Adaptive behavior and later school achievement in children with early-onset epilepsy

Abstract

Aim

To determine whether early measures of adaptive behavior are predictive of later school difficulties and achievement in otherwise neurotypical (unimpaired) children with onset of epilepsy during the pre-school years.

Method

In a prospective cohort study, parents completed the Vineland Adaptive Behavior Scales (VABS) for children who were aged 5 years or less at epilepsy diagnosis. Eight to nine years later, the children were assessed using the Wechsler Intelligence Scales for Children (WISC), the Wide Range Achievement Test (WRAT), and the Child Behavior Checklist (CBCL). Associations of VABS scores with later WRAT and CBCL scores were tested.

Results

A total of 108 neurotypical children (64 males, 44 females; mean age at testing 11y 11mo, SD 2y) were studied. After adjustment for IQ and other factors, there was an increase of 0.15 points (95% confidence interval [CI] 0.03–0.27 points; p=0.03) and 0.14 points (95% CI 0.0–0.28 points; p=0.05) in WRAT reading and spelling scores for each 1-point increment in the VABS communication score. Corresponding numbers for the VABS socialization score were 0.20 (95% CI 0.08–0.32; p=0. 005) and 0.17 (95% CI 0.05–0.29; p=0.005).

Conclusion

In neurotypical preschool children with epilepsy, early social and communication scores predict later school performance. These findings raise questions about opportunities for early identification and intervention for children at greatest risk.

Young people with childhood-onset epilepsy often experience greater difficulties in schooling, employment, and other social outcomes than their peers; this is the case even for neurotypical children (i.e. children in whom clinical examination and clinical imaging are normal and in whom there is no obvious intellectual disability and no obvious cause for the epilepsy).^1,2 In school-aged children with new-onset epilepsy, increased academic difficulties may precede the onset of epilepsy, as evidenced by the higher utilization of special education resources before or at the time epilepsy is initially diagnosed.^1–4 Furthermore, higher levels of behavioral problems are also present from the time of diagnosis.^3–5

Children with epilepsy have poorer cognitive performance than their typically developing peers, even in the absence of obvious structural brain lesions or impairment.^1,3 Furthermore, low IQ and impaired language skills can be associated with behavioral problems^6–8 and poor school achievement.⁹ Consequently, it is unclear whether increased school and behavioral problems simply reflect a mild degree of cognitive and linguistic compromise or whether there are additional difficulties beyond cognitive and language problems. For example, we previously found that school-aged children with epilepsy were more likely than their siblings of the same age to be in a special educational program at the time of diagnosis, even after adjustment for IQ.¹

The majority of studies in this field of research have addressed children already in school when epilepsy was diagnosed. However, half of the cases of childhood-onset epilepsy occur in the preschool years at an age when serious encephalopathic conditions associated with a globally poor prognosis initially manifest.¹⁰ It is not known if otherwise neurotypical children with early-onset epilepsy, once they enter school, experience difficulties similar to those experienced by older children who are already in school when they develop epilepsy and whether there are factors that might foreshadow later educational or behavioral difficulties.

We hypothesized that the level of school achievement and extent of behavioral problems in mid-childhood may be foreshadowed when preschool-aged children are first diagnosed with epilepsy. We examined the predictive value of a standardized behavioral and developmental measure, the Vineland Adaptive Behavior Scales (VABS), for standardized school achievement scores and behavioral problem scores in a community-based study of childhood-onset epilepsy.

