Factors Associated With Seizure Onset in Children With Autism Spectrum Disorder

The Autism Treatment Network Registry Call-Back Assessment (RCBA) study was a longitudinal cohort study of children and adolescents with ASD. Individuals were included in the study if (1) they originally enrolled in the ATN from January 2011 through December 2013 at 1 of 12 sites in the United States and Canada that participated in the RCBA; (2) they met Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition,²⁶  criteria for autism, Asperger syndrome, or pervasive developmental disorder not otherwise specified; and (3) who had complete data for Vineland Adaptive Behavior Scales, Second Edition (Vineland II). Individuals from a 13th site who enrolled in the ATN from September 2015 through November 2016 and who met Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition,¹  criteria for ASD were also eligible for inclusion in the RCBA. Inclusion criteria for the ATN included age 2 years 0 months through 17 years 6 months. There were no exclusion criteria for enrollment.

Among eligible individuals at the first 12 sites (initially restricted to enrollees from January 2011 through December 2012), blocks of 65 participants from each site were identified at random by the coordinating center at Massachusetts General Hospital. Site coordinators selected participants in a randomly generated sequence for participation in the RCBA. Assent from the child or adolescent was obtained as well, as required by the review board at each site. Additional blocks were generated for sites if >15 from the original block could not be located, declined consent to the RCBA, or were determined to be ineligible on the basis of information not known at the time of their selection by the data coordinating center. Because of limited enrollment at 8 sites, eligibility criteria were relaxed to accept individuals who had originally enrolled in the ATN through December 2013. At the 13th site, all participants enrolled in the ATN were eligible to enroll in the RCBA 9 to 15 months after registry enrollment.

A total of 7164 children and adolescents with ASD had enrolled in the ATN by August 31, 2017, from which the pool of eligible RCBA subjects was generated. Of those, 2200 potentially eligible ATN participants were identified, and 658 completed an RCBA assessment. Of those, 559 did not have a diagnosis of seizure disorder or any reported history of seizures at the time of ATN entry. Twenty-nine individuals were excluded because of conflicting history, and 58 were excluded because of lack of adequate follow-up. The final sample included 472 RCBA participants without a seizure disorder or any reported history of seizures at the time of ATN entry who could be classified at the end of at least 2.5 years of follow-up as either individuals who reported development of seizures subsequent to ATN entry or individuals without development of seizures subsequent to ATN entry.

Consented participants in the RCBA returned for a clinic visit or completed assessments by phone. The following standardized assessments were administered: Vineland II, survey interview,²⁷  ABC,^28,29  Child Behavior Checklist (CBCL),³⁰  Child Sleep Habits Questionnaire (CSHQ),³¹  and Pediatric Quality of Life Inventory (PedsQL).³²  Baseline IQ scores were collected on individuals who had obtained testing, which included the Stanford-Binet Intelligence Scales, Mullen Scales of Early Learning, Weschler Intelligence Scale for Children, Bayley Scales of Infant Development, and the Differential Abilities Scales. Scores were reported as standard scores with a mean of 100 and SD of 15. In addition, parents were asked to report on the presence of gastrointestinal symptoms, asthma, allergies, autoimmune disorders, and history of seizures. The site investigator reviewed participants’ medical histories, recorded current medications, and conducted a physical examination if the visit occurred in person.

Baseline characteristics were compared between participants who did or did not experience onset of seizures after ATN enrollment by Fisher’s exact test and t test. Inference from t tests was confirmed by Wilcoxon rank test to protect against the influence of outliers. Change from enrollment to follow-up in Vineland II, ABC, CBCL, CSHQ, and PedsQL were compared between participants who did or did not experience onset of seizures after ATN enrollment by multivariate linear regression, adjusting for age at enrollment, sex, baseline Vineland II composite score, and length of follow-up. Age, sex, and baseline Vineland II, ABC, CBCL, CSHQ, and PedsQL scores were tested as predictors of seizure onset by using a binomial generalized linear model with log link, adjusting for age at enrollment, sex, baseline Vineland II composite score, and length of follow-up (as an offset after log transformation). Analyses were performed by using SAS version 9.4 (SAS Institute, Inc, Cary, NC). Inference is based on 2-tailed tests. Comparisonwise P values are reported without correction for multiple comparison given the small sample size of children who experienced onset of seizures with nominal significance declared for P < .05. For reference, with 19 measures tested, P < .003 would be required for familywise significance on the basis of a Bonferroni correction for multiple comparisons.

