During the conduct of this research Dr. Tian was with the Department of Biostatistics, Joseph L. Mailman School of Public Health and Dr. Leary was with the Department of Pediatrics and the Department of Neurology, Columbia University College of Physicians and Surgeons, New York, NY.
Address correspondence to Dale C. Hesdorffer, Ph.D., GH Sergievsky Center, Columbia University, P&S Unit 16, 630 West 168th Street, New York, NY 10032, U.S.A.. E-mail: email@example.com
Objective: Whether magnetic resonance imaging (MRI) is informative in febrile seizures (FS) is unknown. We undertook a study to determine the frequency of MRI-detected brain abnormalities and to evaluate their association with FS type and with specific features of complex FS.
Methods: A prospective cohort study, from 1999 to 2004, included children with first FS from one Pediatric Emergency Department. MRI of the brain was performed within 1 week of the seizure. FS type was categorized by experts blind to the prior clinical history and MRI results. MRI examinations were read blind to the child's clinical history and FS type, and interviewers were blind to MRI results.
Results: In 159 children with a first FS, imaging abnormalities occurred in 12.6% (N = 20). Eight of the 54 with complex FS had imaging abnormalities compared to 12 of the 105 with simple FS (n.s.). Compared to children with simple FS, children with both focal and prolonged FS (N = 14) were more likely to have imaging abnormality (OR = 4.3, 95% CI = 1.2–15.0), even after adjustment for abnormal neurological examination. Imaging abnormalities included those known to be associated with seizures (e.g., focal cortical dysplasia) and those not typically associated with seizures (e.g., subcortical focal hyperintensities ≥ 5 mm).
Discussion: Our data suggest that brain abnormalities may lower seizure threshold in febrile children, predisposing to the development of FS. Clinical management was unaffected and therefore these data do not alter the recommendation that MRI is unnecessary in children with FS, without some other neurological indication.
Computed tomography (CT) scans are not recommended in FS (NIH, 1980; Al-Qudah, 1995), although preexisting brain abnormality has been proposed (Annegers et al., 1987) to explain the observation that complex, but not simple, FS are associated with increased risk for focal epilepsy. Several pathologies (Mathern et al., 1995; Lado et al., 2000), including mesial temporal sclerosis (MTS), may account for associations between prolonged FS and later epilepsy, but the scarcity of data limit conclusions. While there have been two studies of magnetic resonance imaging (MRI) in febrile status epilepticus (SE) (VanLandingham et al., 1998; Scott et al., 2003), MRI studies are lacking in cohorts with simple and complex FS of shorter duration, where such studies could contribute to understanding why children experience FS.
We undertook a prospective study of children with first FS (1) to determine the proportion with MRI-detected brain abnormalities, and (2) to evaluate whether these abnormalities were more common among children with complex FS than among those with simple FS or were associated with specific features of complex FS.
In this prospective cohort study, we identified children, aged 6 months to 5 years, with first FS by daily screening of The Morgan Stanley Children's Hospital of New York-Presbyterian Pediatric Emergency Department log and by review of pediatric hospital discharges with an ICD-9 code of 780.3 between March 1999 and April 2004. We defined FS as seizures occurring among children with a rectal temperature of at least 101° Fahrenheit (38.3oC), in the absence of a history of unprovoked seizures or concurrent central nervous system infection (NIH, 1980). Children with prior neonatal seizure were included.
Upon identification, we contacted the child's primary care physician to request permission for the study team to contact the child's parents or guardians. After obtaining permission from the parent, the physician notified the study team and the family was contacted to explain the study and offer participation.
Measures and procedures
If initial screening confirmed the first FS, we obtained informed consent and the child received an MRI of the brain within 1 week of the FS. Within 1 month, parents were interviewed regarding the child's medical history, demographics, and family history of FS and epilepsy in first-degree relatives. At the parent interview, children received a neurological examination, which was classified as abnormal if the following findings were identified; hypertonia, hypotonia, or microcephaly. Follow-up MRI examinations were performed at 1 year for those children whose families consented to the follow-up MRI.
