- Top of page
Purpose: No population-based study has investigated the risk of autism spectrum disorders (ASDs) in children after unprovoked seizures with onset in the first year of life. Our objective was to determine whether infantile spasms were related to risk of ASD as compared to unprovoked seizures with onset in the first year of life after adjusting for symptomatic origin of seizures.
Methods: This is a population-based case-control study nested in a cohort of children with unprovoked seizures in the first year of life. The cohort comprised 95 children, 34 boys and 61 girls. Cases were defined as children with ASD, controls were without ASD, and exposure was a history of infantile spasms. The Mantel-Haenszel test and logistic regression were used to calculate the odds ratio (OR) and 95% confidence intervals (CI).
Results: The crude OR for ASD associated with infantile spasms was 5.53 (95% CI 1.25–23.06). Stratification on age and gender did not change the OR. The OR for ASD associated with infantile spasms adjusted for symptomatic seizures was 1.55 (95% CI 0.33–7.37), while the OR for ASD associated with symptomatic seizures adjusted for infantile spasms was 8.73 (95% CI 1.88–40.54). Restriction to mental age 24 months or higher yielded higher ORs.
Discussion: Infantile spasms predicted high risk for ASD, but this was to a large extent explained by the association of ASD with symptomatic origin of seizures.
A recent review cited a number of studies that found increased rates of epilepsy (5% to 38.3%) in individuals with autism and autism spectrum disorders (ASDs) (Tuchman & Rapin, 2002). It is well documented, that in some cases, seizures are present prior to the diagnosis of ASD (Olsson et al., 1988; Wong, 1993; Steffenburg et al., 2003; Danielsson et al., 2005). This sequence is particularly evident in infantile spasms, since their onset is during the first year of life in 94% of cases (Hrachovy & Frost, 2003), but ASD is rarely diagnosed until after 2 years of age, although symptoms may be present earlier.
Population-based studies on infantile spasms that have reported on the prevalence of ASDs show a considerable range in their results (9% to 35%) (Riikonen & Amnell, 1981; Sidenvall & Eeg-Olofsson, 1995; Saemundsen et al., 2007a). A recent population-based study on the association between seizures in the first year of life (other than infantile spasms) and ASDs reported a relatively high prevalence of ASD (7%) (Saemundsen et al., 2007b). According to the above-mentioned studies, the prevalence of ASDs among both children with infantile spasms and those with other unprovoked seizures in the first year of life seems to exceed the prevalence of ASDs in the general population, which ranges from 0.5% to 1% in recent studies (Chakrabarti & Fombonne, 2005; Baird et al., 2006; Fombonne et al., 2006; Ellefsen et al., 2007; Jonsdottir et al., 2007).
Infantile spasms may indicate a more serious encephalopathy leading to more severe cognitive impairment than other seizures in the first year of life, and thus increasing the risk of ASD, since intellectual disabilities and ASDs are closely related (de Bildt et al., 2005). However, children with cryptogenic origin of infantile spasms have a better outcome in general than those with symptomatic origin (Hrachovy & Frost, 2003). This inevitably generates the hypothesis that it is the etiology of early seizures that is associated with ASD, not the seizure type as such (Olsson et al., 1988; Askalan et al., 2003).
Previous electroencephalography (EEG) studies have suggested that the risk of autism and ASD in children with infantile spasms is increased when there is a history of epileptiform temporal lobe focus, both in case series with mixed etiologies and among cases with tuberous sclerosis complex (Riikonen & Amnell, 1981; Bolton et al., 2002).
Our objective was to determine whether infantile spasms were related to the risk of ASD as compared to unprovoked seizures with onset in the first year of life, after adjusting for symptomatic origin of seizures.
- Top of page
This is a population-based case-control study nested in a cohort of children with unprovoked seizures in the first year of life. The cohort was compiled from two studies of Icelandic children, one on children with infantile spasms in the first year of life detected during the period 1981–1998, and the other on children with unprovoked seizures in the first year of life, other than infantile spasms, detected during the period 1982–2000. The sources of children with infantile spasms and unprovoked seizures were hospital records from all three in-patient pediatric facilities in Iceland as reported previously (Saemundsen et al., 2007a, 2007b), thus a brief summary will be given. The cohort in the present study is based on the overlapping period in both studies, from 1 January 1982 to 31 December 1998. Altogether, 121 children made up this cohort, with 25 children having a history of infantile spasms in the first year of life and 96 children having had other unprovoked seizures with onset in the first year. Of these, five had died and one lived abroad. Initially, the parents of the remaining 115 children were invited into a study of possible ASDs, of whom 95 (82.6%) children participated, 34 boys and 61 girls. The parents were contacted and questioned whether they had concerns regarding developmental skills or behavior of the child or whether the child had been diagnosed with a neurodevelopmental disorder. A girl with unprovoked seizures during the first year of life was discovered after the publication of the second study (Saemundsen et al., 2007b). She was included in the present study and later found to be with ASD.
