How common is ictal hypoxemia and bradycardia in children with partial complex and generalized convulsive seizures?


Address correspondence to Brian D. Moseley MD, 200 First Street SW, Rochester, MN 55905, U.S.A. E-mail:


Purpose:  Autonomic effects of seizures, including cardiorespiratory abnormalities, may be involved in sudden unexpected death in epilepsy (SUDEP). The purpose of this study was to determine the prevalence and risk factors for ictal hypoxemia (oxygen saturation <90%) and ictal bradycardia (heart rate < second percentile for age) in children during recorded seizures.

Methods:  The medical records of children admitted to our Epilepsy Monitoring Unit (EMU) between November 1, 2007 and March 13, 2009 were reviewed. Children selected for this study had at least one partial complex or generalized convulsive seizure with recorded oximetry and/or heart rate data.

Results:  Forty-nine children were identified and 225 seizures were analyzed. Ictal hypoxemia was observed in 48.9% of children and 26.8% of seizures. Ictal hypoxemia was significantly more likely to occur during generalized versus nongeneralized seizures (43.9% vs. 18.9%) and when tapering antiepileptic drugs (AEDs) (75% vs. 35.5%). For partial complex seizures, there was an association between ictal hypoxemia and prolonged seizure duration. There was no correlation between ictal hypoxemia and partial seizure onset localization or lateralization.

Ictal bradycardia occurred in 8.2% of children and 3.7% of seizures. Ictal bradycardia was observed solely with partial complex seizures of extratemporal onset. Due to the low prevalence of ictal bradycardia, these findings were not statistically significant.

Discussion:  Ictal hypoxemia is common, particularly in the setting of generalized tonic–clonic seizures, prolonged partial complex seizures, and when AEDs are tapered. In contrast to previous ictal bradycardia studies, ictal bradycardia occurred exclusively in extratemporal partial complex seizures in this cohort.

Sudden unexpected death in epilepsy (SUDEP) is the most important direct epilepsy-related cause of death. In children, the incidence of SUDEP is estimated to be 1–2 per 10,000 patient-years (Camfield & Camfield, 2005). A prospective, community-based cohort study by Berg et al., 2004 reported the incidence to range from 0.52 per 1,000 patient-years in those with cryptogenic/idiopathic epilepsy to 12.6 per 1,000 patient-years in those with symptomatic epilepsy (2004). It is estimated that 12% of all epilepsy deaths in the pediatric population represent cases of SUDEP (Yang et al., 2001). Among all age groups, a greater incidence of SUDEP has been associated with generalized tonic–clonic seizures, seizure frequency >10 per year, duration of epilepsy >10 years, medication noncompliance, mental retardation, psychiatric disease, and alcohol abuse (Opeskin et al., 2000).

Although the mechanisms of SUDEP remain unclear, autonomic effects of seizures may be involved. There is debate regarding whether the primary mechanism of SUDEP is related to cardiac or respiratory effects of seizures. There have been case reports of ictal cardiac arrhythmias recorded immediately prior to death in patients with epilepsy (Dasheiff & Dickinson, 1986). In addition, there are reports of nonfatal tachyarrhythmias in adult patients during seizures (Opherk et al., 2002). In a study of children and adolescents with symptomatic temporal lobe epilepsy, 98% of seizures were reported to have concurrent tachycardia (Mayer et al., 2004).

Although they are reported less frequently than tachyarrhythmias, ictal bradyarrhythmias also occur. This phenomenon was first reported in 1906, when Russell documented asystole in one of his patients during a seizure (Russell, 1906). Since the study by Oppenheimer et al. (1992) involving cortical stimulation, attempts have been made to lateralize ictal bradycardia to the left hemisphere. However, in a recent review of the ictal bradycardia literature, Britton et al. (2006) were unable to make such an association. Instead, ictal bradycardia was most often associated with bilateral hemispheric seizure activity (Britton et al., 2006). An association between ictal bradycardia and left hemispheric onset in pediatric patients has also not been consistently confirmed.

Ictal hypoxemia has been studied extensively in adult patients. A recent study by Bateman et al. (2008) in 57 consecutive adults with intractable localization-related epilepsy showed hypoxemia occurring in 33% of recorded seizures. However, only a limited number of small studies have evaluated the prevalence of ictal hypoxemia in children (with discordant results). A single study of 53 seizures in 10 children aged 1 week to 5 years showed that hypoxemia was present in 42 seizures (79%) (Hewertson et al., 1996). Another study documented apnea and bradypnea on visual inspection of the video ictal recording in 12 (12%) of 100 children age 12 and younger with partial seizures; however, pulse oximetry was not recorded (Fogarasi et al., 2006). Younger age correlated with greater risk of ictal respiratory disturbance in this study. Finally, in a study of oxygen saturation and respiratory patterns in 40 partial onset seizures in children, 16 (40%) were associated with significant oxygen desaturations. In contrast to the study by Bateman et al. (2008) described above, desaturations in this study occurred exclusively during secondarily generalized seizures (O’Regan & Brown, 2005).