Participants were followed for a mean of 3.7 ± 0.6 years (range: 2.5–5.8 years). Of the 472 children and adolescents included in this analysis, 22 patients (4.7%) had no history of seizures before enrollment in the study but reported seizures at the call-back assessment (see Table 1 for patient demographics). The ratio of boys to girls was ∼4.8:1 in the group without seizures and 2.7:1 in the group with seizure development. In the group that developed seizures, 11 (50%) had 1 to 5 seizures, 2 (9.1%) had 11 or more seizures, and 9 (40.9%) did not have any seizures in the 3 months before RCBA. Only 41% reported using antiepileptic medications (18% oxcarbazepine, 18% valproic acid, and 13% lamotrigine, with 2 individuals reporting use of both oxcarbazepine or valproic acid and lamotrigine).

Genetic testing results, including karyotype (111 patients), fragile X (207 patients), and chromosomal microarray (190), were available. No significant differences in genetic test results were seen between individuals with and without seizures. Baseline IQ data were available for 381 patients (81%). A higher percentage of individuals with development of seizures (53%) had IQs <70 compared with the group without seizures (43%), but this difference was not significant. Mean IQ for individuals without seizures was 75.9 ± 22.4 vs 68.1 ± 21.8 in those who developed seizures (P = .19). The calibrated Autism Diagnostic Observation Schedule severity score, BMI, and head circumference were not significantly different between groups.

There were no significant differences between individuals with and without development of seizures in medical history variables collected, including environmental or food allergies, gastrointestinal disturbances, asthma, atopic dermatitis, inflammatory bowel disease, autoimmune thyroiditis, type I diabetes, or other autoimmune disorders.

Cognitive and behavioral measures captured at registration and at pre- and postseizure onset are summarized in Table 2. Adjusted estimates of change in cognitive and behavioral measures are summarized in Table 3.

Individuals with subsequent development of seizures exhibited significantly lower scores on all domains of the Vineland II at entry compared with those who did not develop seizures, although only communication scores were significant with correction for multiple comparisons. At call-back, all domains continued to be lower in the individuals who developed seizures with the exception of motor skills. When comparing change in scores from entry to call-back, adjusting for age, sex, length of follow-up, and baseline Vineland II composite score (which was used as an estimate of cognitive functioning), individuals who developed seizures experienced declines in all domains, with large declines in daily living skills and overall composite scores. Only change in daily living skills was significantly different from in individuals who did not develop seizures (Table 3).

Individuals with subsequent development of seizures exhibited elevation of scores in the hyperactivity subscale at entry compared with individuals who did not develop seizures. No other ABC domains differed at baseline. At call-back, irritability and stereotypy were elevated in those who developed seizures versus those without seizures. When comparing change in scores from entry to call-back, no significant differences were seen when adjusting for age, sex, length of follow-up, and cognitive functioning. In addition, none of the baseline or call-back comparisons were significant after correction for multiple comparisons.

No significant differences were seen in scores at entry, call-back, or in change from entry to call-back.

Individuals with subsequent development of seizures exhibited lower physical functioning scores and total scale score compared with individuals without seizure development. At call-back, social function score, school functioning score, psychosocial health score, and total score were all lower in those who had developed seizures, with differences in school functioning and psychosocial health scores significant after multiple comparisons correction. Physical functioning score continued to be lower, but the difference was not significant. When adjusting for age, sex, length of follow-up, and cognitive functioning, significant declines were observed among individuals with seizure development in school functioning and psychosocial health, although only the decrease in school functioning was greater than that among individuals who did not develop seizures.

Adjusting for age, sex, and length of follow-up, lower baseline Vineland II composite score predicted a greater risk of seizures (risk ratio = 0.95 per unit Vineland II composite score, 95% confidence interval [CI] 0.92 to 0.99; P = .007). No other baseline characteristics predicted development of seizures after also adjusting for Vineland II composite score, although ABC hyperactivity score was weakly associated.

Autism and epilepsy are neurologic disorders with highly genetic etiologies.^4,34–39  Multiple meta-analyses and smaller studies have identified ID as a consistent risk factor for seizures. In addition, ID may form a background on which epilepsy is part of the mosaic of other characteristics present in many, but not all, individuals with ASD and recurrent seizures. In this study, we identify further characteristics of those at risk for seizures in ASD, including lower adaptive functioning, hyperactivity, and lower physical functioning. Onset of seizure development is associated with further changes of clinical characteristics, including lowering of adaptive functioning, more irritability and stereotypy, and lowering of school and social function.