As an initial test of autistic behaviors, the Social Communication Questionnaire (SCQ) (Rutter et al., 2003a) was used for all children with a known neurodevelopmental disorder or parental concern about the development or behavior of the child. A score of 15 points on the SCQ is the recommended cutoff to distinguish between ASDs and other diagnoses (Rutter et al., 2003a). However, in order to minimize the possibility of missing ASD in cases with milder sympomatology or in high functioning individuals, a score of 10 points was chosen. Diagnostic instruments (Saemundsen et al., 2007b) were applied when behavioral scores reached or exceeded this cutoff. The diagnostic instruments consisted of the Autism Diagnostic Interview-Revised (ADI-R) (Lord et al., 1994), the Autism Diagnostic Observation Schedule (Lord et al., 2001), and the Childhood Autism Rating Scale (Schopler et al., 1988). For a diagnosis of autism, autistic behaviors must be present to a certain degree in three symptom domains: (1) impaired social interaction, (2) impaired communication, and (3) stereotyped and/or repetitive behavior. In addition, abnormality in development has to be evident at or before 36 months according to the International Classification of Diseases-Tenth Revision (ICD-10) (WHO, 1993). However, in the case of atypical autism and Asperger's syndrome, a less severe symptomatology is to be expected, and the onset prior to 36 months is not a requirement.
The data gathering regarding possible ASD took place during the calendar years 2002–2005. The mean age at assessment was 11 years and 2 months (sd = 4 years 8 months; range 4–20 years).
Epilepsy was defined as recurrent episodes of unprovoked, paroxysmal seizure activity (Cowan, 2002). Seizures were classified as nonsymptomatic (cryptogenic) if the children had normal neurological and developmental history, normal neurological examination, no known associated etiological factor, and negative diagnostic evaluation. All other cases were classified as symptomatic (Ludvigsson et al., 1994). Symptomatic seizures were present in 22 children (perinatal asphyxia, 4; porencephalic cyst, 2; developmental delay, 2; trisomy 21, 2; cortical dysplasia, 2; seizures prior to infantile spasms, 2; tuberous sclerosis, 1; hypertensive encephalopathy, 1; Aicardi syndrome, 1; subdural hemorrhage, 1; perinatal stroke, 1; congenital toxoplasmosis. 1; Sturge-Weber syndrome, 1; and septic shock, 1). The classification into symptomatic and nonsymptomatic was performed at diagnosis of infantile spasms or unprovoked seizure, except for one child who, 2 years later, was found to have cortical dysplasia on magnetic resonance imaging.
Altogether, 89 children out of 95 had EEGs, and of these, 28 were considered abnormal based on reports. All children diagnosed with infantile spasms had an EEG, which in all cases showed hypsarrhythmia. As previous studies have found association of ASD in infantile spasms with epileptiform temporal lobe focus (Riikonen & Amnell, 1981; Bolton et al., 2002), we looked at the type of EEG pattern recorded during the first year of life and classified the EEG tracings as ever epileptiform temporal focus, either left or right.
All the parents gave their written informed consent, as did the older and more able children. The study was approved by the Data Protection Authority and the National Bioethics Committee in Iceland.
The incidence estimate of unprovoked seizures in the first year of life was based on data from Statistics Iceland using the official population on 31 December of each year of the inclusion period of the cohort 1982 to 1998 (Statistics Iceland, 2007). Total person years of observation were 74,043. Incidence per 100,000 was calculated, and 95% confidence intervals (CI) were found, assuming Poisson distribution (Armitage & Berry, 1991).
Cases in this study are defined as children with ASD, and the controls are all other children without ASD in the cohort. History of infantile spasms, the exposure variable, was compared for cases and controls stratified separately on age (born 1990 and earlier, or later) and gender, using the method of Mantel-Haenszel. If the expected values were less than five, Fisher’s exact test was used to calculate 95% CI, and this was completed in Epi-Info 6. Univariate analysis was performed for ASD associated with infantile spasms and separately for ASD associated with symptomatic origin of seizure. A multivariate case-control analysis was performed using logistic regression. The adjusted odds ratio (OR) and exact computation of 95% CI were calculated using the SPIDA software package (Gebski et al., 1992). Infantile spasms versus seizures other than infantile spasms were treated as a dichotomous variable. Year of birth was treated as a continuous variable expressed in years, and gender was treated as a dichotomous variable. The classification of seizures into symptomatic and nonsymptomatic (cryptogenic) was treated as a dichotomous variable. Another separate regression analysis was performed, however, after a restriction was applied to cases and controls with mental age 24 months or higher. The reason for the restriction was that the diagnosis of autism based on the ADI-R is of constrained validity in individuals with mental age less than 24 months (Rutter et al., 2003b).