Classification of febrile seizures
We classified FS as simple or complex: simple, if they were brief, nonfocal and did not recur within the febrile illness; and complex, if they were prolonged (at least 15 min), focal (including Todd's paralysis) or recurred during the same illness (NIH, 1980).
Study epileptologists (W.A.H., L.D.L.) classified seizures blind to the child's MRI findings and prior clinical history. Medical records and the detailed parent interview were separately examined to arrive at a summary classification, incorporating all information. Disagreements were resolved in consensus discussions.
The MRI examination was performed on a 1.5 Tesla magnet within 72 h of the first FS whenever possible, but no longer than 1 week after the seizure. We selected this window in order to be able to see acute seizure-related changes on the MRI. Animal data suggest that 1 week is the maximal window for detecting such abnormalities (Dubé et al., 2004). Follow-up MRI scans were performed 1 year after the baseline scan.
Children were evaluated, sedated with oral chloral hydrate or IV pentobarbital, and monitored and discharged by a physician and nurse from the Department of Radiology, according to hospital and departmental policies and procedures governing sedation of pediatric patients.
The MRI protocol included the following sequences: T1-weighted sagittal scout localizing images; proton-density and T2-weighted fast spin-echo axial and coronal slices with a 5 mm slice thickness and no gap; fast FLAIR axial images with 4 mm slice thickness and 1 mm gap; high resolution T2-weighted fast spin-echo coronal images (Tien et al., 1993) through the temporal lobes, using contiguous 3 mm sections; volumetric T1-weighted gradient-echo coronal images (Ashtari et al., 1991) of the whole brain using 124 contiguous 1.5 mm sections; ultrahigh-resolution T2-weighted fast multiplanar inversion-recovery axial images (Chan et al., 1998) of the frontal and parietal lobes using 4 mm slice thickness and a 1 mm gap; and axial diffusion images.
A single neuroradiologist with epilepsy expertise (SC) read MRI scans blind to FS type, medical history, and physical examination. Because nothing is known concerning structural abnormalities among children with first FS, the readings incorporated all findings. Findings considered incidental included single small (<5 mm) subcortical focal hyperintensities, cysts or mild cortical atrophy. Definite abnormalities on MRI were varied (Table 2) and included large (≥5 mm) or numerous subcortical and cortical focal hyperintensities or the presence of a structural abnormality likely to be associated with short- or long-term neurological disability. MRI examinations were classified as normal when all findings were normal or incidental and were classified as definitely abnormal when at least one finding was abnormal. Focal cortical dysplasia, type I, was diagnosed by the presence of blurring of the gray-white matter junction of the affected gyrus, whether or not it was accompanied by focal cortical thickening (Chan et al., 1998). Focal cortical dysplasia, type II, was diagnosed by the presence of subcortical white matter hyperintensity, as well as blurring of the gray-white matter junction and focal cortical thickening of the affected gyrus.
Table 2. Types of MRI-detected brain abnormalities in children with a first febrile seizure
Type of abnormality
MRI: magnetic resonance imaging.
aThe number of abnormalities (N = 28) is greater than the number of children with abnormalities (N = 20) because some children had more than one abnormality.
bIncluding one in the hippocampus that was mesial temporal sclerosis on follow-up MRI 1 year later and one bilateral basal ganglia subcortical focal hyperintensity.
cChiari I malformations.
dLeft temporal lobe.
Subcortical focal hyperintensity
Abnormal white matter signal
Focal cortical dysplasia
Gray matter heterotopia
Congenital midline abnormality
Focal decreased volume
Cortical focal hyperintensity
Enlarged lateral ventricle
Intrarater reliability was estimated on rereading of 68 scans, including all scans originally read as definite abnormality. Kappa was 0.65 (95% CI = 0.47–0.83), indicating good reliability for classification as normal or incidental findings versus definite abnormality. To eliminate interviewer bias, MRI readings were stored securely and were separate from the child's research folder.
