Dr Rob Forsyth, Institute of Neuroscience, Newcastle University, Sir James Spence Institute, Royal Victoria Infirmary, Newcastle upon Tyne NE1 4LP, UK. E-mail firstname.lastname@example.org
Aim To describe the frequency and causes of death in children with epilepsy, ascertain the contribution of seizure disorder to cause of death, and compare with rates of sudden unexplained death in children without epilepsy.
Method This study was a retrospective review of clinical and death certificate records. It examined two UK population-based samples of deaths in children with epilepsy from 1 month to 18 years, together comprising the largest reported series of deaths in children with epilepsy (n=265).
Results In approximately two-thirds, the death was not due to the seizure disorder. Rates of unexplained death were similar in the two samples at 7.3% and 9.7%: all were in children with symptomatic or presumed symptomatic epilepsy. There were no unexplained deaths in the children with idiopathic epilepsy. Four per cent of the deaths were of children experiencing acute symptomatic seizures as part of their final illness. The risk of unexpected, unexplained death in children with idiopathic epilepsy is not more than 65 per 100 000 child-years.
Interpretation Epilepsy is associated with an increased risk of death in childhood but this risk is almost entirely confined to those with an associated neurodevelopmental disorder. The risk of unexpected, unexplained death in children with idiopathic epilepsy is extremely small.
Confidential Enquiry into Maternal and Child Health
Sudden unexpected death in epilepsy
What this study adds
• Most deaths in children with epilepsy are not epilepsy-related.
• The risk of unexpected, unexplained death in children with idiopathic epilepsy appears smaller than the risk of seizure-related accidental death, and is not more than 65 per 100 000 child-years.
Awareness of the increased mortality of people with epilepsy and the need for appropriate counselling has grown in the past decade.1,2 In the UK a 2002 National Sentinel Clinical Audit of Epilepsy-Related Deaths3 was particularly influential. It emphasized a risk of premature death for people with epilepsy of two to three times that of the general population, and suggested that most premature deaths are in the context of seizures. It particularly emphasized the phenomenon of sudden unexpected death in epilepsy (SUDEP) and the risk factors for SUDEP in adults, which were poor seizure control, nocturnal seizures, abrupt and frequent changes in medication, and non-adherence to treatment.3
Childhood-onset epilepsies constitute a heterogeneous group of conditions differing greatly in treatability and associated morbidity. Although its limitations are recognized,4 the present International League Against Epilepsy classification of epilepsy aetiology5 has three broad categories. Epilepsy due to identifiable pathology is symptomatic: where that underlying cause is very recent or still active, the term ‘acute symptomatic’ may be used. In contrast to the most common usage of ‘idiopathic’ to mean ‘cause unknown’, the term idiopathic is reserved for a defined list of presumed polygenic primary epilepsy syndromes. In remaining cases where the cause is unknown, the phrase ‘presumed symptomatic’ (previously, ‘cryptogenic’) is used: this group is thus defined by exclusion. Many children with severe epilepsy and associated neurodevelopmental disorder still fall into this group. The increasing use of magnetic resonance imaging and yield of genetic investigation means the presumed symptomatic group is becoming smaller over time: thus historical comparisons of risk of death and standardized mortality ratio by epilepsy aetiology,6 including recently published long-term follow-up of cohorts characterized many years ago,7 must be undertaken with caution. Symptomatic aetiology is a strong risk factor for death in children with epilepsy.8–10 The standardized mortality ratio for symptomatic epilepsies is estimated at 20 to 50.10–13 By contrast all recently published estimates of the standardized mortality ratio for non-symptomatic (i.e. either idiopathic or presumed symptomatic) epilepsies have 95% confidence intervals that include 1.11–13
Previous studies of epilepsy-related mortality in children have largely followed cohorts recruited in life9–12,14 and therefore absolute numbers of deaths are small. By contrast we report an analysis of two large samples of deaths in children with epilepsy for whom information was available about circumstances of death, type of epilepsy, and the contribution if any of epilepsy to death. A larger number of deaths facilitates recognition of rarer patterns. Additionally it allows examination of the usefulness of the concept of SUDEP in children, and allows risk of death in children with epilepsy to be placed in the context of risk of unexpected death in all children,15,16 a perspective largely absent in previous reports.
