You are on page 1of 8

American Journal of Epidemiology

Copyright 2007 by the Johns Hopkins Bloomberg School of Public Health


All rights reserved; printed in U.S.A.

Vol. 165, No. 8


DOI: 10.1093/aje/kwk086
Advance Access publication January 30, 2007

Original Contribution
The Long-Term Risk of Epilepsy after Febrile Seizures in Susceptible Subgroups

Mogens Vestergaard1,2, Carsten Bcker Pedersen3, Per Sidenius4, Jrn Olsen2,5, and Jakob
Christensen4,6
1

Department of General Practice, Institute of Public Health, Aarhus University, Aarhus, Denmark.
Department of Epidemiology, Institute of Public Health, Aarhus University, Aarhus, Denmark.
3
National Centre for Register-based Research, Aarhus University, Aarhus, Denmark.
4
Department of Neurology, Aarhus University Hospital, Aarhus, Denmark.
5
Department of Epidemiology, School of Public Health, University of California Los Angeles, Los Angeles, CA.
6
Department of Clinical Pharmacology, Aarhus University, Aarhus, Denmark.
2

A family history of seizures, preexisting brain damage, or birth complications may modify the long-term risk of
epilepsy after febrile seizures. The authors evaluated the association between febrile seizures and epilepsy in
a population-based cohort of 1.54 million persons born in Denmark (19782002), including 49,857 persons with
febrile seizures and 16,481 persons with epilepsy. Overall, for children with febrile seizures compared with those
without such seizures, the rate ratio for epilepsy was 5.43 (95% condence interval: 5.19, 5.69). The risk remained
high during the entire follow-up but was particularly high shortly after the rst febrile seizure, especially in children
who experienced early (<1 year of age) or late (>3 years of age) onset of febrile seizures. At 23 years of follow-up,
the overall cumulative incidence of epilepsy after febrile seizures was 6.9% (95% condence interval: 6.5, 7.3). In
conclusion, persons with a history of febrile seizures had a higher rate of epilepsy that lasted into adult life, but less
than 7 percent of children with febrile seizures developed epilepsy during 23 years of follow-up. The risk was higher
for those who had a family history of epilepsy, cerebral palsy, or low Apgar scores at 5 minutes.
Apgar score; birth weight; cerebral palsy; Denmark; epilepsy; follow-up studies; premature birth; seizures, febrile

Abbreviations: CI, condence interval; ICD-8, International Classication of Diseases, Eighth Revision; ICD-10, International
Classication of Diseases, Tenth Revision.

Febrile seizures are the most common type of seizures in


childhood, affecting 25 percent of all children, but the
long-term impact of these seizures on the developing brain
is unsettled. Retrospective studies of persons with intractable temporal lobe epilepsy (1, 2), neuroimaging studies of
those with prolonged febrile seizures (2, 3), and experimental studies of laboratory animals with hyperthermia-induced
seizures (4, 5) suggest that some febrile seizures may cause
epilepsy. These results, however, may apply to only a small
proportion of children with febrile seizures. Populationbased studies show that the absolute risk of unprovoked

seizures after febrile seizures is low (27 percent) (610)


and indicate that most febrile seizures are age-specific
markers of seizure susceptibility. Early identification of susceptible subgroups could have clinical implications because
these children may need special attention (11). It has been
suggested that preexisting brain damage, family history of
seizures, and early environmental factors may increase the
susceptibility to seizure-induced brain lesions (11, 12).
However, no studies so far have been large enough to evaluate whether the risk of epilepsy after febrile seizures is
modified by these factors.

Correspondence to Dr. Mogens Vestergaard, Department of General Practice, Institute of Public Health, Aarhus University, Vennelyst Boulevard
6, DK-8000 Aarhus C, Denmark (e-mail: mv@soci.au.dk).

911

Am J Epidemiol 2007;165:911918

Downloaded from http://aje.oxfordjournals.org/ by guest on August 4, 2012

Received for publication January 12, 2006; accepted for publication October 5, 2006.

