Issue Navigator

Volume 15 No. 04
Earn CME
Accepted Papers

Scientific Investigations

Electroencephalographic Patterns During Routine Polysomnography in Childhood and Association With Future Epilepsy Diagnosis

Robert C. Stowe, MD1; Daniel G. Glaze, MD2,3
1Department of Medicine, Division of Sleep Medicine, Perelman School of Medicine at University of Pennsylvania, Philadelphia, PA; 2Department of Pediatrics, Section of Pediatric Neurology and Developmental Neuroscience, Baylor College of Medicine and Texas Children's Hospital, Houston, Texas; 3Department of Neurology, Baylor College of Medicine, Houston, Texas


Study Objectives:

Evaluate the frequency of abnormal electroencephalograph (EEG) records during pediatric polysomnography (PSG) at a tertiary referral center and determine frequency with which these records may predict future seizures and a diagnosis of epilepsy.


Retrospective review of all pediatric PSG reports from 2013 was performed. Demographics, medical history, indications, diagnoses, and EEG reports were collected. Patients were evaluated for follow-up of future diagnosis of seizure or epilepsy over a 30-month period.


A total of 1,969 studies (56.9% males, median age 7 years) were analyzed. Abnormal EEG results were detected in 314 studies (15.9%); abnormalities included slowing in 75 (3.8%) and interictal epileptiform discharges (IEDs) in 239 (12.1%). Incidental abnormal EEG recordings were found in 186 patients (9.4%) without a prior diagnosis of seizure or epilepsy. Incidental IEDs were recorded in 126 (6.4%) and were most commonly focal (66.7%). Ten patients received follow-up EEG without clinical follow-up, 68 received clinical follow-up without follow-up EEG, and 29 received both within a 30-month period. Follow-up EEG was normal in only 30.8% of cases. Thirteen patients in the 30-month window received a new diagnosis of epilepsy. Each new diagnosis occurred in patients with preexisting neurodevelopmental disorders at high risk for seizures.


Abnormal EEG during pediatric PSG without additional history of seizure is a poor prognosticator for a future diagnosis of epilepsy. Abnormalities detected on PSG did not always portend abnormal diagnostic EEG and thus its utility to corroborate findings does not appear to be supported without additional clinical context concerning for seizure.


Stowe RC, Glaze DG. Electroencephalographic patterns during routine polysomnography in childhood and association to future epilepsy diagnosis. J Clin Sleep Med. 2019;15(4):553–562.


Current Knowledge/Study Rationale: Interictal epileptiform discharges are commonly recorded during overnight polysomnography in children. The significance and implication of these may suggest a risk of the development of seizures and a diagnosis of epilepsy, and there are no guidelines on how to further investigate or evaluate.

Study Impact: This study at a tertiary referral center found a high incidence of interictal epileptiform discharges recorded during polysomnography in its population; however, such records had a very low association with future diagnosis of epilepsy, and the patients who received such a diagnosis had a high baseline risk of the development of seizures due to underlying neurodevelopmental disorders. Without additional clinical context or history, incidental abnormalities detected on polysomnography are a poor prognostication for future diagnosis of epilepsy.


Polysomnography (PSG) is performed to assess for a wide variety of complaints in the pediatric population. Evidence-based reviews and American Academy of Sleep Medicine (AASM) practice parameters of respiratory and nonrespiratory indications for pediatric PSG have been published.14 The recommendations on approved electroencephalography (EEG) montages during PSG were updated by the AASM in 2007 to improve interrater reliability5 and this was supported by a comparison of approved montages.6

Although the purpose of the EEG during PSG is to characterize sleep stages, there is a potential for recording interictal epileptiform discharges (IEDs), such as sharp waves or spikes, and the sleep physician is left to interpret the significance, if any, of these discharges in the patient without a prior history of a seizure or epilepsy. Seizures are the clinical manifestation of hypersynchronized aberrant electrical discharges in a population of cortical neurons, and epilepsy may be defined as a disorder of recurrent, unprovoked seizures (ie, not secondary to fever, infection, electrolyte abnormality, etc.). Numerous studies have sought to capture the incidence of IEDs within various populations, typically using full-montage EEGs. Such IEDs have been described in zero to 5.6% of normal, healthy children and zero to 6.6% of normal, healthy adults.7 In a population study of 521 patients without epilepsy, the prevalence rate of IEDs was 12.3% and nearly three-fourths of these occurred in patients with an acute or progressive cerebral disorder that was not a seizure or epilepsy.8 A retrospective study evaluating 889 pediatric patients with autism spectrum disorder with 24-hour ambulatory EEG captured epileptiform activity during sleep in approximately 60%.9