We obtained details of deaths in children with epilepsy from two different populations, hereafter referred to as samples X and Y.
Sample X, derived mainly from a tertiary paediatric neurology service
Sample X comprised children aged under 18 years with epilepsy who died between 1 January 1989 and 31 December 2005 in the former Northern Region of England (total population 3.1 million). Cases were primarily identified from a database of all children seen since 1997 as in-patients or outpatients by the single tertiary paediatric neurology service in the region. Additionally death certificates for the period from 1 January 1989 to 31 December 2005 were hand-searched to identify deaths in children with epilepsy who had never attended the regional centre. We searched on ‘epilepsy’, ‘seizure’, ‘SUDEP’ or synonyms, diagnoses implying epilepsy (such as West or Aicardi syndrome), or diseases likely to be associated with epilepsy. The December 2005 cut-off was chosen to avoid overlap with sample Y to which the Northern Region contributed (see below).
Sample Y, derived from the Confidential Enquiry into Maternal and Child Health (CEMACH)
Sample Y was derived from the dataset obtained by the CEMACH ‘Why Children Die’ enquiry.17 This dataset comprised all deaths during calendar year 2006 in children aged 28 days to 17 years 364 days at death, in a defined geographical area comprising the South West, North East, and West Midlands Regions of England, plus Wales and Northern Ireland. These areas were estimated to have a total population under 18 years of 3 868 617, although this figure includes infants younger than 1 month, who are not in the numerator.17
For sample X, hospital records were obtained from the regional centre and local district hospital, with post-mortem data where available. For sample Y, an anonymized abstracted dataset was provided by CEMACH, which included death certificate data, previous medical conditions (in particular, neurological and neurodevelopmental disorders), and prescribed medications.
Using these data, cases were classified on two axes: the aetiology of the seizure disorder and the relationship of the seizure disorder to death. The criteria for this classification are shown in Table I. One author (VN) classified sample X; three authors (VN, MK, RF) classified sample Y. In sample Y, initial classification was performed independently. Interobserver agreement, assessed by the multiobserver variant of the Cohen kappa (κ) statistic,18,19 found κ=0.92 for seizure aetiology and 0.78 for the relationship of seizures to cause of death. Discrepancies in relation to seizure aetiology were readily resolved. Agreement on the relationship of epilepsy to mechanism of death in some cases was more difficult: where a child was known to have epilepsy but death certificate details were scanty, there was sometimes uncertainty as to whether this constituted an ‘unexplained death in a child with epilepsy’ or ‘insufficient data’; such cases were resolved by consensus. In 11 of these cases we concluded that the death was unlikely to be seizure-related.
Table I. Criteria for classification of seizure disorder aetiology and relationship of seizure disorder to death
Aetiology of seizure disorder
Explicit mention of pre-existing epilepsy, or a diagnosis implying this (e.g. West syndrome); or clear statement of central nervous system disorder likely to be substrate for epilepsy (e.g. brain malformation) with a record of regular anticonvulsant prescription. Subdivided into progressive or static groups, the former comprising neurodegenerative or metabolic conditions.
Explicit mention of a pre-existing idiopathic epilepsy syndrome, either focal (e.g. benign epilepsy of childhood with centro-temporal spikes) or generalized (e.g. juvenile myoclonic epilepsy).
Presumed symptomatic epilepsy
Explicit mention of pre-existing epilepsy, or a diagnosis implying this (e.g. West syndrome) but without clear statement of a symptomatic aetiology or an idiopathic epilepsy.
Likely acute symptomatic seizures as part of final illness
Identified other cause of death (e.g. central nervous system infection) capable of causing acute symptomatic seizures, often without explicit mention of seizures, but with documented prescription of antiepileptic drugs.
Relationship to death
Documentation of an active pathology (e.g. bronchopneumonia) as cause of death. In contrast, a past history of severe perinatal asphyxia or preterm birth was not regarded as an active cause of death and such cases are coded as unexplained.