- Top of page
Altogether, over the 17-year inclusion period, 121 children had unprovoked seizures, regardless of type, during the first year of life. The incidence of all unprovoked seizures in the first year of life was 163.4 per 100,000 person years (95% CI 135.6–195.3).
Among the participants of this study (N = 95), 17 or 17.9% had infantile spasms, and 78 or 82.1% had other types of seizures. The majority of the children had epilepsy (more than one unprovoked seizure) or 87.4%. Most of the seizures were classified as nonsymptomatic (76.8%). Females outnumbered males by 1.8 to one. One-fourth (24/95) of the sample had intellectual disability. Thirteen children (13.7%; 95% CI 7.5–22.3) had ASD, eight females and five males, either childhood autism (ICD-10: F84.0) or atypical autism (ICD-10: F84.1). Six of the children with ASD had infantile spasms, and seven had other types of epilepsy (Table 1). All but one of the children with ASD had intellectual disability, and six had profound intellectual disability (intellectual quotient < 20). The distribution of the EEG classification is also shown in Table 1.
Table 1. Characteristics of cases and controls
| ||N = 13 (%)||N = 82 (%)|
|Infantile spasms|| 6 (46.2)||11 (13.4)|
|Symptomatic seizures|| 9 (69.2)||14 (17.1)|
|Epilepsy (> 1 seizure)||13 (100)||70 (85.4)|
|Males|| 5 (38.5)||29 (35.4)|
|Born 1990 or earlier|| 6 (46.2)||36 (43.9)|
|Mental age < 24 months|| 4 (30.8)||8 (9.8)|
|EEG recorded||12 (92.3)||77 (93.9)|
| Abnormal EEG|| 11 (91.7)a|| 17 (22.1)a|
| Epileptiform, right TLb|| 1 (8.3)a|| 3 (3.9)a|
| Epileptiform, left TLb|| 4 (33.3)a|| 4 (5.2)a|
The crude OR for ASD associated with infantile spasms was 5.53 (95% CI 1.25–23.06). The Mantel-Haenszel OR for ASD associated with infantile spasms was 5.52 (95% CI 1.24–22.30) stratified on year of birth (born 1990 and earlier or later) and 5.88 (95% CI 1.26–27.02) stratified on gender. The OR for ASD associated with infantile spasms adjusted for year of birth and gender was 5.81 (95% CI 1.52–22.25), so the differences between crude, stratified, and adjusted estimates were minor. Thus, age and gender were not adjusted for in the remainder of the regression analysis.
The OR for ASD associated with infantile spasms in the regression analysis, unadjusted, was 5.53 (95% CI 1.57–19.54). The OR for ASD associated with symptomatic origin of seizures in the regression analysis, unadjusted, was 10.93 (95% CI 2.95–40.53).
Table 2 shows the adjusted OR taking into account whether the seizures were of symptomatic origin. The OR was 1.55 (95% CI 0.33–7.37) for children with infantile spasms compared to those without infantile spasms, adjusted for symptomatic origin of seizures. The OR was 8.73 (95% CI 1.88–40.54) for children with symptomatic origin of seizures compared to those with nonsymptomatic seizures, adjusted for infantile spasms.
Table 2. Adjusted OR and 95%CI from logistic regression of autism risk among cases and controls according to infantile spasms and symptomatic origin of seizures
| ||Cases||Controls|| |
| ||N = 13||N = 82||ORa||95% CI|
Table 3 shows the results after restriction to cases and controls with mental age 24 months or higher. The OR was 2.59 (95% CI 0.34–19.96) for children with infantile spasms compared to those without infantile spasm, adjusted for symptomatic origin of seizures. The OR was 18.00 (95% CI 3.33–97.29) for children with symptomatic origin of seizures compared to those with nonsymptomatic seizures, adjusted for infantile spasms.
Table 3. Adjusted OR and 95%CI from logistic regression of autism risk among cases and controls according to infantile spasms and symptomatic origin of seizures after restriction to individuals with mental age at or above 24 months
| ||Cases||Controls|| |
| ||N = 9||N = 74||ORa||95% CI|
|Infantile spasms||3|| 5|| 2.59||0.34–19.96|
|Symptomatic seizures||6|| 6||18.00||3.33–97.29|
Excluding from the calculations the girl who was discovered later, the crude OR for ASD associated with infantile spasms was 6.45 (95% CI 1.41–28.49). The exclusion of this case did not materially affect the estimates of the ORs or the confidence limits in the other analyses.
