SEIZURE RECURRENCE AFTER A FIRST FEBRILE SEIZURE: A MULTIVARIATE APPROACH Martin Offringa Gerarda Derksen-Lubsen Patrick M. Bossuyt Jacobus Lubsen
The probability of seizure recurrence after an initial febrile seizure in children remains uncertain (Nelson and Ellenberg 1978, Knudsen 1985, Verity et a/. 1985, El-Radhi et al. 1986, Annegers et al. 1987). Population-based studies have indicated that three or more recurrent seizures occur in only 4.1 to 9 per cent (Nelson and Ellenberg 1978, Verity et al. 1985). On the other hand, studies from referral centres have found the risk of three or more recurrences to be as high as 35 per cent (Tsuboi 1986). This uncertainty is due both to limited follow-up and to the absence of information about the combined predictive value of multiple socalled ‘risk factors’. Most studies have limited their followup to the first recurrence, and have evaluated the predictive value of single-risk indicators through univariate analyses. In previous studies, complex initial seizures (Nelson and Ellenberg 1978, Knudsen 1985), the presence of a first-degree family history of seizures of any kind (Nelson and Ellenberg 1978, Knudsen 1985), body temperature at first febrile seizure (Aicardi 1986, El-Radhi et al. 1986, El-Radhi and Banajeh 1989) and age (Nelson and Ellenberg 1978, Verity el a/. 1985, Shirts et al. 1987) have been found to be associated with an increased risk of recurrences. Yet a recent metaanalysis has indicated that none of these
risk factors alone could identify children at high or low risk of recurrent seizures (Berg et a/. 1990). This conclusion is consistent with the suggestion of others that the risk of recurrent febrile seizures could be predicted better by a combination of risk indicators (Knudsen 1985, Verity et a / . 1985, Shirts et a/. 1987). Simple recurrent febrile seizures usually d o not lead to permanent damage. However, multiple recurrent seizures during subsequent febrile episodes are disturbing and-although to date little information on the subject is available-potentially may result in prolonged seizures with a risk of subsequent neurological impairment. Therefore prediction of recurrences -particularly multiple-after a first febrile seizure could identify children who may benefit from preventive treatment, and could spare others from the inconvenience and risk of retarded cognitive development associated with long-term administration of anti-epileptic drugs (Farwell et al. 1990). In this respect the critical time-period during which the child is at risk of further seizures after the initial febrile seizure is of particular relevance. More insight into this matter may lead to adjustment of the duration of prophylactic treatment for individual children. We describe the course and outcome after a first febrile seizure of a cohort of
No\
Q‘
15
i
.
Dutch children, with special emphasis on the changes in risk of recurrence over time. The combined predictive value of multiple clinical variables at first seizure for future seizure recurrence is assessed, as well as the role of the child’s age in the risk of recurrent seizures.
aJ
.-2 v,
aJ aJ C L
2
B
16
Patients and method
The study was carried out at the emergency room of the Sophia Children’s Hospital in Rotterdam. This is a teaching hospital, which keeps full patient records and has a 24-hour emergency service. Febrile seizures were defined according to the N I H consensus statement (Millichap 1981). All previously healthy children aged between six months and five years, presenting with a first febrile seizure and a body temperature of 3 8 . 5 ~(101 OF) or higher, but without evidence of intracranial infection or defined cause (e.g. metabolic dysregulation), between March 1985 and March 1987 were included. Children with a history of non-febrile seizures or with neurological abnormalities such as cerebral palsy a n d mental retardation were excluded, since their subsequent course is determined mainly by their basic disorder. Furthermore, those being treated with continuous phenobarbitone o r sodium valproate were excluded because we wanted t o study the natural history of previously healthy, untreated children. Seizures were defined as complex if they had lasted for more than 15 minutes, had focal features o r recurred within 24 hours. History a n d clinical d a t a on admission were retrieved from the patient’s record. The outcome variable was the recurrence of seizures during the follow-up period ending 1st September 1989. In August 1989, all children were contacted by telephone after a mailed announcement. Addresses and telephone numbers of families who had changed their address were retrieved from municipal registries: if contact could not be established the family physician provided follow-up information. All followup interviews were carried out by one investigator (M.O.). The number a n d type of recurrent seizures-if any-as well as the time intervals between subsequent seizures were recorded. Seizures were
considered non-febrile i f they had occurred at a body temperature of 3 7 . 5 ” ~or below. A cumulative proportion of children with recurrences after each seizure was estimated using the product limit method of Kaplan and Meier (1958); strata were compared using the log-rank test. Differences in age between children with and without recurrences were assessed with the Student I test after log-transformation of the values (Gardner and Altman 1989); p i 0.01 was considered significant. The risk was assessed of seizure recurrence and of three or more recurrent seizures during follow-up. To evaluate potential risk factors, we used univariate and multivariate Cox proportional hazards regression models (Cox 1970). All risk factors were dichotomized; children lacking the attribute represented the reference category. Age at onset was divided into three groups, with age between 12 and 30 months as the reference category. To compare rates in children with and without a certain factor, hazard-rate ratios with 95 per cent confidence intervals were calculated from the factors’ regression coefficients and their standard errors; in the univariate analyses they are referred to as ‘crude’ rate ratios. A factor with a rate ratio value significantly higher than 1 would be associated with a n increased risk of recurrent seizures, while a factor with a rate ratio significantly lower than 1 would be associated with a decreased risk. Multivariate proportional hazard models were used to examine the combined influence of the risk factors on the probability of subsequent febrile seizures. Covariates were eliminated from the model using the partial likelihood ratio test (Kalbfleish a n d Prentice 1980), with p = O *10 as the p limit for exclusion. Rate ratios, ‘adjusted’ for the presence of other factors retained in the multivariate model, and their 95 per cent confidence intervals were calculated using regression coefficients a n d their standard errors. The combined effect of all relevant risk factors on the probability of at least one recurrence was assessed. A composite risk score for each child in the study was calculated, using the factors’ weights (i.e. the regression coefficients from the final
parsimonious multivariate model), and a coding of 0 if the attribute was absent and 1 i f i t was present. Patients were arranged in order of increasing score value and then divided into three risk groups: a score of 0 implied ‘medium’ risk (no factors present), a score below 0 ‘low’ risk (mainly ‘protecting’ factors) and above 0 ‘high’ risk. Stratified Kaplan-Meier analysis for these three levels of composite risk was performed. Similar scores were calculated using the regression coefficents from the multivariate model for three or more recurrent seizures. T w o groups were formed: children scoring < O (‘low’ risk) and 2 0 (‘high’ risk). This multiple-risk score was used to to assess the risk of any seizure recurrence for the entire cohort. Univariate statistics a n d Kaplan-Meier product limit estimators were calculated, using the SAS-PC statistical package (SAS Institute, Inc., Cary, NC); for the proportional hazard models the B M D P package was used ( B M D P Statistical Software, Inc., Los Angeles, CA).
Results Altogether 175 children met the inclusion criteria, of whom six had a known neurological abnormality; these were excluded. Eight children who had received longterm anti-epileptic treatment after the initial febrile seizure were also excluded from the risk analyses. Six of them apparently were treated for complex seizures with age at onset of less than one year; the other two children had recurring febrile seizures within 24 hours, with partial features in one. Despite monitored treatment, three of these eight children had recurrent seizures, one of whom had frequent non-febrile seizures. Six were lost to follow-up shortly after the initial seizure. Median age at onset of the remaining 155 children was 18 (range six to 54) months. In most instances their seizures were associated with upper respiratorytract infections, diarrhoea o r fever of short duration. They all recovered without complications within several days. Follow-up information was obtained from the parents in 139 instances, and from the family doctor in 16. In all instances, parents a n d family physicians could recall the dates of the recurrent
seizures. Median follow-up time was 38 (range 27 to 60) months.