Unlikely to be seizure-related
Consensus on basis of clinical discussion that death was unlikely to be seizure-related (e.g. at high risk of severe respiratory disease).
Clear documentation of seizures at time of death (e.g. in status epilepticus) or of seizure-related accidental death (e.g. drowning).
A child with epilepsy in whom, despite considerable data being available, there was no understanding of the mechanism of death.
Few or no data.
Analyses used χ2 or Fisher’s exact tests, as appropriate, with a significance threshold of p<0.05
The sample X study was approved as an audit by Newcastle upon Tyne Hospitals NHS Foundation Trust. Access to sample Y was granted under approval from the North West Research Ethics Committee (reference 05/MRE08/51).
From the tertiary centre database and death certificate search we identified 113 children with epilepsy who had died (Table II). A further 22 children for whom medical records were untraceable were not included, although death certificate data suggested 12 of these had had a seizure disorder.
Table II. Tabulation of epilepsy aetiology against contribution of epilepsy to death in sample X
Symptomatic epilepsy: progressive
Symptomatic epilepsy: static
Likely acute symptomatic seizures
aCar passenger in road traffic accident.
Of the 113 children, there were insufficient data to determine the mechanism of death in 16 although it was established that all 16 had severe underlying neurodevelopmental disorders. In 65 of the remaining 97 the cause of death could be confidently directly attributed to the disorder causing the epilepsy: typical examples included primary neurodegenerative disease (leukodystrophies, neuronal ceroid lipofuscinoses, mitochondrial encephalopathies, etc.) and respiratory complications of severe cerebral palsy. A further 25 had seizure-related deaths, including deaths in status epilepticus (n=7), multi-organ failure secondary to status epilepticus (n=6), respiratory failure secondary to status (n=5), seizure-associated aspiration (n=4), seizure-related drowning (n=1), and status-related severe cerebral oedema (n=2).
In the remaining seven, the death was ‘unexplained’ in that clinical data were insufficient to establish if the death was seizure-related or not. Of these seven children, six had severe neurodevelopmental disorder and there were insufficient data for the seventh.
There was no statistically significant association between sex and mechanism of death (χ2=4.23, df=3, p=0.24).
The full CEMACH dataset had details on 955 deaths (Table III). Sample Y comprises the 259 children with epilepsy or acute symptomatic seizures as part of their final illness. Thirty per cent occurred in children less than 1 year old (Fig. 1).
Table III. Tabulation of epilepsy aetiology against contribution of epilepsy to death in sample Y
Symptomatic epilepsy: progressive
Symptomatic epilepsy: static
Likely acute symptomatic seizures
aDrowning, presumed seizure-related (see text). bHanging, but circumstances unclear (see text).
Unlikely to be seizure-related
Of the 25 unexplained deaths, 20 had symptomatic epilepsy, four had presumed symptomatic epilepsy, and one likely acute symptomatic seizures. None had idiopathic epilepsy.
There were two deaths in children with idiopathic epilepsy, one a drowning and the other a hanging. The available data did not establish a cause of the hanging but the drowning was presumed accidental and seizure-related. The remaining eight seizure-related deaths were all in the context of status epilepticus in children with symptomatic epilepsy associated with severe neurodevelopmental disorder.
There was no statistically significant association between sex and mechanism of death (χ2=4.05, df=3, p=0.25).
Six children in the full CEMACH dataset of deaths were identified as highly likely to have suffered a sudden cardiac death without known cardiac disease in life and without epilepsy. There were an additional three probable cases.
The combination of 97 deaths in sample X and 168 deaths in sample Y makes this the largest published series of deaths in children with epilepsy (n=265), and the first population-based description of patterns of death for children with epilepsy in nearly 20 years.13 The two-sample approach was taken in anticipation that the cohorts would have complementary strengths. By ascertainment in part through a regional paediatric neurology service, sample X is biased towards children with more severe epilepsy; however, the availability of hospital medical records, and in some cases coroner post-mortem reports, provided greater clinical detail on the relationship between the child’s seizure disorder and cause of death. Sample Y has a well-defined sampling frame and ascertainment methodology, but clinical data are more limited. Sample Y has a higher proportion of infants.