METHOD

The Connecticut Study of Epilepsy is a prospective, longitudinal study of children (aged 1mo to 16y) in the State of Connecticut first diagnosed with epilepsy between 1993 and 1997 by participating pediatric neurologists. Initially, the parents of the affected children completed interviews covering family history, schooling, and socio-demographic information. Every 3 to 4 months the parents were asked to report seizure occurrence in order for it to be tracked, and a periodic review of medical records was also performed. From 2002 to 2006, 8 to 9 years after each child entered the study, families were invited to participate in a comprehensive reassessment.^11–13

Measurements and definitions

Baseline

The VABS was administered by a trained interviewer to all children aged 5 years or less at the time of study entry. The VABS provides scores standardized to a mean score of 100 and SD of 15 for four domains (socialization, communication, daily living, and motor) and an overall composite score.¹⁴

Reassessment

To assess behavior, the parent-completed version of the Child Behavior Checklist (CBCL) was used.¹⁵ The CBCL provides an overall problem score as well as internalizing and externalizing problem scores, which were all considered as part of this study. The CBCL also provides a total social competence score based on the sum of three subscales: activities (e.g. sports, hobbies), social (e.g. friendships, interpersonal skills), and school (e.g. performance, ability, school problems). We did not use the eight narrower scores that can also be obtained from this instrument. CBCL scores are reported as T-scores with a mean score of 50 and SD of 10. The cognitive battery included the Wechsler Intelligence Scale for Children (WISC)¹⁶ to measure full-scale IQ (FSIQ) and the Wide Range Achievement Test (WRAT)¹⁷ – a standardized test of achievement in arithmetic, spelling, and reading. WISC and WRAT scores are standardized to a mean score of 100 and SD of 15.

Five-year remission status was determined from the date of last seizure to the date of assessment. Medication status was defined as taking one or more antiepileptic drugs (AEDs) or none at reassessment. A structured parent interview covered information regarding significant school problems (i.e. held back a year or in receipt of special services) since onset of epilepsy. Parental education was categorized as college graduate or non-college graduate, as this almost divided the group in half.

Analytic sample

Eligibility requirements for this analytic sample were diagnosis of epilepsy at 5 years of age or less and normal clinical neurological examination. Children had to have an IQ of 70 or above and no known structural or other identified cause for the epilepsy based on assessments during 9 years of follow-up.

Statistical Analysis

Statistical analyses were conducted using Statistical Analysis Software, version 9.2 (SAS Institute, Cary, NC, USA) with correlations, t-tests and χ² tests for bivariate associations and multiple linear regression for multivariable analyses. We examined the correlations between baseline VABS scores and 9-year cognitive WRAT and CBCL scores. t-tests were used to test the relation between each of these scores and 5-year remission status and concurrent AED use. The primary outcome variables were the three WRAT scores and the total problem, internalizing, externalizing, and social competence scores from the CBCL. The baseline composite VABS was used as the primary predictor of these scores when measured 8 to 9 years later. When significant associations were found, they were further explored by considering the VABS domain scores. Multiple linear regression analysis was employed to determine the independent associations of early adaptive measures (VABS) with behavioral (CBCL) and achievement (WRAT) measures later in childhood, with adjustment for FSIQ. Although we examined multiple bivariate associations, the variables were inter-correlated. Accordingly, we did not use a set correction for multiple comparisons. We interpreted p-values between 0.05 and 0.01 as representing borderline or moderate statistical significance and p<0.01 as significant. Residual analyses were conducted to ensure that data fit the assumptions for the linear regression model.

RESULTS

Of 613 children originally recruited into the cohort, 300 were aged 5 years or less when epilepsy was diagnosed. Of these, a baseline VABS was available for 292, and 243 subsequently participated in the reassessment 8 to 9 years later. One child was later excluded for confidential reasons. Another child, diagnosed with autism only, was also excluded for the purposes of these analyses. Of the remainder, 178 had normal examinations and imaging based on examinations performed at epilepsy onset and all information accumulated since. Cognitive testing was performed for 125 of these children, of whom 108 (64 males, 44 females; mean age at testing 11y 11mo, SD 2y) had an IQ of 70 or above and formed the analytic sample (Fig. S1, online supporting information).

The characteristics of the analytic sample are provided in Table SIa (online supporting information). In this group, selected for not having intellectual or neurological disability, scores were about average for the various test norms. Seventy-five children (69.4%) were in 5-year remission and 26 (24.1%) were taking seizure medications (Table SIb, online supporting information).