912 Vestergaard et al.

We evaluated the association between febrile seizures and


epilepsy in a large population-based birth cohort with up to
25 years of follow-up. We examined whether age at first
febrile seizure, time since first febrile seizure, number of
febrile seizures, family history of febrile seizures, family
history of epilepsy, cerebral palsy, low Apgar score, low
birth weight, and preterm birth modify the association between febrile seizures and epilepsy.
MATERIALS AND METHODS
Study population

Study variables

RESULTS

Information on febrile seizures, cerebral palsy, and epilepsy for the study population, their parents, and their siblings was obtained from the National Hospital Register (14),
which contains data on all discharges from Danish hospitals
since 1977. Diagnoses resulting from visits to emergency
rooms and outpatient clinics have been included in the register since 1995. Diagnostic information in the National
Hospital Register is based on the International Classification of Diseases, Eighth Revision (ICD-8) from 1978 to
1993 and the International Classification of Diseases, Tenth
Revision (ICD-10) from 1994 to 2002. We classified children as having a febrile seizure when they were registered
with ICD-8 code 780.21 or ICD-10 code R56.0, were between 3 months and 5 years of age at the time of discharge,
and had no recorded history of epilepsy or intracranial infection (ICD-8 codes 320323; ICD-10 codes G00G09).
Persons were categorized as having epilepsy when they
were registered with ICD-8 code 345 or ICD-10 code
G40 or G41 and as having cerebral palsy when they were
registered with ICD-8 code 343.99 or 344.99 or ICD-10
code G80. Information on Apgar score at 5 minutes, birth
weight, and gestational age at birth was obtained from the
Danish Medical Birth Register.

In the cohort of 1,540,725 persons, 49,857 children were


diagnosed with febrile seizures and 16,481 persons were
diagnosed with epilepsy, including 2,149 who developed
epilepsy after febrile seizures. Overall, for persons with
a history of febrile seizures compared with those without
such seizures, the rate ratio for epilepsy was 5.98 (95 percent confidence interval (CI): 5.71, 6.26) when we adjusted
for age and its interaction with gender and calendar year.
The rate ratio decreased slightly after further adjustments
for history of febrile seizures in siblings, epilepsy in parents
or siblings, cerebral palsy, Apgar score at 5 minutes, birth
weight, and gestational age at birth (rate ratio 5.43, 95
percent CI: 5.19, 5.69) (table1).
The association between febrile seizures and epilepsy
was modified by time since first febrile seizure. The rate
of epilepsy was 26-fold higher during the first 3 months
after onset of febrile seizure and declined to a threefold
higher rate 8 years after the first febrile seizure and throughout the study period (figure 1). The rate ratios shown in
figure 1 were adjusted for several factors, but the adjustments caused little change in the estimate of interest (data
not presented). Furthermore, excluding persons with a history of febrile seizure in siblings, epilepsy in parents or
siblings, cerebral palsy, low Apgar score (<7), low birth
weight (<2,500 g), and low gestational age (<37 weeks)
had no impact on the results presented in figure 1 (11,419
persons developed epilepsy during 14,498,957 person-years
at risk) (data not presented). Information on visits to emergency rooms and outpatient clinics was not included in the
National Hospital Register until 1995. However, we found
little change in the estimate of interest when restricting the
cohort to children born between 1995 and 2002 (2,712 persons developed epilepsy during 2,134,875 person-years at
risk) (figure 2).

Statistical analyses

The study population was followed from birth until onset


of epilepsy, death, emigration, or December 31, 2002,
whichever occurred first. The rate ratio of epilepsy was
estimated by log-linear Poisson regression (15) with the
GENMOD procedure in SAS version 8.1 software (SAS
Institute, Inc., Cary, North Carolina). All rate ratios were
adjusted for age and its interaction with gender and calendar
year. Age, calendar year, number of febrile seizures, time