The main limitation of this study is the high proportion of unascertainable cases in both samples: we considered it preferable to record these as unascertainable rather than assume diagnoses such as ‘possible SUDEP’.20 Despite the availability of hospital medical records in sample X, data were often incomplete, especially when a child died outside hospital. Data from death certificates in isolation were usually insufficient to determine mechanisms of death and the contribution of coexistent epilepsy.
In both samples most deaths in children with identified epilepsy were found to be unrelated to the seizure disorder (66% in sample X and 79% in sample Y), comparable to the 63% reported by Harvey et al.13 The limited clinical data in sample Y inevitably led to a much higher unascertainable rate (almost 40%) than in sample X, but the rates of unexplained death are comparable at 7/96 (7.3%) in sample X and 25/259 (9.7%) in sample Y. Again the finding that all these unexplained deaths were in children with symptomatic epilepsy is common to both datasets (data missing for one child in sample X). Of the two deaths in children with an idiopathic epilepsy in sample Y, one died by drowning and one by a hanging (it was not possible to distinguish accidental death from suicide from the available data). The one death of a child with idiopathic epilepsy in sample X was as a car passenger in a road traffic accident. We could not replicate the finding of a previous study21 of an association between male sex and excess risk of seizure-related death.
Even a conservative estimate that in sample Y 129 of all 955 CEMACH deaths (14%) were associated with epilepsy in life is a much higher proportion than official figures of 1% based on death certificate data.22 The inadequacies of death certificate data have been noted previously13 and motivated the CEMACH study (which found that death certificate data were inaccurate or insufficient in 35% of child deaths17).
By studying samples of paediatric deaths rather than deaths in cohorts of children with epilepsy recruited in life,7,9–12 the absolute number of deaths in this study is much larger, allowing identification of rarer events. There have been three similar studies of paediatric epilepsy-related mortality in cohorts identified post mortem. Donner et al.21 identified sudden epilepsy-related deaths over a 10-year period in Ontario, Canada ascertained largely through the coroner’s office. Of a total of 27 sudden deaths, eight were in children with idiopathic epilepsy, which is at odds both with our findings and other idiopathic epilepsy series.14 Since this sample was ascertained largely through coroners’ services, children with particularly unusual and unanticipated deaths may be overrepresented at the expense of children with known severe neurodisability and epilepsy, whose deaths may have been regarded as less unexpected. Weber et al.23 published a single-hospital series of epilepsy-related deaths, identifying four ‘SUDEPs’ and estimating a rate of 0.43 SUDEPs per 1000 patient-years; however, it is not clear how these four cases were distinguished from other causes of death in children with epilepsy. All four children were described as having severe cryptogenic epilepsies. The paper by Harvey et al.13 is most similar in design to the sample Y study, identifying 93 with a previous history of epilepsy of a total of 1095 deaths from a population-based paediatric mortality surveillance system. In this group of 93, deaths were directly attributed to epilepsy in 20 of which 11 were classified as SUDEPs. Ninety-four per cent of the children in this series had ‘secondary’ (i.e. symptomatic) epilepsy.
Discussions of death in epilepsy often emphasize SUDEP, defined as a ‘sudden, unexpected, witnessed or unwitnessed, non-traumatic and non-drowning, death in a patient with epilepsy, with or without evidence for a seizure, and excluding documented status epilepticus, in which post-mortem examination does not reveal a toxicological or anatomical cause for death’.24 The literature9,10,25 confirms that any paediatric counterpart of the adult SUDEP example of the sudden death of an otherwise healthy individual living a fully independent life, is a very rare form of epilepsy-related death in childhood. The reason for this is unclear although epidemiological study shows an increase in epilepsy-related mortality in early adulthood.26 It should be noted that early childhood onset of epilepsy is associated with an increased SUDEP risk in adulthood,27 although this probably reflects symptomatic aetiology. It is clear from Tables II and III that most deaths in children with epilepsy are in those with symptomatic epilepsy related to underlying neurodevelopmental disorders such as cerebral palsy.