Bivariate associations of baseline VABS with later cognitive, achievement, and behavioral scores

The VABS composite score was correlated with FSIQ (r=0.29, p=0.003) and WRAT scores for reading (r=0.35, p=0.002) and spelling (r=0.35, p=0.004), but not arithmetic (r=0.09, p=0.33; Table Ia). Similar associations were found for the VABS communication and socialization scores. The VABS motor score correlated only with FSIQ (r=0.27, p=0.005) and the VABS daily living score was not correlated with any of these later childhood scores. None of the baseline VABS scores was significantly associated with CBCL scores at the 8- to 9-year follow-up, with the exception of the total social competency score (r=0.28, p=0.003).

Table I.

a: Correlations (and p-values) between baseline Vineland Adaptive Behavior Scale (VABS) scores with cognitive and behavioral scores measured 8 to 9 years later
	VABS scores at study entry
Scores measured at 8–9y after diagnosis of epilepsy	Composite	Communication	Daily Living	Socialization	Motor
FSIQ	0.29 (0.003)	0.27 (0.005)	0.08 (0.28)	0.17 (0.07)	0.27 (0.005)
WRAT – reading	0.35 (0.002)	0.35 (0.002)	0.10 (0.28)	0.37 (<0.001)	0.15 (0.11)
WRAT – spelling	0.34 (0.004)	0.35 (0.002)	0.11 (0.26)	0.32 (0.007)	0.16 (0.09)
WRAT – arithmetic	0.09 (0.33)	0.11 (0.23)	−0.06 (0.56)	0.06 (0.51)	0.13 (0.18)
CBCL – total problems	0.01 (0.90)	0.05 (0.63)	−0.03 (0.76)	−0.03 (0.73)	0.05 (0.63)
CBCL – internalizing problems	0.08 (0.39)	0.08 (0.40)	−0.01 (0.88)	0.06 (0.55)	0.09 (0.33)
CBCL – externalizing problems	0.06 (0.54)	0.10 (0.29)	−0.02 (0.87)	−0.03 (0.80)	0.09 (0.36)
CBCL – total competence score	0.17 (0.07)	0.10 (0.30)	0.06 (0.51)	0.28 (0.003)	0.03 (0.72)

b: Correlations of cognitive and achievement scores with contemporaneous Child Behavior Checklist scores
	FSIQ	WRAT – reading	WRAT – spelling	WRAT – arithmetic
CBCL – total problems	−0.20 (0.03)	−0.19 (0.04)	−0.26 (0.006)	−0.27 (0.006)
CBCL – internalizing problems	−0.09 (0.35)	−0.11 (0.24)	−0.17 (0.08)	−0.19 (0.05)
CBCL – externalizing problems	−0.11 (0.24)	−0.14 (0.14)	−0.20 (0.04)	−0.21 (0.03)
CBCL – total competence score	0.49 (<0.001)	0.46 (<0.001)	0.42 (<0.001)	0.39 (<0.001)

FSIQ, full-scale IQ from the Wechsler Intelligence Scale for Children; WRAT, Wide Range Achievement Test; CBCL, Child Behavior Checklist.

Values are correlation coefficients (and p-values). FSIQ, full-scale IQ; WRAT, Wide Range Achievement Test; CBCL, Child Behavior Checklist.

Bivariate associations among cognition, achievement scores, and behavior at 9-year follow-up

FSIQ was strongly correlated with all three WRAT achievement scores (reading, r=0.59; spelling, r=0.58; arithmetic, r=0.62; all p<0.001) as well as CBCL total social competency score (r=0.49, p<0.001; Table Ib). The three WRAT scores had a similar degree of association with the total competency score.

