The NEW ENGLAND JOURNAL of MEDICINE

FEBRILE seizures are the most common type of seizure and occur in 2 to 4 percent of all children.1 2 3 4 Approximately one third of children who have a febrile seizure have a recurrence.2 , 4 5 6 Only the age at the time of the first febrile seizure is a consistent predictor of the risk of recurrent febrile seizures,5 although recent studies have found an increase in risk associated with a family history of febrile seizures.7 8 9 A family history of epilepsy, a complex febrile seizure as the first seizure, and neurodevelopmental abnormalities have been identified as indications for anticonvulsant prophylaxis to prevent recurrence of febrile seizures,6 but these factors are not consistently associated with a higher risk of recurrence.5 Few other predictors of recurrence have been identified.

We conducted this study to identify predictors of recurrent febrile seizures. In addition to reexamining factors that have been studied in previous reports, this study was specifically designed to examine the episode of illness during which the initial febrile seizure occurred in order to identify other predictors of recurrent febrile seizures.

Methods

Children with first febrile seizures were enrolled in an ongoing prospective cohort study from June 1989 through February 1991 and were followed to determine whether febrile seizures recurred. The children were identified through the pediatric emergency departments of Bronx Municipal Hospital Center, North Central Bronx Hospital, and Montefiore Medical Center in Bronx, New York, and Yale–New Haven Hospital in New Haven, Connecticut. The logs for all visits to the four emergency departments were reviewed twice each week to identify potentially eligible patients. Nine children seen after a first febrile seizure in the private practice of one of the authors were also included and are grouped with the patients at Montefiore Medical Center.

Eligibility

A febrile seizure was defined as a seizure that occurred while the child had a rectal temperature of at least 101°F (38.3°C) or an axillary temperature of at least 100°F (37.8°C) documented either in the emergency department or in the history. Children with previous febrile or unprovoked seizures were excluded, as were children with intracranial infections. Some definitions of febrile seizures have lower and upper age limits (for example, three months to five years6). Such limits reflect conventions but are not founded on an etiologic basis and are not universally accepted.7 In this study, children between the ages of 1 month and 10 years were considered eligible. For a child to be included, his or her parents or guardian had to have a telephone and had to speak either English or Spanish.

Interview with Parents

The parents of children who were identified in the emergency department logs as having had first febrile seizures received a letter that explained the study and invited them to participate. A research assistant trained in interviewing techniques by the investigators then called and obtained their informed consent. The parents were interviewed with use of a standardized questionnaire in either English or Spanish. The interviewer first asked a brief series of screening questions to verify that the child had never previously had a febrile or unprovoked seizure; children with such a history were excluded. A history of neonatal seizures was not grounds for exclusion,10 but such a history was reported only once. The records of all interviews were reviewed for errors, and the responses were coded on an ongoing basis by the principal investigator.

The interviewer elicited a complete description of the seizure from the parent or, if the parent did not witness the seizure, from someone who did. Descriptions of seizures were compiled from the medical records and the interview, and each seizure was classified on the basis of this information by an expert in epileptology and pediatric neurology. If necessary, the parent or other witness was called back for clarification. As in previous studies,2 , 7 seizures were considered complex if they were focal in onset, if they were prolonged, or if there was more than one seizure during the course of the illness. Any seizure with secondary generalization was classified as a focal-onset seizure.11 Todd's paresis was taken as evidence of a focal seizure. A seizure was classified as prolonged if it lasted more than 10 minutes.7 If a child had more than one seizure without fully regaining consciousness between seizures, the seizures were counted as one continuous seizure.11 A febrile seizure could have more than one complex feature.

Information was also obtained about the illness during which the seizure occurred, including the duration of fever before the seizure (<1 hour, 1 to 24 hours, or >24 hours), the highest body temperature that was recorded before or just after the seizure, and other symptoms of illness. Parents were asked whether there was a family history of febrile and unprovoked seizures and whether the child had any preexisting neurodevelopmental problems.