For the child with strictly defined focal or generalized idiopathic epilepsy, assuming this comprises approximately 30% of children with epilepsy,28 the at-risk population of children with idiopathic epilepsy in the denominator for sample Y is approximately 4600 child-years. Given no occurrence of SUDEP in this sample, a rate cannot be estimated but the upper 95% confidence interval on the true SUDEP risk is approximately 65 per 100 000 child-years.29 In comparison, rates of sudden unexplained death in type 1 diabetes are estimated at 45 per 100 000 person-years (albeit typically in adult life).15 Sudden death in children with no health problems identified either in life or at autopsy is thought primarily to relate to conditions predisposing to cardiac rhythm disturbances. Wren et al. estimated that in children over 2 years (i.e. excluding sudden unexpected death in infancy, cot death) this situation accounted for about 7% of all identified sudden deaths, a rate of approximately 2 per million children per annum.16 Our observation of at least six sudden cardiac deaths in the CEMACH dataset in children without previously identified cardiac disease equates to a comparable 1.5 per million child-years. It is striking that two of the eight deaths of children with idiopathic epilepsy in the Donner et al. 21 series were witnessed collapses in the context of physical activity. Although a previous diagnosis of epilepsy does increase the risk for sudden death, on occasions the association may be coincidental.
Children comprised only 12% of the cases in the National Sentinel Clinical Audit3 and its findings summarized in the Introduction need important qualifications to reflect the paediatric situation. First, quoting an overall risk for people with epilepsy misses the enormously different risks of symptomatic and idiopathic epilepsies. Second, the suggestion that most premature deaths in people with epilepsy are seizure-related is not confirmed either in this paper or others13 where most paediatric deaths are due to the underlying neurological disease of which the seizure disorder is a manifestation. This may reflect the fact that the National Sentinel Clinical Audit specifically pre-selected the subset of cases (33%) where epilepsy was judged the probable cause of death based on death certificate data.
The definition of SUDEP includes deaths both in and without the immediate context of a seizure. The seizure-related deaths group in sample X comprises complications of severe and prolonged seizures; however, it could be argued that some of the seizure-related deaths in sample Y might have met SUDEP criteria. This lack of specificity in the definition of SUDEP limits its usefulness in relation to informed discussion of preventability; although improved seizure control does appear to reduce mortality,30,31 it is unclear what proportion of SUDEPs is theoretically preventable. An additional problem in applying the current concept of SUDEP to the paediatric age group is the value judgement involved in deciding whether or not a death in a severely disabled child with epilepsy was ‘expected’: this was a regular theme of discussions between the authors and deceased children’s local paediatricians that has also been noted by others.9
Current clinical guidelines advocate sensitive and informed discussion of SUDEP risks with young people with epilepsy and carers1,2 although the practicalities of how to do this remain controversial.32–34 Proponents of early discussion emphasize the right to know, that such information will be gathered anyway from other less balanced sources, and that appropriate information may lead to greater adherence to therapy, avoidance of risk factors, and informed lifestyle choices. Sceptics worry about the psychological consequences of informing someone in a rushed outpatient clinic of a threat to life of unknown mechanism and uncertain preventability so that the fact of a ‘conversation’ can be documented in the notes.
The contribution of this study is that rather than emphasizing a spectre of inexplicable sudden death, such conversations should be framed in the broader context of the mortality risks of incomplete seizure control, responses to which will address all known SUDEP risk factors. For severely disabled children with symptomatic epilepsy, this can lead to informed, shared decision-making; for example, balancing the aspiration risks of uncontrolled seizures against those of aggravating impaired bulbar function by excessive dosing with anticonvulsants such as benzodiazepines.
The grief of those who lose loved ones to epilepsy is often reportedly greater because they had not realized it could happen. Epilepsy in children in association with underlying neurological disease undoubtedly poses an increased mortality risk and families should be sensitively advised accordingly: in our experience, families are often already aware and accepting of these risks. Children with idiopathic epilepsies can be reassured that their mortality risk largely relates to addressable risk factors such as unsupervised bathing.
We are grateful to Mary Bythell, Marjorie Renwick, Martin Ward-Platt, Sandeep Dhamaraj, and Louise Potts for support and assistance with the sample X study, and the CEMACH office in granting access to sample Y data.