These studies were typically based on conventional EEG recordings of limited duration. Overnight PSG including EEG, especially when an expanded EEG montage is utilized, has the potential to enhance identification of epileptiform abnormalities. However, such information is more limited. One study reported 1.45% of 970 healthy children 5 to 8 years of age had evidence of epileptiform activity during PSG and suggested a correlation with attention and behavioral abnormalities but made no mention of seizure or epilepsy risk.10 A study of 42 pediatric patients with attention deficit hyperactivity disorder (ADHD) found that 53% had IEDs during video PSG with full 18-lead EEG montage.11 The prevalence rate of IEDs in patients younger than 1 year undergoing PSG was 28%, although 40% had trisomy 21 and only 4% had no comorbid medical condition.12

Few studies have sought to report rates of seizure occurrence or epilepsy diagnosis after incident IED observation, even in patient populations with high IED rates.7,13 With these considerations, we sought to delineate the frequency of incidentally recorded IEDs during PSG in our sleep study population, the frequency in which they occur in patients without a history of seizure or epilepsy, and specifically if such patients with incidentally recorded IEDs would go on to have a seizure and receive an epilepsy diagnosis.


A retrospective review of all reports from overnight PSG studies between 7 and 9 hours in duration conducted on patients between 0 and 18 years of age at Texas Children's Hospital during 2013 was performed. Of note, Texas Children's Hospital's sleep center utilizes a 6-channel EEG montage (Fp1-C3, Fp2-C4, C3-O1, C4-O2, F7-T3, F8-T4, Fz-Cz, Cz-Oz, C4-M1), more than is recommended by the AASM. The entire PSG was reviewed by a board-certified sleep medicine physician. During the study period three neurology-trained (including author DGG) and two non-neurology sleep medicine physicians were responsible for generating PSG reports. Studies were scored according The AASM Manual for the Scoring of Sleep and Associated Events: Rules, Terminology and Technical Specifications.5 The PSG report was then evaluated by the first author (RCS). Studies were excluded if the patient was age 19 years or older, if the study was a repeat from the same year in the same patient, or if the study failed secondary to absence of sleep or study was aborted.

From the final PSG reports, the following information was recorded: age, sex, body mass index (BMI), past medical history, medications, indication(s) for the PSG, apnea-hypopnea index (AHI), respiratory disturbance index (RDI), periodic limb movement index (PLMI), EEG characteristics, and final diagnosis(es) provided. The AHI was calculated as the number of scored apneas and hypopneas per hour of sleep. The RDI was the number of scored apneas, hypopneas, and respiratory effort-related arousal events per hour of sleep. The PLMI was the number of periodic limb movements per hour of sleep. One hypopnea or apnea per hour of sleep was considered significant and qualifiers of obstructive and central sleep apnea were determined via indices of obstructive and central events per hour. Diagnoses of obstructive sleep apnea or central sleep apnea greater than minimal (1 to < 2 events/h) were batched for ease of reporting and analysis. Of note, higher central indices (two to three events per hour) can be normal in younger children and this diagnosis was adjusted dependent on age of patient by primary PSG reader.

EEGs were determined to be abnormal if they included background slowing, an abnormal posterior dominant rhythm (PDR), IEDs, or seizure activity. In the development of the normal EEG, the occipital dominant rhythm is 3 Hz at 3 months, 4 Hz at 6 months, 6 Hz at 12 months, 7 Hz at 3 years, and 9 Hz (approximately the adult PDR frequency) by 9 years of age;14 background slowing was determined as slower frequency PDRs than normative age-expected values. Epileptiform discharges were further characterized by being focal (occurring in one or two independent brain regions), multifocal (occurring in three or more independent brain regions), or generalized (occurring in all regions of the brain simultaneously), as well as morphology of IED (ie, sharp wave or spike, etc.). Patients with generalized discharges and focal discharges were recorded as having both. Sharp waves are defined by a general contour morphology as well as a duration of 70-200 msec, whereas spikes have a duration less than 70 msec. Polyspikes demonstrate multiple up-and-down sweeps to the spike component. An accompanying slow wave may occur after sharp waves or spikes and is of unclear significance.

For those patients with an abnormal EEG, their medical charts were reviewed for occurrence of seizures and/or epilepsy and recording of diagnostic EEGs during the subsequent 30 months after completion of PSG. Follow-up during that 30-month window was included from diverse multiple clinical settings, including emergency room visits, subspecialist consultative outpatient visits (including but not limited to neurology and sleep medicine), and primary care physician visits. The absence or presence of seizures and/or epilepsy described in those clinical encounters and visit diagnoses dichotomized our patients.

Other patient elements were explored to determine their influence on the recording of EEG abnormalities. These includes patient age, due to the classic associations of benign focal epilepsies and sleep-related EEG abnormalities and seizures, and prior medical history, as various neurodevelopmental disorders such as autism and trisomy 21 can be associated with EEG abnormalities. In addition, we analyzed how sleep breathing indices and diagnoses of sleep-disordered breathing were influenced by EEG findings.

