Phenomenology of prolonged febrile seizures

Population.

Eligible were children, ages 1 month through 5 years, who presented with FSE, defined as a seizure or a series of seizures without full recovery in between lasting ≥30 minutes^12,13 that also met the definition of FS.^12,14 FS was defined as a provoked seizure where the sole acute provocation was fever (temperature >38.4°C, 101.0°F) without prior history of afebrile seizures and with no evidence of an acute CNS infection or insult.^14,16 Children with known severe neurologic disability prior to entry were excluded. The five recruiting sites were Montefiore and Jacobi Medical Centers in the Bronx, Children’s Memorial Hospital in Chicago, Duke University Medical Center, Virginia Commonwealth University Hospital, and Eastern Virginia Medical School. In addition, the International Epilepsy Consortium at Virginia Commonwealth University was the Data Coordinating Center and the Epidemiology/Biostatistics Core was based at Columbia University.

Recruitment.

Emergency department (ED) and hospitalization records were reviewed daily to identify potential subjects., The trigger to screen was any child with FS reportedly lasting ≥15 minutes. However, to be eligible, the study team had to determine that seizure duration was ≥30 minutes. Families of eligible subjects were approached for consent to enroll their child in the study. The procedures were approved by the Institutional Review Boards at all participating institutions.

Procedures.

Subjects were recruited within 72 hours of the episode of FSE. As part of the diagnostic evaluation during the initial hospitalization, the following procedures were performed:

1. Detailed history of the circumstances of the seizure with a focus on duration and focality

2. A structured interview including developmental milestones, past medical history, family history of FS, and epilepsy in first-degree relatives and characteristics of the seizure

3. Neurologic and physical examination

4. Laboratory data summarized including CSF results with >20 WBCs in the CSF considered an exclusionary criterion. However, lumbar punctures were done for clinical indication and absence of a lumbar puncture was not an exclusionary criterion

5. MRI examination

6. EEG

7. Specimens for acute human herpesvirus (HHV)-6 and HHV-7 assays

In addition, serum samples from probands and both parents are sent to the National Institute of Neurological Disorders and Stroke repository at Coriell Institute for Medical Research¹⁵ for future genetic analysis.

Consensus process for seizure semiology and duration.

In addition to local coding, FSE was classified by three independent central raters (J.M.P., D.N., S.S.). Materials available for the review included the medical record, including the ED records and the ambulance call sheets, if available, the structured interview, and the comments from the local study team. FSE was coded as continuous or behaviorally intermittent. Seizures were then classified in accordance with the International League Against Epilepsy (ILAE) seizure classification as generalized tonic clonic, simple partial, complex partial, or partial with secondary generalization.¹⁶ Seizure classification was based exclusively on clinical features of the seizures and did not rely on imaging or EEG. Finally, the raters classified, based on the medical record, whether the treating clinicians in the ED recognized that the child had SE or not. Disagreements across classifications were discussed and consensus achieved in all cases. A detailed description of the overall methodology and the classification and consensus process can be found in the supplemental detailed methodology file on the Neurology® Web site at www.neurology.org.

Statistics.

Descriptive analysis was conducted through the determination of a percentage, mean, or median. Standard deviations are presented for means and the interquartile range (IQR) for the median. For the comparison of intermittent vs continuous seizures, comparison of means was performed using the t test, comparison of medians using the Wilcoxon Rank Sum test, and comparison of frequencies using the χ² test.¹⁷ For the comparison of each rater to the consensus determination, inter-rater reliability was calculated using the κ statistic for categorical variables and a 95% CI was constructed.¹⁸ A κ ≥0.75 was considered to be excellent agreement beyond chance, a κ of 0.41 to 0.74 was considered to be good agreement, and values of ≤0.40 were considered to be poor agreement. For continuous variables, inter-rater reliability was determined using the intraclass correlation coefficient comparing each of the three raters.

Seizure duration was modeled as a Weibull distribution with two parameters, α or the scale parameter, and β or the shape parameter.¹⁹ A best fit was obtained for the function

S(T) = e^−(t/α)β)

Fit of the Weibull model was assessed with the Log Likelihood statistic.¹⁹ The predicted survival curve based upon the computed shape and scale parameters was compared to the Kaplan Meier survival curve.¹⁹

Population characteristics.

As of April 15, 2007, we have recruited 146 subjects. Among the 119 children whose seizure semiology has been evaluated in consensus, the median age was 1.3 years (table 1). Development was normal in 86% and clearly abnormal in only 8%. Approximately 24% had a prior history of FS. A first degree family history of FS was present in 25% and of epilepsy in 9%. The racial and ethnic distribution reflected the populations served by these medical centers, 34% Caucasian, 38% African American, 23% Hispanic, and 6% other.

Illness characteristics.

Most of the children had not sought prior medical care for the illness associated with the FSE episode (76%, table 2). Clear viral illnesses predominated (54%) followed by otitis media (18%), pneumonia (5%), and unknown source (14%).

Fever.

