Abstract
Background: Febrile seizures affect 2 to 4% of children, and 2 to 10% develop subsequent unprovoked seizures. Secondary analyses of two large cohorts identified neurodevelopmental abnormalities, complex febrile seizures, and a family history of epilepsy as predictors of unprovoked seizures. We present an analysis of children prospectively followed from their first febrile seizure to reassess these three factors, examine factors of equivocal importance, and assess the importance of some new factors that we identified as predictors of recurrent febrile seizures. Methods: Children (N = 428) were prospectively identified for a first febrile seizure through pediatric emergency departments of four hospitals. Information was collected from medical records and interviews with parents. Children were followed for 2 years or more. Results: Unprovoked seizures occurred in 26 (6%). Neurodevelopmental abnormalities, complex febrile seizures, and a family history of epilepsy were associated with an increased risk of unprovoked seizures. Recurrent febrile seizures and brief duration of fever before the initial febrile seizure were also risk factors. A family history of febrile seizures, temperature and age at the initial febrile seizure, sex, and race were not associated with unprovoked seizures. Conclusions: We confirmed the increased risk associated with traditionally accepted predictors of epilepsy following febrile seizures. Also, the risk clearly increased with recurrent febrile seizures. In general, predictors of subsequent unprovoked seizures differ from predictors of recurrent febrile seizures. One notable exception, brief duration of fever before the initial febrile seizure, predicts both types of outcome and may be a marker for an increased susceptibility to seizures.
NEUROLOGY 1996;47: 562-568
Febrile seizures are the most common type of seizure and occur in 2 to 4% of all children. [1-4] The greatest concern for children with febrile seizures is the possibility of unprovoked seizures or epilepsy. Prior to the two large epidemiologic studies of the 1970s of the prognosis of febrile seizures, [2,5] this risk was thought to be very high. Many considered febrile seizures a form of epilepsy. [6-10] These epidemiologic studies demonstrated that the risk of epilepsy in children with febrile seizures is low, 2 to 10%, and identified neurodevelopmental abnormalities, family history of epilepsy, and complex febrile seizures as the primary predictors of unprovoked seizures in children with febrile seizures. [2,5] Both of these studies involved retrospective, secondary analyses of data.
We previously reported the risk and predictors of recurrent febrile seizures in a cohort of children prospectively identified and followed from the time of their initial febrile seizure. [11] The current report examines the risk and predictors of developing unprovoked seizures. Study factors include the three identified above, some factors of unclear significance--family history of febrile seizures and the number of recurrent febrile seizures--and two new factors that are important predictors of recurrent febrile seizures in this cohort, height of temperature and duration of fever prior to the initial febrile seizure. [11]
Methods.
Children with first febrile seizures were prospectively identified and followed to determine the risk and predictors of recurrent febrile seizures and of unprovoked seizures. Children were identified through the pediatric emergency departments of Bronx Municipal Hospital Center, North Central Bronx Hospital, and Montefiore Medical Center in Bronx, NY, from June 1989 through May 1991, and Yale-New Haven Hospital in New Haven, CT, from June 1989 through January 1992. The logs for all visits to the four emergency departments were reviewed twice weekly for potentially eligible patients. The details of this cohort have been published previously. [11]
Eligibility.
A febrile seizure was defined as a seizure that occurred with a rectal temperature >or=to101 degrees F (38.3 degrees C) or axillary temperature >or=to100 degrees F (37.8 degrees C) documented in the emergency department or by history. Children with prior febrile or unprovoked seizures were excluded as were children who had intracranial infections as the underlying illness accompanying the index seizure. Although some definitions of febrile seizures have age limits (e.g., 3 months to 5 years [12]), such limits reflect conventions but are not founded on an etiologic basis and are not universally used. [5] For this study, children between the ages of 1 month and 10 years were eligible. Because interviews were done by phone, a patient's parents or guardians had to have a phone. Children were excluded if the parents could not be interviewed in either English or Spanish. A history of prior provoked seizures, including neonatal seizures, was not grounds for exclusion; however, this occurred in only two children.
Interview.
The parents of patients who were identified in the emergency department logs received a letter that explained the study and invited them to participate. A research assistant who had been trained by the investigators called and obtained informed consent. Parents were interviewed, usually by phone, with a standardized questionnaire in English or Spanish. The questionnaire had been translated into Spanish and then checked for accuracy by a second translator.