Power and Statistical Analysis

Power analysis was conducted to ascertain the sample size needed to compare three proportions (likelihood of normal, slow, and IED EEGs). For this clinical study, the incidence of abnormal EEGs was expected to be low; to address these potential limitations, power was set at 95% with a 1% type I error rate, suggesting a benchmark of 627 patients as appropriate for detecting small effects.

For study outcomes, in order to determine whether the observations were consistent with the normal distribution, a Shapiro-Wilk test was performed on age, BMI, AHI, RDI, and PLMI. Results indicated that age (W239 = 0.96, P < .05), BMI (W237 = 0.88, P < .05), AHI (W237 = 0.56, P < .05), RDI (W237 = 0.64, P < .05), and PLMI (W237 = 0.32, P < .05) deviated from normal distributions. Visual inspections via Q-Q plots indicated similar findings.

Rather than the parametric analysis of variance, Kruskal-Wallis analyses were utilized to determine differences among patients with normal, slow, and IED EEG. Median and median average deviation (MAD) were computed as central tendency and dispersion, respectively. For each omnibus test, follow-up post hoc analyses were utilized to determine which groups differed at a pairwise level. All post hoc analyses were subject to a Bonferroni correction in order to minimize type 1 error. Significance was determined at P < .05.


Division of Study

Figure 1 demonstrates the breakdown of study participants and results. All PSG reports from 2013 were reviewed (n = 2,045). Seventy-six reports were excluded because the patient was outside the age range of 0-18 years of age (n = 6), the patient underwent PSG that was repeated during the same year (n = 63), and for study failures (n = 7). A total of 1,969 studies were evaluated and 314 PSG tests contained abnormal EEG results (15.9%). Epileptiform discharges were recorded in 239 tests (12.1%) and isolated slow-for-age EEGs were recorded in 75 studies (3.8%). There were 186 patients without a history of seizures or epilepsy who had abnormalities on the EEG, resulting in an incidental abnormal EEG detection rate of 9.4%. Sixty were slowed PDR, resulting in an incidental IED detection rate of 6.4%.

Flow chart of study.

EEG = electroencephalography, PDR = posterior dominant rhythm, PSG = polysomnography


Figure 1

Flow chart of study.

(more ...)

Follow-Up EEGs, Clinical Follow-Up, and Epilepsy Diagnosis

From the 186 patients without a history of seizure or epilepsy who had EEG abnormalities on the PSG, 39 (21.0%) had a follow-up diagnostic EEG (Figure 2). A normal follow-up EEG was recorded in 12 patients (30.8%), although one-third of these normal studies did not capture sleep. Epileptiform abnormalities were detected on 61.5% of follow-up diagnostic EEGs; there were almost even rates between focal, generalized, and multifocal IEDs (20.5%, 17.9%, and 15.4%, respectively). Twenty-nine of the patients undergoing diagnostic EEGs also had clinical follow-up within a 30-month window with neurology (Figure 1). Of the 10 patients without clinical follow-up, 8 had normal diagnostic EEGs, 1 had mild slowing of the PDR, and 1 died secondary to viral illness-induced respiratory failure complicating congenital merosin deficiency.

EEG results.

* = all IEDs is a subsumed category of focal, multifocal, generalized, and generalized + focal or multifocal IED subheadings. EEG = electroencephalography, IED = interictal epileptiform discharge, PDR = posterior dominant rhythm, PSG = polysomnography.


Figure 2

EEG results.

(more ...)

Ninety-seven of the 186 patients (52.2%) had clinical follow-up documented in the electronic medical record within 30 months of the PSG (Figure 1). Thirteen (13.4%) received a diagnosis of epilepsy at the end of the 30-month study period. Each had a comorbid neurodevelopmental condition that increased the risk of seizures. These conditions included 3 patients with severe hydrocephalus secondary to prematurity-related intraventricular hemorrhage, 3 with aneuploidy (2 trisomy 18, 1 trisomy 21), 2 with brain malformations and global developmental delays, 1 with congenital cytomegalovirus infection, 1 with Kabuki syndrome, 1 with Bohring-Opitz syndrome, and 1 patient with autism associated with a 15q11.2 microdeletion. Of the 8 patients with an eventual diagnosis of epilepsy who received diagnostic follow-up EEG, the initial EEG abnormalities on the PSG were slowed PDR in 3, focal IEDs in 2, generalized IEDs in 2, and multifocal IEDs in 1. The follow-up EEG results were similarly distributed.