Am J Epidemiol 2007;165:911918

Downloaded from http://aje.oxfordjournals.org/ by guest on August 4, 2012

We used data from the Danish Civil Registration System


(13) to identify all persons born in Denmark between January
1, 1978, and December 31, 2002 (1,540,725 persons). All
liveborn children and new residents of Denmark are assigned a unique personal identification number, which is
stored in the Danish Civil Registration System together with
information on vital status, emigration, disappearance, and
personal identification numbers of mothers, fathers, and siblings. The personal identification number is used as a key to
individual information in all other national registries.

since first febrile seizure, history of epilepsy in a parent or


sibling, and history of febrile seizures in siblings were treated
as time-dependent variables, whereas all other variables
were treated as variables independent of time. Time since
first febrile seizure was classified in 3-month periods from
0 to 1 year, in 6-month periods from 1 to 5 years, in 1-year
periods from 5 to 20 years, and as 20 or more years. Age was
categorized in 3-month periods from 0 to 11 months, in
1-year periods from 1 to 21 years, and as 22 or more years.
Calendar year was categorized in 1-year periods. p values
were based on likelihood ratio tests, and 95 percent confidence limits were calculated by Walds test.
The cumulative incidence was calculated as the proportion of persons registered with epilepsy in the population at
a given time. We accounted for death by other causes by
using competing-risk Cox regression (16). Because the association between febrile seizures and epilepsy depends
strongly on time since first febrile seizure, we used this time
scale in the competing-risk Cox regression among persons
with a history of febrile seizures. By contrast, we estimated
the cumulative incidence according to age among children
with no history of febrile seizures.

Long-Term Risk of Epilepsy after Febrile Seizures

913

TABLE 1. Estimated rate ratios of epilepsy according to febrile seizures and the risk factors investigated
based on data for the 1,540,725 persons born in Denmark between 1978 and 2002 and followed for epilepsy
until 2002

Overall

First adjustmenty

Full adjustmentz

Febrile
seizures
in study
subject

No. of personyears at risk

No. of
cases*

No

17,496,653

14,332

1.00

Reference

1.00

Reference

497,477

2,149

5.98

5.71, 6.26

5.43

5.19, 5.69

17,016,095

13,790

1.00

Reference

1.00

Reference

Yes

Rate
ratio

95% CI

Rate
ratio

95% CI

Febrile seizures in siblings


No

No
Yes

444,736

1,914

6.04

5.75, 6.34

5.62

5.35, 5.90

Yes

No

480,559

542

1.41

1.29, 1.53

1.31

1.20, 1.43

Yes

52,740

235

6.17

5.42, 7.03

5.15

4.52, 5.87
Reference

Epilepsy in parents or siblings


No

17,111,777

13,365

1.00

Reference

1.00

Yes

479,623

1,974

6.01

5.73, 6.30

5.60

5.34, 5.88

No

384,876

967

3.37

3.15, 3.60

3.22

3.02, 3.44

Yes

17,854

175 14.41 12.41, 16.73 12.70 10.93, 14.77

Epilepsy after febrile seizures


in the same sibling{
No
Yes

No

17,478,619

14,272

1.00

Reference

1.00

Reference

Yes

495,142

2,123

5.96

5.69, 6.24

5.62

5.36, 5.89

No

18,034

60

4.27

3.32, 5.51

4.21

3.26, 5.42

Yes

2,335

26 15.65 10.65, 23.00 14.80 10.07, 21.75

Cerebral palsy
No
Yes

No

17,467,858

13,838

1.00

Reference

1.00

Reference

Yes

495,303

2,042

5.95

5.68, 6.24

5.65

5.38, 5.93

No

28,796

494 24.71 22.58, 27.04 16.94 15.40, 18.64

Yes

2,173

107 71.22 58.87, 86.16 49.24 40.60, 59.70

Apgar score at 5 minutes


7
<7

No

17,142,887

13,649

1.00

Reference

1.00

Reference

Yes

487,423

2,085

6.09

5.81, 6.38

5.59

5.33, 5.86

No

97,074

370

4.69

4.23, 5.20

2.97

2.67, 3.30

Yes

3,337

7.20

5.19, 9.97

1.00

Reference

37 15.68 11.35, 21.65

Birth weight (g)


