OUP user menu

Three to four years after diagnosis: cognition and behaviour in children with ‘epilepsy only’. A prospective, controlled study

K. J. Oostrom , H. van Teeseling , A. Smeets-Schouten , A. C. B. Peters , A. Jennekens-Schinkel
DOI: http://dx.doi.org/10.1093/brain/awh494 1546-1555 First published online: 7 April 2005


A 3.5-year follow-up study of cognition and behaviour in 42 children with newly diagnosed idiopathic or cryptogenic epilepsy (‘epilepsy only’) attending mainstream education and 30 healthy gender-matched classmate controls was carried out to identify differences between groups, to detect factors that contribute to the difference and its change over time, and to establish the proportion of poorly performing children. The neuropsychological battery covered the major domains of cognition, mental and motor speed and academic language skills. Children were tested at the time of diagnosis (before any anti-epileptic drug treatment started) and 3, 12 and approximately 42 months later. Parents and teachers completed behaviour checklists, for which the scoring was adapted to prevent any influence of epilepsy-related ambiguity. Based on parental interviews at the time of diagnosis, children with epilepsy were categorized as having longstanding behavioural and/or learning problems, as belonging to a troubled family, as being exposed to ‘off-balance’ parenting starting at the time of epilepsy onset and/or as reacting maladaptively to the changes in relation to the onset of epilepsy. Throughout follow-up, the group of children with epilepsy only performed less well than healthy classmates on measures of learning, memory span for words, attention and behaviour. After controlling for school delay, proactive interference (number of responses to the same images as in the learning trials, but now presented in reordered locations) was the only remaining variable that distinguished the group of children with epilepsy only. Group-wise, no changes in cognitive and behavioural differences over time were found, but instability in individual performances appeared to characterize children with epilepsy only. Rather than intrinsically epilepsy-related variables, such as idiopathic versus cryptogenic aetiology, seizure control or anti-epileptic drug treatment, the child's prediagnostic learning and behavioural histories and the parents' ability to continue their habitual parenting in the face of the diagnosis of epilepsy only were shown by both group-wise and case-by-case analyses to be important for understanding the cognitive and behavioural functioning of the children with epilepsy only.

  • child
  • epilepsy
  • cognition
  • behaviour
  • psychosocial


At least 70% of childhood epilepsies consist of idiopathic or cryptogenic epilepsy, also referred to as non-symptomatic epilepsy. We refer to this well-defined subset of epilepsies as ‘epilepsy only’ (EO) (Sillanpää et al., 1998). The majority of these children are otherwise healthy, have normal intelligence, attend mainstream schools and become seizure-free within the first 2 years after diagnosis, either spontaneously or with anti-epileptic drug (AED) treatment (Berg et al., 1995).

Nevertheless, the school career of children with EO is at risk (Austin et al., 1999; Bailet and Turk, 2000). We previously reported that, notwithstanding average intelligence levels, no less than 22% of children with EO had repeated a year at school either prior to or in the first year after the diagnosis (Schouten et al., 2001). That is appreciably more than the 11% reported for mainstream primary schools in The Netherlands in the corresponding period (van de Grift, 2000). Moreover, in the group of children researched, special educational assistance had been required for more children who were later diagnosed with EO (51%) than for their healthy classmates (27%). Consequences may be lifelong, as epidemiological follow-up studies involving adults who stopped having seizures and were able to discontinue AED treatment report lasting educational and psychosocial deprivation (Jalava et al., 1997). However, there may be a cultural influence, as Wakamoto and colleagues could not establish social disadvantages in Japanese persons who had had non-symptomatic epilepsy in childhood and who had normal intelligence (Wakamoto et al., 2000).

Seeking to understand the educational risks, the present authors studied cognition and behaviour during the first year after diagnosis and reported poorer scores across several measures in patients with EO than in healthy gender-matched classmate controls (Oostrom et al., 2003). Group-wise, cognitive and behavioural differences existed already at baseline; the proportion of children with deficits did not change during the first year after diagnosis, but the children who had the deficits did change. Moreover, psychosocial context rather than characteristics of the epilepsy were related to patients' deficits. This prompted an extension of the study. In the present article we report the course of cognitive and behavioural differences between children with EO and healthy controls during the first 3–4 years after diagnosis. Again, the impact of epilepsy variables and that of variables pertaining to psychosocial context are evaluated.