Parental education association with VABS and 9-year child behavioral, cognitive, and achievement scores

Parental education was only modestly associated with VABS composite and motor scores. With the exception of the total social competence score (college vs non-college graduate, 52.6 [SD 10.5] vs 45.5 [SD 9.9] respectively; p<0.005) there was no association with CBCL scores. There was a substantial difference in FSIQ between children with a college graduate (106.1, SD 14.1) and those with a non-college graduate (96.2, SD 14) parent (p=0.0004). Reading (p=0.04) and spelling (0.01) achievement scores were only modestly associated with parent education; however, there was a somewhat larger difference in the arithmetic scores for college graduate (105.0, SD 15.4) than for non-college graduate (93.7, SD 21.1) parents (p=0.002).

Seizure remission and AED status versus baseline VABS and 9-year behavioral, cognitive, and achievement scores

Not being in remission was associated with significantly and substantially lower FSIQ and achievement scores as well as a lower total competency score (Table II). Children taking AEDs at the time of the assessment had higher attention problem scores, lower FSIQ, and slightly lower WRAT scores. There were no substantial differences, however, in baseline VABS scores at study entry associated with remission or treatment 8 to 9 years later.

Table II.

Mean differences in baseline Vineland Adaptive Behavior Scales (VABS) and 8- to 9-year cognitive, achievement, and behavioral scores as a function of 5-year remission and medication status

Score/testa	Mean difference (95% CI)b between those not in 5y remission (n=33) and those in remission (n=75)	p	Mean difference (95% CI)b between those on AEDs (n=26) and those not on AEDs (n=82)	p
VABS scales
Composite	1.4 (−6.3 to 9.2)	0.69	0.2 (−7.5 to 7.8)	0.97
Communication	1.8 (−5.7 to 9.3)	0.63	1.4 (−6.8 to 9.5)	0.74
Socialization	2.4 (−5.8 to 10.6)	0.55	1.2 (−7.6 to 10.0)	0.79
Daily living	3.1 (−3.7 to 9.8)	0.35	0.7 (−6.3 to 7.7)	0.84
Motor	−3.5 (−12.6 to 5.6)	0.40	−2.9 (−11.7 to 6.0)	0.52
Cognitive and achievement (WRAT)
FSIQ	−11.5 (−17.3 to −5.8)	<0.001	−11.0 (−17.3 to −4.7)	0.008
Reading	−4.6 (−10.8 to 1.5)	0.14	−6.1 (−12.7 to 0.5)	0.07
Spelling	−7.7 (−14.2 to −1.2)	0.02	−7.6 (−14.6 to −0.5)	0.04
Arithmetic	−11.6 (−19.8 to −3.4)	0.004	−12.9 (−21.3 to −4.6)	0.03
CBCL – behavior
Total problems	4.9 (−0.4 to 10.3)	0.07	2.7 (−2.5 to 7.9)	0.30
Internalizing problems	1.6 (−2.9 to 6.1)	0.47	−0.5 (−5.2 to 4.2)	0.82
Externalizing problems	5.0 (0.2 to 9.7)	0.03	2.6 (−2.3 to 7.5)	0.29
Total competence score	−6.8 (−10.9 to −2.6)	0.002	−3.7 (−8.5 to 1.1)	0.13

^aThe Vineland Adaptive Behavior Scales, full-scale IQ (FSIQ; from the Wechsler Intelligence Scales for Children) and the Wide Range Achievement Test (WRAT) scores are standardized to a mean of 100 (SD=15). The Child Behavior Checklist (CBCL) scores are standardized to a mean of 50 (SD 10). Higher CBCL problem scores are worse with the exception of the competency score for which lower scores are worse.

^bThose not in remission and those on antiepileptic drugs (AEDs) tended to have poorer (lower) scores. For the CBCL total, internalizing, and externalizing problem domains, higher scores indicate more problems. CI, confidence interval.

Baseline VABS scores and later school performance

Following initial diagnosis of epilepsy, 59 (55%) children in the group went on to have school problems after entering school. VABS scores, particularly for communication, were somewhat lower in those who did than in those who did not experience school problems (Table III).

Table III.