2,500

No
Yes

461,356

1,927

6.09

5.80, 6.39

5.74

5.46, 6.03

<2,500

No

856,897

1,376

2.08

1.97, 2.20

1.54

1.43, 1.65

Yes

33,468

212

9.34

8.15, 10.70

5.61

4.85, 6.49

15,407,024

12,236

1.00

Reference

1.00

Reference

16,525,191

12,827

1.00

Reference

Gestational age (weeks) at birth


37

No
Yes

434,971

1,841

6.07

5.78, 6.38

5.68

5.40, 5.97

<37

No

796,126

1,226

1.89

1.78, 2.00

1.22

1.13, 1.32

Yes

31,027

203

9.35

8.14, 10.75

4.96

4.28, 5.76

* Because of missing values, the numbers of cases of epilepsy in each category do not always add to 16,481.
y Estimates were adjusted for calendar period and for age and its interaction with gender.
z Estimates were adjusted for calendar period, age and its interaction with gender, febrile seizures in siblings,
epilepsy in parents or siblings, cerebral palsy, Apgar score, birth weight, and gestational age.
CI, condence interval.
{ These estimates were adjusted for neither epilepsy in parents or siblings nor febrile seizure in a sibling.

Am J Epidemiol 2007;165:911918

Downloaded from http://aje.oxfordjournals.org/ by guest on August 4, 2012

Yes

No

914 Vestergaard et al.

Children who experienced early (<1 year) or late (>3


years) onset of febrile seizures had a higher rate of epilepsy
within the 2 years after the first seizure compared with children whose onset occurred between ages 1 and 3 years. Age
at onset of febrile seizures had no effect on the rate of epilepsy 2 or more years after the first febrile seizures (figure 3).
The rate ratio for epilepsy after febrile seizures increased
with increasing number of admissions for febrile seizures.
The rate ratios were 4.52 (95 percent CI: 4.27, 4.79), 8.10
(95 percent CI: 7.36, 8.92), and 19.67 (95 percent CI: 17.88,
21.65) for children registered one, two, and three or more
times with febrile seizures compared with children with no
history of febrile seizures.
Persons with a family history of epilepsy, a family history
of febrile seizures, cerebral palsy, an Apgar score of less
than 7 at 5 minutes, a birth weight of less than 2,500 g, or
a gestational age at birth of less than 37 completed weeks
had a higher rate of epilepsy than that for the general population (table 1). After we adjusted for several confounding
factors, the rate ratio of epilepsy was particularly high for
persons with both a history of febrile seizures and a family
history of epilepsy, cerebral palsy, or low Apgar score. The
combined effect seemed to be higher than the sum but lower
than the product of the two separate effects (table 1).
The cumulative incidence of epilepsy after febrile seizures was 1.3 percent (95 percent CI: 1.2, 1.4), 3.2 percent
(95 percent CI: 3.0, 3.4), 4.6 percent (95 percent CI: 4.4,

4.8), 5.6 percent (95 percent CI: 5.3, 5.8), 6.4 percent (95
percent CI: 6.1, 6.7), and 6.9 percent (95 percent CI: 6.5,
7.3) 1, 5, 10, 15, 20, and 23 years after the first febrile
seizures, respectively. Among persons with no history of
febrile seizures, the cumulative incidence of epilepsy was
1.8 percent (95 percent CI: 1.8, 1.9) at the age of 25 years,
that is, the mean age of persons followed up for 23 years
after the first febrile seizures. The cumulative incidence
varied according to the subgroups under study. The highest
risk was found among children with cerebral palsy, Apgar
score of less than 7 at 5 minutes, or a family history of
epilepsy, especially in siblings of children in whom epilepsy
developed after febrile seizures (table 2).
DISCUSSION

We found that febrile seizure was associated with an increased risk of epilepsy that lasted into adulthood. The rate
was highest shortly after the first febrile seizure, especially
for children who experienced early (<1 year) or late onset
(>3 years) of febrile seizures. The overall cumulative incidence of epilepsy after febrile seizures was 6.9 percent at
23 years of follow-up. The risk was particular high for persons with cerebral palsy, low Apgar scores, or a family history of epilepsy.
The association between febrile seizures and epilepsy
may be explained by at least four different mechanisms.
Am J Epidemiol 2007;165:911918