Information about the seizure, the symptoms of the illness, the diagnosis, the child's temperature (both as reported by the parents and on arrival at the hospital), and the results of laboratory tests was obtained from the medical records. Temperatures measured at the hospital were considered the most reliable because they were measured and recorded by medically trained personnel. With few exceptions, temperatures were recorded in degrees Fahrenheit. There was a predominance of temperatures recorded in whole degrees, suggesting a preference on the part of care givers for rounded measurements and hence error in the recording of temperatures. Because this error is linked to the Fahrenheit scale, temperatures were not converted and analyzed in degrees Celsius. Celsius equivalents are provided in parentheses. Occasionally, the temperature was not recorded or was less than 101°F despite a diagnosis of febrile seizure. When it was available, the temperature measured at home and recorded in the chart while the child was in the hospital was considered an adequate substitute. There were some children (n = 14) who were included in the cohort on the basis of the parents' report of a temperature of at least 101°F and a diagnosis of febrile seizure in the medical record but for whom no temperature was

recorded in the chart. These 14 children were, however, excluded from the analysis of temperature and risk of recurrence. Children were ineligible for the study if the sole documentation of fever was a report based on tactile assessment.

Follow-up

After the initial interview, the patents were called every three months to determine whether the child had had any further seizures. Whenever possible, documentation of the recurrence was obtained from the medical record. The follow-up period for this report extended from study entry through March 1992.

Statistical Analysis

We used statistical methods that take into account the variable length of follow-up.12 13 14 15 The product-limit method was used to calculate the risk of recurrence at various times after the first seizure.12 Standard errors and 95 percent confidence intervals were calculated with an approximate Greenwood formula.14 The results are displayed by means of Kaplan–Meier curves. Univariate and multivariate analyses were performed for both continuous and discrete variables with the Cox proportional-hazards model.12 13 14 The rate ratio was used as a measure of the magnitude of the association between a variable and the risk of recurrence. For the purposes of this report, data were censored (treated as though follow-up had terminated) when a child had an unprovoked seizure.

Results

There were 356 children who met all the entry criteria and whose parents were interviewed. One parent declined further participation in the study. Eight children were excluded because no follow-up information could be obtained after the initial interview. This report is therefore based on the 347 children for whom at least one follow-up call was made: 119 (34 percent) from Yale–New Haven Hospital, 154 (44 percent) from Bronx Municipal Hospital Center, 48 (14 percent) from North Central Bronx Hospital, and 26 (7 percent) from Montefiore Medical Center. The median length of follow-up was 20 months (range, 2 to 31). Of these 347 children, 335 (97 percent) were followed for 6 months or more, 283 (82 percent) for 12 months or more, and 203 (59 percent) for 18 months or more. Seven children were treated continuously with phenobarbital, and phenobarbital was prescribed for three children to be administered only during a febrile illness, despite the acknowledged ineffectiveness of this kind of therapy.6 , 16 So far, 32 children (9 percent) have been lost to follow-up after less than one year.

Overall Risk of Recurrence

Ninety-four children (27 percent) had recurrent febrile seizures. The risk of recurrence was 14 percent (95 percent confidence interval, 10 to 18 percent) at 6 months, 25 percent (95 percent confidence interval, 20 to 30 percent) at 12 months, 27 percent (95 percent confidence interval, 22 to 32 percent) at 18 months, and 30 percent (95 percent

confidence interval, 24 to 35 percent) at 24 months (Fig. 1Figure 1Kaplan–Meier Curve for the Risk of a Recurrence after a First Febrile Seizure.). These results are comparable to those of most other studies.5 There were no appreciable differences in the risk of recurrence among the four hospitals.

Predictors of Recurrence

Table 1Table 1 Univariate Association between Each Variable and the Risk of Recurrence.* lists the risk factors that were examined, the number of children with each risk factor, the number and percentage who had recurrences, and the univariate rate ratio (with 95 percent confidence intervals) from the proportional-hazards model.