- Top of page
The incidence of all unprovoked seizures of 163.4 per 100,000 person years in the first year of life in the present study (period 1982 to 1998) was somewhat higher than the estimate reported in a prospective study in the population with the same geographical definition (period 1995–1999), in which the incidence was 130.2 per 100,000 person years (95% CI 77.1–205.7) (Olafsson et al., 2005). The similarity of these incidence estimates, found with different surveillance methods, supports the view that our cohort reflects the status in the population.
Our calculations show a significant association between seizures of symptomatic origin and risk of ASD adjusted for whether the seizures were infantile spasms or not. The OR was high, and the 95% CI did not include unity. The association between infantile spasms and risk of ASD adjusted for symptomatic seizures was not statistically significant, although the ORs were 1.55 or higher in the analyses. The observed association is supported by evidence from other studies (Bolton et al., 2002; Askalan et al., 2003), suggesting a causal association of ASD with neuropathological phenomena rather than the seizures themselves. The results are also in agreement with a recent study on children with active epilepsy and intellectual disability, where no difference in the distribution of infantile spasms was found between groups with or without ASDs (Steffenburg et al., 2003).
The restriction in this study to those with mental age 24 months or higher was made to increase the diagnostic accuracy of ASD. After the restriction, the ORs for symptomatic seizures and infantile spasms were higher, further strengthening the association between ASD and symptomatic origin of seizures and infantile spasms.
All the cases had epilepsy, so in this study it was not possible to evaluate the role of epilepsy in association with ASD. As EEG is clinically indicated in epilepsy, and the EEG pattern is often found abnormal in epilepsy, introducing EEG tracings into the multivariate analysis would automatically invite circular reasoning.
Population studies of autism find three to four males to every female, but in present cases of ASD, derived from a cohort of children with seizure in the first year of life, the females outnumbered the males, indicating a unique position of these children as compared with other children with ASD ( Saemundsen et al., 2007b).
The primary information through a 17-year period on children with seizures in the first year of life, including the seizure type (i.e., whether infantile spasms or not and whether symptomatic or not) as well as the EEG tracings, was obtained from the comprehensive health care system in Iceland, ensuring the inclusion of seizure cases and the recognition of epilepsy cases. Through a cross-sectional study with an 83% participation rate, the SCQ was systematically used as the initial step in the search for autistic behaviors. Ascertainment and diagnosis of ASD, according to ICD-10, were based on current diagnostic instruments and clinical judgment by trained and experienced professionals, ensuring reliable diagnosis and diminishing the risk of misclassification of cases and controls (Saemundsen et al., 2007a, 2007b). The use of the comprehensive population register in Iceland enabled us to ascertain the vital and emigration status and the addresses of the children in the study. The data on the exposure variable in this study (i.e., whether the seizures were infantile spasms or not and whether the seizures were symptomatic or not) were obtained from the medical records of the cases and the controls. The medical records were noted before and unrelated to the formulating of the hypothesis of the present study and diagnosing of the cases, which eliminates the possibility of recall bias concerning the exposure variable.
The present study has some limitations. First, although the material is population-based, only children with known neurodevelopmental disorders or parental concern regarding developmental skills or behavior of the child received the SCQ as an initial test of autistic behaviors. A more thorough approach would be to test the whole cohort with the SCQ. Second, it was not possible to keep the professionals who were engaged in the diagnostic process blind to the results of the SCQ or other sources of information. Third, there was no formal reliability check between the different professionals who applied the diagnostic instruments, although all of them had undergone proper training in their use (Saemundsen et al., 2007b).
To our knowledge, this is the first population-based case-control study of ASD among children with seizures in the first year of life, adjusted for whether symptomatic in origin or not. Our results indicate that seizures of symptomatic origin may be a causative factor in ASD, which is in agreement with the understanding that multiple factors are associated with the risk of ASD. We cannot conclude on the possible role of infantile spasms in the etiology of ASD, as in this study, infantile spasms were not an independent risk factor for ASD. The clinical implication is that a history of seizures in the first year of life should attract attention to the possibility of subsequent ASD, particularly if seizures are of symptomatic origin. Further studies are indicated on the association of ASD with different types of epilepsy in the first year of life, as well as the possible contribution of pre- and postnatal risk factors.
- Top of page
The authors want to extend their gratitude towards the children and their parents for their participation in the study. Also, the authors thank the many professionals who assisted in the data gathering process. We thank Helgi Sigvaldason, ME, for his assistance in the statistical analysis. This work was supported in part by the Memorial Fund of Helga Jonsdottir and Sigurlidi Kristjansson and the Freemasons Fund of the Icelandic Order of Freemasons.
Conflict of interest: 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 have nothing to declare.