Subjects and methods


Sixty-nine children with EO had been enrolled consecutively between January 1997 and November 1998 and they had agreed to participate three times in the first year after diagnosis. Inclusion criteria at study entry had been at least two unprovoked non-febrile seizures or status epilepticus, idiopathic or cryptogenic aetiology (ILAE, 1989), age between 7 and 16 years, and attendance of mainstream education. Exclusion criteria had been any associated neurological disorder (identified by history, physical examination or neuroimaging), having been diagnosed with another chronic physical disease or previous use of AEDs. In case of doubt with respect to the non-symptomatic nature of the epilepsy, the physician had made clear to the parents and the child that neuroimaging (MRI) was needed to rule out a minor chance of an underlying structural cause for the child's epilepsy. Only after having communicated the normal result of MRI did the doctor introduce the present study.

Because one of the patients could not find a suitable classmate in time, the control group consisted of 68 children, matched for gender and educational level, who provided data to control for normal development and the effects of retesting. Excluded were classmates with chronic conditions such as (slight) asthma or a history of neurological conditions, as they may influence cognition and/or behaviour. Further details on recruitment and characteristics of the initial cohort have been reported previously (Oostrom et al., 2003). The findings reported here concern the subset of children who positively reacted to a telephone request for participation in a fourth assessment. The group under study included 42 children with EO and 30 gender-matched healthy classmate controls.

Idiopathic versus cryptogenic aetiology and syndrome classification were based on data recorded by the children's neurologists at diagnosis, according to a standard protocol formulated by the Dutch Study of Epilepsy in Childhood (DuSECh) (Arts et al., 1999), which were double-checked and, if necessary, changed after a year (Stroink et al., 2004). In this study there was no standard protocol for discontinuation of medications after a seizure-free interval. However, most child neurologists collaborating in DuSECh consider medication withdrawal after a 2-year seizure-free interval.

The study was approved by the ethics and research committees of the participating hospitals. All children participated on the basis of written informed consent by their parents. Older children (aged 12 years and over) also gave their own informed assent.

The multicentre study was prospective, longitudinal, and controlled. All children were assessed within 48 h after diagnosis, before they started AED treatment (if prescribed), and reassessed 3, 12 and 42 (±6) months after first assessment.

Parental interviews

At the first assessment, a well-trained psychologist (A.S.-S., K.J.O.) conducted extensive semistructured interviews with the patients' parents. The interviewer asked preformatted questions addressing the medical, mental and social histories of their child. The interviews were transcribed and processed according to Chi's guidelines (Chi, 1997). Every phrase containing information that pertained to one of the interview variables was labelled according to the key content of the utterance (e.g. ‘fear of stigma’, ‘worry about the future’, ‘experience of child's death’). Analyses of 42 interviews resulted in a grid consisting of a row for every label (79 in total) and a column for every interviewee. Then, labels could be grouped into five broader dichotomous domains (called ‘interview variables’) (Box 1) (Oostrom et al., 2001a). If, per interview, more than half of the labels within a domain reflected the presence of a difficulty the case was allocated to the negative category for that domain (e.g. family, ‘problems’; parenting, ‘off balance’; child's reaction, ‘maladaptive’); if labels did not reflect difficulties/problems, the case was allocated to the positive (‘no problems’, ‘in balance’ or ‘adaptive’) category of that interview variable. At follow-up assessments, parents were interviewed briefly in order to trace changes in the psychosocial context of the patients.

Box 1 Interview variables

Qualitative analysis of the interviews (Chi, 1997) yielded the following five dimensions, to which contextual problems were linked (coding between parentheses).

  1. Longer existing behavioural problems dating from before the diagnosis of epilepsy, i.e. difficult behaviour perceived by the parent as still present but dating from before the first recognition of signs and symptoms of epilepsy (yes/no). E.g. Being an outsider at school, noticeably sad, noticeably hyperactive and dominant.

  2. Long-standing learning problems, i.e. disappointing school results perceived by the parent as still present but dating from before the first recognition/appearance of signs and symptoms of epilepsy (yes/no). E.g. Due to lack of concentration, to school absenteeism or with no clear cause.