r4 e
-
rn
e-
Recurrent seizures Of the 155 initially untreated children 58 (37 per cent) had at least one, 47 (30 per cent) at least two and 27 (17 per cent) at least three recurrent seizures. Five (3 per cent) had unprovoked non-febrile seizures, in one case as the first recurrence; in the other four the seizures occurred after two recurrent febrile seizures. Four of the five had more than five recurrences. Only one child experienced a complex recurrent seizure. None had neurological sequelae. Risk factors Univariate analysis revealed that a family history of febrile or unprovoked recurrent seizures and a multiple initial febrile seizure were associated with an increased risk of recurrent seizures, while a body temperature of 4 0 ” ~or higher at the initial seizure and a n age of 2 3 0 months were associated with a reduced risk. These factors were all retained in the multivariate model (Table I). Children with 2 3 recurrences had a lower median age at the initial seizure compared with those with none or one or two recurrent febrile seizures (13, 19 and 20 months, respectively). For this outcome a family history of febrile o r unprovoked seizures was selected as a risk factor for the multivariate model; a temperature 2 4 0 ” and ~ age 2 30 months at initial febrile seizure were also selected and associated with decreased risk, but reached n o statistical significance (Table I). Age at first o r second recurrence differed between children with and without mutiple recurrences: median 20 vs. 28 months for age at first recurrence and median 26 vs. 32 months for age at second recurrence; only the difference between ages at first recurrence was significant (p= 0.01). The time interval between the initial seizure and the first and second recurrences did not differ between children with and without multiple recurrences (median 5 - 1 vs. 5 . 2 months and 4 - 8 vs. 7.0 months, respectively). All five children with recurrent non-febrile seizures had a complex initial febrile seizure, four of whom had a temperature below 40”c.
m
17
Seizure Recurrence after First Febrile Seizure Martin Offringa el al
4
00
TABLE 1 Number of recurrences, crude and adjusted rate ratios for at least one recurrence and 2 3 recurrences in 155 children, initially not treated with anticonvulsant prophylaxis
Factor
Gender Female Male Age at onset < 12 mths 12-30 mths 2 30 mths Family seizure historyt None Unprovoked Febrile Initial seizure Duration 5 15 min Duration > 15 min Generalized Focal Single Multiple Temperature < 40°C 2 40°C
A t-risk
A t least one recurrence
23
recurrences
N
Hazard rate ratio* with 95% CI Univariaie analysis Multivariate analysis
N
Hazard rate ratio* with 95% CI Univariate analysis Multivariaie analysis
65 90
24 34
rc 1.03 (0'61-1.73)
**
14 13
rc 0.63 (0.29-1.34)
**
25 99 31
11 38 9
1.25 (0.64-2.46) rc 0.68 (0.32-1.40)
7 18
1.56 (0.65-3.74) rc 0.34 (0.34-1.47)
**
rc 0.43 (0.19-0.94)
rc 0.30 (0.07-1.25)
128 7 20
39 5 14
rc 3.76 (1.50-9.41) 3.41 (1.87-6.24)
rc 3.62 (1.40-9.21) 3.92 (2'08-7.27)
18
3 6
rc 3.46 (1.04-11.50) 1.90 (0.77-4.72)
rc 5.04 (1.42-17.50) 2.98 (1.23-7.18)
136 19 147 8 130 25
53 5 55 3 44 14
rc 0.74 (0.30-1.86) rc 0.85 (0.27-2.72) rc 2.07 (1.13-3.76)
**
rc 1.13 (0.34-3.76) rc 1.49 (0.35-6.30) rc 1.22 (0.46-3.23)
**
rc 2.45 (1.27-4.64)
24 3 25 2 22 5
96 59
41 17
rc 0.63 (0'35-1.12)
rc 0.46 (0.25-0.82)
21 6
rc 0.57 (0.22-1.48)
*Compared with reference category (rc). **Not retained in stepwise multivariate proportional hazard model tOnly first-degree family.
**
**
2
**
** rc 0.47 (0.19-1.11)
Two were younger than 12 months at initial seizure and none was older than 30 months.
I
loo 80
A
60
Risk of recurrence Kaplan-Meier curves for the first, second and third recurrence (155, 58 and 47 children at risk, respectively) with 95 per cent confidence intervals indicated that risk of recurrence was highest during the first months after a seizure and declined rapidly after several months without seizures (Fig. 1). T h e recurrence rate among children without seizures during the first six months declined from 21 to 11 per cent; during the first 12 months at follow-up the recurrence rate declined to 4 per cent. After a first recurrent seizure the rate of further occurrences was 37 per cent in the first six months and-after six months without seizures-22 per cent in the second six months. After the second seizure recurrence the rate of further recurrences was 34 per cent in the first and 25 per cent in the second six months. Kaplan-Meier analysis of children, using family history of seizures a n d body temperature at initial seizure, revealed different two-year recurrence rates for children who had none (35 per cent), either (80 per cent with positive history, temperature < 4 0 ° c ; 18 per cent with negative family history, temperature 1 4 0 " ~ or ) both of these factors ( 5 5 per cent). First febrile seizure at a body temperature above 4 0 " ~in children with a positive family history appeared t o lower the over-all recurrence rate compared with those with a positive family history but who had a seizure at a lower temperature. Risk score A composite risk score was calculated for each child o n the basis of the presence of each of the five variables. in the final multivariate model for the risk of any seizure recurrence. Weights a n d coding are given in Table 11, with a calculated risk score f o r four hypothetical children. Patients were divided into three risk groups: low (N=66), medium (N=47) and high ( ~ = 4 2 ) .Figure 2 shows the Kaplan-Meier curves for children a t these three levels of composite risk of first recurrence. At 12 months after the initial
40 20 0
e
0
6
12
18
24
30
36
y
6
12
18
24
30
36
100
ri-
m
0'
h
8
80
v
100
8ol
c
Months since previous seizure
Fig. I . Probability of seizure recurrence in initially unfreated children: (A) affer initial febrile seizure (I55 at risk), (B) after first recurrence (58 at risk), (C) after second recurrence (47 at risk). Solid line = Kaplan-Meier estimate; dotted lines = 95 per cent confidence limits.
100 1
I r -_
.
0
6
12
18
24
-30
36
Months since first seizure
Fig. 2. Probability of seizure recurrence according to risk score: (A) high risk ( N = 4 2 ) , (B) medium risk (N = 47). ( C ) low risk ( N = 66). Kaplan-Meier estimates.
19
4 :
...
TABLE I I Weights for risk factor5 for at least one recurrence, and for 2 3 recurrences retained in rnultirariate C o \ regrcssion model. Ri\k le\el for fnur hypothetical children (A, B, C , D) with ( + ) and without ( - ) factors
-0.8
Age 2 30 n i o i i t l i s Fan1iIy ti ist o r y uI1p rovo k ed Febrile \ei/ure
-.ai
: I-
I
1.2 1.3
t
0.8
-
0.7
+
I
I n i t i a l sei/ui-e
M 11 It i 17 le Tcmpcraturc
2
40 ' C '
.-
Lr.
Ki+score f o r any recurrence R i d o f any recurrenccf
aJ Ir'
Risk-scores l o r 2 3 recurrence\ Kisk o f 2 3 recurrence\$
Im
t t
0 0.8 0
M edi uiii 0 High
* I n d i c a t o r variabler coded as 0 = a t t r i b u t e absent; I = a t t r i b u t e prerent. .lRegi-e\sion coefficient f r o m m u l t i v a r i a t e Cox model lor at least one a n d f o r t 3 rccuri-eiiccs. $Categories defined i n tmt.
seizure there is a cumulative recurrence rate of 48 per cent for the high-risk group, as opposed to 15 per cent for the group at low risk. There is a n obvious difference in recurrence patterns between groups @=0.01). Only four of 45 children in the low-risk group developed three recurrences. Similar scores were calculated from the results of the predictive model for the occurrence of three or more subsequent febrile seizures. Patients were divided according to score values into two risk groups: 116 low risk and 39 high risk. Three o r more recurrences occurred ir. 13 ( 1 1 per cent) children in the low-risk group and 14 (36 per cent) in the high-risk group. Using this risk score to address the risk of multiple recurrence, assessment of risk of any seizure recurrence for the whole cohort provided good differentiation between children with a high a n d low first-year risk of recurrence. The observed cumulative recurrence rate was 38 per cent and 14 per cent after six months, a n d 48 per cent and 23 per cent after 12 months for high- a n d low-risk children, respectively; all differences were statistically significant.
Discussion
20
The aim of the present study was t o determine the influence of potential risk factors o n seizure recurrence rate, according to the presence o r absence of other
risk factors. Also studied were the course and outcome of children after a first seizure and evolution of the recurrence risk over time since last seizure. We concentrated on predictors of multiple recurrences, since we feel this to be a relevant clinical outcome of febrile seizures. Children at risk of multiple recurrences may need continuous prophylactic treatment for some time, while those at risk of just one o r two uncomplicated recurrent seizures would probably not need such treatment.