The characteristics of the fever are summarized in table 2. The mean peak temperature was 103.2°F (median 103.0°F). Temperatures were normally distributed. In 30 (25%) children, the seizure was the first manifestation of the febrile illness as duration of recognized fever was less than 1 hour. Twenty-five children (21%) were not febrile in the ED but developed their fever within 24 hours of the episode of FSE.

Characteristics of the episode of FSE.

The characteristics of the FSE are summarized in table 3. The median duration was quite long at 68.0 minutes, even though the entry criteria specified ≥30 minutes and the criteria to screen for eligibility was ≥15 minutes. A substantial proportion had seizures lasting more than 2 hours (24%). In 87% seizures did not stop spontaneously, ceasing only after the administration of a benzodiazepine class drug. Despite the relatively long seizure durations and the active intervention required to stop them, the ED clinicians, while providing appropriate therapy in most cases, consistently underestimated the duration and failed to recognize SE in over a third of the cases, including children with continuous seizures.

A logistic regression was performed to assess the association between clinical characteristics and seizure duration using ≤68 minutes vs >68 minutes for the analysis. There was no significant association between seizure duration and peak temp (OR = 0.97), temperature in the ED (OR = 0.97), FSE focality (OR = 0.95), and family history of FS (OR = 1.6) or of epilepsy (OR = 0.98).

A plot of the distribution of seizure durations is shown in the figure, A. The data are best fit by a Weibull distribution with parameters:

• Download figure
• Open in new tab
• Download powerpoint

Figure Duration of febrile status epilepticus in FEBSTAT Cohort

(A) Distribution of the duration of febrile status epilepticus (n = 119). Circles represent the actual data points which is equivalent to a Kaplan-Meier curve. The smooth line represents the best fit by a Weibull distribution with shape parameter = 1.68. (B) Probability that a seizure that has continued to x minutes will not stop between x minutes and x + Δx minutes. Probability, or hazard function, is calculated from the Weibull best fit curve with a shape parameter of 1.68 shown in (A). It is the derivative of the Weibull best fit curve.

S(T) = e^{−(t/95.0)1.68}

These parameters, and in particular the shape parameter of 1.68, suggest that in this population the longer a seizure lasts, the less likely it is to spontaneously stop (figure, B).

Continuous and behaviorally intermittent seizures were equally common. Intermittent seizures consisted, in almost all cases of repeated convulsions, without recovery in between. They were somewhat longer (median 82 minutes, IQR = 55–120) than continuous seizures (median 55 minutes, IQR = 45–85, p = 0.002) but median duration of total convulsive time was less (38 minutes, IQR = 26–60 for intermittent seizures and 55 minutes, IQR = 45–85 for continuous seizures, p < 0.0001). The median duration of the longest identifiable convulsion in the intermittent group was 25 minutes (IQR = 15–38) and the median duration of the total estimated convulsive time was 38 minutes (IQR = 26–60). There were no differences between continuous and intermittent FSE with respect to characteristics of the children (age, family history), characteristics of the illness or fever, or seizure type (focal vs generalized).

Prior work has shown that FSE is more likely to be focal^5,20 and indeed the majority of seizures were focal (table 3). Focality was not related to duration of FSE (67% of 60 children with FSE lasting ≥68 minutes and 68% of 59 children with FSE lasting < 68 minutes). Almost all FSE (99%) is convulsive, whether generalized tonic clonic or partial with secondary generalization. In the 77 children with partial seizures, focal onset consisted of focal jerking of an extremity or face in 33 (43%). Other features at onset included staring (n = 51, 66%), eye deviation to one side (n = 36, 47%), head turning to one side (n = 33, 43%), and a period of blankness or impaired consciousness prior to the convulsion (n = 29, 38%). Focal features not present at onset included asymmetry of the convulsive seizures (n = 30, 39%) and a Todd’s paresis (n = 10, 13%).

Mortality and morbidity.

Two children have died to date; both were developmentally delayed. One child died a few weeks after FSE from complications of pulmonary disease and the other child returned to baseline after FSE and died a few months later. In neither case was the death believed to be related to the episode of FSE.

Inter-rater reliability.

We compared the preconference readings for each of the three raters with consensus, which was achieved in all cases after conferencing. There was unanimity on whether this was a FS and on whether it was FSE. Inter-rater reliability was excellent for continuous vs behaviorally intermittent FSE (κ = 0.90, 0.90, 0.90). There was also excellent agreement on seizure duration with an intraclass correlation coefficient of 0.75 among the three observers. Inter-rater reliability concerning recognition of SE in the emergency department was good, ranging from 0.60 to 0.81.

As expected, there was less agreement on seizure focality with κ values of 0.54, 0.74, and 0.89 for each rater compared to consensus. Disagreements were occasionally due to the rater missing a statement in the multiple source documents that was clearly indicative of focality. More often, disagreements centered on the validity of various observations, particularly when information was discrepant or incomplete. Disagreements were not random as one observer was less likely to call a seizure focal. Agreement on definite lateralization in focal seizures was somewhat better with κ values of 0.62, 0.82, and 0.84 compared with consensus.