Group Comparisons: Normal, Slow, and IED for Study Outcomes

There was a significant difference in age between patients with normal EEG (median 7.00, MAD 4.45), slow EEG (4.00, 2.97), and IED EEG (7.00, 4.45), H2 = 22.03, P < .05 (Table 1). Patients with slow EEG were significantly younger, P < .05, but both patients with normal and IED EEG typically were the same age. For BMI, Kruskal-Wallis test results suggested a difference between patients with normal EEG (median 18.65, MAD 4.74), slow EEG (16.71, 4.45), and IED EEG (17.76, 4.45), H2 = 39.52, P < .05. Bonferroni-adjusted post hoc analysis indicated that patients with normal EEG had a higher BMI than those with slow and IED EEG, which were not significantly different from each other, P < .05. For AHI, Kruskal-Wallis test results suggested a difference among patients with normal EEG (median 4.16, MAD 3.77), slow EEG (6.00, 5.68), and IED EEG (3.60, 3.45), H2 = 13.58, P < .05. Post hoc analyses indicated that patients with slow EEG had a significantly higher AHI than the other two groups, P < .05. For RDI, initial Kruskal-Wallis test results suggested a significant difference among patients with normal EEG (median 6.60, MAD 4.60), IED EEG (5.90, 4.74), and slow EEG (7.90, 4.46), H2 = 16.05, P < .05. Post hoc analyses indicated that all groups were significantly different, with patients with slow EEG having the most respiratory disturbance and patients with IED EEG the least.

Demographic data, PSG data, and final diagnoses.


table icon
Table 1

Demographic data, PSG data, and final diagnoses.

(more ...)

For PLMI, there appeared to be no significant difference between groups; the typical patient in all groups had no periodic limb movements. When evaluating subsets of patients with a PLMI > 5 events/h, there was no significant difference between all groups χ22 = 1.73, P > .05.

Probability: Obstructive Sleep Apnea and Central Sleep Apnea Final Diagnosis

For incidence of obstructive sleep apnea diagnosis, chi-square analysis indicated that there was a difference between groups, χ22 = 12.66, P < .05 (Table 1). All three groups were significantly different from each other, even controlling for familywise error with the Holm method. Patients with slow EEG (94.7%) were most likely and patients with IED EEG (74.9%) were least likely to have an obstructive sleep apnea diagnosis.

For incidence of central sleep apnea diagnosis, a chi-square indicated a significant difference in likelihood between groups, χ 22 = 11.60, P < .05. Patients with slow EEG (41.3%) were significantly more likely to have a final diagnosis of central sleep apnea than patients with normal EEG (25.9%), P < .05 (Table 1). After controlling for familywise error, patients with IED EEG (30.9%) did not have an incidence significantly different from either group.

Abnormal EEG Records and Patient Age

Table 2 demonstrates the proportions of normal, abnormal, and IED EEG records divided among age groups: neonates (younger than 1 month, n = 7), infants (1-23 months of age, n = 110), young children (2 to 6 years of age, n = 842), school-aged children (7 to 11 years of age, n = 637), and adolescents (12 to 18 years of age, n = 373). Chi-square test was utilized to determine whether the incidence of IED EEG was equally probable between all groups. Because of few neonates and infants in the abnormal EEG groups, these age groups were omitted in analysis, so as not to violate the assumptions of the chi-square approximation. In comparison of all those with recorded slow EEG between age groups, young children (5.5%) were more likely than school-aged children (1.9%) and adolescents (2.1%) to have a recorded slow PDR, P < .05. When comparing those with incidentally slow EEG between age groups, young children were again more likely to have a slow PDR, P < .05. There was no significant difference in incidence of all IED recordings attributable to age group, χ22 = 2.58, P > .05; however, when comparing only those with incidental IED EEG, young (7.5%) and school-aged (6.9%) children were at greater risk for IEDs to be recorded during PSG compared to adolescents, P < .05.

Division of normal, slow, and IED EEG records based on age groups and inclusion of incidental abnormalities.


table icon
Table 2

Division of normal, slow, and IED EEG records based on age groups and inclusion of incidental abnormalities.

(more ...)

Further analysis compared frequency of incidental EEG abnormality (slow versus IED) by age group (Table 2). Due to low numbers in neonate and infant groups, these again were omitted from chi-square analyses. School-aged children were more likely to have incidentally recorded IEDs than a slow EEG (χ22 = 9.37, P < .05). There were no differences between frequency of incidental slow or incidental IED EEG in young children (χ22 = 3.67, P = .055) or adolescents (χ22 = 0.185, P > .05), although the low frequency of adolescents with incidental EEG abnormalities may skew these results.

To examine potential age effects, continuous patient age in years was also utilized as a covariate in an omnibus ordinal logistic regression wherein age, AHI, BMI, PLMI, and RDI were regressed on incidence of EEG outcomes and a separate reduced model was generated with age removed. There were no discernible differences in results with age excluded. In the original omnibus model (Akaike information criterion = 2025.18), age was not a significant predictor of IED EEG incidence (β = 0.01, P = .45) and there were no significant changes when removing age (Akaike information criterion = 2023.76) from this model. As such, no further analyses controlled for age.