  3. Family trouble, such as marital distress, divorce, psychopathology in another member of the family (yes/no).

  4. Parents' perceptions of discontinuity in their parenting habits in relation to the onset of epilepsy and hearing about the diagnosis (thrown off balance/not thrown off balance). E.g. Being emotionally overwhelmed by the signs and symptoms and diagnosis, undifferentiated fear, excessive fear of stigmatisation, steering clear of relevant information.

  5. Parent's perception of the child's maladaptive reaction to changes caused by/associated with the onset of epilepsy (maladaptive/adaptive). E.g. Excessive feelings of shame and withdrawal interfering with going to school or other social activities, excessive fear of recurrence of seizures or of adverse effects of anti-epileptic drugs, death wish or benefit from being ill.

Neuropsychological assessment

The comprehensive neuropsychological assessment covered major domains of cognition, academic language skills, and motor and mental speed (see Box 2 for tests, dependent variables and their definitions). Available instruments that were downward extensions of adult psychological tests not suitable for children were adapted. If no proper tests were available, new tasks were developed; these pertained to Word Span forwards and backwards and aspects of learning and memory (learning locations; Schouten et al., 2002), several aspects of attention (balloon piercing; Oostrom et al., 2002), and behaviour regulation (Computerised Colour Trails I and II; Oostrom et al., 2002).

Box 2 Neuropsychological Assessment Battery


Tests/tasks (parameter)

General Intelligence

Coloured Progressive Matrices. Computerised (c-CPM) (Raven et al., 1990; Schufried, 1996a) (children aged <11 years) [norm score (IQ)]

Standard Progressive Matrices. Computerised (c-SPM) (Raven et al., 1992; Schufried, 1995) (children aged ≥11 years) [norm score (IQ)]

Vocabulary (Wechsler Intelligence Scale for Children—Revised, Dutch edition) (WISC-RN) (Bruyn et al., 1986) (standard score)


Colour Trails I and II. Computerized adaptation of Color Trails (Maj et al., 1993) (Oostrom et al., 2002) In Part I the child was requested to connect, sequentially and as fast as possible, numbered circles that were scattered over the touch-screen randomly. All odd numbers were printed in a yellow circle; all even numbers were printed in a pink circle. In Part II each number was printed twice: once in a pink and once in a yellow circle. The child was requested to connect circles in consecutive order from 1 to 25, by alternating between pink and yellow circles. The times needed for Parts I and II and the difference in performance time between Parts I and II were recorded electronically (ms).

Balloon Piercing (Oostrom et al., 2002). Seated in front of the touch screen the child watched a sequence of 375 yellow balloons entering the screen with randomly varying interstimulus intervals and wandering over the screen according to a fixed pattern of tracks. The balloons were provided with complete (target) or incomplete (distracter) faces (ratio 3 : 10). The child had to eliminate the targets as quickly as possible by piercing them (a touch on the target made the balloon pop). The time (ms) for piercing targets was recorded. The task lasted 10.3 min.


Learning Locations (computerized non-verbal task). Learning locations of visually presented images of natural objects in a 4 × 4 matrix (Schouten et al., 2002) [total immediate (immed.) recall: summed number of correctly recalled locations in each of the five successive, identical learning trials (maximum = 80); response times: electronically recorded time elapsed between stimulus presentation and the child's touching the location of choice during the learning phase (ms); proactive interference (interf.): number of responses after one presentation of the same images as in the learning trials, but now presented in reordered locations (the higher the number of correctly recalled locations, the lower the sensitivity to proactive interference)1].


Word span forward and backward: repeating orally presented sequences of nouns with imaginable theme (Schouten et al., 2002) (number of items in largest correctly repeated sequence).

Learning Locations (computerized non-verbal task). Learning of visually presented images of natural objects (4 × 4 matrix) (Schouten et al., 2002) (delayed recall: the number of correctly recalled locations after a delay of 30 min filled with other tasks).


Reaction Time (Vienna Test System) (Schufried, 1996b) simple and choice conditions, light, beep, light + light, light + beep targets [electronically recorded mean time (RT) (ms) of correct reactions].

Manual tapping [electronically recorded number of taps per 30 s (dominant hand)].