Association of baseline Vineland Adaptive Behavior Scales scores (mean [SD]) for later school problems and, t-test on the differences of the means

Held back or received services	Composite	Communication	Socialization	Daily living	Motor
No (n=49)	107.2 (14.1)	109.8 (16.7)	111.0 (17.6)	95.8 (13.7)	105.5 (18.8)
Yes (n=59)	99.5 (18.7)	99.5 (18.1)	105.5 (21.2)	93.3 (17.0)	100.0 (20.4)
Difference between the two groups (p)	0.02	0.003	0.15	0.42	0.15

Multivariable association: baseline VABS with 9-year achievement and behavioral measures

We performed multiple linear regression analyses to determine whether early markers of development (VABS) in otherwise neurotypical children with epilepsy might predict school achievement or performance above and beyond their association with FSIQ and independent of the influence of parental education and epilepsy-related factors. The VABS scores were the independent (predictor) variables. Because there was little or no evidence of an association between VABS scores and WRAT arithmetic or total social competence after adjustment for FSIQ, only the models for WRAT reading and spelling are shown (Table IV). FSIQ was strongly correlated with WRAT achievement scores, and school problems were associated with lower reading and spelling scores. Parental education, 5-year remission status, and medication status were not significant predictors of achievement scores after taking FSIQ and school problems into account. After adjustment for FSIQ and school problems, the VABS composite, and in a separate model the VABS communication score, were modestly associated with the WRAT reading and spelling scores (p=0.03 to p=0.05). The VABS socialization score had stronger associations (p<0.005 and p=0.005 respectively). The effect estimates were nearly half the size of that for FSIQ.

Table IV.

Results of multiple linear regression models for prediction of achievement and total competency scores by baseline Vineland Adaptive Behavior Scales (VABS) scores after adjustment for full scale IQ

Model	Reading	Spelling
(a) Model 1: Achievement scores predicted by the VABS composite score after adjustment for FSIQ and school problems
FSIQ	0.45 (0.09), p<0.001	0.41 (0.09), p<0.001
School problems	−5.87 (2.54), p=0.03	−10.48 (2.67), p=0.002
Composite – VABS	0.15 (0.07), p=0.03	0.14 (0.07), p=0.05
Intercept	44.11 (10.80), p<0.001	52.91 (11.34), p<0.001
Model R2/adjusted R2	0.42/0.40	0.45/0.43
(b) Model 2: Achievement scores predicted by the VABS communication score after adjustment for FSIQ and school problems
FSIQ	0.46 (0.09) p<0.001	0.42 (0.09), p<0.001
School problems	−5.41 (2.57) p=0.04	−10.10 (2.70), p=0.002
Communication – VABS	0.15 (0.06), p=0.03	0.14 (0.07), p=0.05
Intercept	43.44 (10.97), p<0.001	52.31 (11.52), p<0.001
Model R2/adjusted R2	0.42/0.40	0.45/0.43
(c) Model 3: Achievement scores as a function of the VABS socialization score after adjustment for FSIQ and school problems
FSIQ	0.45 (0.08), p<0.001	0.41 (0.09), p<0.001
School problems	−5.89 (2.48), p=0.02	−10.54 (2.61), p<0.001
Socialization – VABS	0.20 (0.06), p=0.005	0.17 (0.06), p=0.005
Intercept	38.12 (10.4), p=0.004	48.69 (11.09), p<0.001
Model R2/adjusted R2	0.46/0.44	0.47/0.45
(d) Model 4: Achievement scores as a function of VABS socialization scores after adjustment for verbal comprehension and problems
Verbal Comprehension – WISC	0.51 (0.08), p<0.001	0.39 (0.09), p<0.001
School problems	−5.97 (2.29), p=0.01	−11.48 (2.57), p<0.001
Socialization – VABS	0.15 (0.05), p=0.009	0.13 (0.06), p=0.03
Intercept	36.50 (9.46), p=0.002	53.82 (10.62), p<0.001
Model R2/adjusted R2	0.50/0.49	0.46/0.44

Values are estimate (standard error) p-value. Separate models are presented for reading and spelling scores. FSIQ, full-scale IQ; WISC, Wechsler Intelligence Scale for Children.