Probability: Prior Medical History

A series of chi-squares were used to indicate differences in incidence of the following disorders between normal, slow, or IED EEG classifications: trisomy 21, autism, ADHD, behavior problems, seizure or epilepsy, or developmental delay (Table 3). Behavioral problems were defined here as aggression, disruptive behaviors, oppositional defiance disorder, but excluded specific entries such as ADHD or explicitly stated psychiatric disorders (eg, bipolar disorder). Developmental delay was defined as undefined developmental delay, global developmental delay, gross motor delay, fine motor delay, and/ or speech delay. Patients without a prior medical history were more likely to have a normal EEG compared to patients who had slow or IED EEGs, P < .05.

Frequency of past medical history diagnoses within the study population.


table icon
Table 3

Frequency of past medical history diagnoses within the study population.

(more ...)

Trisomy 21 results indicated a difference in incidence between groups, χ22 = 30.87, P < .05. A Fisher exact test indicated that those with slow EEG (18.7%) had far greater incidence of trisomy 21 than those with normal EEGs (4.0%), P < .05. Those with normal EEG were also more likely than those with IED EEG (2.9%) to have a diagnosis of trisomy 21; however, a small number (n = 7) of patients with IED EEG also had trisomy 21, which can overinflate the chi-squared approximation.

For autism, the chi-square indicated a difference in groups, χ 22 = 12.17, P < .05. The post hoc analysis suggested patients with IED EEG (4.2%) and slow EEG (4.0%) were more likely to have autism than patients with normal EEG (1.3%), P < .05. Few patients with slow EEG had autism (n = 3), which can skew the chi-squared approximation.

For ADHD, there was no significant difference in likelihood between all groups χ22 = 1.09, P > .05. Few individuals had ADHD reported in the past medical history in the study, especially in the slow EEG group (n = 2) which presumptively skews statistical analysis outcomes

There was a marginal effect between groups for behavioral problems, χ22 = 5.93, P = .05. It appeared that there was a difference between the normal (7.6%) and IED (12.1%) groups, but this difference was eliminated when controlling for familywise error in the post hoc tests.

For seizure or epilepsy, there was significant difference in incidence in all three groups, χ22 = 290.41, P < .05. A Fisher exact test indicated all groups were significantly different from each other, despite familywise error correction. Patients with normal EEG (4.0%) were less likely than patients with slow EEG (20.0%) and patients with IED EEG (47.3%) were more likely than the other two groups to have history of seizure or epilepsy, all Ps < .05.

For developmental delay, there was a difference between all groups, χ22 = 52.98, P < .05. Patients with normal EEG (8.1%) were least likely to have developmental problems in their medical history, followed by patients with IED (17.6%), and slow EEG (36.0%). All groups were significantly different from one another, Ps < .05.

EEG Abnormalities Defined

Figure 3 demonstrates how all electrographic abnormalities were characterized. Isolated sharp waves or spikes were the most common IEDs. Seizures were recorded in 10 patients (9 of whom had a prior history of seizure or epilepsy), accounting for 0.5% of all patients. Seizure onset was focal in 4 of 10, generalized in 2 of 10, and unspecified in 4 of 10. In those with unspecified seizure onset, the described semiology was consistent with focal seizures. Each patient who experienced a seizure had otherwise abnormal EEG characteristics: two had focal sharp waves or spikes, one had focal sharp and slow waves, three had generalized sharp or spike and slow wave, and four had multifocal sharp or spike and slow-wave discharges.

Frequency of EEG abnormality in abnormal EEG PSG records and characterization of IED abnormalities.

* = other comprised undefined epileptiform abnormalities in 10 patients (5 focal, 3 multifocal, 2 generalized), 2 patients with excessive fast activity, 2 with unexplained periods of generalized voltage attenuation, 1 with voltage asymmetry, 1 without discernible normal activity, 1 with bilateral frontal-predominant 3-4 Hz rhythmic activity without clinical change, and 1 with a possible seizure defined by episode of abnormal movement lasting 2 seconds with irregular whole body clonic movement associated with generalized slowing in theta range with moderately high voltage following arousal from stage N3 sleep. EEG = electroencephalography, IED = interictal epileptiform discharge, PDR = posterior dominant rhythm, PSG = polysomnography.


Figure 3

Frequency of EEG abnormality in abnormal EEG PSG records and characterization of IED abnormalities.

(more ...)

Epileptiform discharges were identified in 239 patients and characterized by brain region onset zone as focal, multifocal, or generalized (Figure 4). These were most frequently focal (44.6%), followed by generalized (15.3%) and multifocal (9.9%). Incident IEDs recorded in patients without a history of seizures (n = 126, 6.4% of entire study) were most frequently focal (66.7%) or generalized (20.6%). Patients with a history of seizures and IEDs noted during PSG (n = 113) typically had focal (49.6%) or multifocal (22.1%) IEDs. Patients without a history of seizures were more likely to have focal IEDs, P < .05, and patients with a history of seizures were more likely to have multifocal IEDs, P < .05. Focal abnormalities were most commonly recorded from the centrally-located electrodes (Table S1 in the supplemental material).