Academic linguistic skills (standard educational packages; selection based on the child's educational level) (Bulthuis-deVeer, 1970; Schippers and Sixma, 1974)

Reading [errors (% of number of text words)].

Writing to dictation [errors (% of number of dictated words)].


Child Behavior Checklist (CBCL), Dutch version (Verhulst et al., 1996), adapted scoring (Oostrom et al., 2001) on the Total Problems scale: raw scores obtained through standard computerized processing (Achenbach et al., 1991).

Teacher's Report Form (TRF), Dutch version (Verhulst et al., 1997), adapted scoring (Oostrom et al., 2001) (Raw scores on the Total Problems scale obtained through standard computerised processing (Achenbach, 1991)].

The patient and control child were assessed simultaneously in a van that was equipped with two assessment rooms and that was stationed in the grounds of the hospital where the child with epilepsy was treated.

Behaviour questionnaires

At every assessment parents and teachers were asked to complete behaviour checklists [CBCL (Achenbach and Edelbrock, 1991) and TRF (Achenbach, 1991)]. Ratings on the seven items that were identified as possibly eliciting ambiguous answers, because the rater could interpret the items in terms of both epilepsy and behaviour, were treated as missing values (Perrin et al., 1991; Oostrom et al., 2001b). For the interested reader, the items were: ‘confused or seems to be in a fog’ (13), ‘daydreams or gets lost in his/her thought’ (17), ‘nervous movements or twitching’ (46), ‘stares blankly’ (80), ‘strange behaviour’ (84), ‘wets self during the day’ (107) and ‘wets the bed’ (108). Raw scores rather than T-scores were used for data analysis in order to maximize differences within the normal range (Perrin et al., 1991). Only the Total Problems scale was studied due to concerns about the construction of the CBCL and TRF scales (Hartman et al., 1999).

Statistical analysis

Data were analysed by means of SPSS 10.0 for Windows. We used χ2 analyses to test the interdependency of categorical data that were subsequently used as between subject factors in general linear model repeated measures analyses of variance (GLM–ANOVA; full factorial) with neuropsychological and behavioural data as dependent variables and time after diagnosis as within-subject factor. Independent epilepsy-related variables were: medical status (epilepsy versus healthy); the level of aetiological classification (idiopathic versus cryptogenic); seizure remission (remission versus ongoing seizures); and AED treatment (AED versus no AED). Variables that were not intrinsically epilepsy-related were school achievement (with school delay or remedial teaching versus without school delay) and five interview variables (see Box 1 for more detail). Since the modest number of subjects could not support analysis of all variables as factors in mixed design analyses, for the most part separate analyses were carried out. Age was a covariate in all but age-normed (CPM/SPM and WISC-R Vocabulary) measures. Tests were two-sided, with a 5% significance interval. For multiple comparisons Bonferroni correction was applied.

In case-by-case analyses, for each parameter the criterion for underperformance was fixed at 2 SD worse than the mean performance in the control group. If, within the same assessment, underperformance concerned at least two parameters within the same domain it was considered a deficit in that domain. If a deficit persisted throughout at least 1 year it was called a persistent deficit.


Group analyses

Characteristics of the groups studied

Statistically significantly more patients (82%) than controls (62%) decided to reparticipate (χ2 = 11.43, P = 0.01). Comparison of epilepsy-related, demographic, school and interview variables did not yield any statistically significant differences between reparticipating children with EO and dropouts. Nor did reparticipating control children differ from dropouts in demographic or school variables.

Characteristics of the children are displayed in Table 1. Epilepsy was localization-related in 11 children with idiopathic epilepsy (ILAE 1.1.a) and in 19 children with cryptogenic epilepsy (ILAE 1.3.a) (eight frontal, four temporal, and seven with no definable focus). Generalized epilepsies included childhood absence epilepsy (ILAE 2.1.d) in seven and other generalized epilepsy syndromes (ILAE 2.1.h) in five children. Immediately after the first neuropsychological assessment, 25 children started AED treatment (12 with cryptogenic epilepsy and 13 with idiopathic epilepsy), of whom seven were withdrawn from medication (four after being free of seizures within 3 months of treatment; two within 6 months of treatment; one girl decided to withdraw in spite of ongoing sporadic tonic–clonic insults). Of the 17 children who did not immediately start AED medication after diagnosis, seven initiated treatment within the first 2 years; of the remaining 10 children, five children were free of seizures within at the most 18 months after diagnosis and the other five continued to have seizures at low frequency. Interdependencies between parameters of epilepsy appeared not to be statistically significant.