Recurrences The percentage of all 155 children with at least one a n d two or more recurrences after the first febrile seizure was 37 and 30 per cent, respectively. This is consistent with general rates in clinic-based studies (Berg 1990). The occurrence of three o r more subsequent seizures among initially untreated children was 17 per cent (27 of 155). Only one other clinic-based study reports a figure-35 per cent-for this separate clinical point (Tsuboi 1986). Population-based studies have reported rates of 4 . 1 t o 9 per cent (Nelson and Ellenberg 1978, Verity et al. 1985). Nonfebrile seizures occurred in the present study in 3 per cent of children; previous clinic-based studies with varying followu p have reported rates of 2 - 6 t o 7 6 . 9 per cent (Ellenberg and Nelson 1980). In the present study there were n o neurological
12
24
36
60
P
Age (months)
sequelae from recurrent seizures and only one child had a complex recurrence. However, children with aberrant previous neurological status and acquired sequelae at initial febrile seizure-a postulated main predictor of epilepsy (Nelson and Ellenberg 1978)-were excluded.
Risk factors First-degree family history of any type of seizure was associated with both single and multiple recurrences (see Table I): this is consistent with other studies (Berg et at. 1990). A body temperature of 2 4 0 " ~at initial seizure appeared to be associated with a lower risk, as was a n age of 2 30 months. The association between elevated temperature a n d recurrence risk has also been reported by El-Radhi and co-workers (1986, 1989). A high body temperature at the initial seizure seems to modify the effect of a positive family history by lowering the recurrence rate. These findings are compatible with a model of decreased seizure threshold at a certain age which, as suggested by Aicardi (1986), may be mainly genetically determined. A long or focal initial seizure, and even status epilepticus without neurological sequelae, were not associated with an increased risk of further febrile seizures. Again, these findings are consistent with those in Berg's review (1990). Since several factors appear t o act together on the risk of recurrent seizuressometimes in opposite ways-a composite recurrence risk score was assigned to each patient. Using all relevant factors in combination, it was possible retrospectively to
identify subgroups of children with a oneyear recurrence risk as low as 15 per cent (high temperature, n o risk factors) and as high as 48 per cent (low temperature, family history o r multiple initial seizure) (see Table 11, Fig. 2). In addition, groups of children at high and low risk for frequent recurrences could be identified, with one-quarter of all children (39) having a three-fold risk (36 vs. 1 I per cent).
Recurrence risk over time: the role of age Of great relevance to the decision to start (and stop) continuous prophylaxis is the finding of a rapidly declining over-all recurrence risk during the three to six months after the previous seizure (see Fig. 1); this decline seems to be even greater among high-risk children (see Fig. 2). In the light of recently described effects of long-term phenobarbital prophylaxis on children's cognitive development (Farwell et al. 1990), we feel that only short-term treatment-if any-after a first febrile seizure in children with several risk factors would be justified. The effect of age was studied in more detail. In contrast t o previous reports (Berg et a/. 1990), children whose initial febrile seizure was before one year of age were not at higher risk of recurrent seizures than those whose age at onset was between 12 and 30 months (see Table I). O n the other hand, children with multiple recurrences had their first a n d second seizures at younger ages than those with none, one or two recurrences. However, the time interval between seizures-often
21
-
i
E
.-2 r A
.L
i4’ L
M
d M
L
s
22
used as a clinical basis on which t o start anticonvulsant prophylaxis treatment when the interval is short (e.g. less than three o r six months)-did not provide any additional information. We hypothesized that there is a critical age-period in which children with febrile seizures are most vulnerable to recurrent seizures. To determine this age-period, we estimated the recurrence hazard ratedefined as number of recurrences per unit person-time of follow-up-as a function of the attained age for all children in the study (Kalbfleisch and Prentice 1980). In this analysis the recurrence hazard rate is calculated for each age-group at risk for further seizures, containing only children who had their first seizure before that age and have not (yet) had a recurrence. Thus, each child enters the risk set at its age of onset and leaves the set at a higher age, either through the occurrence of a subsequent seizure o r at the end of follow-up. Using the technique described by Thaler (1984) a ‘smoothed’ recurrence hazard function was constructed (Fig. 3). The recurrence hazard was highest between 12 and 24 months of age, after which it dropped t o a much lower level beyond the ages of 30 t o 36 months. The recurrence hazard curve is very similar t o recently published age-at-onset distribution curves (Wallace 1988, Forsgren et al. 1990, Offringa et al. 1991), a n d is consistent with the theory of increased seizure susceptibility during particular stages of cerebral development during the second year of life (Wallace 1988). It would be interesting t o examine the effect of the various risk factors o n this agebased recurrence hazard; however, in the present study t o o little d a t a were available for stratification. It is likely that, in addition t o the factors studied, other factors during the period after the initial febrile seizure, such as number of febrile episodes (Knudsen 1988), parental management of fever and type of infection (Lahat el al. 1990, Rantala et al. 1990, Zvulunov et al. 1990), may influence the recurrence risk. These factors need further investigation. The. results of the present analysis are relevant t o decisions about preventive anticonvulsive treatment after a first febrile seizure. The decision is based o n
inconvenience and side-effects of prolonged treatment with anti-epileptic drugs against the risk of multiple recurrences, with possible subsequent neurological impairment. O u r finding that the risk of recurrent seizures is highest between 12 and 24 months of age a n d declines rapidly after six months from the previous seizure suggests that any prophylactic treatment after a first febrile seizure should be restricted in duration a n d recommended only for children with multiple risk factors. Since most follow-up studies have shown that (first) recurrent seizures are not threatening, it may even be advocated that ‘high-risk’ children be treated only after the first recurrence. In order to prevent lengthy recurrences, it seems t o be of greater importance t o counsel the parents properly, and to instruct them how to manage fever and recurrent seizures by administration of rapidly acting anticonvulsants, such as a diazepam solution by rectum.
Conclusions The main findings of the present study are that the combined predictive value of age at onset, first-degree family history of seizures of any kind, nature of first seizure (single o r multiple) a n d degree of fever at first seizure (more o r less than 40”c) for both single and multiple recurrences is superior t o that of single variables. The child’s age at each subsequent febrile seizure has predictive value for further seizures: the younger the child at the second and third seizures, the higher the likelihood of further recurrences. The hazard of recurrence after a first febrile seizure is highest between the ages of 12 and 24 months. In contrast, time between seizures varies little between children with o r without 2 3 recurrences. Therefore, when assessing the prognosis, it seems more appropriate t o consider only the age of the child at each recurrent seizure. Accepted for publication 13th August 1991 A ckno wledgements
The authors thank Dr. Phil Eskes and Dr. Jochannan Benbassat for their helpful comments. This study was supported by a grant from the Netherlands’ Health Research Promotion Programme (SGO) and the Sophia Foundation for the Sick Child.
Authors’ Appointments
*Martin Offringa, M.D., Department of Paediatrics, Univenity Hospital and Sophia Children’s
Hospital, and Center for Clinical Decision Analysis; Gerarda Derksen-Lubsen, M.D., Ph.D., Department of Paediatrics, University Hospital and Sophia Children’s Hospital; Patrick M . Bossuyt, Ph.D.;
Jacobus Lubsen, Ph.D.; Center for Clinical Decision Analysis; Erasmus University, Rotterdam, The Netherlands.
N d
2 d-
m
*Correspondence tofirst author at Room EE 2171, Erasmus University, P.O. Box 1738, 3000 DR Rotterdam, The Netherlands.