Duration of the Recognized Fever

One year after the initial febrile seizure, the risk of recurrence was 44 percent (95 percent confidence interval, 32 to 55 percent) for children whose parents had been aware of the fever for less than 1 hour before the seizure occurred, 23 percent (95 percent confidence interval, 17 to 29 percent) for those with fever lasting 1 to 24 hours, and 13 percent (95 percent confidence interval, 5 to 20 percent) for those with fever lasting more than 24

hours (Fig. 2Figure 2 Kaplan–Meier Curves for the Risk of a Recurrence after a First Febrile Seizure as a Function of the Duration of Recognized Fever before the First Febrile Seizure.). The unadjusted rate ratio for the duration of fever before the seizure (<1 hour, 1 to 24 hours, or >24 hours) was 0.48, indicating an approximate halving of the recurrence rate with each reduction from one category to the next shorter one.

Temperature

A rectal temperature of at least 101°F (38.3°C) was recorded at the hospital for 312 (90 percent) of the children. Twenty-one children (6 percent) had temperatures below 101°F when measured in the emergency department; however, the temperature according to the history as recorded in their medical records was 101°F or more. For 14 children, information about the temperature was missing or it was below 101°F. In the interview, the parents of these children reported measuring a temperature of 101°F or more, and the

diagnosis in the medical record was febrile seizure. These 14 children were included in the cohort but were excluded from the analyses of temperature.

The risk of recurrence decreased with increasing temperature (Fig. 3Figure 3

Kaplan–Meier Curves for the Risk of a Recurrence after a First Febrile Seizure as a Function of the Child's Temperature When Evaluated for the First Febrile Seizure.). The risk of recurrence one year after the initial febrile seizure was 35 percent (95 percent confidence interval, 17 to 56 percent) for children whose temperature was recorded as 101°F (38.3°C), 30 percent (95 percent confidence interval, 18 to 41 percent) for those with temperatures of 102°F (38.9°C), 26 percent (95 percent confidence interval, 17 to 36 percent) for those with temperatures of 103°F (39.4°C), 20 percent (95 percent confidence interval, 12 to 29 percent) for temperatures of 104°F (40.0°C), and 13 percent (95 percent confidence interval, 2 to 23 percent) for temperatures of 105°F (40.6°C) or higher (P for trend = 0.024). The unadjusted rate ratio was 0.82 for every increase of 1 degree Fahrenheit above 101°F, indicating an average 18 percent decrease in the recurrence of febrile seizures for each additional 1 degree Fahrenheit.

The association between temperature and the risk of recurrence persisted after stratification according to the duration of recognized fever, especially among children whose seizures occurred less than one hour after the recognition of fever. Among these children the risk of recurrence at 12 months according to temperature was 71 percent for children with temperatures of 101°F, 48 percent for 102°F, 44 percent for 103°F, 26 percent for 104°F, and 0 percent for 105°F or more.

Age at First Febrile Seizure

The median age at the time of the first febrile seizure was 18 months (range, 1 month to 9.5 years). Eleven children (3 percent) were more than five years of age when the first febrile seizure occurred. The risk of recurrence one year after the first febrile seizure in children less than 18 months old was 30 percent (95 percent confidence interval, 23 to 37 percent), and that in children 18 months old or older was 18 percent (95 percent confidence interval, 12 to 24 percent).

Family History

Among the 340 children for whom information on family history was available, 81 (24 percent) each had a first-degree relative (parent or full sibling) who had had febrile seizures. At one year, the risk of recurrence was 36 percent (95 percent confidence interval, 25 to 46 percent) for those with a family history of febrile seizures and 20 percent (95 percent confidence interval, 15 to 26 percent) for those without such a family history.

Fifteen children (4 percent) each had a first-degree relative with a single unprovoked seizure or epilepsy. At one year the risk of recurrence was 36 percent (95 percent confidence interval, 11 to 61 percent) for those with a family history of unprovoked seizures or epilepsy and 24 percent (95 percent confidence interval, 19 to 28 percent) for those without such a family history.

Complex Febrile Seizures

There was no evidence of an increased risk of recurrence among children whose initial febrile seizure was complex. Furthermore, no single complex feature (including status epilepticus) or combination of complex features was associated with a substantial increase in the risk of recurrence.

