This is the first paediatric population-based study reporting mortality beyond 30 days following an episode of CSE. The main findings of our study are that: (i) children with CSE are 46 times more likely to die within 8 years after their CSE compared with the rest of the unaffected population; (ii) a history of pre-existing neurological impairment, with or without accompanying epilepsy, at the time of CSE (remote or cryptogenic CSE) is a major risk factor for death with no evidence that other clinical or demographic factors increase risk of death; (iii) there were no deaths following prolonged febrile convulsions and idiopathic CSE; (iv) cardio-respiratory complications and/or intractable seizures are the most commonly reported causes of death; and (v) the attributable role of CSE itself on mortality, rather than the underlying cause of CSE, could be less than generally perceived.
Although the overall mortality in our cohort was 46 times higher than expected in the reference population, this effect was predominantly seen in children with pre-existing clinically significant neurological impairments at the time of CSE (remote symptomatic and cryptogenic CSE). Almost all deaths in children with acute symptomatic CSE occurred within 30 days of their episode of CSE, and those who survive beyond this period were not at a significantly increased risk of death compared with the reference population.
The overall risk of death following CSE was almost seven times higher in children with pre-existing neurological impairments (motor and/or cognitive) compared with those who were previously neurologically normal; and this risk increased to 19 times in those who survived beyond 30 days of their episode of CSE. All deaths in previously neurologically normal children occurred in those with acute symptomatic CSE; three children with acute bacterial meningitis died during the acute episode and one with viral encephalitis died 5 years later. There were no deaths in children following prolonged febrile convulsions, an observation similar to the previous reports of very low risk of mortality following prolonged febrile convulsions, and similar to the mortality in simple febrile seizures (Verity et al., 1993
; Logroscino et al., 2002
; Vestergaard et al., 2008
We found that none of the clinical and demographic factors at the time of CSE were independent predictors of subsequent mortality except pre-existing neurological impairment. Previous studies have reported higher mortality in individuals with prolonged status epilepticus (>60
min), continuous status epilepticus (compared with intermittent) and higher age, but this observation was valid only in adults and not for the paediatric population (Towne et al., 1994
; DeLorenzo et al., 1999
; Waterhouse et al., 1999
). Also, none of the previously reported predictors of short-term mortality following CSE in children, such as acute symptomatic cause of CSE, young age at onset, longer duration of CSE and focal-onset of CSE, were independently associated with mortality beyond 30 days in our cohort (Raspall-Chaure et al., 2006
; Sadarangani et al., 2008
). The observed differences in our study and other reported hospital-based studies may at least in part be related to the inclusion of more severe cases in hospital-based studies. The frequency of CSE and medication non-compliance may also potentially increase overall mortality. STEPSOUT is an ongoing study and the surviving cohort members are currently being recruited to determine the neurological, neuroimaging, cognitive and behavioural outcomes. The details on medication compliance, epilepsy control and frequency of CSE during follow-up are not available on all cohort members and we were not therefore able to adjust for these factors in our modelling.
We report the cause(s) of death in our cohort as stated on their death certificates. Most deaths within 30 days of an episode of CSE were as a consequence of complications of the underlying cause, such as acute bacterial meningitis or a progressive neurological disorder. Even among the 30-day survivors, the majority of deaths seem to have been associated with complications of their underlying brain disorder, rather than as a direct consequence of CSE. Only four deaths were seizure related; three as a result of CSE and one as a result of intractable seizures. There were no sudden unexpected deaths in epilepsy (SUDEP) or accidental deaths. This is in contrast to the observation by Aicardi and Chevrie (1970)
, who reported that about half the deaths in their cohort with CSE were seizure related. There are several possible explanations for this observed difference. The first was that the Aicardi study (1970)
was hospital based, and CSE was defined as continued seizure activity lasting at least 60
min rather than the 30
min used in our study. Thus, the children in the Aicardi (1970)
study may reflect children with more severe conditions compared with our cohort. It is also possible that improvements in acute management of CSE in children in recent years may explain reduced risk of death as a direct consequence of CSE. Another explanation may be inaccuracies in the cause of death as stated on the death certificates as only 3 out of 23 in our study had autopsies, and the stated cause of death on certificates may not reflect circumstances surrounding death (Sillanpaa and Shinnar, 2010
). It is possible that the actual cause of death may be due to seizure activity although not listed as such, e.g. unobserved seizure activity may lead to aspiration and subsequent respiratory failure.
As the main predictor of mortality is presence of pre-existing clinically significant neurological impairment at the time of CSE, it becomes important to ask whether CSE itself has any additional impact on subsequent mortality. There are no reported paediatric studies designed to address the independent influence of CSE on long-term mortality. Ideally, a direct comparison of the mortality in children with a pre-existing significant neurological disorder and no CSE, with a similar population of children with CSE, or mortality between idiopathic epilepsy without CSE and with CSE may help to determine the additional impact of CSE on mortality. However, even such an approach may not provide a satisfactory answer to this important question as the outcome may still be confounded by the severity of the underlying structural/functional brain disorder and CSE itself could be a marker of the severity.