N-
m
0‘
SUMMARY The results are presented of a follow-up study of 155 Dutch children after the first febrile seizure. Of these initially untreated children 37 per cent had had at least one, 30 per cent at least two and 17 per cent at least three subsequent seizures. The vulnerable period for recurrent seizures after the first febrile seizure was between 12 and 24 months, whereafter the risk was four to five times lower; after any seizure the risk was highest within the first six months, declining steadily after six months without seizures. A first-degree family history of any type of seizure predicted multiple recurrences; an age of at least 30 months and a temperature of 2 4 0 ° C at initial seizure were associated with reduced risk. Factors in combination influenced the risk of recurrent seizures, sometimes in opposite ways. RESUME Recidives de crises comiiiales apres une premiere crise hyperthertnique: une approche niuliivariee L’article rapporte les resultats d’une etude longitudinale de I55 enfants hollandais apres une premiere crise hyperthermique. Parmi ces enfants initialement non traites, 37 pour cent avaient eu au moins une crise ulterieure, 30 pour cent a u moins deux crises et 17 pour cent au moins trois crises. La periode sensible pour le retour de crises etait comprise entre 12 et 24 mois, le risque ulterieur etant quatre ou cinq fois plus faible; apres n’importe quelle crise, le risque maximal se situait dans les six mois suivants, declinant rapidement apres six mois sans crises. Une histoire familiale de n’importe quel type de crise dans les relations au premier degre conduisait a prevoir de multiples recidives; a un i g e d’au moins 30 mois et a une temperature egale ou superieure a 40°C lors de la crise initiale, le risque de recidive etait reduit. Les facteurs associes modifient le risque de recidive, agissant parfois de facon opposee. ZUSAMMENFASSUNG Wiederauftrelen von A nfallen nach dern ersten Fieberkrampf: eine Multivarianzanalyse Es werden die Ergebnisse einer Kontrollstudie bei 155 hollandischen Kindern nach ihrem ersten Fieberkrampf dargestellt. Von diesen zunachst unbehandelten Kindern hatten 37 Prozent mindestens einen, 30 Prozent zwei und 17 Prozent drei weitere Anfalle. Die vulnerable Phase fur Anfallsrezidive nach dem ersten Fieberkrampf war zwischen 12 und 24 Monaten, danach war das Risiko vier bis funf Ma1 geringer; nach jedem Anfall war das Risiko innerhalb der ersten sechs Monate a m hochsten und fie1 nach sechs anfallfreien Monaten stetig ab. Ein Anfallsleiden bei Familienangehorigen ersten Grades bedeutete multiple Rezidive. Ein Alter von mindestens 30 Monaten und eine Temperatur von 2 4 0 ° C beim ersten Krarnpfanfall waren mit einem geringeren Risiko verbunden. Die Kombination von Faktoren beeinfluate das Risiko fur Anfallsrezidive, gelegentlich in kontrarer Wiese. RESUMEN Recurrencia de convulsiones despues de una convulsion febril. Enfoque multivariado Se presenta 10s resultados d e un estudio sobre el curso de 155 nillos holandeses despues d e una convulsion febril. De estos nillos, no tratados inicialmente, el 37 por ciento habian tenido por lo menos una, el 30 por ciento por lo menos dos y el 17 por ciento por lo menos tres convulsiones consecutivas. El period0 vulnerable para las convulsiones recurrentes despues de la primera convulsion febril fue de 12 a 24 meses. Mas adelante el riesgo era d e cuatro a cinco veces menor. Despues de cualquier convulsion el riesgo era maximo dentro d e 10s primeros seis meses, para disminuir despues de forma constante despues de seis meses sin convulsiones. Una historia familiar de primer grado de cualquier tip0 d e convulsion era predictiva de multiples recurrencias. Una edad de por lo menos 30 meses y una temperatura de 2 4 0 ” en la convulsion inicial iban asociados a un riesgo reducido. La combinacion de diversos factores influye sobre el riesgo de convulsiones recurrentes a veces en sentidos opuestos. References Aicardi, J . (1986) ‘Febrile convulsions.’ In Epilepsy in Children. New York: Raven Press. pp. 212-23 1. Annegers, J . F., Hauser, W . A., Shirts, S. B., Kurland, L. T. (1987) ‘Factors prognostic of unprovoked seizures after febrile convulsions.’ New England Journal of Medicine, 316, 493-498. Berg, A. T., Shinnar, S., Hauser, W . A,, Leventhal,
J . M . (1990) ‘Predictors of recurrent febrile seizures: a meta-analytic review.’ Journal of Pediatrics, 116, 329-337. Cox, D. R. (1970) The Analysis of Binary Data. London: Chapman & Hall. Ellenberg, J. H., Nelson, K. B. (1980) ‘Sample selection and the natural \istory of disease. Studies of febrile seizures. Journal of the American Medical Association, 342, 1337-1 340.
23
. i
...
2
.-N W
rA
El-Radhi, A . S., Banajeh. S. (1989) ‘Effect of fever o n recurrence rate of febrile convulsionb.’ -
Archives of Disease in Childhood, 64, 869-870. Withana, K., Banajeh, S. (1986) ‘Recurrence
seizures.’ It7 Nel5on. K . B., Ellenbere, J . H . ( E d \ . ) Febrile Seiaires. New York: Raven Pre\\. p. 2 Nelson, K . B., Ellenberg, J . H . (1978) ‘Prognosi\ i i i children with febrile seizure,.’ Pediurric.y. 61.
rate of febrile convulsions related to the degree o f pyrexia during the first attack.’ Clinical Pediatrics, 25, 3 1 1-3 13. Farwell, J . R.. Lee, Y. J . , Hirtz, D. G., Sulrbacher, S. I . , Ellenberg, J . H., Nelson, K. B. (1990) ‘Phenobarbital for febrile seizures-effects on intelligence a n d o n seizure recurrence.’ NeM’ England Journal of Medicine, 332, 364-369. Forsgren, L., Sidenvall, R., Blomquist, H . K., Heijebel, J . (1990) ‘A prospective study of the incidence of febrile convulsions.’ Acia Paediarrica
Offringa, M., Hazebroek-Kainschreiir, A . A . J . M.. Derksen-Lubsen, G . (1991) ‘Prevalence of febrile seizures in Dutch schoolchildren.’ Paediutric. und Perinaral Epidemiology, 5, I 8 I- 188. Rantala, H . , Uhari, M.. T u o k k o , H . (1990) ‘Viral infections a n d recurrence of febrile convul\ioiis.’ Journal of Pedia/rics, 1 16. 195- 199. Shirts, S. B., Annegers, J . I:., Hauser, W .A. (1987) ‘The relation of age at first febrile scizure to recurrence of febrile seizures.’ Epilepsiu, 28, 6 2 5 .
Scandinavica. 79, 550-557.
W
rA
-
Press. p. 24. Kalbfleish, J. D., Prentice, R. L. (1980) The Statistical Analysis qf Fuilure Ti~rieData. N ~ M
population-based cohort o f children, Rochester. M N . ’ Neurology, 37 (Suppl. I ) , 149. (Ahstruct.) Thaler, H . T . (1984) ‘Nonparametric estimation of the hazard ratio.’ Journal of the A n ~ e r i c ~ n
Journul of the Americon Slalistical Associalion, 53, 203-223. Knudsen, F. U . (1985) ‘Recurrence risk after first
febrile seizure a n d effect o f short-term diazepam prophylaxis.’ Archives of Disease in Childhood,
60, 1045-1049. -
(1988) ‘Frequent febrile episodes a n d recurrent febrile convulsions. ’ A ctu Neurologicu Scundinavico, 78, 414-417. Lahat, E., Katz, V., Bistritzer, T., Eshel, G . , Aladjem, M . (1990) ‘Recurrent seizures in children with Shigella-associated convulsions.’ Annals of Neurologv, 28, 393-395. Millichap. I . G . (1981) ‘The definition of febrile
24
( Abstract.)
Gardner, M . J . , Altman, D. G . (1989) Statistics with Confidence. London: British Medical Journul
York: J o h n Wiley. Kaplan, E. L., Meier, P . (1958) “on-parametric estimation from incomplete observations.’
.-2
720-727.
Hauser, A , , Annegers. J . F., Kurland. L. T.
(1987) ‘Risk of recurrence of febrile seizures i n
:I
Stutistical Association, 19, 290-293.
Tsuboi, T. (1986) ‘Seizures of childhood. A population-based a n d clinic-based study.’ Actu Neurologica Scandinavica, Suppl. 110, 130. Verity, C. M., Butler, N . R., Golding, J . (1985) ‘Febrile convulsions in a national cohort followed up from birth. 1: Prevalence a n d recurrence in thc first year of life.’ Britisl7 Medical Journal, 290. 1307-1 3 10. Wallace, S. (1988) The Child wiih Febrile Sei,-irre.s. Guildford: Butterworth Scientific. pp. 15-23. Zvulunov, A . , Lerman, M., Ashkenazi, S., Weitz. R., Nitzan, M . , Dinari, G. (1990) ‘The prognosis of convulsions during childhood shigellosis.’ Eirropear7 Journal of Pediatrics, 149, 293-294.