The purpose of our study was to determine the prevalence and risk factors for ictal hypoxemia (defined as a measured oxygen saturation of <90%) and ictal bradycardia (defined as a heart rate less than the second percentile for age) in children who have partial complex or generalized convulsive seizures. Such risk factors could potentially have implications for understanding the mechanisms and early detection of patients at risk for SUDEP.


The electroencephalography (EEG) reports for all children (aged 1 month to 18 years) who were consecutively admitted to the Pediatric Epilepsy Monitoring Unit (EMU) at Mayo Clinic Rochester between November 1, 2007 and March 13, 2009 were reviewed. Children were identified if they experienced at least one complex partial, secondarily generalized, primarily generalized tonic–clonic, or generalized tonic seizure during which pulse oximetry and/or single-channel electrocardiography (ECG) data were successfully acquired. All eligible pediatric EMU patients were evaluated with computer-assisted continuous 30-channel scalp EEG recordings. The International 10-20 system was used for electrode placement. Children admitted to our Pediatric EMU routinely have digital pulse oximetry monitored; this is displayed every second on a dedicated channel on the video-EEG recording. Pediatric EMU patients also have cardiac rhythm (ECG) recorded continuously on a single channel.

In children meeting the preceding criteria, clinical and electrophysiologic data were reviewed by the authors. Data acquired included age at monitoring, gender, cognitive status (normal/borderline vs. mentally retarded), family history of epilepsy, seizure etiology, age at seizure onset, duration of epilepsy prior to admission, and presence/absence of magnetic resonance imaging (MRI) brain abnormalities. The number of antiepileptic drugs (AEDs) at the time of admission, whether or not AEDs were tapered during the admission to elicit seizures, the number of prior failed AEDs, current or prior use of the ketogenic diet, current or prior use of a vagus nerve stimulator (VNS), and previous history of epilepsy surgery were also examined. Baseline seizure frequency (categorized as daily, less than daily to weekly, less than weekly to monthly, or less than monthly) and seizure type(s) per history were recorded.

For seizures meeting the preceding criteria, the video-EEG recordings were reviewed by the authors to determine the following: seizure duration (defined as the time in seconds from electrographic onset to electrographic offset), diurnal versus nocturnal onset, onset localization (temporal, extratemporal, or generalized), and onset lateralization (right vs. left). Oxygen saturation levels preceding, during, and after seizures were examined. This involved recording oxygen saturation readings 120 s prior to electrographic onset of seizures, the lowest saturation during the seizure, and the lowest saturation for the 120 s following electrographic seizure offset. If ictal desaturations occurred, the latency to and durations of readings <90%, <80%, <70%, and <60% were recorded. If oxygen saturation had not recovered to ≥90% by electrographic seizure offset, the latency of postictal oxygen saturation recovery to ≥90% was documented. Finally, the presence or absence of new postictal oxygen desaturations was recorded. Heart rate data were similarly reviewed. We defined bradycardia as a heart rate less than the second percentile for age as defined in The Harriet Lane Handbook, 11th edition (Rowe, 1987).

For children having >10 recorded seizures, only the first 10 seizures were assessed. All data entry and statistical analysis were performed using SPSS Version 16.0 (SPSS Inc., Chicago, IL, U.S.A.). Correlations with ictal hypoxemia and bradycardia were examined using chi-square analysis (Pearson’s chi-square) for categorical data and independent samples t-tests for continuous data. The p-values <0.05 were considered statistically significant. Variables with significant correlation were subsequently entered into a linear regression model.


Forty-nine children were identified who met inclusion criteria. Of these, 27 (55.1%) were monitored for presurgical evaluation, 9 (18.4%) for seizure quantification, 7 (14.3%) for spell classification, and 6 (12.2%) for epilepsy syndrome classification. Most children (67.3%) did not have their AEDs tapered on admission. A majority (89.8%) were refractory to AED therapy (i.e., had failed at least two AEDs); 85.7% had never tried the ketogenic diet, 89.8% had never undergone VNS implantation, and 87.8% had never undergone resective surgeries for epilepsy. The remaining patient characteristics are shown in Table 1.