An indirect way of analysing the attributive role of CSE on mortality could be by comparing the mortality in idiopathic CSE in our cohort with the reported mortality in children with idiopathic epilepsy, and mortality in children with neurological impairments in our cohort with that of remote symptomatic epilepsy and cerebral palsy as these are the main groups that make up the subcategory of children with pre-existing neurological problem in our cohort. There were no deaths in children who had idiopathic CSE in our cohort (SMR 0, 95% CI 0.0–3.68), an observation similar to the low mortality reported in children with idiopathic epilepsy (Harvey et al., 1993
; Callenbach et al., 2001
; Camfield et al., 2002
; Berg et al., 2004
; Geerts et al., 2010
; Sillanpaa and Shinnar, 2010
). The SMR for children with significant neurological impairment prior to their episode of CSE in our study was 91.4 (95% CI 53.2–146.3). In comparison, reported SMRs in children with symptomatic epilepsies range between 22.9 (95% CI 7.9–37.9) and 49.7 (95% CI 31.7–77.9) (Harvey et al., 1993
; Callenbach et al., 2001
; Berg et al., 2004
; Geerts et al., 2010
). The mortality in our cohort with significant neurological impairment prior to CSE appears to be higher than that reported for remote symptomatic epilepsy. However, children classified as having remote symptomatic epilepsy by definition do not necessarily have major neurological impairments and therefore the comparison with our group is not very robust (Engel, 2001
). The mortality in our cohort is comparable with the reported mortality in children with cerebral palsy, a cohort likely to have similar neurological impairments (SMR 36.4 for mild and 102.8 for severe cerebral palsy, 95% CI not provided); however, a significant proportion of children with severe cerebral palsy probably had epilepsy and some also experienced episodes of CSE (Strauss et al., 1999
). All these comparisons need to be interpreted with caution, as they do not take into consideration the heterogeneity of the study populations, the severity of underlying conditions in different cohorts and the unreported possible confounders affecting the outcome.
Long-term follow-up studies have reported no significant difference in mortality in children with epilepsy with or without status epilepticus (Sillanpaa and Shinnar, 2002
; Stroink et al., 2007
). Although a history of prior CSE was associated with increased long-term mortality on univariable analysis in some studies, this was no longer significant after adjusting for other clinical factors, and the underlying diagnosis was the major predictor of mortality (Callenbach et al., 2001
; Camfield et al., 2002
; Sillanpaa and Shinnar, 2002
; Berg et al., 2004a
The available data therefore suggest that CSE does not substantially increase the risk of death in neurologically normal children who survive the acute episode of CSE, and any impact it may have on increasing long-term mortality is seen primarily in children with significant underlying brain disorder. This, however, needs confirmation from studies designed to investigate the additional impact of CSE on mortality without being confounded by severity of underlying brain disorder, and a possible approach could be by comparing the mortality associated with short seizures against that with CSE in children without an identifiable underlying aetiology (e.g. febrile seizures, idiopathic epilepsy). One such study done in adult population showed a non-significant 2.4-fold increased risk of death at 10 years in incident idiopathic/cryptogenic CSE compared with short seizures (adjusted relative risk 2.4, 95% CI 0.9–6.3) (Logroscino et al., 2008
). However, the study was limited by a small sample size, and the increased risk was restricted to the elderly and those who later developed epilepsy.
Our study is the largest prospective paediatric population-based cohort of CSE in resource-rich countries. Despite this, it is possible that lack of statistical significance for some of the clinical and demographic factors could be due to our modest sample size. As is inevitable in follow-up studies, there will be some degree of attrition but our 9% attrition is within the acceptable range for follow-up cohort studies (Fewtrell et al., 2008
). Furthermore, the children who were lost to follow-up were similar to those for whom we had follow-up data suggesting that our results are representative of the entire CSE cohort. This study, from a resource-rich country, may well not reflect the longer term mortality in a resource-poor setting where it is already clear that the mortality within 3 years of an episode of CSE is substantially higher (Sadarangani et al., 2008
In summary, our study provides novel information that will be useful for prognostication. There is a substantially increased risk for death within 8 years after CSE with having pre-existing clinically significant neurological disorder at the time of CSE being the major predictor. The risk of death is not increased in children with prolonged febrile convulsions and idiopathic CSE, and in those with acute symptomatic aetiology who survive the acute episode of CSE. Most deaths during follow-up occur as a complication of underlying brain disorder rather than seizures themselves. In addition, the study has generated the hypothesis that there is minimal additional impact of CSE itself on mortality. Case–control studies specifically designed to investigate this hypothesis are now required. Although it may be feasible to detect a significant difference in mortality among those with and without CSE, the challenge is identifying an appropriate control population. As CSE is associated with increased morbidity as well as mortality, research into morbidity following CSE in childhood is needed.