Downloaded from http://aje.oxfordjournals.org/ by guest on August 4, 2012

FIGURE 1. Adjusted rate ratios of epilepsy after febrile seizures according to time since rst febrile seizure among persons born in Denmark.
Estimates were based on data for the 1,540,725 persons born in Denmark between 1978 and 2002 and followed until 2002. Rate ratios were
adjusted for age, gender, calendar period, history of febrile seizures in siblings, history of epilepsy in parents or siblings, cerebral palsy, birth weight,
gestational age at birth, and Apgar score at 5 minutes. Point estimates are given, with error bars representing 95% condence intervals.

Long-Term Risk of Epilepsy after Febrile Seizures

915

First, seizures occurring during fever may be the first manifestation of epilepsy and not febrile seizures. Dravets syndrome (severe myoclonic epilepsy of infancy) is an example
of a rare condition that appears within the first year of life,
often during a febrile episode and with unprovoked seizures
following shortly after (17). Misclassification of epilepsy as
febrile seizures could explain at least some of the high rate
of epilepsy we and others (6) found shortly after the first
febrile seizures. It is difficult to avoid this misclassification
in the clinical setting because no other factors besides fever
may separate the two conditions. However, our data indicate
that children may need special attention if the first febrile
seizure occurs at an uncommon age: before 1 year of age or
after 3 years of age (18). Others have found a higher rate of
epilepsy associated with febrile seizures occurring before
1 year of age, perhaps because the more severe febrile seizures tend to occur early (6, 7).
Second, febrile seizure may be an age-specific marker of
seizure susceptibility if febrile seizures and epilepsy share
causes. We adjusted for age, gender, calendar period, history
of febrile seizures in siblings, family history of epilepsy,
cerebral palsy, birth weight, gestational age at birth, and
Apgar score at 5 minutes and found little change in the
estimates, but other confounders may be present. Genetic
analyses of familial epilepsies have identified mutations in
ion channel genes that result in a wide range of phenotypes
from febrile seizures to severe forms of childhood epilepsy
Am J Epidemiol 2007;165:911918

(19). We adjusted for family history of seizures, but that is


only a crude proxy for genetic susceptibility. We found that
siblings of children with febrile seizures had a higher rate of
epilepsy even if they had no personal history of febrile seizures, indicating that genetic and environmental factors
shared by family members may play a role.
Third, prolonged febrile seizures may damage the developing brain and cause epilepsy. Neuroimaging studies have
demonstrated acute swelling and edema of the hippocampus
after prolonged febrile seizures (2, 3), and animal studies
suggest that hyperthermia-induced seizures may cause longlasting modifications of channels, synapses, and neuronal
networks within the hippocampus, leading to sustained dysfunction of these cells and a decreased seizure threshold
(4, 5). One would expect a particularly high risk of complex
partial epilepsy after febrile seizures if febrile seizure damages the hippocampus. Unfortunately, we did not have sufficiently detailed information on type of epilepsy to address
this discussion. However, the rate of epilepsy after febrile
seizures remained high into adulthood, where partial seizures are the most common type of seizures. Our data and
those of others (6, 7, 9) indicate that the rate of epilepsy
increases with increasing number of febrile seizures. A
dose-response-like pattern need not imply causality, however. (20) The number of febrile seizures may be a marker of
seizure susceptibility, and children admitted several times
for febrile seizures may be a selected group experiencing

Downloaded from http://aje.oxfordjournals.org/ by guest on August 4, 2012

FIGURE 2. Adjusted rate ratios of epilepsy after febrile seizures according to time since rst febrile seizure. Estimates were based on data for all
children born in Denmark between 1995 and 2002 and followed until 2002. Rate ratios were adjusted for age, gender, calendar period, history of
febrile seizures in siblings, history of epilepsy in parents or siblings, cerebral palsy, birth weight, gestational age at birth, and Apgar score at
5 minutes. Point estimates are given, with error bars representing 95% condence intervals.