This study was approved by the IRB at Columbia University. Parents gave written informed consent.
Data were analyzed with SAS (SAS Institute, Cary, NC), using Student's t-test for continuous variables and χ2 or Fisher's exact test for categorical variables with testing at the two-tailed level of 0.05. For all analyses, we classified MRI abnormalities as definite abnormality or no abnormality. The association between FS type and imaging abnormality was evaluated using logistic regression. Analyses compared imaging abnormality in complex versus simple FS, and evaluated each feature of complex FS compared to simple FS. As suggested by studies of febrile SE, (VanLandingham et al., 1998; Scott et al., 2003) we also examined the association between imaging abnormality and focal and prolonged FS compared to simple FS.
We evaluated several factors that are associated with the occurrence of FS as potential confounders of the association between FS type and imaging abnormalities. These were selected based upon information on FS risk factors as associations with imaging abnormalities were unknown and included rectal temperature in the ED (<104 °F vs. ≥104 °F), age (<18 months vs. ≥18 months), abnormal neurological examination, duration of febrile illness, and first-degree family history of FS and epilepsy. We also evaluated gender and ethnicity.
We identified 397 children, aged 6–60 months, with a first FS. Parents of 118 children (29.7%) refused participation, 39 (9.8%) could not be contacted, 80 (20.2%) agreed to participate, but the child remained too sick to sedate within the time frame for the MRI examination, and for 1 (0.3%) the MRI examination was not technically adequate. Participants (N = 159) did not differ from nonparticipants (N = 238) in gender, age, FS type, or temperature at the time of the FS, suggesting that participants are a representative sample of children with first FS from our community.
Among the 159 children included in this analysis (Table 1), 72 (45.3%) were female, 75 (47.2%) under 18 months of age, 3 (1.9%) with known developmental disability (i.e., mental retardation or cerebral palsy), 33 (20.7%) with first degree family history of FS, 10 (6.3%) with family history of epilepsy, and 134 (84.3%) were Hispanic, reflecting 2,000 census information for the community surrounding the hospital. Fifty-four children (34.0%) presented with complex FS (Table 1).
Table 1. Characteristics of children with a first febrile seizure (FS)
No brain imaging abnormality N (%)
Definite brain imaging abnormality N (%)
All cases N (%)
ap = 0.06, excluding the one child with missing examination.
bAbnormal neurological examination included any of the following: hypertonia, hypotonia, and microcephaly.
cFamily history in first degree relatives only.
dReported as the percentage of complex FS.
Number of subjects
Age in months
Family history of FSc
Family history of epilepsyc
Prolonged duration onlyd
Prolonged duration and repeatedd
Prolonged duration and focald
Focal and repeatedd
Prolonged duration, focal, and repeatedd
The baseline MRI examination was normal (N = 104) or revealed incidental findings (N = 35) in 139 children (87.4%) and definite abnormalities in 20 (12.6%). The most common type of definite imaging abnormality was subcortical focal hyperintensity (N = 9 of 28, 32.1%; Table 2 and Fig. 1). Next most common were abnormalities of white matter signal (Fig. 2) followed by focal cortical dysplasia.
Among children with definite baseline imaging abnormality, 14 (70.0%) had one abnormality, 4 (20.0%) had two abnormalities, 1 (5.0%) had three abnormalities, none had four abnormalities, and 1 (5.0%) had five abnormalities (two of which were considered incidental). Among children with definite abnormalities, only the child with five abnormalities had incidental abnormalities. When the distribution of number of abnormalities per child was compared in subgroups, there was no statistical difference for complex versus simple FS or for subtypes of complex FS, although numbers were small.