View this table:
Table 1

Characteristics of children with ‘epilepsy only’ (EO) and controls (healthy classmates)

Children with EOControlsTotalP*
Age at diagnosis (years; mean ± SD)8.4 ± 1.99.2 ± 3.38.8 ± 2.68.6 ± 2.88.7 ± 2.7NS
IQ at intake (mean ± SD)102 ± 11108 ± 16102 ± 17101 ± 16104 ± 14NS
Having repeated a school grade (n) before intake4482–410–12NS
Having repeated a school grade at 4th assessment (n)66123–715–19
Using AED at 4th assessment111425
Seizure remission of >1.5 years at 4th assessment28
    With AED71219
    Without AED909
  • * Patients compared with controls; missing data for

  • two and

  • four controls. IQ = intelligence quotient; AED = anti-epileptic drug; NS = not significant.

In the interviews, parents of 11 children with EO claimed freedom from any contextual problem. Single clues for contextual problems (Box 1) were found in 15 children and multiple clues were found in 16 children. Overall, the parents of 21 children (50%) were thrown off balance as far as their habitual parenting was concerned and they attributed this to the diagnosis of EO or to the epilepsy in a wider sense (e.g. clearly distorted expectations regarding the child's future including marriage and driving licence, overprotectiveness due to excessive fear of adverse AED effects or fear of recurrence of seizures), eight children (19%) were themselves maladapted (e.g. excessive feelings of shame and withdrawal, interfering with going to school or other social activities), 11 (26%) belonged to socially or relationally disrupted families, 12 (29%) had behavioural problems antedating the diagnosis of epilepsy and eight (19%) had learning problems antedating the diagnosis of epilepsy.


The neuropsychological test results of children with EO and controls are summarized in Table 2. Table 3 shows the results of statistical analyses. Children with EO performed statistically significantly worse than healthy classmates on measures of learning (learning locations: proactive interference, response times, total immediate recall), word span backwards and sustained attention (balloon piercing) (Table 3, column ‘Medical status’; i.e. patients versus healthy controls). Mixed design analysis with both medical status and school achievement (i.e. repeated versus did not repeat a grade at school) as independent variables showed children with EO to be significantly more sensitive to proactive interference, whereas children who repeated a grade at school once (whether a patient or a control child) performed worse across measures of intelligence, attention, working memory, writing to dictation and simple reaction time to light [Table 3, column ‘MS × school achievement (SA) main effects]. The slight statistical significances of the separate analyses suggest that the disadvantage of the children with EO is not bound to idiopathic or cryptogenic aetiology, to seizure control or to AED treatment (Table 3, columns ‘Aetiology’, ‘Seizure remission’ and ‘AED treatment’).

View this table:
Table 2

Results of neuropsychological assessment in children with EO and healthy classmates, shown separately for each assessment