916 Vestergaard et al.

more severe febrile seizures. Moreover, randomized studies


have shown that preventing recurrent febrile seizures does
not alter the subsequent risk of epilepsy (21).
Fourth, results may be biased by systematic errors in the
selection of study subjects or in the information on febrile

seizures and epilepsy. Clinical case series tend to overestimate the frequencies of adverse outcomes after febrile
seizures (22), but our study was based on a nationwide cohort with virtually complete follow-up. Bias due to selection
of study participants and nonresponse can therefore not

TABLE 2. Cumulative incidence* of epilepsy according to selected characteristics


among persons born in Denmark between 1978 and 2002 and followed until 2002
Febrile seizures
No

All children

Yes

95% CIy

95% CI

1.8

1.8, 1.9

6.9

6.5, 7.3

Siblings of children who had febrile seizures

2.3

2.1, 2.6

7.3

6.3, 8.2

Children who had parents or siblings with epilepsy

5.5

5.1, 5.8

16.7

14.6, 18.8

Siblings of children who developed epilepsy after


febrile seizures

6.9

5.7, 8.1

21.0

14.1, 27.9

21.0

18.8, 23.3

56.6

48.3, 64.8

Children with an Apgar score of <7 at 5 minutes

5.3

4.4, 6.2

20.0

10.4, 29.6

Children whose birth weight was <2,500 g

3.0

2.8, 3.2

8.7

7.4, 10.1

Children whose gestational age at birth was


<37 weeks

2.8

2.5, 2.9

8.7

7.2, 10.2

Children with cerebral palsy

* The cumulative incidence was assessed 23 years after the rst febrile seizure for persons
with a history of febrile seizures and at the 25th birthday for persons with no history of febrile
seizures.
y CI, condence interval.

Am J Epidemiol 2007;165:911918

Downloaded from http://aje.oxfordjournals.org/ by guest on August 4, 2012

FIGURE 3. Rate ratios of epilepsy after febrile seizures according to age at rst febrile seizure and time since rst febrile seizure. Estimates were
based on data for the 1,540,725 persons born in Denmark between 1978 and 2002 and followed until 2002. Point estimates are given, with error
bars representing 95% condence intervals.

Long-Term Risk of Epilepsy after Febrile Seizures

Am J Epidemiol 2007;165:911918

Thus, our estimates are too high for some children and too
low for others. Previous studies have shown that the risk of
epilepsy tends to increase with increasing duration of febrile
seizures, that children with focal febrile seizures are prone
to develop partial-onset epilepsy, and that children with recurrent febrile seizures are prone to develop generalizedonset epilepsy (7). One of the most controversial issues
within the field of epileptology is whether prolonged febrile
seizures damage the hippocampus and cause temporal lobe
epilepsy. To address this important issue, we need neuroimaging studies with long-term follow-up of children with
a history of prolonged febrile seizures (28).
In this study, we found that persons with a history of
febrile seizures had a higher rate of epilepsy that lasted into
adult life, but less than 7 percent of children with febrile
seizures developed epilepsy during 23 years of follow-up.
The risk was higher for those who had a family history of
epilepsy, cerebral palsy, or low Apgar scores at 5 minutes.

ACKNOWLEDGMENTS

This study was supported by the Danish Research Agency


(grant 22-02-0207), P. A. Messerschmidt and Wifes Foundation, Managing Director Kurt Bnnelycke and Mrs. Grethe
Bnnelyckes Foundation, and Aase and Ejnar Danielsens
Foundation. The Danish National Research Foundation
funds the activities of the Danish Epidemiology Science
Center and the National Center for Register-based Research.
The sponsors of the study had no role in study design,
data collection, data analysis, data interpretation, writing
of the report, or the decision to submit the report for
publication.
Conflict of interest: none declared.