The probability of seizure recurrence after an initial febrile seizure in children remains uncertain (Nelson and Ellenberg 1978, Knudsen 1985, Verity et a/. 1985, El-Radhi et al. 1986, Annegers et al. 1987). Population-based studies have indicated that three or more recurrent seizures occur in only 4.1 to 9 per cent (Nelson and Ellenberg 1978, Verity et al. 1985). On the other hand, studies from referral centres have found the risk of three or more recurrences to be as high as 35 per cent (Tsuboi 1986). This uncertainty is due both to limited follow-up and to the absence of information about the combined predictive value of multiple socalled ‘risk factors’. Most studies have limited their followup to the first recurrence, and have evaluated the predictive value of single-risk indicators through univariate analyses. In previous studies, complex initial seizures (Nelson and Ellenberg 1978, Knudsen 1985), the presence of a first-degree family history of seizures of any kind (Nelson and Ellenberg 1978, Knudsen 1985), body temperature at first febrile seizure (Aicardi 1986, El-Radhi et al. 1986, El-Radhi and Banajeh 1989) and age (Nelson and Ellenberg 1978, Verity el a/. 1985, Shirts et al. 1987) have been found to be associated with an increased risk of recurrences. Yet a recent metaanalysis has indicated that none of these
risk factors alone could identify children at high or low risk of recurrent seizures (Berg et a/. 1990). This conclusion is consistent with the suggestion of others that the risk of recurrent febrile seizures could be predicted better by a combination of risk indicators (Knudsen 1985, Verity et a / . 1985, Shirts et a/. 1987). Simple recurrent febrile seizures usually d o not lead to permanent damage. However, multiple recurrent seizures during subsequent febrile episodes are disturbing and-although to date little information on the subject is available-potentially may result in prolonged seizures with a risk of subsequent neurological impairment. Therefore prediction of recurrences -particularly multiple-after a first febrile seizure could identify children who may benefit from preventive treatment, and could spare others from the inconvenience and risk of retarded cognitive development associated with long-term administration of anti-epileptic drugs (Farwell et al. 1990). In this respect the critical time-period during which the child is at risk of further seizures after the initial febrile seizure is of particular relevance. More insight into this matter may lead to adjustment of the duration of prophylactic treatment for individual children. We describe the course and outcome after a first febrile seizure of a cohort of
No\
Q‘
15
i
.
Dutch children, with special emphasis on the changes in risk of recurrence over time. The combined predictive value of multiple clinical variables at first seizure for future seizure recurrence is assessed, as well as the role of the child’s age in the risk of recurrent seizures.
aJ
.-2 v,
aJ aJ C L
2
B
16
Patients and method
The study was carried out at the emergency room of the Sophia Children’s Hospital in Rotterdam. This is a teaching hospital, which keeps full patient records and has a 24-hour emergency service. Febrile seizures were defined according to the N I H consensus statement (Millichap 1981). All previously healthy children aged between six months and five years, presenting with a first febrile seizure and a body temperature of 3 8 . 5 ~(101 OF) or higher, but without evidence of intracranial infection or defined cause (e.g. metabolic dysregulation), between March 1985 and March 1987 were included. Children with a history of non-febrile seizures or with neurological abnormalities such as cerebral palsy a n d mental retardation were excluded, since their subsequent course is determined mainly by their basic disorder. Furthermore, those being treated with continuous phenobarbitone o r sodium valproate were excluded because we wanted t o study the natural history of previously healthy, untreated children. Seizures were defined as complex if they had lasted for more than 15 minutes, had focal features o r recurred within 24 hours. History a n d clinical d a t a on admission were retrieved from the patient’s record. The outcome variable was the recurrence of seizures during the follow-up period ending 1st September 1989. In August 1989, all children were contacted by telephone after a mailed announcement. Addresses and telephone numbers of families who had changed their address were retrieved from municipal registries: if contact could not be established the family physician provided follow-up information. All followup interviews were carried out by one investigator (M.O.). The number a n d type of recurrent seizures-if any-as well as the time intervals between subsequent seizures were recorded. Seizures were
considered non-febrile i f they had occurred at a body temperature of 3 7 . 5 ” ~or below. A cumulative proportion of children with recurrences after each seizure was estimated using the product limit method of Kaplan and Meier (1958); strata were compared using the log-rank test. Differences in age between children with and without recurrences were assessed with the Student I test after log-transformation of the values (Gardner and Altman 1989); p i 0.01 was considered significant. The risk was assessed of seizure recurrence and of three or more recurrent seizures during follow-up. To evaluate potential risk factors, we used univariate and multivariate Cox proportional hazards regression models (Cox 1970). All risk factors were dichotomized; children lacking the attribute represented the reference category. Age at onset was divided into three groups, with age between 12 and 30 months as the reference category. To compare rates in children with and without a certain factor, hazard-rate ratios with 95 per cent confidence intervals were calculated from the factors’ regression coefficients and their standard errors; in the univariate analyses they are referred to as ‘crude’ rate ratios. A factor with a rate ratio value significantly higher than 1 would be associated with a n increased risk of recurrent seizures, while a factor with a rate ratio significantly lower than 1 would be associated with a decreased risk. Multivariate proportional hazard models were used to examine the combined influence of the risk factors on the probability of subsequent febrile seizures. Covariates were eliminated from the model using the partial likelihood ratio test (Kalbfleish a n d Prentice 1980), with p = O *10 as the p limit for exclusion. Rate ratios, ‘adjusted’ for the presence of other factors retained in the multivariate model, and their 95 per cent confidence intervals were calculated using regression coefficients a n d their standard errors. The combined effect of all relevant risk factors on the probability of at least one recurrence was assessed. A composite risk score for each child in the study was calculated, using the factors’ weights (i.e. the regression coefficients from the final
parsimonious multivariate model), and a coding of 0 if the attribute was absent and 1 i f i t was present. Patients were arranged in order of increasing score value and then divided into three risk groups: a score of 0 implied ‘medium’ risk (no factors present), a score below 0 ‘low’ risk (mainly ‘protecting’ factors) and above 0 ‘high’ risk. Stratified Kaplan-Meier analysis for these three levels of composite risk was performed. Similar scores were calculated using the regression coefficents from the multivariate model for three or more recurrent seizures. T w o groups were formed: children scoring < O (‘low’ risk) and 2 0 (‘high’ risk). This multiple-risk score was used to to assess the risk of any seizure recurrence for the entire cohort. Univariate statistics a n d Kaplan-Meier product limit estimators were calculated, using the SAS-PC statistical package (SAS Institute, Inc., Cary, NC); for the proportional hazard models the B M D P package was used ( B M D P Statistical Software, Inc., Los Angeles, CA).
Results Altogether 175 children met the inclusion criteria, of whom six had a known neurological abnormality; these were excluded. Eight children who had received longterm anti-epileptic treatment after the initial febrile seizure were also excluded from the risk analyses. Six of them apparently were treated for complex seizures with age at onset of less than one year; the other two children had recurring febrile seizures within 24 hours, with partial features in one. Despite monitored treatment, three of these eight children had recurrent seizures, one of whom had frequent non-febrile seizures. Six were lost to follow-up shortly after the initial seizure. Median age at onset of the remaining 155 children was 18 (range six to 54) months. In most instances their seizures were associated with upper respiratorytract infections, diarrhoea o r fever of short duration. They all recovered without complications within several days. Follow-up information was obtained from the parents in 139 instances, and from the family doctor in 16. In all instances, parents a n d family physicians could recall the dates of the recurrent
seizures. Median follow-up time was 38 (range 27 to 60) months.