Neurodevelopmental Abnormalities

There was no evidence of an increased risk of recurrent febrile seizures among the 21 children (6 percent) with neurodevelopmental abnormalities. Seven (33 percent) of these children had unprovoked seizures, at which point data on them were censored from further analysis of recurrent febrile seizures. Five of these seven began therapy with anticonvulsant medications.

Sex and Race or Ethnic Group

The risk of recurrence was not significantly higher among boys than among girls. The racial and ethnic distribution of the children in this study reflects the fact that the populations from which the patients were drawn were largely black and Hispanic. There was no indication that the risk of recurrence was greater in any of the three ethnic or racial groups studied (blacks, whites, and Hispanics).

Multivariable Analysis

To control for individual effects of related variables, the Cox proportional-hazards model was used for multivariable analyses. Children for whom data were missing on one or more variables (n = 22) were excluded from this analysis. All the variables described above were tested simultaneously in the model. Only the variables with a statistically significant association with recurrence were retained. These were temperature, duration of fever before the seizure, age at onset, and family history of febrile seizures (Table

2Table 2 Multivariable Cox Proportional-Hazards Model of Predictors of Recurrent Febrile Seizures.*). The nonsignificant risk factors were entered into the final model one at a time to generate an estimate of the rate ratio after adjustment for the significant risk factors (Table 2).

Discussion

Febrile seizures were once considered a form of epilepsy.17 18 19 Largely as a result of two epidemiologic studies,2 , 10 febrile seizures are now recognized as a benign syndrome distinct from, although associated with, epilepsy.20 In the wake of an era in which continuous phenobarbital prophylaxis was routinely used even for a single simple febrile seizure21 , 22 and in the light of the growing evidence of the adverse side effects of phenobarbital,23 24 25 the 1980 Consensus Development Panel devised guidelines for decreasing the use of drug prophylaxis and identified limited circumstances in which it might still be appropriate to consider prophylactic treatment.6 Of the factors that were identified by the panel, only a young age at onset has been consistently associated with an increased risk of recurrence.5 The other factors — a family history of epilepsy, complex features, and neurodevelopmental abnormalities — are predictors of later epilepsy in children who have febrile seizures.2 , 10 , 26 , 27 Current data indicate that anticonvulsant prophylaxis against febrile seizures does not alter the risk of unprovoked seizures.6 , 26 , 27 Hence, although the guidelines were meant to limit the use of anticonvulsant agents to the children most likely to benefit from therapy, for the most part they did not identify those at highest risk for recurrent febrile seizures.

In addition to the factors identified in the consensus-development conference, we examined the circumstances in which the first febrile seizure occurred, specifically the duration of recognized fever before the seizure and the temperature. Temperature has been correlated with the risk of recurrence in two smaller studies.9 , 28 In a much larger sample, we found a trend toward decreasing risk with increasing temperature. This finding is consistent with a threshold model of febrile seizures and suggests that different children have different temperature thresholds above which they may have a seizure. Temperatures measured at the hospital after the seizure are only approximations of the children's temperatures at the time of the seizures. Furthermore, there was evidence suggesting inaccuracies in the recording of temperature. In general, such errors attenuate the association between temperature and the risk of recurrence.29 Thus, it is likely that the association is even stronger than we have described. Of course, it is also unlikely that the height of the fever is the sole factor triggering a febrile seizure.

The duration of recognized fever before the seizure was another strong predictor of recurrence. The biologic mechanisms underlying this association are not yet clear. Longer follow-up may help clarify whether this pattern is consistent over the course of several recurrences and may determine how this information can best be used in planning and evaluating strategies for prevention.

Most previous studies of febrile seizures have reported an association between young age at onset and an increased risk of recurrence.5 , 9 This relation appears to be due to the longer period during which a younger child is at risk, rather than a greater tendency to have seizures.9 , 30

All studies that have focused on a history of febrile seizures in a first-degree relative, including this one, have found such a history to be associated with an increase in the risk

of recurrence.7 8 9 The results of studies that examined the effect of a family history of epilepsy are conflicting. A large study in Rochester, Minnesota, found virtually no difference in the risk of recurrence among children with a family history of epilepsy (25 percent) and those without such a history (23 percent).7 Our study results tend to agree with those of the Rochester study, although ours are less conclusive.