The interviewer elicited a complete description of the seizure from the parent or person who witnessed the seizure. Descriptions of seizures from medical records and interviews were compiled. If necessary, the witness was called back for clarification. Consistent with previous studies and guidelines, seizures were considered complex if they were of focal onset, prolonged, or if there was >1 seizure during the illness episode. [2,4,5] Complex features were classified by one of the authors (S.S.). Seizures of partial onset that secondarily generalized were considered focal. [13] A Todd's paresis was evidence of a focal seizure. Seizures were prolonged if they lasted >or=to10 minutes. Duration was also classified as >or=to15 minutes and as status epilepticus (>or=to30 minutes). [13] If a child had >1 seizure without fully regaining consciousness between seizures, the seizures were counted as one continuous seizure. A febrile seizure could have more than one complex feature. Complex features of recurrent febrile seizures were classified in the same way. Medical documentation of recurrences was obtained whenever possible. For the purpose of analysis, complex features were considered for the initial febrile seizure only, as well as cumulated across the initial and all recurrent febrile seizures that occurred prior to the first unprovoked seizure.
A history of febrile seizures, single unprovoked seizures and epilepsy, and other provoked seizures was obtained for parents and siblings, as well as for other relatives. A history of clear neurodevelopmental delays or frank neurologic abnormalities identified by a pediatrician or other professional was obtained from the parent. Parents were also questioned about the illness during which the initial febrile seizure occurred. In particular, they were asked about the duration of recognized illness symptoms and specifically about duration of recognized fever prior to the seizure. They were asked to estimate this duration as precisely as possible. We then categorized the answers as <1 hour, 1 to 24 hours, and >24 hours.
Further information about the initial seizure, diagnosis, temperature measured and recorded at the hospital, and laboratory results were obtained from medical records. The criteria for temperature have been detailed previously. [11] Temperatures measured at the hospital were considered the most reliable because they were measured and recorded by medically trained individuals. With few exceptions, temperatures were recorded in degrees Fahrenheit and thus were not converted into the Celsius scale.
Follow-up.
After the initial interview, parents were called every 3 months to ascertain if there had been further seizures and the circumstances under which such seizures occurred. Information about complex features was elicited for all recurrent febrile seizures in the same manner as for initial febrile seizures. Whenever possible, documentation of the recurrence was obtained from the medical record. We attempted to follow all children for at least 2 years from the initial febrile seizure. A seizure was considered unprovoked if there were no acute precipitating circumstances to which the seizure might reasonably be attributed, such as fever or head trauma.
Analysis.
Statistical methods were used that account for the variable length of follow-up. [14-17] The product-limit method was used to calculate the risk at 2 years after the first febrile seizure. [14] Other analyses were performed with the Cox proportional hazards model. [16] The rate ratio was used as a measure of association between individual study factors and risk of unprovoked seizures. Children who had no unprovoked seizures were censored from further analysis when they were lost to follow-up prior to 2 years or when follow-up was terminated at the end of the study. Their time in the study was the interval from the initial febrile seizure to the date of the last follow-up call. Febrile seizures occurring before the first unprovoked seizure were treated as predictors of unprovoked seizures.
Results.
There were 445 children who met all entry criteria and whose parents were interviewed. Four parents refused further participation. Thirteen were excluded because no follow-up was obtained after the initial interview. This report is based on the 428 children for whom at least one follow-up call was made. Of these, 174 (41%) were from Yale-New Haven Hospital, 170 (40%) from Bronx Municipal Hospital, 58 (14%) from North Central Bronx Hospital, and 26 (6%) from Montefiore Medical Center. The median length of follow-up was 29 months (range, 2 to 44 months). Three-hundred ninety-eight children (93%) were followed for >or=to12 months, 372 (87%) for >or=to18 months, and 355 (83%) for >or=to24 months.
Overall risk of an unprovoked seizure.
Twenty-six children (6.1%) had >or=to1 unprovoked seizure. The risk of an unprovoked seizure with 95% confidence intervals (CI) was 3.8% (2.0, 5.7) and 6.2% (3.9, 8.6) at 1 and 2 years. There were no appreciable differences in the risk among the four hospitals.
Description of unprovoked seizures.
