Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
n e w e ng l a n d j o u r na l
of
m e dic i n e
original article
A BS T R AC T
Background
From the Departments of Pediatric Neurology and Public Health, University of
Turku and Turku University Hospital
both in Turku, Finland (M.S.); and the
Departments of Neurology, Pediatrics,
and Epidemiology and Population Health
and the Comprehensive Epilepsy Management Center, Montefiore Medical Center,
Albert Einstein College of Medicine, Bronx,
NY (S.S.). Address reprint requests to Dr.
Shinnar at the Comprehensive Epilepsy
Management Center, Montefiore Medical
Center, 111 E. 210th St., Bronx, NY 10467,
or at sshinnar@montefiore.org.
N Engl J Med 2010;363:2522-9.
Copyright 2010 Massachusetts Medical Society.
There are few studies on long-term mortality in prospectively followed, well-characterized cohorts of children with epilepsy. We report on long-term mortality in a
Finnish cohort of subjects with a diagnosis of epilepsy in childhood.
Methods
Sixty subjects died (24%); this rate is three times as high as the expected age- and
sex-adjusted mortality in the general population. The subjects who died included 51
of 107 subjects (48%) who were not in 5-year terminal remission (i.e., 5 years
seizure-free at the time of death or last follow-up). A remote symptomatic cause of
epilepsy (i.e., a major neurologic impairment or insult) was also associated with an
increased risk of death as compared with an idiopathic or cryptogenic cause (37% vs.
12%, P<0.001). Of the 60 deaths, 33 (55%) were related to epilepsy, including sudden,
unexplained death in 18 subjects (30%), definite or probable seizure in 9 (15%),
and accidental drowning in 6 (10%). The deaths that were not related to epilepsy
occurred primarily in subjects with remote symptomatic epilepsy. The cumulative
risk of sudden, unexplained death was 7% at 40 years overall and 12% in an analysis that was limited to subjects who were not in long-term remission and not receiving
medication. Among subjects with idiopathic or cryptogenic epilepsy, there were no
sudden, unexplained deaths in subjects younger than 14 years of age.
Conclusions
2522
Me thods
Subjects and Study Design
The study population included all children younger than 16 years of age who were living in the
catchment area of Turku University Hospital,
Turku, Finland, at the end of 1964 and who had
epilepsy, which was defined as at least two unprovoked seizures. The subjects were identified
from the files of Turku University Hospital, other
hospitals and institutions in southern Finland, and
private surgeons and from Finlands National
Health Service registry, which covers the entire
population of that country. Of the 245 patients
identified, the majority (223 patients) were seen
in Turku University Hospital. In the 1960s, all
pediatric patients who had one or more epileptic
seizures were referred for inpatient hospital evaluation. The study population has been described
previously.20-22
The 245 subjects included 150 subjects with
incident cases of epilepsy (61%) whose initial visit
for the evaluation of new-onset seizures occurred
between 1961 and 1964. The remaining 95 subjects with prevalent cases of epilepsy (39%) were
seen before 1961, but they were seen again during the 19611964 study period for active epilepsy
(at least one epileptic seizure during the preceding 3 years and a prior diagnosis of epilepsy). Excluded were children who had febrile seizures only
or other acute symptomatic seizures and children
with a single unprovoked seizure. Also excluded
were children with an onset of epilepsy before
1961 who were either in remission or who died
before 1961.
A follow-up examination was performed every
fifth year up to 2002. Data on deaths were obtained from continuous follow-up of the patients
2523
The
n e w e ng l a n d j o u r na l
of
m e dic i n e
Variable
Total deaths no.
All Subjects
(N=245)
60
15
45
Age at death yr
Median
23
26
21
Range
150
1150
149
8692
4638
4054
All
6.90 (5.38.9)
3.23 (1.95.4)
11.10 (8.314.9)
Men
7.33 (5.210.2)
2.69 (1.26.0)
11.63 (8.016.8)
Women
6.41 (4.49.4)
3.74 (1.97.2)
10.33 (6.416.6)
51/60 (85)
11/15 (73)
40/45 (89)
No. of person-yr
No. of deaths/1000 person-yr (95% CI)
9/60 (15)
4/15 (27)
5/45 (11)
nitive impairment, clinically significant developmental delay [developmental quotient <70 in children younger than 7 years of age], cerebral palsy,
or autism) or a history of a major neurologic insult (e.g., head trauma, stroke, or meningitis). The
remaining subjects were determined to have cryptogenic or idiopathic epilepsy. The classification
of epilepsy was made in all subjects, and the detailed seizure types and epilepsy syndromes in this
cohort have been reported previously.22 These classifications were determined before the new classifications of epilepsy syndromes26 and are used
here for consistency with previous studies.
Sudden, unexpected death in a person with
epilepsy was defined, in accordance with the definition by Annegers,19 as sudden death with no
evidence of a seizure and no other identified cause
of death, with confirmation of these findings by
autopsy. In almost all cases of sudden, unexplained death in a person with epilepsy, pulmonary edema was detected on autopsy. We also examined the rates of sudden, unexplained death
among persons with epilepsy in accordance with
the alternative definition by Nashef,27 which
includes deaths for which there is evidence of
probable or definite seizures, although it excludes
deaths in persons with known status epilepticus.