The VABS socialization is, in part, language driven, as are the reading and spelling scores on the WRAT. We repeated the model for VABS socialization but substituted the verbal comprehension score from the WISC for FSIQ. The VABS socialization score remains significantly predictive of WRAT reading (p=0.009) and spelling scores (p=0.03).

To examine the impact of low VABS scores in predicting low achievement scores, VABS socialization and WRAT reading and spelling scores were categorized as less than 85 (low/impaired), 85 to 99 (low average), 100 to 115 (high average), or greater than 115 (above average). There was a striking correlation between a low VABS socialization score and a later low reading and spelling score (Fig. S2, online supporting information).

DISCUSSION

Our findings suggest that, in neurotypical preschool-aged children with newly diagnosed epilepsy, early adaptive behavior predicts later school achievement. This is partly because these early measures predict overall cognitive ability; however, the associations persist even after considering factors such as IQ and parental education and above and beyond recognition by the school system that a child is having difficulties sufficient to lead to being held back a grade or referral to special education services.

Past studies of neurotypical children with epilepsy have suggested that social, employment, and educational outcomes in later adulthood are relatively impaired despite good seizure control.^18,19 Although having had epilepsy could potentially affect these outcomes, considerable literature emphasizes that attention, behavioral, and cognitive problems may predate the onset of epilepsy in neurotypical school-aged children.^3,4,20 This raises questions regarding whether these long-term adult outcomes could have roots very early in life that could be discerned at the time when epilepsy first presents.

In contrast to most other studies, we have focused on children who were not yet in school at the time of their initial diagnosis of epilepsy. By virtue of being neurotypical, the study group that we selected had a relatively mild course of epilepsy, and most children were in long-term remission and off medication at the time of reassessment (Table SIa, online supporting information). Nonetheless, at 8 to 9 years after initial diagnosis, over half the group had experienced school problems (i.e. held back a grade, referred to special services, or both). This is similar to the proportion seen in the older segment of the cohort (i.e. children already in school when first diagnosed with epilepsy) and about 2.5 times the levels seen in the matched sibling comparisons (45% vs 17.6%).²⁰ In that analysis, IQ scores did not explain differences in receipt of services.

Our neurotypical sample with an IQ of 70 and above is the group implicated in some of the older cohorts reported in earlier literature.^18,19 These studies did not benefit from the same detailed information that we had available regarding IQ and neuroimaging, and, potentially, some children considered ‘normal’ (neurotypical) in those earlier studies would not be considered such by the criteria we employed. Had we extended eligibility to include a broader (lower) range of IQ scores, or had we included individuals with relatively subtle magnetic resonance imaging findings that would not have been detected in these earlier studies, we might have seen stronger associations between early VABS scores and later WRAT and FSIQ scores; however, we were specifically concerned with whether early measures (VABS) provided information about later school outcome in children for whom intellectual disability alone could not explain low performance.

We did not find that early VABS scores were correlated with later problem behavior scores measured by the CBCL. The CBCL scores were also not strongly associated with cognitive scores; only the total social competency score, which includes school performance, was consistently associated with FSIQ and all three WRAT scores. While these are likely to be important associations, the only significant finding was for the VABS socialization and CBCL total social competence scores. This suggests that the VABS in early childhood does not predict problem behaviors assessed by the CBCL.