DomainAt diagnosis3 months12 months42 ± 6 months later
Test/taskPatients Mean (SD)Controls Mean (SD)Patients Mean (SD)Controls Mean (SD)Patients Mean (SD)Controls Mean (SD)Patients Mean (SD)Controls Mean (SD)
General intelligence
c-CPM/c-SPM (IQ)102 (17)101 (16)104 (15)104 (15)103 (13)104 (17)94 (15)97 (15)
Vocabulary (standard score)9.8 (2.8)10.3 (2.9)9.7 (2.7)9.8 (2.7)9.4 (2.2)10.1 (2.6)8.2 (2.2)9.6 (2.6)
c-Colour trails (ms)
    Part I136 835 (65 225)128 116 (55 017)118 909 (46 935)113 817 (45 206)113 755 (41 927)104 269 (39 399)93 153 (33 471)85 435 (26 219)
    Part II224 314 (10 2287)223 501 (96 773)203 423 (122 307)202 305 (102 005)195 734 (80 645)190 214 (85 332)167 128 (68 323)131 592 (38 137)
    Interference (II–I)79 468 (74 023)79 388 (75 054)101 456 (116 131)88 488 (68331)81 979 (53296)85 945 (57858)73 976 (47320)46 157 (23662)
Balloon piercing4000 (1194)3662 (1011)3731 (1063)3443 (650)3651 (983)3218 (473)3463 (787)3181 (539)
Learning locations
    Total immediate recall (no.)53 (12)54 (12)58 (12)62 (9)60 (10)63 (11)60 (12)65 (6)
    Response time2314 (677)2338 (632)2170 (693)1774 (296)2050 (773)1792 (335)2068 (788)1613 (442)
    Proactive interference (no.)5.0 (2.5)6.1 (2.1)5.5 (2.7)6.9 (2.9)5.8 (2.4)6.9 (3.0)6.7 (3.3)7.9 (3.3)
Word span (no.)
    Forwards3.8 (1.2)3.6 (1.2)4.3 (1.3)4.3 (1.0)4.5 (1.0)4.5 (1.1)4.7 (0.9)5.0 (0.9)
    Backwards2.9 (1.0)3.2 (0.8)2.8 (0.8)3.1 (0.8)3.1 (0.9)3.4 (1.1)3.4 (0.9)3.8 (1.0)
Learning locations (no.)
    Delayed recall12.4 (2.6)13.3 (2.7)12.9 (2.9)14.2 (1.6)13.1 (2.8)13.9 (2.8)13.7 (2.0)14.6 (1.4)
Speed (ms)
Reaction times
    Simple Rtlight544 (104)562 (118)521 (86)528 (93)542 (114)517 (82)521 (183)479 (89)
    Simple Rtbeep486 (111)487 (102)467 (93)453 (84)477 (90)445 (77)437 (86)423 (89)
    Choice Rtlight+light670 (120)674 (101)767 (168)647 (112)641 (93)640 (101)604 (94)575 (86)
    Choice Rtlight+beep666 (151)691 (86)759 (469)654 (108)658 (104)652 (108)599 (112)575 (106)
Manual tapping (no.)139 (25)140 (23)140 (23)135 (31)143 (22)142 (21)147 (28)149 (29)
Academic linguistic skills (% errors)
Reading4.2 (3.0)3.9 (3.1)5.0 (5.3)4.9 (4.4)4.0 (3.3)3.7 (3.3)5.4 (3.8)4.5 (3.5)
Writing to dictation14.1 (13.8)14.7 (15.3)9.8 (8.4)13.5 (8.2)14.2 (13.4)12.5 (12.6)13.1 (10.4)10.8 (9.0)
Behaviour (total score)
CBCL27.1 (21.9)14.3 (11.5)23.0 (21.3)10.7 (9.4)25.4 (22.4)8.7 (9.5)18.2 (14.7)11.9 (12.2)
TRF21.4 (21.2)8.1 (9.3)21.1 (22.3)7.1 (10.3)26.9 (22.7)10.9 (11.8)23.8 (19.2)5.6 (4.7)
  • Abbreviations: see Box 2.

View this table:
Table 3

Statistically significant differences between (subgroups of) children with EO and healthy controls over time (P value, F value) (blanks: not statistically significant)

Medical status (MS)MS × school achievement (SA)AetiologySeizure remissionAED treatment
Main effectMain effectsInteraction effectsMain effect (to the disadvantage of)
EO children vs controls*MS*SAMS × SAMS × SA × timeIdiopathic£Cryptogenic¥RemissionOngoing seizuresNo AEDAED$
General intelligence
Vocabulary0.05, 4.19
Part I0.00, 13.880.03, 3.48
Part II0.00, 16.070.02, 3.55
Interference (II–I)0.00, 13.58
Balloon piercing0.03, 4.780.03, 4.700.03, 3.760.02, 4.12
Learning locations
    Total immediate recall0.05, 4.190.04, 3.32
    Response time0.03, 5.070.01, 4.90
    Proactive interference0.01, 6.450.01, 6.890.02, 4.340.04, 3.33
Word span forward0.04, 8.84
Backward0.03, 4.860.03, 4.920.03, 3.59
Learning location
Delayed recall0.02, 4.490.03, 3.59
Reaction times (RT)
    Simple RTlight0.01, 17.94
    Simple RTbeep
    Choice RTlight+light
    Choice RTlight+beep
Manual tapping
Academic linguistic skills
Writing to dictation0.04, 4.780.02, 3.74
CBCL total score0.01, 8.380.01, 4.65
TRF total score0.00, 17.17
  • Differences are to the disadvantage of: *children with EO