The prospective collection of data and random selection of participants from the larger ATN cohort permits an unbiased characterization of individuals with ASD before the development of seizures. Results of this exploratory study suggest that individuals with ASD prospectively followed before the age of seizure development are already distinct from their age- and sex-matched compatriots. Previous studies have documented that those with ASD, ID, and epilepsy have lower motor function, social and/or adaptive function, poorer communication (or are less verbal), more repetitive behaviors, and more hyperactivity.^7,23  This study suggests that it is not the seizures and/or neuronal activity that are responsible for more severe symptoms. Careful studies of developmental trajectory in the synapsin I and Contactin Associated Protein 2 mutant mice, animal models of autism and epilepsy, support this hypothesis. For example, impaired social behavior preceded both seizure activity on EEG and clinical seizures of the mutant mice.^40,41 

The Vineland II composite score, a close correlate with IQ, has previously been shown to be strongly associated with risk for epilepsy in a multiplex ASD sample.⁴  Probands with Vineland adaptive composite scores <70 were 3 times as likely to have epilepsy as those with composite scores ≥70. In this study, Vineland II composite score was the only significant predictor of seizures after adjusting for age, sex, and length of follow-up. Interestingly, biomarkers such as sex, BMI, and head circumference shown to be predictive of seizures in ASD were not predictive of seizures in this cohort.^10,42–44  Associations of these biomarkers with epilepsy risk in ASD may be greater for adolescents.⁴² 

The etiology of comorbid ID, ASD, and epilepsy is complex and likely involves both biological and environmental factors. Several pathophysiologic mechanisms have been proposed: (1) accidental co-occurrence of ID, ASD, and epilepsy, (2) altered organization of minicolumns associated with defects in local gamma aminobutyric acid circuits and their gamma aminobutyric acid type A receptors, (3) common genetic or neurodevelopmental factors, and (4) induction of autistic behaviors by early-life epilepsy, such as infantile spasms. The latter hypothesis, although definitely relevant to specific disorders, such as TSC and Dravet syndrome, is less likely in our current population. Indeed, ID is an integral part of each disorder. Genetic disorders of high effect size (TSC, fragile X, phosphatase and tensin homolog, Phelan-McDermid) are high risk for the co-occurrence of ID, epilepsy, and ASD.^14–21  Careful prospective tracking of neural development of at-risk ASD populations may provide the data needed to evaluate these pathophysiological mechanisms.

As an observational study, the associations identified in our analysis are hypothesis generating and should not be interpreted to imply causation. The small number of patients with onset of seizures (n = 22) reduced our power to detect associations. Future studies could be improved by applying a more rigorous definition of epilepsy and more complete description of epilepsy reports, including description of seizure semiology (type[s] and frequency), MRI results, EEG results, comorbidities, and response to antiepileptic drugs.⁴⁵  The age range of subjects in this study was skewed toward younger ages. Future studies would be improved by including broader age ranges to determine more fully the roles of sex, age, macrocephaly, and BMI as biomarkers for seizures in ASD.^10,42  This study also lacked data on prenatal environmental exposures associated with epilepsy risk. Finally, most measures reported here are parental reported measures, which may not accurately predict seizure-sensitive phenotypes.⁴⁶  We also lack comprehensive IQ data on each subject. New direct measures, such as the National Institutes of Health Toolbox Cognition Battery, may more accurately define parameters sensitive to onset of seizures.⁴⁷ 

ASDs are known to be associated with an increased risk of seizures. Prospective studies of genetic, histopathological, EEG, imaging studies, and extensive clinical phenotyping comparing individuals with ASD with and without epilepsy are needed to delineate well-defined clinical subpopulations and specific pathogenic mechanisms of ASD and epilepsy comorbidity. With caveats, this prospective pilot study represents a pragmatic first step toward identification of this subgroup. Those individuals with ASD at risk for seizures (before onset of seizures) have lower adaptive functioning. Onset of seizures is associated with decreased adaptive functioning, increased irritability and stereotypies, plus possible lower school and social functioning. In this prospective sample, only the Vineland adaptive composite score was a significant independent predictor for the development of seizures. Given the limited information in the literature on comorbid ASD and epilepsy, this study lays the groundwork for further investigation in this area.

ABC

Aberrant Behavior Checklist–Community

ASD

autism spectrum disorder

ATN

Autism Treatment Network

CBCL

Child Behavior Checklist

CI

confidence interval

CSHQ

Child Sleep Habits Questionnaire

ID

intellectual disability

PedsQL

Pediatric Quality of Life Inventory

RCBA

Autism Treatment Network Registry Call-Back Assessment

TSC

tuberous sclerosis complex

Vineland II

Vineland Adaptive Behavior Scales, Second Edition