The VABS was originally chosen in order to provide a standardized measure of development in preschool-aged children who had not yet been evaluated, to some extent, for cognitive preparedness and intellectual level. It was initially selected because it can be administered by lay interviewers and is suitable for use in large-scale studies. It was a well-known and respected tool at the time (1997–2003), as it remains so at the time of writing. Early VABS scores may be useful in predicting school achievement scores in children who are being followed for language impairment.²¹ They may also correlate with receipt of special services in children with borderline or very mild intellectual impairment.²² In children with epilepsy, other studies point to subtle language dysfunction that is not readily revealed in IQ tests. These include impairments in syntactic processing and higher-level language skills as well as in communication competence.^7,23 The WRAT taps into these higher-level skills to a certain degree. Notably, even after adjustment for FSIQ and WISC verbal comprehension, aspects of early adaptive behavior, communication to a modest extent, and socialization in particular, still contributed to variation in two of the three standardized achievement measures (reading and spelling). Thus, the VABS provides some information that is not entirely contained in standard IQ scores.

In this neurotypical group, school problems (held back a grade and special services) were not strongly correlated with baseline VABS scores. School problems were strongly correlated with the WRAT spelling component, only modestly associated with reading, and not associated with arithmetic. This again suggests that the comprehension domain tapped by the WRAT reading component may not be as salient for identifying neurotypical children with academic difficulties. Others have noted that the VABS, a parent-reported measure, appears to be sensitive to communication skills for which standardized IQ tests may be relatively insensitive. Thus, parents provide an important perspective on their children’s abilities that complements other testing.²¹

The reading component of the WRAT assesses the ability to extract meaning from words and understand thoughts and information that are expressed through written sentences. It tests more than vocabulary. Such early communication competence and language skills are related to long-term functional and behavioral outcome in children.²⁴ Potentially, the VABS and WRAT information may provide us with a unique opportunity to assess whether early markers (VABS), through their association with intermediate markers (WRAT), could herald poorer outcomes once children reach adulthood.

The VABS may not be the most sensitive measure to predict later learning and behavior difficulties. Our findings suggest, however, that such difficulties may be predicted based on preschool assessments and might be even better predicted with a different measure.

Weaknesses in our design include the lack of neurotypical comparison children without epilepsy who were similarly assessed during the preschool period. When the cohort was initially recruited, the understanding of pre-existing cognitive and behavioral comorbidities was just developing, and comparison children were not part of the initial research plan. It is likely that similar associations (between preschool adaptive behavior and school-aged achievement scores) may be found in children who never develop epilepsy. Nonetheless, ample evidence demonstrates that young people with epilepsy have a higher burden of a variety of problems relative to their typically developing peers that are not readily explained by seizure factors or medications. The cohort was community based, and the children representative of the state’s population. Follow-up was excellent, with 86% of children participating in testing. Nonetheless, the cohort comes from a single state in the USA with a relatively high parental education level and may not represent children elsewhere in the country or the world.

National professional societies such as the American Academies of Pediatrics already recommend basic developmental and autism screening for all preschool-aged children. A recent Institute of Medicine report emphasized the need for screening of cognitive and behavioral difficulties in people with epilepsy.²⁵ Questions remain about whether interventions could make a difference in this setting, and these would have to be addressed in a different series of studies. Our findings, however, raise questions and concerns about the roots of educational, employment, and social outcomes in later childhood, adolescence, and even adulthood. These questions include how early such problems may be detected in young children with epilepsy, and whether there are opportunities for early intervention that could improve outcomes and provide justification for developing a comprehensive approach to routine screening in these children.

Supplementary Material

Supp Fig S1. Figure S1.

Derivation of analytic sample.

Supp Fig S2. Figure S2.

Association between VABS socialization score (categorized by standard deviations from the mean) and low (<85) scores on the WRAT reading and spelling tests 8–9 years later.

What this paper adds.

• Early measures of adaptive behavior may identify children with epilepsy at risk for low school achievement.

• Precursors to difficulties may be detectable in younger children soon after initial diagnosis.

ABBREVIATIONS

AED

Antiepileptic drug

CBCL

Child Behavior Checklist

FSIQ

Full-scale IQ

VABS

Vineland Adaptive Behavior Scales

WISC

Wechsler Intelligence Scale for Children

WRAT

Wide Range Achievement Test