REFERENCES
1. Abou-Khalil B, Andermann E, Andermann F, et al. Temporal
lobe epilepsy after prolonged febrile convulsions: excellent
outcome after surgical treatment. Epilepsia 1993;34:87883.
2. Cendes F, Andermann F, Dubeau F, et al. Early childhood
prolonged febrile convulsions, atrophy and sclerosis of mesial
structures, and temporal lobe epilepsy: an MRI volumetric
study. Neurology 1993;43:10837.
3. VanLandingham KE, Heinz ER, Cavazos JE, et al. Magnetic
resonance imaging evidence of hippocampal injury after prolonged focal febrile convulsions. Ann Neurol 1998;43:41326.
4. Dube C, Chen K, Eghbal-Ahmadi M, et al. Prolonged febrile
seizures in the immature rat model enhance hippocampal excitability long term. Ann Neurol 2000;47:33644.
5. Chen K, Baram TZ, Soltesz I. Febrile seizures in the developing brain result in persistent modification of neuronal excitability in limbic circuits. Nat Med 1999;5:88894.
6. Nelson KB, Ellenberg JH. Predictors of epilepsy in children
who have experienced febrile seizures. N Engl J Med 1976;
295:102933.
7. Annegers JF, Hauser WA, Shirts SB, et al. Factors prognostic
of unprovoked seizures after febrile convulsions. N Engl J Med
1987;316:4938.

Downloaded from http://aje.oxfordjournals.org/ by guest on August 4, 2012

explain our findings. We recently evaluated the quality of the


diagnoses in the Danish National Hospital Register and
found that the predictive value of a febrile seizure diagnosis
was 93 percent (95 percent CI: 89, 96) (23) and that the
predictive value of an epilepsy diagnosis was 81 percent
(95 percent CI: 75, 87) (Christensen et al., Department of
Neurology, Aarhus University Hospital, Denmark, unpublished manuscript). Forty percent of those who did not fulfill
the criteria for epilepsy (at least two unprovoked seizures on
separate days) had experienced a single unprovoked seizure.
Because some of these seizures signify yet-undiagnosed epilepsy, the true validity of the epilepsy diagnosis is even
higher. We found that the completeness of febrile seizure
registration was 72 percent (95 percent CI: 66, 76) (23),
but we lacked data on the completeness of the epilepsy diagnosis. Incomplete registration of epilepsy causes underestimation of the cumulative incidence of epilepsy, but rate
ratios are affected only if the completeness of the registration is different for persons with a history of febrile seizures compared with those with no such history. In Denmark,
most inhabitants have easy access to a hospital, and all hospital treatment is paid by taxes and is free of charge for the
patient. The National Hospital Register holds information
on all public hospitals and on the only private center in
Denmark treating epilepsy patients. Thus, our study included
all persons with epilepsy who had been hospitalized (1978
2002) or treated as outpatients (19952002) in Denmark.
The age-specific incidence rates of epilepsy in our study
(Christensen et al., Department of Neurology, Aarhus University Hospital, Denmark, unpublished manuscript) are
similar to those found in other population-based studies
(24). We found no significant difference in epilepsy rate
ratios with regard to time since first febrile seizure when
restricting the cohort to persons born from 1995 through
2002 and followed exclusively during a time period when
information on both inpatients and outpatients was available.
Information in the National Hospital Register is collected on
a routine basis for administrative purposes, which means that
our study did not affect the diagnostic process or introduce
surveillance bias (25). Given this background, we find it unlikely that the data quality biased our results significantly.
To determine the need for treatment of febrile seizures,
the natural history of the condition should be understood. In
keeping with previous population-based studies (610), our
study showed that the majority of children with febrile seizures do not develop epilepsy. However, the risk of epilepsy
was significantly higher for persons with a family history of
epilepsy, cerebral palsy, or low Apgar score shortly after
birth. Even after adjusting for several confounding factors,
these characteristics seem to contribute independently to the
rate of epilepsy; the combined effect seems more than additive but less than multiplicative. Recent studies suggest
that the cause of the febrile illness may modify the subsequent risk of epilepsy (26). We have shown previously that
febrile seizures provoked by vaccination against measles,
mumps, and rubella were associated with a similar rate ratio
of epilepsy compared with febrile seizures provoked by
fever of a different etiology (27).
Our study was limited by lack of clinical information
such as on duration of febrile seizures and type of epilepsy.