Of the 26 with unprovoked seizures, 13 had only one unprovoked seizure during the follow-up period. The others had two or more, thereby meeting the definition of epilepsy. [18] Eight children had partial-onset seizures, and 18 had generalized-onset seizures. These included three with absence epilepsy. Three children developed difficult-to-control epilepsy. All three were tried on multiple drugs without success during the time they were followed.
Predictors of unprovoked seizures.
(Table 1, Table 3) summarizes the results of all univariate comparisons.
Table 1. Univariate associations between each study factor and the risk of an unprovoked seizure. The number of children and the number (N) and percent (%) who had unprovoked seizures and the rate ratio with its 95% confidence interval (CI) derived from a proportional hazards model are shown for each study factor. (Table 3)
Table 3. Continued from (Table 1).
(1) Neurodevelopmental delay or abnormality.
Of 24 children with neurodevelopmental abnormalities, 8 (33%) had >or=to1 unprovoked seizure during the follow-up period. The risk of an unprovoked seizure at 2 years was 4.6% (95% CI 2.4, 6.7) among neurodevelopmentally normal children and 34.9% (95% CI 15.2, 54.6) among neurodevelopmentally abnormal children.
(2) Complex febrile seizures.
Complex febrile seizures were associated with an increased risk of unprovoked seizures. Compared with children with simple febrile seizures only, each complex feature was associated with an increased risk. The association was stronger when complex features were cumulated across recurrent febrile seizures rather than when characteristics of the initial seizure only were considered.
(3) Family history of epilepsy.
Information about family history was missing for seven children who were in foster care or adopted. Three of 17 children (17.7%) who had a first-degree relative with epilepsy had unprovoked seizures compared with 22 of 404 (5.5%) without. A history of epilepsy in any relative (regardless of degree of relatedness) was not strongly associated with unprovoked seizures.
(4) Family history of febrile seizures.
A family history of febrile seizures in either first-degree or any relative was associated with a small, nonsignificant decrease in risk of an unprovoked seizure. The confidence intervals tend to rule out a substantially increased risk associated with this factor.
(5) Number of febrile seizures.
There was a clear trend between the number of febrile seizures and the risk of unprovoked seizures. The risk and 95% confidence intervals at 2 years for an unprovoked seizure in children with 1, 2, 3, and 4 or more febrile seizures was 4.2% (1.9, 6.6), 6.1% (0.3, 1.18), 8.1% (0, 17.0), and 20.4% (4.4, 36.4).
(6) Duration of recognized fever.
A briefer duration of recognized fever prior to the initial febrile seizure was associated with an increased risk of unprovoked seizures. At 2 years, the risk and confidence intervals in children with duration of fever <1 hour, 1 to 24 hours, and >24 hours were 10.7% (4.1, 17.4), 6.1% (3.0, 9.2), and 1.1% (0, 3.4).
(7) Other factors.
There was no consistent association between temperature during the initial febrile seizure and unprovoked seizures. There was also no evidence of an association between age at the time of the initial febrile seizure and unprovoked seizures. Finally, no differences could be appreciated between male and female children or between children of black, Hispanic, or white origin.
Multivariable analysis.
To assess the independent effects of related variables, the Cox proportional hazards model was used for multivariable analyses. This was especially important in light of the facts that (1) the number of febrile seizures and the occurrence of complex febrile seizures accumulated over recurrences were both associated with the risk of unprovoked seizures, and (2) brief duration of fever prior to the initial febrile seizure is also a risk factor for recurrent febrile seizures in this cohort. [11] Children missing data on one or more variables (N = 21) were dropped from some analyses. In addition, separate models were developed for neurodevelopmentally normal and abnormal children.
Overall (*Table 2*, part A).
(Table 2) Neurodevelopmental abnormality, the number of febrile seizures, and complex febrile seizures were strong, statistically significant predictors of unprovoked seizures. Duration of recognized fever prior to the initial febrile seizure retained borderline significance (p = 0.059). Part of this factor's initial importance was explained by its association with recurrent febrile seizures; however, the multivariable results suggest it has additional importance beyond its association with recurrences. A family history of epilepsy was no longer statistically significant, although its adjusted rate ratio is consistent with other studies. [2,5] No individual complex feature was strongly enough associated with unprovoked seizures to be an independent predictor.
Table 2. Multivariable Cox Proportional Hazards Model
Neurodevelopmentally normal children (*Table 2*, part B).
