of sleep-disordered breathing events among the subjects. Finally, we excluded subjects based on history of sleep disorders, and history alone may be unreliable in excluding sleep disorders, especially in children, which is another shortcoming of our study.


Sleep architectural is abnormal in children with idiopathic generalized epilepsy. Although the abnormalities could have been due to respiratory events and or medication effects, it also is possible that they are due to the epilepsy itself. Further studies are needed to determine whether abnormal sleep architecture could lead to poor daytime behavior and attention in these patients. The findings of this study offer preliminary data of regarding sleep, attention, and behavior in children with idiopathic generalized epilepsy and appear to support the undertaking of a prospective study with a larger sample size, the better to delineate sleep and its relation to abnormal behavior and attention in these children. Controlling for medication effects also should be considered an integral part of such a study design. Studies examining sleep architecture before and after treatment with AEDs also should be considered. Management strategies that focus not only on controlling seizures, but also on optimizing quality of sleep may result in improvements in behavior and cognition.


Acknowledgment:  This study was supported by the Victor Horsley Clinical Research Fellowship grant from the Epilepsy Foundation of America and The University of Wisconsin–Madison General Clinical Research Center Clinical Research Feasibility grant. Grant supervisor, R.D.S.