r4 e
-
rn
e-
Recurrent seizures Of the 155 initially untreated children 58 (37 per cent) had at least one, 47 (30 per cent) at least two and 27 (17 per cent) at least three recurrent seizures. Five (3 per cent) had unprovoked non-febrile seizures, in one case as the first recurrence; in the other four the seizures occurred after two recurrent febrile seizures. Four of the five had more than five recurrences. Only one child experienced a complex recurrent seizure. None had neurological sequelae. Risk factors Univariate analysis revealed that a family history of febrile or unprovoked recurrent seizures and a multiple initial febrile seizure were associated with an increased risk of recurrent seizures, while a body temperature of 4 0 ” ~or higher at the initial seizure and a n age of 2 3 0 months were associated with a reduced risk. These factors were all retained in the multivariate model (Table I). Children with 2 3 recurrences had a lower median age at the initial seizure compared with those with none or one or two recurrent febrile seizures (13, 19 and 20 months, respectively). For this outcome a family history of febrile o r unprovoked seizures was selected as a risk factor for the multivariate model; a temperature 2 4 0 ” and ~ age 2 30 months at initial febrile seizure were also selected and associated with decreased risk, but reached n o statistical significance (Table I). Age at first o r second recurrence differed between children with and without mutiple recurrences: median 20 vs. 28 months for age at first recurrence and median 26 vs. 32 months for age at second recurrence; only the difference between ages at first recurrence was significant (p= 0.01). The time interval between the initial seizure and the first and second recurrences did not differ between children with and without multiple recurrences (median 5 - 1 vs. 5 . 2 months and 4 - 8 vs. 7.0 months, respectively). All five children with recurrent non-febrile seizures had a complex initial febrile seizure, four of whom had a temperature below 40”c.
m
17
Seizure Recurrence after First Febrile Seizure Martin Offringa el al
4
00
TABLE 1 Number of recurrences, crude and adjusted rate ratios for at least one recurrence and 2 3 recurrences in 155 children, initially not treated with anticonvulsant prophylaxis
Factor
Gender Female Male Age at onset < 12 mths 12-30 mths 2 30 mths Family seizure historyt None Unprovoked Febrile Initial seizure Duration 5 15 min Duration > 15 min Generalized Focal Single Multiple Temperature < 40°C 2 40°C
A t-risk
A t least one recurrence
23
recurrences
N
Hazard rate ratio* with 95% CI Univariaie analysis Multivariate analysis
N
Hazard rate ratio* with 95% CI Univariate analysis Multivariaie analysis
65 90
24 34
rc 1.03 (0'61-1.73)
**
14 13
rc 0.63 (0.29-1.34)
**
25 99 31
11 38 9
1.25 (0.64-2.46) rc 0.68 (0.32-1.40)
7 18
1.56 (0.65-3.74) rc 0.34 (0.34-1.47)
**
rc 0.43 (0.19-0.94)
rc 0.30 (0.07-1.25)
128 7 20
39 5 14
rc 3.76 (1.50-9.41) 3.41 (1.87-6.24)
rc 3.62 (1.40-9.21) 3.92 (2'08-7.27)
18
3 6
rc 3.46 (1.04-11.50) 1.90 (0.77-4.72)
rc 5.04 (1.42-17.50) 2.98 (1.23-7.18)
136 19 147 8 130 25
53 5 55 3 44 14
rc 0.74 (0.30-1.86) rc 0.85 (0.27-2.72) rc 2.07 (1.13-3.76)
**
rc 1.13 (0.34-3.76) rc 1.49 (0.35-6.30) rc 1.22 (0.46-3.23)
**
rc 2.45 (1.27-4.64)
24 3 25 2 22 5
96 59
41 17
rc 0.63 (0'35-1.12)
rc 0.46 (0.25-0.82)
21 6
rc 0.57 (0.22-1.48)
*Compared with reference category (rc). **Not retained in stepwise multivariate proportional hazard model tOnly first-degree family.
**
**
2
**
** rc 0.47 (0.19-1.11)
Two were younger than 12 months at initial seizure and none was older than 30 months.
I
loo 80
A
60
Risk of recurrence Kaplan-Meier curves for the first, second and third recurrence (155, 58 and 47 children at risk, respectively) with 95 per cent confidence intervals indicated that risk of recurrence was highest during the first months after a seizure and declined rapidly after several months without seizures (Fig. 1). T h e recurrence rate among children without seizures during the first six months declined from 21 to 11 per cent; during the first 12 months at follow-up the recurrence rate declined to 4 per cent. After a first recurrent seizure the rate of further occurrences was 37 per cent in the first six months and-after six months without seizures-22 per cent in the second six months. After the second seizure recurrence the rate of further recurrences was 34 per cent in the first and 25 per cent in the second six months. Kaplan-Meier analysis of children, using family history of seizures a n d body temperature at initial seizure, revealed different two-year recurrence rates for children who had none (35 per cent), either (80 per cent with positive history, temperature < 4 0 ° c ; 18 per cent with negative family history, temperature 1 4 0 " ~ or ) both of these factors ( 5 5 per cent). First febrile seizure at a body temperature above 4 0 " ~in children with a positive family history appeared t o lower the over-all recurrence rate compared with those with a positive family history but who had a seizure at a lower temperature. Risk score A composite risk score was calculated for each child o n the basis of the presence of each of the five variables. in the final multivariate model for the risk of any seizure recurrence. Weights a n d coding are given in Table 11, with a calculated risk score f o r four hypothetical children. Patients were divided into three risk groups: low (N=66), medium (N=47) and high ( ~ = 4 2 ) .Figure 2 shows the Kaplan-Meier curves for children a t these three levels of composite risk of first recurrence. At 12 months after the initial
40 20 0
e
0
6
12
18
24
30
36
y
6
12
18
24
30
36
100
ri-
m
0'
h
8
80
v
100
8ol
c
Months since previous seizure
Fig. I . Probability of seizure recurrence in initially unfreated children: (A) affer initial febrile seizure (I55 at risk), (B) after first recurrence (58 at risk), (C) after second recurrence (47 at risk). Solid line = Kaplan-Meier estimate; dotted lines = 95 per cent confidence limits.
100 1
I r -_
.
0
6
12
18
24
-30
36
Months since first seizure
Fig. 2. Probability of seizure recurrence according to risk score: (A) high risk ( N = 4 2 ) , (B) medium risk (N = 47). ( C ) low risk ( N = 66). Kaplan-Meier estimates.
19
4 :
...
TABLE I I Weights for risk factor5 for at least one recurrence, and for 2 3 recurrences retained in rnultirariate C o \ regrcssion model. Ri\k le\el for fnur hypothetical children (A, B, C , D) with ( + ) and without ( - ) factors
-0.8
Age 2 30 n i o i i t l i s Fan1iIy ti ist o r y uI1p rovo k ed Febrile \ei/ure
-.ai
: I-
I
1.2 1.3
t
0.8
-
0.7
+
I
I n i t i a l sei/ui-e
M 11 It i 17 le Tcmpcraturc
2
40 ' C '
.-
Lr.
Ki+score f o r any recurrence R i d o f any recurrenccf
aJ Ir'
Risk-scores l o r 2 3 recurrence\ Kisk o f 2 3 recurrence\$
Im
t t
0 0.8 0
M edi uiii 0 High
* I n d i c a t o r variabler coded as 0 = a t t r i b u t e absent; I = a t t r i b u t e prerent. .lRegi-e\sion coefficient f r o m m u l t i v a r i a t e Cox model lor at least one a n d f o r t 3 rccuri-eiiccs. $Categories defined i n tmt.
seizure there is a cumulative recurrence rate of 48 per cent for the high-risk group, as opposed to 15 per cent for the group at low risk. There is a n obvious difference in recurrence patterns between groups @=0.01). Only four of 45 children in the low-risk group developed three recurrences. Similar scores were calculated from the results of the predictive model for the occurrence of three or more subsequent febrile seizures. Patients were divided according to score values into two risk groups: 116 low risk and 39 high risk. Three o r more recurrences occurred ir. 13 ( 1 1 per cent) children in the low-risk group and 14 (36 per cent) in the high-risk group. Using this risk score to address the risk of multiple recurrence, assessment of risk of any seizure recurrence for the whole cohort provided good differentiation between children with a high a n d low first-year risk of recurrence. The observed cumulative recurrence rate was 38 per cent and 14 per cent after six months, a n d 48 per cent and 23 per cent after 12 months for high- a n d low-risk children, respectively; all differences were statistically significant.
Discussion
20
The aim of the present study was t o determine the influence of potential risk factors o n seizure recurrence rate, according to the presence o r absence of other
risk factors. Also studied were the course and outcome of children after a first seizure and evolution of the recurrence risk over time since last seizure. We concentrated on predictors of multiple recurrences, since we feel this to be a relevant clinical outcome of febrile seizures. Children at risk of multiple recurrences may need continuous prophylactic treatment for some time, while those at risk of just one o r two uncomplicated recurrent seizures would probably not need such treatment.