Definite baseline imaging abnormality was detected on the MRI of 8 children (14.8%) with complex FS and 12 (11.4%) with simple FS (n.s.). However, imaging abnormalities were related to a subtype of complex FS that included a combination of individual features. Specifically, children with both focal and prolonged FS (N = 14) were more likely to have imaging abnormality compared to children with simple FS (N = 105) (OR = 4.3, 95% CI = 1.2–15.0 for definite abnormality). Adjustment for abnormal neurological examination did not substantially decrease this association (adjusted OR = 4.0, 95% CI = 1.1–14.4 for definite abnormality). Separate adjustment for other potential confounders did not change the results, nor did exclusion of the three children with neurodevelopmental abnormality since birth (i.e., mental retardation or cerebral palsy).
We examined whether other factors, aside from the phenomenology of the febrile seizure, predicted the presence of imaging abnormality and found that no factor was significantly associated with imaging abnormality (Table 1). In a further exploratory analysis stratified by FS type, abnormal neurological examination was associated with imaging abnormality in complex FS (OR = 6.3; 95% CI = 1.1–36.6) but not in simple FS (OR = 1.3; 95% CI = 0.1–11.8).
Focal cortical dysplasia was Taylor I in all cases; one in the left medial occipital lobe, one in the left posteriomedial temporal lobe and one in the left posterioinferior temporal lobe. Both heterotopias were single, right-sided and periventricular.
We further examined the significance of the subcortical focal hyperintensities, which were the most common abnormality found in first febrile seizure. Abnormalities were found in the subcortical white matter (N = 1; 12.5%), deep white matter (N = 5; 62.5%), medial temporal structures (N = 0) and basal ganglia/thalamus (N = 2, 25.0%). Two were bilateral in the basal ganglia, one bilateral in the thalamus, and one bilateral in the corona radiate/centrum semiovale. The remaining five were right-sided in the periventricular space (N = 1), frontal lobe (N = 1), and corona radiate/centrum semiovale (N = 3). Among the five right-sided subcortical focal hyperintensities, focal febrile seizures occurred in three. Among the four bilateral subcortical focal hyperintensities in three children, two had focal febrile seizures. Overall, subcortical focal hyperintensities were more common in children with complex (N = 7, 13.0%) compared to simple FS (N = 1, 0.9%, p = 0.001).
Overall 1-year MRI examinations were completed in 47.2% of the cohort (N = 75). Among the children with definite abnormalities at baseline (N = 20), 9 (45.0%) had 1-year MRI examinations, all of which showed persistence of the baseline abnormality. One year follow-up MRI scans were performed in five of the eight children with subcortical focal hyperintensities and all showed persistence of the abnormality. Among the four children with abnormalities of white matter signal, only two had follow-up scans at 1 year. Both scans showed persistence of the abnormal white matter signal.
Among the 139 children without definite abnormality at baseline, 66 (47.5%) received 1-year MRI examinations, and 62 (93.9%) were read as normal at 1 year. Scans of the remaining four children were read as: abnormal white matter signal in a child whose initial scan was read as normal because the child was too young to regard the pattern as abnormal; subcortical focal hyperintensity in a child with a possible abnormality at baseline; cortical focal hyperintensity; and midline abnormality that was not observed on the baseline scan but during the interval growth of the brain and skull, the inferior tip of the cerebellar tonsils relocated in the posterior spinal space at the C1 level.
Although we found a higher than anticipated rate of MRI imaging abnormalities in children with first FS (12.6%), there was a low rate of imaging abnormalities that are expected to increase seizure risk (e.g., focal cortical dysplasia). We were unable to include an imaged control group, but we obtained comparison information from a the NIH Study of Normal Brain Development, a study that employs population-based sampling to collect neuroanatomical and clinical/behavioral data from normal children aged 7 days to 18.3 years (Almli et al., 2007). To date, among the children 6 months to 5 years of age in the Normal Brain Development study for whom brain imaging was performed, there were no brain abnormalities on baseline MRI scans (Personal communication).