(Table 2) The most important predictor of unprovoked seizures in this group was the number of febrile seizures. Complex febrile seizures were significant on univariate analysis, but only marginally so on multivariable analysis (p = 0.09). The adjusted rate ratio of 2.57 is not markedly different from that found in the overall analysis, so this may still be an important predictor in this group. A family history of epilepsy also was not significant. The number of patients with such a history is too small to have sufficient power to examine this factor satisfactorily. The adjusted rate ratio of 2.07 is consistent with other reports. [2,5] Duration of fever before the initial febrile seizure did not approach statistical significance, and the rate ratio was substantially closer to unity compared with what was found in the overall analysis. This tends to suggest it is not particularly important in this group.
Neurodevelopmentally abnormal children (*Table 2*, part C).
(Table 2) Despite the small number of children in this subgroup, three factors were substantial independent predictors of unprovoked seizures in this group. These were duration of fever, complex febrile seizures, and the number of recurrent febrile seizures. Status epilepticus was significant on univariate analysis. Above and beyond its contribution as a complex feature, it did not influence the risk of subsequent unprovoked seizures.
Paradoxical relation between risk factors for febrile seizure recurrence and unprovoked seizures.
A family history of febrile seizures, young age at onset, and low temperature at the time of the initial febrile seizure have been repeatedly identified as predictors of recurrent febrile seizures. [11,19-21] As recurrent febrile seizures were associated with an increased risk of unprovoked seizures in this study, it seemed paradoxical that these three would not be associated with unprovoked seizures, at least on univariate analysis. The reason for this was that neurodevelopmental abnormalities, the strongest predictor of unprovoked seizures, was strongly and negatively associated with family history of febrile seizures and young age at onset. There was no consistent association with temperature. In the neurodevelopmentally normal group, family history, age, and temperature were not associated with unprovoked seizures. The association between the number of recurrent febrile seizures and unprovoked seizures was examined after stratification on these three risk factors. This exploration suggested that, in children who are at high risk for recurrent febrile seizures (based upon the presence of these three risk factors), there is no increased risk associated with having recurrences. On the other hand, in children at low risk for recurrent febrile seizures, there is a substantially increased risk for unprovoked seizures if a child has recurrent febrile seizures.
Discussion.
The findings from our prospective cohort are similar to earlier studies with respect to the importance of neurodevelopmental abnormalities, complex febrile seizures, and a family history of epilepsy, even though this last factor was not statistically significant.
The total number of febrile seizures has been associated with an increase risk of unprovoked seizures in previous studies, although each time the association was limited to a subgroup of children. In one study, the number of recurrences was associated with an increased risk in developmentally suspect or abnormal children. [2] This designation was based upon standardized 8-month exams and included about 20% of that cohort. The Rochester study [5] found the risk of generalized-onset unprovoked seizures increased with the number of febrile seizures and decreased for partial-onset unprovoked seizures. We found a strong independent trend, not restricted to neurodevelopmentally abnormal children or to partial-onset unprovoked seizures. These data do not indicate that recurrent febrile seizures cause epilepsy. In fact, evidence from three randomized trials has demonstrated that prevention of febrile seizures does not alter the risk of subsequent epilepsy. [22-24] Additional evidence from related areas in epilepsy indicates that, in most cases, seizures do not cause further seizures. [25] Thus, our data should not be used to advocate treatment of febrile seizures for the purpose of preventing subsequent epilepsy.
The duration of fever before the first febrile seizure is the one factor clearly associated with both recurrent febrile seizures and subsequent unprovoked seizures. This goes against the earlier notion that seizures with fever that occur during the initial rise in temperature are the most benign. [26] This association awaits replication in independent studies. If, as we suspect, it represents a marker for overall seizure susceptibility, it could have significance for counseling parents, planning management strategies, and ultimately shedding light on the nature of the link between febrile seizures and epilepsy.
Previous studies of febrile seizures found young age at onset to be associated with an increased recurrence risk for febrile seizures. [2,4,5,11,19,26,27] This appears to be due to a longer period during which a child is susceptible and does not indicate a greater propensity toward seizures. [28] Our findings also suggest early age at first febrile seizure is not indicative of a general seizure susceptibility.