Frequency and characterization of IED abnormality in PSG records based on brain origin as compared between patients with and without a history of seizures.

Significant values are denoted as calculated by chi-square test. EEG = electroencephalography, IED = interictal epileptiform discharge, PDR = posterior dominant rhythm, PSG = polysomnography.


Figure 4

Frequency and characterization of IED abnormality in PSG records based on brain origin as compared between patients with and without a history of seizures.

(more ...)

Indications for PSG

There were multitudinous indications for PSG; there were a total of 2,599 indications for 1,969 studies, an average of 1.32 indications per PSG. Overall, the 1,969 patients had indications subsumed into meaningful categories: sleepiness (n = 261), abnormal breathing (n = 1,775), abnormal movements (n = 112), and seizures (n = 45). Patients may have had multiple indications related to abnormal breathing (eg, obstructive sleep apnea and snoring as independent indications), but these were subsumed into a single “abnormal breathing” indication per patient. The principal aim was to characterize the likelihood of indication between the three groups (Table S2 in the supplemental material). Patients with slow PDR were significantly more likely to have a PSG indication of abnormal breathing compared to patients with normal and IED EEG, χ22 = 6.18, P < .05.

For an indication of seizure, there was a significant difference in likelihood between normal, slow, and IED EEG groups, χ22 = 90.81, P < .05. Fisher exact tests indicated that all groups were significantly different in incidence; patients with IED EEG (11.3%) were most likely and patients with normal EEG (1%) were least likely, P < .05. However, this result should be taken with a grain of salt, given only two individuals in the slow EEG group had seizure as an indication. As such, this chi-square analysis result may be inflated. Of the 45 patients who had an indication of nocturnal seizure, only 10 patients did not possess a prior diagnosis of seizure or epilepsy and only 2 of those patients were found to have EEG abnormality.


In a single year at a high-volume pediatric tertiary referral center, 15.9% of all PSG studies included some EEG abnormality. By examining for only features suggestive of, or consistent with, an epileptiform process that number reduces slightly to 12.1%. This was a large and heterogeneous referral population and 10.1% had a prior history of seizure or epilepsy. By removing this population, the incidental discovery of IEDs during PSG becomes 6.4% of all studies performed. Some form of clinical follow-up within 30 months of PSG occurred in 97 patients with abnormal EEG and 13 received a new diagnosis of seizure or epilepsy. These patients by past medical history alone were all at high risk of the development of seizures. No patient with an otherwise normal or negative past medical history found to have incidental IEDs went on to develop seizures within a 30-month window.

There was no clear age predilection for EEGs containing IEDs between ages 2 to 18 years (Table 2), although the numbers in the youngest age groups prevented appropriate statistical comparisons. Children aged 2 to 6 years were more likely to have a slow EEG in comparison to older age groups, although this developmental age period experiences significant changes in PDR14 and may reflect some delays in normal development that may normalize with time. Current guidelines also recommend patients with trisomy 21, in whom a slow PDR is common, first undergo PSG during this time.15 Although patients with trisomy 21 have reported increased frequency of IEDs,12 this was not observed in our study. Of the incidental EEG abnormality comparisons on age, school-aged children were more likely to have incidental IEDs compared to incidentally slow EEGs. Young children and school-aged children were both more likely to have incidental IEDs compared to adolescents. This corresponds to the finding of a higher incidence of focal abnormalities in patients without a history of seizure or epilepsy (Table S1), which may relate to those in benign rolandic epilepsy; however, these patients lacked a clinical history to support such a diagnosis. No family history data were collected to assess the possible association of genetic central spikes on EEG. The high incidence of multifocal IEDs in those with history of epilepsy is understandable as it reflects a more irritable and widely distributed brain cortex at risk of seizure.

Notable findings from this study demonstrated that patients with slowed EEG background rhythms tended to be younger, have a higher AHI/RDI, and receive a diagnosis of obstructive sleep apnea and central sleep apnea more frequently than patients with normal and IED EEG (Table 1). There has been no association between a slow PDR and incidence of sleep apnea, although a slow EEG record was more frequently associated with past history of developmental delay and trisomy 21 (Table 3). These patients often have concomitant hypotonia that may predispose them to significant sleep apnea indices. Patients with normal EEGs were more likely to have higher BMIs, no past medical history, and be referred for PSG for sleepiness. Patients with IEDs detected on EEG were understandably more likely to have seizure as a PSG indication or in their past medical history, but interestingly have lower AHI and less frequent diagnosis of obstructive sleep apnea. Some of our statistical subanalyses were limited by small numbers so some caution must be used in interpretation.