Table 1.   Cohort demographics and characteristics
  1. PS, partial simple; PC, partial complex; 2nd Gen, secondarily generalized; GTC, generalized tonic–clonic; AEDs, antiepileptic drugs.

Age (years)9.34 ± 5.49 (range 0.5–18.3)
Gender61.2% male, 38.8% female
Cognitive status51% impaired, 49% normal/borderline
Age at seizure onset (years)3.94 ± 3.89 (range 0.08–12.1)
Duration of epilepsy (years)5.5 ± 4.84 (range 0–17.5)
Seizure syndrome80.9% localization related, 19.1% generalized
Epilepsy etiology79.6% symptomatic, 20.4% idiopathic/cryptogenic
Seizure type by history (note: patients could have one or more seizure types)20.4% PS, 73.5% PC, 34.7% 2nd Gen, 4.1% partial undifferentiated, 16.3% GTC, 14.3% tonic, 10.2% other
Reported seizure frequency at time of admission61.2% daily, 28.6% <daily–weekly, 6.1% <weekly–monthly, 4.1% <monthly
No. of current AEDs2.39 ± 1.35, range 0–6
No. of failed AEDs secondary to efficacy3.2 ± 2.72, range 0–9
MRI Brain abnormalities55.1% no, 44.9% yes
Family history of epilepsy77.6% no, 16.3% yes, 6.1% unknown

A total of 225 seizures were analyzed. The mean number of seizures reviewed per subject was 4.63 ± 2.99 (range 1–10). Of the 225 seizures, 156 (69.3%) were partial complex, 26 (11.6%) were secondarily generalized tonic–clonic, 22 (9.8%) were primary generalized tonic–clonic, and 21 (9.3%) were generalized tonic. Details of the timing and duration of these various seizure types are shown in Table 2. Details of seizure localization and lateralization are shown in Table 3.

Table 2.   Timing and duration of seizures
Seizure typeTiming (% nocturnal)Mean duration (s)
Partial complex56.4103.76 ± 171.59 (7–1,261)
Secondarily generalized23.192.65 ± 25.97 (34–144)
Primary generalized tonic–clonic36.457 ± 21.49 (21–96)
Generalized tonic38.117 ± 15.79 (8–64)
Table 3.   Focus of onset and lateralization of partial onset seizures (with clear lateralization)
Seizure typeLateralizationFocus at onset
Partial complex37.8% right, 62.2% left26.7% temporal, 73.3% extratemporal
Secondarily generalized47.8% right, 52.2% left21.7% temporal, 78.3% extratemporal

Ictal hypoxemia

Pulse oximetry data were available in 47 of 49 patients and 209 of 225 seizures. Ictal hypoxemia was observed in 48.9% of children and 26.8% of seizures (18.9% of partial complex seizures, 69.6% of secondarily generalized seizures, 59.1% of generalized tonic–clonic seizures, and 0% of generalized tonic seizures). Of those with ictal oxygen desaturations, the lowest saturation level was 80–89% in 39.3%, 70–79% in 17.9%, 60–69% in 10.7%, and <60% in 32.1% of seizures. The mean latency between seizure onset and onset of desaturation and the mean duration of desaturation are summarized in Table 4.

Table 4.   Mean latency to and duration of desaturations
Desaturation (%)Latency to desaturation from seizure onset (s) (range)Duration of desaturation (s)
80–89.965.91 ± 55.47 (1–305)65.8 ± 110.39 (2–599)
70–79.977.03 ± 77.61 (1–434)52.82 ± 67.6 (5–343)
60–69.995.52 ± 115.23 (1–585)48.67 ± 55.7 (4–243)
<60112.83 ± 140.87 (1–600)42.94 ± 59.82 (1–219)

Of those seizures with recorded desaturations, a minority (33.9%) were characterized by recovery to ≥90% prior to seizure offset. The majority (66.1%) had desaturations continuing beyond seizure offset. A minority (26.8%) had new desaturations <90% after seizure offset. The mean duration of postictal desaturation was 32.12 ± 23.68 s (range 2–133 s).