Recurrences The percentage of all 155 children with at least one a n d two or more recurrences after the first febrile seizure was 37 and 30 per cent, respectively. This is consistent with general rates in clinic-based studies (Berg 1990). The occurrence of three o r more subsequent seizures among initially untreated children was 17 per cent (27 of 155). Only one other clinic-based study reports a figure-35 per cent-for this separate clinical point (Tsuboi 1986). Population-based studies have reported rates of 4 . 1 t o 9 per cent (Nelson and Ellenberg 1978, Verity et al. 1985). Nonfebrile seizures occurred in the present study in 3 per cent of children; previous clinic-based studies with varying followu p have reported rates of 2 - 6 t o 7 6 . 9 per cent (Ellenberg and Nelson 1980). In the present study there were n o neurological
12
24
36
60
P
Age (months)
sequelae from recurrent seizures and only one child had a complex recurrence. However, children with aberrant previous neurological status and acquired sequelae at initial febrile seizure-a postulated main predictor of epilepsy (Nelson and Ellenberg 1978)-were excluded.
Risk factors First-degree family history of any type of seizure was associated with both single and multiple recurrences (see Table I): this is consistent with other studies (Berg et at. 1990). A body temperature of 2 4 0 " ~at initial seizure appeared to be associated with a lower risk, as was a n age of 2 30 months. The association between elevated temperature a n d recurrence risk has also been reported by El-Radhi and co-workers (1986, 1989). A high body temperature at the initial seizure seems to modify the effect of a positive family history by lowering the recurrence rate. These findings are compatible with a model of decreased seizure threshold at a certain age which, as suggested by Aicardi (1986), may be mainly genetically determined. A long or focal initial seizure, and even status epilepticus without neurological sequelae, were not associated with an increased risk of further febrile seizures. Again, these findings are consistent with those in Berg's review (1990). Since several factors appear t o act together on the risk of recurrent seizuressometimes in opposite ways-a composite recurrence risk score was assigned to each patient. Using all relevant factors in combination, it was possible retrospectively to
identify subgroups of children with a oneyear recurrence risk as low as 15 per cent (high temperature, n o risk factors) and as high as 48 per cent (low temperature, family history o r multiple initial seizure) (see Table 11, Fig. 2). In addition, groups of children at high and low risk for frequent recurrences could be identified, with one-quarter of all children (39) having a three-fold risk (36 vs. 1 I per cent).
Recurrence risk over time: the role of age Of great relevance to the decision to start (and stop) continuous prophylaxis is the finding of a rapidly declining over-all recurrence risk during the three to six months after the previous seizure (see Fig. 1); this decline seems to be even greater among high-risk children (see Fig. 2). In the light of recently described effects of long-term phenobarbital prophylaxis on children's cognitive development (Farwell et al. 1990), we feel that only short-term treatment-if any-after a first febrile seizure in children with several risk factors would be justified. The effect of age was studied in more detail. In contrast t o previous reports (Berg et a/. 1990), children whose initial febrile seizure was before one year of age were not at higher risk of recurrent seizures than those whose age at onset was between 12 and 30 months (see Table I). O n the other hand, children with multiple recurrences had their first a n d second seizures at younger ages than those with none, one or two recurrences. However, the time interval between seizures-often
21
-
i
E
.-2 r A
.L
i4’ L
M
d M
L
s
22
used as a clinical basis on which t o start anticonvulsant prophylaxis treatment when the interval is short (e.g. less than three o r six months)-did not provide any additional information. We hypothesized that there is a critical age-period in which children with febrile seizures are most vulnerable to recurrent seizures. To determine this age-period, we estimated the recurrence hazard ratedefined as number of recurrences per unit person-time of follow-up-as a function of the attained age for all children in the study (Kalbfleisch and Prentice 1980). In this analysis the recurrence hazard rate is calculated for each age-group at risk for further seizures, containing only children who had their first seizure before that age and have not (yet) had a recurrence. Thus, each child enters the risk set at its age of onset and leaves the set at a higher age, either through the occurrence of a subsequent seizure o r at the end of follow-up. Using the technique described by Thaler (1984) a ‘smoothed’ recurrence hazard function was constructed (Fig. 3). The recurrence hazard was highest between 12 and 24 months of age, after which it dropped t o a much lower level beyond the ages of 30 t o 36 months. The recurrence hazard curve is very similar t o recently published age-at-onset distribution curves (Wallace 1988, Forsgren et al. 1990, Offringa et al. 1991), a n d is consistent with the theory of increased seizure susceptibility during particular stages of cerebral development during the second year of life (Wallace 1988). It would be interesting t o examine the effect of the various risk factors o n this agebased recurrence hazard; however, in the present study t o o little d a t a were available for stratification. It is likely that, in addition t o the factors studied, other factors during the period after the initial febrile seizure, such as number of febrile episodes (Knudsen 1988), parental management of fever and type of infection (Lahat el al. 1990, Rantala et al. 1990, Zvulunov et al. 1990), may influence the recurrence risk. These factors need further investigation. The. results of the present analysis are relevant t o decisions about preventive anticonvulsive treatment after a first febrile seizure. The decision is based o n
inconvenience and side-effects of prolonged treatment with anti-epileptic drugs against the risk of multiple recurrences, with possible subsequent neurological impairment. O u r finding that the risk of recurrent seizures is highest between 12 and 24 months of age a n d declines rapidly after six months from the previous seizure suggests that any prophylactic treatment after a first febrile seizure should be restricted in duration a n d recommended only for children with multiple risk factors. Since most follow-up studies have shown that (first) recurrent seizures are not threatening, it may even be advocated that ‘high-risk’ children be treated only after the first recurrence. In order to prevent lengthy recurrences, it seems t o be of greater importance t o counsel the parents properly, and to instruct them how to manage fever and recurrent seizures by administration of rapidly acting anticonvulsants, such as a diazepam solution by rectum.
Conclusions The main findings of the present study are that the combined predictive value of age at onset, first-degree family history of seizures of any kind, nature of first seizure (single o r multiple) a n d degree of fever at first seizure (more o r less than 40”c) for both single and multiple recurrences is superior t o that of single variables. The child’s age at each subsequent febrile seizure has predictive value for further seizures: the younger the child at the second and third seizures, the higher the likelihood of further recurrences. The hazard of recurrence after a first febrile seizure is highest between the ages of 12 and 24 months. In contrast, time between seizures varies little between children with o r without 2 3 recurrences. Therefore, when assessing the prognosis, it seems more appropriate t o consider only the age of the child at each recurrent seizure. Accepted for publication 13th August 1991 A ckno wledgements
The authors thank Dr. Phil Eskes and Dr. Jochannan Benbassat for their helpful comments. This study was supported by a grant from the Netherlands’ Health Research Promotion Programme (SGO) and the Sophia Foundation for the Sick Child.
Authors’ Appointments
*Martin Offringa, M.D., Department of Paediatrics, Univenity Hospital and Sophia Children’s
Hospital, and Center for Clinical Decision Analysis; Gerarda Derksen-Lubsen, M.D., Ph.D., Department of Paediatrics, University Hospital and Sophia Children’s Hospital; Patrick M . Bossuyt, Ph.D.;
Jacobus Lubsen, Ph.D.; Center for Clinical Decision Analysis; Erasmus University, Rotterdam, The Netherlands.
N d
2 d-
m
*Correspondence tofirst author at Room EE 2171, Erasmus University, P.O. Box 1738, 3000 DR Rotterdam, The Netherlands.