2524
Statistical Analysis
Our statistical methods took into account the timedependent nature of the mortality data. The rate
of death was calculated as the number of deaths
divided by the person-years at risk during the
study; direct adjustment for age and sex was performed on the basis of the 2002 Finnish population. The product-limit method was used to calculate the risk of death from the onset of
epilepsy.28,29 Data on subjects were censored at
the time of the last follow-up or on January 1,
2003. Standard errors and 95% confidence intervals were calculated with the use of a modification of the Greenwood formula.29 Univariate and
multivariate analyses were performed with the use
of the Cox proportional-hazards model, and remission status was treated as a time-dependent covariate.29-31 Rates of death according to remission
status were adjusted for the number of personyears at risk during the remission period. Rate
ratios with 95% confidence intervals were calculated with the use of the Cox regression models.
When the cohort with incident epilepsy and the
cohort with prevalent epilepsy were combined, adjustment was made for the left truncation due to
the delayed entry of the cohort with prevalent epilepsy.32 A direct age- and sex-adjusted standard
Variable
Total deaths no.
Death related to epilepsy no./total
no. of deaths (%)
60
15
45
33/60 (55)
9/15 (60)
24/45 (53)
Drowning no.
18
11
26 (43)
6 (40)
20 (44)
12
12
Suicide no.
* A remote symptomatic cause of epilepsy indicates epilepsy associated with a major neurologic abnormality or insult.
R e sult s
Characteristics of the Study Population
abnormality or insult). The median duration of epilepsy before achieving 5-year long-term remission
was 17 years (range, 1 to 52) for surviving subjects and 22 years (range, 1 to 50) for those who
died during the follow-up period. At the time of
the last follow-up contact or death, 110 patients
(45%) were in 5-year terminal remission and were
not receiving medication, 28 (11%) were in 5-year
terminal remission and were receiving medication, and 107 (44%) were not in 5-year remission.
Overall Mortality
2525
The
n e w e ng l a n d j o u r na l
50
40
Remote symptomatic cause
30
20
Idiopathic or cryptogenic cause
10
10
15
20
25
30
35
40
of
m e dic i n e
Univariate Analysis
Hazard Ratio
(95% CI)
Multivariate Analysis
P Value
Hazard Ratio
(95% CI)
P Value
5.3 (2.611.0)
<0.001
4.7 (1.514.9)
0.007
3.4 (1.96.1)
<0.001
1.5 (0.73.6)
0.31
1.9 (1.23.2)
0.01
1.5 (0.73.0)
0.28
1.4 (0.82.3)
0.20
1.7 (0.83.5)
0.13
6.4 (2.218.8)
<0.001
4.7 (1.514.9)
0.007
3.1 (1.46.7)
0.004
1.5 (0.73.6)
0.31
2.1 (1.14.2)
0.03
1.5 (0.73.0)
0.28
1.9 (0.963.8)
0.06
1.7 (0.83.5)
0.13
5.2 (1.418.5)
0.01
5.0 (1.220.1)
0.02
2.8 (1.17.0)
0.03
2.6 (0.97.5)
0.07
2.1 (0.85.5)
0.11
1.9 (0.75.2)
0.20
1.9 (0.74.8)
0.20
0.8 (0.32.4)
0.72
Localization-related epilepsy
0.8 (0.32.0)
0.60
0.5 (0.21.5)
0.21
All deaths
Epilepsy-related deaths
* A Cox proportional-hazards model was used to assess the risk of death associated with an absence of 5-year terminal
remission, with remission status treated as a time-dependent covariate. A remote symptomatic cause of epilepsy indicates epilepsy associated with a major neurologic abnormality or insult. CI denotes confidence interval.
Discussion
The strength of this study is that it involves a cohort that was prospectively followed for 40 years,
2527
The
n e w e ng l a n d j o u r na l
0.5
0.4
0.3
All epilepsy-related deaths
0.2
0.1
Sudden, unexplained deaths
0.0
10
15
20
25
30
35
40
of
m e dic i n e
among children younger than 14 years of age who lower among those who become seizure-free after
had idiopathic or cryptogenic epilepsy; however, surgery than among those who do not,35 sugthe risk of sudden, unexplained death and other gesting that the risk is potentially modifiable.
epilepsy-related deaths was relatively high among
adults with active epilepsy.
Supported by a grant from the Finnish Epilepsy Research
Our data alone do not provide support for ag- Foundation.
Disclosure forms provided by the authors are available with
gressive treatment to prevent sudden, unexplained the full text of this article at NEJM.org.
death in patients with epilepsy. However, studies
We thank Hannu Kalimo, M.D., Ph.D., professor of neuropaof epilepsy surgery in adults with medically re- thology, and Pekka Saukko, M.D., Ph.D., professor of forensic
medicine, for collecting the autopsy data; Hans Helenius, M.Sc.,
fractory partial epilepsy have shown that the rates for statistical advice; Olli Kaleva, B.A., for computations; and
of sudden, unexplained death are substantially Dale Hesdorffer, Ph.D., for her constructive suggestions.
References
1. Leestma JE, Walczak T, Hughes JR,
2529