The strength of this study is its large population of pediatric patients to avoid any selection bias on patient population subtypes. Patients were referred from multiple clinical sites, including but not limited to sleep and neurology clinics. We did not evaluate whether the specialty of the referring provider changed the catch rate of EEG abnormality detection on PSG, nor the influence of specialty of the provider ordering follow-up diagnostic EEGs. Limitations to the study include its retrospective design and the potential selection bias that may be inherent to a tertiary referral center that predisposes to more medically complex patients than is representative of a community sample. Most indications for PSG favored sleep-related breathing disorders (Table S2) and only approximately 5% of indications related to abnormal movements; this latter, small population may not effectively characterize the incidence of abnormal EEG with such indications. Only two patients with an indication of nocturnal seizure were found to have an incidental EEG abnormality (1.1% of all incidental EEG abnormalities); this is an unlikely source of bias to our findings.

The use of a more elaborate EEG montage than is recommended by the AASM presumptively increased the capture rate of EEG abnormalities. Some experts have advocated for a more elaborate, but still abbreviated, montage to aid in seizure detection during PSG.16 The preponderance of neurology-trained sleep physicians interpreting PSG during the study period likely increased the catch rate of EEG abnormalities; we did not collect primary discipline of the reading physician to observe the differences of reported EEG abnormalities.

Final PSG reports often stated “clinical correlation advised” when an incidental EEG abnormality was described, but never explicitly recommended EEG or neurology follow-up. Physicians ordering the follow-up EEG typically were primary care pediatricians or child neurologists to whom the patient was referred, and sometimes the latter referral occurred only after receipt of an abnormal follow-up EEG. Multiple types of clinical encounters were included to capture as many patients as possible. There was a common pattern, although not specifically tabulated, of patients being referred to neurology after the discovery of incidental EEG abnormalities in whom no follow-up EEG was recommended due to the absence of antecedent seizure activity. These patients may be one cause of the low incidence of follow-up diagnostic EEG. We did not analyze reproducibility or diagnostic accuracy between PSG and follow-up EEG due to the inherent differences between the two studies and the bias of all diagnostic follow-up EEGs being performed in patients with a previously detected EEG abnormality. Rather, this is descriptive of a nonperfect correlation between studies and that follow-up EEG but was not an essential factor in the diagnosis of epilepsy.

Another limitation was that only a review of the EEG description from the PSG report was conducted rather than a direct review of each record's EEG. This was approached so as to determine how practitioners might also react and adjust treatment plans based on incidentally discovered EEG abnormalities described in a PSG report. Although it has been well described that slow-wave sleep is activating to IEDs,17 this was not assessed in the current study nor other EEG facets such as spike-wave index. Notably, there is no consensus on how to report or describe EEG abnormalities in a PSG report, let alone if it is at all necessary to do so.

Future directions for similar studies to understand risk stratification of incidentally discovered IEDs on PSG or EEG include prospective and blinded studies and concerted and timed clinical follow-up with an evaluation of neuropsychological, learning, behavioral, and sleep aspects and how they may correspond to incidentally recorded IEDs. It is also left to determine how to best follow up on incidental IEDs on PSG. Is a “routine EEG,” typically a 30- to 60-minute study that may not even capture sleep in such a short recording, sufficient or would an overnight EEG be more effective? Furthermore, specific EEG components to better stratify risk may be necessary, such as spike-wave indices or tabulation of prevalence of particular IEDs during particular sleep stages.

What is of particular interest to physicians is the consideration of using EEG as a treatment biomarker—would treatment with antiepileptic drugs (AEDs) improve the EEG results and influence affected neuropsychological, learning, behavioral, or sleep components? The determination of best course after incident IED on EEG is unclear. Some have advocated treating with AEDs, although the determination of treatment effect and explicit goal (seizure prevention, normalization of EEG, etc.) has not been explicitly studied. Perhaps most challenging in the context of future directions from our study is determination of best management in a patient with incidentally discovered IEDs at high risk of the development of seizures (eg, history of cerebral injury or genetic disorder with known association to epilepsy)—does initiation of treatment with an AED prevent or delay onset of seizures and how does one gauge treatment efficacy? A common dilemma in child neurology clinics regards the benefits of initiating AEDs versus the possibility of side effects of a daily medication in a child with one lifetime epileptic seizure and an abnormal EEG.18,19 Absent the antecedent history of seizure, the argument to initiate AEDs for an EEG abnormality becomes more nebulous.


It remains true that standard PSG review is not ideal for detection of IEDs because of the limited recording montages and compressed time scale of EEG display. However, incidentally recorded IEDs are not uncommon in pediatric PSG and may represent up to 6.4% of all records. We found no clear association with the location or morphology of the IED incidentally recorded on PSG and future diagnosis of seizure or epilepsy. The true risk of future seizures and a diagnosis epilepsy after uncovering an otherwise incidental EEG abnormality remains to be determined and referral to a pediatric neurologist or sleep physician with prior neurology training is a reasonable initial clinical decision. Further research is needed to prospectively evaluate utility and cost-effectiveness of future diagnostic evaluations for incidental EEG abnormalities during sleep.