The presence of ictal hypoxemia did not correlate with age (p = 0.17), preadmission reported seizure frequency (p = 0.08), cognitive function (p = 0.31), age of seizure onset (p = 0.14), duration of epilepsy (p = 0.47), presence/absence of MRI abnormalities (p = 0.11), symptomatic versus cryptogenic/idiopathic etiology (p = 0.87), or family history of epilepsy (p = 0.08). However, ictal hypoxemia did correlate with seizure type; it was significantly more likely to occur during generalized tonic–clonic seizures (either primary or secondarily) versus partial complex seizures (27 of 143 partial offset, 29 of 66 generalized offset, p < 0.001) or generalized tonic seizures (0 of 21 tonic seizures, 29 of 66 generalized tonic–clonic seizures, p < 0.001). Prolonged seizure duration also correlated with greater risk of ictal desaturation (p < 0.002). Ictal desaturation was significantly more likely to occur in children who underwent AED tapering in the EMU (12 of 16 children who underwent tapering vs. 11 of 31 children who did not, p = 0.01). There was no statistically significant association between the total number of seizures recorded during EMU admission and the occurrence of at least one episode of ictal hypoxemia (4.96 ± 3.04 seizures recorded in those with hypoxemia, 8.69 ± 11.39 seizures recorded in those without, p = 0.12). When entered into a linear regression model, both prolonged seizure duration (p < 0.004) and generalized tonic–clonic seizure type (p < 0.009) were predictive of higher risk of ictal desaturation. However, tapering of AEDs was no longer predictive (p = 0.18).

For the partial complex seizures recorded, there was no significant correlation between the occurrence of ictal hypoxemia and temporal versus extratemporal seizure localization (p = 0.907). Of the 127 partial complex seizures with clear onset lateralization and pulse oximetry data, there was a trend favoring a correlation between ictal hypoxemia and right hemisphere lateralization (14 of 46 right, 13 of 81 left, p = 0.057); however, this trend was not statistically significant. Partial complex seizures with ictal hypoxemia were longer in duration than seizures without desaturations (mean duration was 295.44 ± 314.11 s vs. 59.31 ± 70.72, p = 0.001). Although desaturations <90% tended to be longer during seizures that remained partial at offset (vs. secondarily generalized and generalized tonic–clonic seizures), this was not statistically significant (106.95 ± 173.92 s during partial offset seizures, 31.61 ± 25.86 s during generalized offset seizures, p = 0.08).

There were 18 seizures characterized by severe ictal hypoxia (defined as a pulse oximeter reading <60%). There was no correlation between the presence of severe ictal hypoxemia and the presence/absence of seizure generalization (p = 0.49) or temporal versus extratemporal onset (p = 0.15).

Ictal bradycardia

Interpretable ECG data were available for all 49 children and in 219 of 225 seizures. Ictal bradycardia was noted in 8.2% of children and 3.7% of seizures. There was no correlation between ictal bradycardia and age (p = 0.41), preadmission seizure frequency (p = 0.75), cognitive status (p = 0.97), duration of epilepsy (p = 0.75), presence/absence of MRI abnormalities (p = 0.83), symptomatic versus cryptogenic/idiopathic etiology (p = 0.13), family history of epilepsy (p = 0.21), recorded seizure duration (p = 0.78), taper of AEDs at the time of admission to the epilepsy monitoring unit (p = 0.06), or total number of seizures recorded in the EMU (p = 0.29).

Ictal bradycardia was associated with younger age at seizure onset (1.2 ± 1.57 years with ictal bradycardia, 4.18 ± 3.95 years without, p = 0.02). Ictal bradycardia was exclusively observed with partial onset seizures (8 of 176 partial onset seizures vs. 0 of 43 primary generalized seizures, p = 0.15), and in those with extratemporal onset (8 of 115 extratemporal seizures vs. 0 of 36 temporal, p = 0.10). There was no correlation between the side of onset of partial complex seizures and ictal bradycardia (p = 0.99).


Ictal hypoxemia was common in our cohort of pediatric epilepsy patients; 48.9% of children admitted to our EMU experienced hypoxemia during one or more of their seizures. Although hypoxemia was observed during all seizure types, it was statistically more likely to occur during primary or secondarily generalized tonic–clonic seizures. It was also associated with longer seizure duration and the tapering of AEDs during EMU admission. There was no statistical association between ictal hypoxemia and seizure localization or lateralization.

The oxygen desaturations captured in our cohort were not trivial. A majority (60.7%) of recorded episodes of ictal hypoxemia were characterized by oxygen saturations <80%. More than one-fourth (32.1%) had oxygen saturations <60%. These desaturations were not brief. The mean duration of desaturation <90% was 65.8 s. In one of our monitored children, the desaturation lasted almost 10 min. The majority (66.1%) of oxygen desaturations extended into the postictal period. A smaller percentage (26.8%) experienced new oxygen desaturations after seizure termination.

Ictal hypoxemia could be a mechanism of SUDEP. In a review of 15 cases of witnessed SUDEP, 12 (80%) were marked by respiratory difficulty (Langan et al., 2000). Of five reported cases of SUDEP or near-SUDEP in an epilepsy monitoring unit, a respiratory mechanism was suggested in two (Tavee & Morris, 2008).