Our study demonstrated interesting differences in the prognostic values of a family history of epilepsy versus one of febrile seizures. As previously reported, family history of febrile seizures predicted recurrent febrile seizures in this cohort, whereas a family history of epilepsy did not. [11,27] For unprovoked seizures, the findings appear to be reversed. A family history of epilepsy, but not of febrile seizures, seems to predict unprovoked seizures.
We deliberately designed this study to assess complex features, and we prospectively collected this information in a standardized manner, soon after the events occurred. In addition, we classified complex features prior to determination of outcomes. The high (relative to other studies) proportion of complex seizures was largely due to the methods for obtaining seizure descriptions. [29] Inter-rater agreement for classifying each complex feature was good to excellent. [30] Thus, that the high proportion of complex seizures was due only to recruitment through emergency departments or idiosyncrasies in classification criteria is unlikely. Complex features were clearly associated with the risk of unprovoked seizures, although no single complex feature was more important than the others. Unlike the findings from Rochester, [5] the effects of individual complex features were not cumulative; that is, the expression of two or three instead of one feature was not associated with a greater risk.
The stronger association between unprovoked seizures and complex features in the neurodevelopmentally abnormal compared with normal children has not been previously reported. The confidence intervals for the rate ratios in these two groups overlap substantially. Thus these data do not demonstrate a clear difference in the prognostic value of complex seizures in neurodevelopmentally normal and abnormal children. Further attempts to replicate this finding are needed to fully appreciate its significance, if any.
Only one half of the children who experienced an unprovoked seizure had two or more unprovoked seizures. In previous studies from our group, we underscored that the risk of recurrence after a first unprovoked seizure in patients followed from the time of the first seizure is low. In interpreting our findings of the present study, there are two considerations. First, one third of the children with unprovoked seizures had neurodevelopmental delays or abnormalities and thus were at high risk for recurrent unprovoked seizures. [31] Second, although prospectively followed from the time of their first febrile seizure, unprovoked seizures were sometimes not recognized until they were already recurrent (e.g., children with absence seizures). Hence, these children presented with new onset epilepsy and would not have entered a study of prognosis following a first unprovoked seizure. Thus, the fact that 50% had recurrent unprovoked seizures cannot be used to infer the risk of recurrence after a first unprovoked seizure.
Large gaps remain in our understanding of the link between febrile seizures and epilepsy. Genetic factors are assumed to be important. Our findings for family history of febrile seizures and of epilepsy as predictors of recurrent febrile seizures versus as predictors of unprovoked seizures suggest the underlying genetics have some fairly specific components. The presence of neurodevelopmental abnormalities may be a susceptibility factor for seizures of any kind. [32,33] The role of complex features is not understood.
In children with febrile seizures, unprovoked seizures are still the exception, especially in neurodevelopmentally normal children. We believe that the widely accepted approach of not treating children with febrile seizures is correct. [34,35] The medications used to prevent seizures have side effects and are not entirely effective. Treatment does not prevent epilepsy, although the drugs may suppress seizures while the patient is on medication. To our knowledge, febrile seizures, even if prolonged, do not cause brain damage. Their most serious consequence is parental anxiety, which is best treated with education. [35,36]
Acknowledgments
A very special acknowledgment is owed Dr. W. Allen Hauser who collaborated with us on the initial phases of this project. Drs. Mervyn Susser and Zena Stein also provided necessary support and encouragement during the earliest phases. Dr. Theodore Holford kindly provided biostatistical consultation throughout the study. We thank the directors of the Pediatric Emergency Departments at the participating hospitals, Drs. Ellen Crain, Morton Salomon, and Eugene Shapiro. We also acknowledge the cooperation of the New York City Health and Hospitals Corporation and Bronx Municipal Hospital Center, North Central Bronx Hospital, Montefiore Medical Center, and Yale-New Haven Hospital.
- Copyright 1996 by Advanstar Communications Inc.
REFERENCES
Disputes & Debates: Rapid online correspondence
NOTE: All authors' disclosures must be entered and current in our database before comments can be posted. Enter and update disclosures at http://submit.neurology.org. Exception: replies to comments concerning an article you originally authored do not require updated disclosures.
- Stay timely. Submit only on articles published within the last 8 weeks.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- 200 words maximum.
- 5 references maximum. Reference 1 must be the article on which you are commenting.
- 5 authors maximum. Exception: replies can include all original authors of the article.
- Submitted comments are subject to editing and editor review prior to posting.