All authors have reviewed the manuscript and have approved it in its final form. The authors completed this work at Texas Children's Hospital, Houston, TX without funding from Texas Children's Hospital, Baylor College of Medicine, or any third-party sponsors. The authors report no conflicts of interest.



American Academy of Sleep Medicine


antiepileptic drug


apnea-hypopnea index




interictal epileptiform discharge


periodic limb movement index




respiratory disturbance index


median average deviation


posterior dominant rhythm


The authors thank Thomas Tibbett, PhD for his assistance and expertise in statistical analysis.



Wise MS, Nichols CD, Grigg-Damberger MM, et al. Executive Summary of respiratory indications for polysomnography in children: an evidence-based review. Sleep. 2011;34(3):389–398. [PubMed Central][PubMed]


Kotagal S, Nichols CD, Grigg-Damberger MM, et al. Non-respiratory indications for polysomnography and related procedures in children: an evidence-based review. Sleep. 2012;35(11):1451–1466. [PubMed Central][PubMed]


Aurora RN, Zak RS, Karippot A, et al. Practice parameters for the respiratory indications for polysomnography in children. Sleep. 2011;34(3):379–388. [PubMed Central][PubMed]


Aurora RN, Lamm CI, Zak RS, et al. Practice parameters for the non-respiratory indications for polysomnography and multiple sleep latency testing for children. Sleep. 2012;35(11):1467–1473. [PubMed Central][PubMed]


Iber C, Ancoli-Israel S, Chesson AL Jr, Quan SF; for the American Academy of Sleep Medicine. The AASM Manual for the Scoring of Sleep and Associated Events: Rules, Terminology and Technical Specifications. Westchester, IL: American Academy of Sleep Medicine; 2007.


Duce B, Rego C, Milosavljevic J, Hukins C. The AASM recommended and acceptable EEG montages are comparable for the staging of sleep and scoring of EEG arousals. J Clin Sleep Med. 2014;10(7):803–809. [PubMed Central][PubMed]


So EL. Interictal epileptiform discharges in persons without a history of seizures: what do they mean? J Clin Neurophysiol. 2010;27(4):229–238. [PubMed]


Sam MC, So EL. Significance of epileptiform discharge in patients without epilepsy in the community. Epilepsia. 2001;42(10):1273–1278. [PubMed]


Chez MG, Chang M, Krasne V, Coughlan C, Kominsky M, Schwartz A. Frequency in epileptiform EEG abnormalities in a sequential screening of autistic patients with no known clinical epilepsy from 1996 to 2005. Epilepsy Behav. 2006;8(1):267–271. [PubMed]


Capdevila OS, Dayyat E, Kheirandish-Gozal L, Gozal G. Prevalence of epileptiform activity in healthy children during sleep. Sleep Med. 2008;9(3):303–309. [PubMed]


Silvestri R, Gagliano A, Calarese T, et al. Ictal and interictal EEG abnormalities in ADHD children recorded over night by video-polysomnography. Epilepsy Res. 2007;75(2-3):130–137. [PubMed]


Adeleye A, Ho AW, Nettel-Aguirre A, Kirk V, Buchhalter J. Prevalence of interictal epileptiform discharges in children less than one year of age referred for polysomnography. Neurodiagn J. 2015;55(2):122–135. [PubMed]


Shelley BP, Trimble MR, Boutros NN. Electroencephalographic cerebral dysrhythmic abnormalities in the trinity of nonepileptic general population, neuropsychiatric, and neurobehavioral disorders. J Neuropsychiatry Clin Neurosci. 2008;20(1):7–22. [PubMed]


Hrachovy RA. Development of the normal electroencephalogram. In: Levin K, Luders H, eds. Comprehensive Clinical Neurophysiology. Philadelphia, PA: WB Saunders; 2000:387–413.


Bull MJ; Committee on Genetics. Health supervision for children with Down syndrome. Pediatrics. 2011;128(2):393–406. [PubMed]


Marzec ML, Malow BA. Approaches to staging sleep in polysomnographic studies with epileptic activity. Sleep Med. 2003;4(5):409–417. [PubMed]


Bazil CW. Nocturnal seizures and the effects of anticonvulsants on sleep. Curr Neurol Neurosci Rep. 2008;8(2):149–154. [PubMed]


Camfield PR, Camfield CS, Dooley JM, Tibbles JA, Fung T, Garner B. Epilepsy after a first unprovoked seizure in childhood. Neurology. 1985;35(11):1657–1660. [PubMed]


Kanemura H, Sano F, Ohyama T, Mizorogi S, Sugita K, Aihara M. EEG characteristics predict subsequent epilepsy in children with their first unprovoked seizure. Epilepsy Res. 2015;115:58–62. [PubMed]

Supplemental Material

Supplemental Material

(.pdf | 88 KB)