There has been much discussion about the potential mechanisms resulting in periictal respiratory dysfunction and resulting SUDEP. Although laryngospasm was postulated to play a role in one case (Tavee & Morris, 2008), such laryngospasm may be secondary to aspiration (of saliva, mucus, and/or stomach contents) rather than mediated directly by epileptiform activity. In contrast, pulmonary edema has been commonly reported in postmortem examinations of SUDEP victims (Salmo & Connolly, 2002). Although rare, neurogenic pulmonary edema has also been documented following generalized tonic–clonic seizures and status epilepticus (Darnell & Jay, 1982). However, this edema is often not severe and is unlikely to represent the primary etiology of respiratory dysfunction (So, 2008). Neurogenic pulmonary edema can take hours to develop following a cerebral insult (such as mild to moderate head trauma). This timing contrasts with most witnessed cases of SUDEP, where patients die within minutes (Surges et al., 2009).

Instead, apnea may be the first respiratory event in those with SUDEP. Bateman et al. found oxygen desaturations <90% in 33.2% of all seizures, regardless of whether electrographic onset was focal or generalized. When measured, oxygen desaturations were always accompanied by increases in end-tidal CO2, as would be expected with hypoventilation (Bateman et al., 2008). Whether such apnea is central or the result of tonic and/or clonic activity of respiratory muscles has not been clearly elucidated.

Given the above accounts and our findings, we believe that pulse oximetry should be standard in pediatric epilepsy monitoring units, particularly for children in whom medication adjustments are planned. In addition, care should be taken when reducing or withdrawing AEDs in an unmonitored setting (e.g., at the child’s home) where oxygen cannot be readily administered, as tapering AEDs too rapidly may increase the risk of breakthrough seizures with associated hypoxemia. Although home monitoring of oxygen saturation can be considered in children with documented desaturations and nocturnal seizures, the disruption caused by frequent false alarms limits the utility of this technology for many families.

In contrast to ictal hypoxemia, ictal bradycardia was less common in our cohort; it was recorded in only 8.2% of children and 3.7% of seizures. Due to the small number of ictal bradycardia cases in our cohort (n = 8), no statistically significant correlations could be made between ictal bradycardia and seizure-onset localization/lateralization.

There is still controversy as to whether or not bradyarrhythmias are associated with SUDEP. In a study of 20 patients with refractory partial seizures implanted with loop recorders, Rugg-Gunn et al. reported bradycardia (HR < 40) or asystole in four patients. Three of these four had potentially life-threatening asystole, and all of them required permanent pacemaker placement (Rugg-Gunn et al., 2004). However, when reviewing ECG and EEG data from 21 patients with definite/probable SUDEP versus 43 patients with refractory partial epilepsy, Nei et al. found significantly higher ictal heart rates in the SUDEP group. In addition, a majority (56%) of the SUDEP patients had documented ictal cardiac repolarization and rhythm abnormalities versus only 39% of those in the non-SUDEP cohort (Nei et al., 2004).

Our study was not without limitations. Nearly 90% of children in our study had medically intractable seizures, compared to only 8–23% in population-based studies of children with epilepsy (Sillanpaa, 1993; Camfield & Camfield, 1996). Our cohort was also unique given that it was comprised of children referred for special video-EEG monitoring in a tertiary center. However, reported incidence rates of SUDEP are greatest in cohorts of patients with refractory epilepsy (Hitiris et al., 2007). Therefore, our population represents those at highest risk. Although we were able to measure ictal hypoxemia via pulse oximetry, we did not measure chest movement or end-tidal CO2. This prevented us from commenting on the presence or absence of true apnea. The combination of ictal hypoxemia and apnea (rather than hypoxemia alone) may be a stronger predictor of risk for SUDEP. Our smaller sample size, particularly of those with hypoxemia <60% (n = 18 seizures) and bradycardia (n = 8 seizures), prevented us from making definitive correlations with seizure lateralization and localization. Finally, we did not investigate the prevalence of tachyarrhythmias in our study population. Although malignant tachyarrhythmias are not routinely reported during seizures (Nashef et al., 2007), their potential contribution to SUDEP cannot be ignored.


This study was not sponsored by any private, governmental, or institutional grants. The authors would like to thank the following individuals for their help in accessing archived video-EEG recordings: Cindy Nelson, Charlene Harstad, Jeffrey Goihl, Susan Senjem, Jean Varner, Randy Berge, and Eric Marshall.

We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.


None of the authors has any conflict of interest to disclose.