Some of the abnormalities found in children in our study (e.g., focal cortical dysplasia and gray matter heterotopia) are abnormalities associated with epilepsy in the literature (Guerrini & Carrozzo, 2001; d'Orsi et al., 2004; Alonso-Nanclares et al., 2005). Other abnormalities (e.g., subcortical focal hyperintensities and possible delayed myelination), while common in our study, have not been linked to epilepsy. Because they have not been found in normal children in this age group (Personal communication), our findings suggest that these abnormalities, although not traditionally considered abnormal in children with seizures, may be associated with a lower seizure threshold in febrile children. As suggested by studies of febrile SE (VanLandingham et al., 1998; Scott et al., 2003), we found a strong relationship between first focal and prolonged FS and preexisting brain abnormality on MRI that was not explained by other factors, including abnormal neurological examination.
In animal studies, seizures are more easily induced in hyperthermic environments than in normothermic or hypothermic environments (Tancredi et al., 1992; Lui et al., 1993). Furthermore, hyperthermia potently lowers seizure threshold in immature rats with experimentally induced neuronal migration disorders compared to control rats (Germano et al., 1996). Similarly, fever may lower seizure threshold in children, but preexisting imaging abnormalities may exacerbate this effect, increasing the risk for FS.
In one developmentally delayed child with febrile SE in our study, the baseline MRI showed subcortical focal hyperintensity in the hippocampus and mesial temporal sclerosis on the follow-up scan 1 year later (Fig. 3), by which time the child had developed afebrile complex partial seizures. Interestingly, this was the only child to show an acute hippocampal abnormality. This is probably because the frequency of febrile SE in this cohort was only 5%, consistent with other studies of first FS (Berg et al., 1995). Others have identified acute hippocampal oedema following prolonged FS (VanLandingham et al., 1998; Scott et al., 2003), with infrequent subsequent hippocampal atrophy (VanLandingham et al., 1998).
Our study has multiple strengths, including MRI imaging of children within 72 h of a first febrile seizure, the detailed data collected, and the efforts to minimize bias through blinding. Most importantly, the study examined a common childhood condition, FS, using a method that has not previously been systematically applied. The study examined brain structure in first FS and found that MRI abnormalities are associated with focal and prolonged febrile seizure.
Potential limitations of the study include the low participation rate, the absence of concurrent MRI in a matched control group, and bias. None of these issues are believed to have had a significant impact on the results. First, the comparability of participants and nonparticipants makes it unlikely that selection bias existed. Second, for ethical reasons we were unable to obtain MRI examinations with sedation from an age matched control group of normal children; however, preliminary data from the NIH Study of Normal Brain Development suggest that normal children, aged 6 months to 5 years, have normal brains. Third, blinding was maintained to eliminate other major sources of bias.
Brain abnormalities occurred in 12.6% of children with first FS; 11.4% with first simple FS and 14.8% with first complex FS. Among children with complex FS, only focal and prolonged febrile seizures were associated with the occurrence of definite imaging abnormality. Our data suggest that brain abnormalities may be associated with a lower seizure threshold, predisposing to the development of FS. These abnormalities may also be associated with the development of recurrent febrile seizures and of afebrile seizures. These data did not affect clinical management and therefore do not alter the recommendation that MRI is unnecessary in children with FS, without some other neurological indication.
This work was supported by a grant from the National Institute of Child Health and Human Development (5R01 HD 36867). The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines. The authors would like to thank the families who participated in the study, as well as all those who worked on the study but do not qualify for authorship: Douglas Nordli Jr., MD; Thania Perez, MD; Martha Orbe, MD; Vivian Santiago, MS; Eva Gomez; Kathleen O'Neil, MS; Beatriz Plaza, MS; and Maria DeLaPaz-Debes, MD.
Conflict of interest statement: There are no conflicts of interest on behalf of any of the authors. Dr. Hauser is a consultant to Abbott and to Schwartz Bioscience, although his work in both cases is unrelated to the topic of this manuscript. The remaining authors report no competing interests.