917

918 Vestergaard et al.

18. Hauser WA. The prevalence and incidence of convulsive disorders in children. Epilepsia 1994;35(suppl 2):S16.
19. Burgess DL. Neonatal epilepsy syndromes and GEFS:
mechanistic considerations. Epilepsia 2005;46(suppl 10):518.
20. Rothman KJ, Greenland S, eds. Modern epidemiology. 2nd ed.
Philadelphia, PA: Lippincott-Raven, 1998.
21. Knudsen FU, Paerregaard A, Andersen R, et al. Long term
outcome of prophylaxis for febrile convulsions. Arch Dis
Child 1996;74:1318.
22. Ellenberg JH, Nelson KB. Sample selection and the natural
history of disease. Studies of febrile seizures. JAMA 1980;
243:133740.
23. Vestergaard M, Obel C, Henriksen TB, et al. The Danish
National Hospital Register is a valuable study base for epidemiological research in febrile seizures. J Clin Epidemiol
2005;59:616.
24. Hauser WA, Annegers JF, Rocca WA. Descriptive epidemiology of epilepsy: contributions of population-based studies
from Rochester, Minnesota. Mayo Clin Proc 1996;71:57686.
25. Sorensen HT, Sabroe S, Olsen J. A framework for evaluation
of secondary data sources for epidemiological research. Int J
Epidemiol 1996;25:43542.
26. Suga S, Suzuki K, Ihira M, et al. Clinical characteristics of
febrile convulsions during primary HHV-6 infection. Arch Dis
Child 2000;82:626.
27. Vestergaard M, Hviid A, Madsen KM, et al. MMR vaccination
and febrile seizures: evaluation of susceptible subgroups and
long-term prognosis. JAMA 2004;292:3517.
28. Shinnar S, Pellock JM, Berg AT, et al. Short-term outcomes
of children with febrile status epilepticus. Epilepsia 2001;
42:4753.

Am J Epidemiol 2007;165:911918

Downloaded from http://aje.oxfordjournals.org/ by guest on August 4, 2012

8. Verity CM, Goldberg J. Risk of epilepsy after febrile convulsions: a national cohort study. BMJ 1991;303:13736.
9. Berg AT, Shinnar S. Unprovoked seizures in children with
febrile seizures: short-term outcome. Neurology 1996;47:
5628.
10. Van der Berg BJ, Yerushalmy J. Studies on convulsive disorders in young children. I. Incidence of febrile and nonfebrile
convulsions by age and other factors. Pediatr Res 1969;3:
298304.
11. Berkovic SF, Scheffer IE. Febrile seizures: genetics and relationship to other epilepsy syndromes. Curr Opin Neurol
1998;11:12934.
12. Shinnar S. Do febrile seizures lead to temporal lobe epilepsy?
Prospective and epidemiological studies. In: Baram TZ,
Shinnar S, eds. Febrile seizures. New York, NY: Academic
Press, 2002:87101.
13. Pedersen CB, Gotzsche H, Moller JO, et al. The Danish Civil
Registration System. A cohort of eight million persons. Dan
Med Bull 2006;53:4419.
14. Andersen TF, Madsen M, Jorgensen J, et al. The Danish National Hospital Register. A valuable source of data for modern
health sciences. Dan Med Bull 1999;46:2638.
15. Breslow NE, Day NE. Statistical methods in cancer research.
Volume IIThe design and analysis of cohort studies. IARC
Sci Publ 1987;(82):1406.
16. Rosthoj S, Andersen PK, Abildstrom SZ. SAS macros for
estimation of the cumulative incidence functions based on
a Cox regression model for competing risks survival data.
Comput Methods Programs Biomed 2004;74:6975.
17. Dravet C, Bureau M, Oguni H, et al. Severe myoclonic epilepsy
in infancy: Dravet syndrome. Adv Neurol 2005;95:71102.