The literature on the relation between complex febrile seizures and the risk of recurrence is also inconsistent. The findings of our study are similar to those of the National Collaborative Perinatal Project.2 Others, however, have reported that either specific complex features9 or complex features in certain subgroups of patients7 were associated with an increased risk of recurrence.

A large proportion of seizures in our study were complex. It is possible that the procedure for recruiting patients through emergency departments led us to identify children with more severe seizures and caused the underrepresentation of children with simple seizures who were treated in a pediatrician's office and sent home. Although this may have happened to a small degree, it is uncommon for a child with a first febrile seizure not to be referred to an emergency department for a full sepsis workup, including a lumbar puncture. The high proportion of complex seizures was due largely to our methods of obtaining the descriptions of seizures.31 Information from parents about features of complex seizures was frequently missing from medical records; such records have generally been the primary source of information about seizures in other studies,7 although some investigators have not specified how this information was collected.9 , 32 Furthermore, the degree of agreement among three independent pediatric neurologists in classifying each seizure as complex or simple was good to excellent.33 Thus, it is unlikely that the high proportion of complex seizures was due to the selection of the patients or to idiosyncrasies in our classification criteria.

Finally, children with neurodevelopmental abnormalities were not at increased risk for recurrent febrile seizures, although several had unprovoked seizures. Other studies have already demonstrated that this is a group at high risk for epilepsy.2 , 10 There is, however, no evidence that treating the first febrile seizure reduces that risk.26 , 27 Moreover, treatment is associated with a high incidence of cognitive and behavioral side effects.23 24 25 Thus, there does not appear to be any justification for treating a child after a single febrile seizure, even if neurodevelopmental problems are present.

Some pediatric neurologists argue strongly against long-term therapy, even for complex or recurrent febrile seizures.34 , 35 A recently proposed alternative is intermittent treatment with agents such as benzodiazepines or the use of antipyretics only when fever is present.36 , 37 The risks and benefits of these approaches have yet to be fully explored.

Because of the tendency in the past to consider febrile seizures a form of epilepsy, most studies of recurrent febrile seizures have focused on predictors of epilepsy. With the exception of age, these studies, as a group, have not identified consistent, strong predictors of recurrence. We focused on factors related to the occurrence of the initial febrile seizure, not just those related to the later development of epilepsy. The duration of

recognized fever and the temperature appear to provide useful information about which children are at very high risk and which at very low risk for recurrence.

Supported by a FIRST award (1R29 NS27728) to Dr. Berg and a grant (1R01 NS26151) to Dr. Shinnar from the National Institute of Neurological Disorders and Stroke.

Presented in part at the American Epilepsy Society Meeting, Philadelphia, December 8–11, 1991.

We are indebted to the interviewers, Nancy Bassett and Toby Gillman, for their dedicated work on this project and to all the parents of the children who participated in this study; to Drs. Mervyn Susser and Zena Stein, who supported this project in its earliest phases; to Dr. Theodore Holford, who kindly provided consultation throughout the study; and to the New York City Health and Hospitals Corporation, Bronx Municipal Hospital Center, North Central Bronx Hospital, Montefiore Medical Center, and Yale–New Haven Hospital for their cooperation.

Source Information

From the Department of Pediatrics, Yale School of Medicine, New Haven, Conn. (A.T.B., E. D.S.); the Departments of Neurology (S.S., M.A.) and Pediatrics (S.S., M.E.S., E.F.C.) and the Montefiore—Einstein Epilepsy Management Center (S.S., M.A.), Montefiore Medical Center, Albert Einstein College of Medicine, Bronx, N.Y.; and the Gertrude Sergievsky Center, Columbia University College of Physicians and Surgeons, New York (S.S., W.A.H.). Address reprint requests to Dr. Berg at the Department of Pediatrics, Yale University School of Medicine, 333 Cedar St., New Haven, CT 06510