N-
m
0‘
SUMMARY The results are presented of a follow-up study of 155 Dutch children after the first febrile seizure. Of these initially untreated children 37 per cent had had at least one, 30 per cent at least two and 17 per cent at least three subsequent seizures. The vulnerable period for recurrent seizures after the first febrile seizure was between 12 and 24 months, whereafter the risk was four to five times lower; after any seizure the risk was highest within the first six months, declining steadily after six months without seizures. A first-degree family history of any type of seizure predicted multiple recurrences; an age of at least 30 months and a temperature of 2 4 0 ° C at initial seizure were associated with reduced risk. Factors in combination influenced the risk of recurrent seizures, sometimes in opposite ways. RESUME Recidives de crises comiiiales apres une premiere crise hyperthertnique: une approche niuliivariee L’article rapporte les resultats d’une etude longitudinale de I55 enfants hollandais apres une premiere crise hyperthermique. Parmi ces enfants initialement non traites, 37 pour cent avaient eu au moins une crise ulterieure, 30 pour cent a u moins deux crises et 17 pour cent au moins trois crises. La periode sensible pour le retour de crises etait comprise entre 12 et 24 mois, le risque ulterieur etant quatre ou cinq fois plus faible; apres n’importe quelle crise, le risque maximal se situait dans les six mois suivants, declinant rapidement apres six mois sans crises. Une histoire familiale de n’importe quel type de crise dans les relations au premier degre conduisait a prevoir de multiples recidives; a un i g e d’au moins 30 mois et a une temperature egale ou superieure a 40°C lors de la crise initiale, le risque de recidive etait reduit. Les facteurs associes modifient le risque de recidive, agissant parfois de facon opposee. ZUSAMMENFASSUNG Wiederauftrelen von A nfallen nach dern ersten Fieberkrampf: eine Multivarianzanalyse Es werden die Ergebnisse einer Kontrollstudie bei 155 hollandischen Kindern nach ihrem ersten Fieberkrampf dargestellt. Von diesen zunachst unbehandelten Kindern hatten 37 Prozent mindestens einen, 30 Prozent zwei und 17 Prozent drei weitere Anfalle. Die vulnerable Phase fur Anfallsrezidive nach dem ersten Fieberkrampf war zwischen 12 und 24 Monaten, danach war das Risiko vier bis funf Ma1 geringer; nach jedem Anfall war das Risiko innerhalb der ersten sechs Monate a m hochsten und fie1 nach sechs anfallfreien Monaten stetig ab. Ein Anfallsleiden bei Familienangehorigen ersten Grades bedeutete multiple Rezidive. Ein Alter von mindestens 30 Monaten und eine Temperatur von 2 4 0 ° C beim ersten Krarnpfanfall waren mit einem geringeren Risiko verbunden. Die Kombination von Faktoren beeinfluate das Risiko fur Anfallsrezidive, gelegentlich in kontrarer Wiese. RESUMEN Recurrencia de convulsiones despues de una convulsion febril. Enfoque multivariado Se presenta 10s resultados d e un estudio sobre el curso de 155 nillos holandeses despues d e una convulsion febril. De estos nillos, no tratados inicialmente, el 37 por ciento habian tenido por lo menos una, el 30 por ciento por lo menos dos y el 17 por ciento por lo menos tres convulsiones consecutivas. El period0 vulnerable para las convulsiones recurrentes despues de la primera convulsion febril fue de 12 a 24 meses. Mas adelante el riesgo era d e cuatro a cinco veces menor. Despues de cualquier convulsion el riesgo era maximo dentro d e 10s primeros seis meses, para disminuir despues de forma constante despues de seis meses sin convulsiones. Una historia familiar de primer grado de cualquier tip0 d e convulsion era predictiva de multiples recurrencias. Una edad de por lo menos 30 meses y una temperatura de 2 4 0 ” en la convulsion inicial iban asociados a un riesgo reducido. La combinacion de diversos factores influye sobre el riesgo de convulsiones recurrentes a veces en sentidos opuestos. References Aicardi, J . (1986) ‘Febrile convulsions.’ In Epilepsy in Children. New York: Raven Press. pp. 212-23 1. Annegers, J . F., Hauser, W . A., Shirts, S. B., Kurland, L. T. (1987) ‘Factors prognostic of unprovoked seizures after febrile convulsions.’ New England Journal of Medicine, 316, 493-498. Berg, A. T., Shinnar, S., Hauser, W . A,, Leventhal,
J . M . (1990) ‘Predictors of recurrent febrile seizures: a meta-analytic review.’ Journal of Pediatrics, 116, 329-337. Cox, D. R. (1970) The Analysis of Binary Data. London: Chapman & Hall. Ellenberg, J. H., Nelson, K. B. (1980) ‘Sample selection and the natural \istory of disease. Studies of febrile seizures. Journal of the American Medical Association, 342, 1337-1 340.
23
. i
...
2
.-N W
rA
El-Radhi, A . S., Banajeh. S. (1989) ‘Effect of fever o n recurrence rate of febrile convulsionb.’ -
Archives of Disease in Childhood, 64, 869-870. Withana, K., Banajeh, S. (1986) ‘Recurrence
seizures.’ It7 Nel5on. K . B., Ellenbere, J . H . ( E d \ . ) Febrile Seiaires. New York: Raven Pre\\. p. 2 Nelson, K . B., Ellenberg, J . H . (1978) ‘Prognosi\ i i i children with febrile seizure,.’ Pediurric.y. 61.
rate of febrile convulsions related to the degree o f pyrexia during the first attack.’ Clinical Pediatrics, 25, 3 1 1-3 13. Farwell, J . R.. Lee, Y. J . , Hirtz, D. G., Sulrbacher, S. I . , Ellenberg, J . H., Nelson, K. B. (1990) ‘Phenobarbital for febrile seizures-effects on intelligence a n d o n seizure recurrence.’ NeM’ England Journal of Medicine, 332, 364-369. Forsgren, L., Sidenvall, R., Blomquist, H . K., Heijebel, J . (1990) ‘A prospective study of the incidence of febrile convulsions.’ Acia Paediarrica
Offringa, M., Hazebroek-Kainschreiir, A . A . J . M.. Derksen-Lubsen, G . (1991) ‘Prevalence of febrile seizures in Dutch schoolchildren.’ Paediutric. und Perinaral Epidemiology, 5, I 8 I- 188. Rantala, H . , Uhari, M.. T u o k k o , H . (1990) ‘Viral infections a n d recurrence of febrile convul\ioiis.’ Journal of Pedia/rics, 1 16. 195- 199. Shirts, S. B., Annegers, J . I:., Hauser, W .A. (1987) ‘The relation of age at first febrile scizure to recurrence of febrile seizures.’ Epilepsiu, 28, 6 2 5 .
Scandinavica. 79, 550-557.
W
rA
-
Press. p. 24. Kalbfleish, J. D., Prentice, R. L. (1980) The Statistical Analysis qf Fuilure Ti~rieData. N ~ M
population-based cohort o f children, Rochester. M N . ’ Neurology, 37 (Suppl. I ) , 149. (Ahstruct.) Thaler, H . T . (1984) ‘Nonparametric estimation of the hazard ratio.’ Journal of the A n ~ e r i c ~ n
Journul of the Americon Slalistical Associalion, 53, 203-223. Knudsen, F. U . (1985) ‘Recurrence risk after first
febrile seizure a n d effect o f short-term diazepam prophylaxis.’ Archives of Disease in Childhood,
60, 1045-1049. -
(1988) ‘Frequent febrile episodes a n d recurrent febrile convulsions. ’ A ctu Neurologicu Scundinavico, 78, 414-417. Lahat, E., Katz, V., Bistritzer, T., Eshel, G . , Aladjem, M . (1990) ‘Recurrent seizures in children with Shigella-associated convulsions.’ Annals of Neurologv, 28, 393-395. Millichap. I . G . (1981) ‘The definition of febrile
24
( Abstract.)
Gardner, M . J . , Altman, D. G . (1989) Statistics with Confidence. London: British Medical Journul
York: J o h n Wiley. Kaplan, E. L., Meier, P . (1958) “on-parametric estimation from incomplete observations.’
.-2
720-727.
Hauser, A , , Annegers. J . F., Kurland. L. T.
(1987) ‘Risk of recurrence of febrile seizures i n
:I
Stutistical Association, 19, 290-293.
Tsuboi, T. (1986) ‘Seizures of childhood. A population-based a n d clinic-based study.’ Actu Neurologica Scandinavica, Suppl. 110, 130. Verity, C. M., Butler, N . R., Golding, J . (1985) ‘Febrile convulsions in a national cohort followed up from birth. 1: Prevalence a n d recurrence in thc first year of life.’ Britisl7 Medical Journal, 290. 1307-1 3 10. Wallace, S. (1988) The Child wiih Febrile Sei,-irre.s. Guildford: Butterworth Scientific. pp. 15-23. Zvulunov, A . , Lerman, M., Ashkenazi, S., Weitz. R., Nitzan, M . , Dinari, G. (1990) ‘The prognosis of convulsions during childhood shigellosis.’ Eirropear7 Journal of Pediatrics, 149, 293-294.