  • who once repeated a grade at school

  • £ with idiopathic epilepsy

  • ¥ with cyptogenic epilepsy

  • in seizure remission

  • $ using AED

  • this significant difference can not be interpreted as adverse AED-effect because differences already existed at pre-treatment baseline and no interaction effects with time were found. AED = anti-epileptic drug; EO = ‘epilepsy only’; MS = medical status; SA = school achievement; RT = reaction time. For other abbreviations see Box 2.

When considering only patients, those who repeated a grade once had statistically significantly worse scores than those without educational delay on measures of attention [Colour Trails Part I (P < 0.001, F = 10.21), Part II (P < 0.001, F = 12.26) and interference (P < 0.00, F = 12.74)], memory [learning locations delayed recall (P < 0.05, F = 6.29)]), reading (P < 0.05, F = 4.13) and simple reaction time to light (P < 0.01, F = 7.83). Epilepsy characteristics (idiopathic versus cryptogenic aetiology of epilepsy, seizure remission [defined as seizure freedom of at least 1.5 years] and AED treatment) were statistically insignificant factors. However, worse scores in measures of attention [response time for piercing balloons (P < 0.05, F = 5.81)] and learning locations [total immediate recall (P < 0.00, F = 21.36), response time (P < 0.05, F = 5.90) and proactive interference (P < 0.01, F = 7.04)] were found in children who were exposed to off-balance parenting. As could be expected, children with long-standing learning problems performed worse than children without learning problems on intelligence [CPM/SPM (P < 0.00, F = 13.11)], word span backwards (P < 0.04, F = 4.63), response time when piercing balloons (P < 0.00, F = 13.12) and reading (P < 0.00, F = 9.11). Patients with longstanding behavioural problems, and those belonging to troubled families (P values < 0.00, F values 10.95 and 12.68 respectively) were reported to have more behavioural problems (CBCL) than patients without a history of longstanding behavioural problems and patients belonging to happy families.


Mean CBCL and TRF total problems scores are shown in Table 2. In general, parents as well as teachers reported more behavioural problems in children with EO than in healthy controls (Table 3, column ‘Medical status’). Differences could not be related to school achievement, idiopathic versus cryptogenic aetiology of epilepsy or AED treatment. Parents of children in seizure remission reported the most behavioural problems (Table 3, column ‘Seizure remission’).

However, within the group of patients, having or not having reached seizure remission was not statistically significant. Patients who belonged to troubled families were rated by parents as having more behavioural problems than those growing up in happy families. Of course, parents reported more problems in children with a history of behavioural problems than in children with no history of behavioural problems.

Time effects

In the complete sample (patients and controls), statistically significant improvement of test performances (all P values < 0.05) during follow-up was found in computerised colour trails parts I and II, word span (forwards), balloon piercing (response time), tapping and learning locations (total immediate recall and response time). Age was statistically significant in all relevant measures except for response time in learning locations and for measures of behaviour. It was only in both parts of colour trails and in writing to dictation that medical status seemed important: school repeaters with EO continued to perform poorly whereas initial differences between healthy school repeaters and healthy non-repeaters disappeared over time (Table 3, column ‘MS × school achievement (SA)’).

Cases with persisting deficits

Summated over the four assessments, virtually all patients (41/42) and controls (28/30) underperformed, that is, obtained a score that was at least 2 SD worse than the mean of the controls, on at least one parameter within a domain and in at least one assessment. No statistically significant difference was found between the number of patients (16/42) and the number of controls (6/30) qualifying for a deficit, defined as at least two underperformances within the same domain (38 and 20% respectively). In the complete sample, statistically significantly more patients (n = 8;19%) than controls (n = 1; 3%) had persistent deficits (χ2 = 4.58, P = 0.03). Their characteristics are presented in Table 4.

View this table:
Table 4

Characteristics of the eight children with epilepsy and one control child with persistent cognitive deficits

PatientSex (M,F)Age at intake (years)IQ at intakeIndex seizureILAE*Seizure remissionDeficit
1M5104GTC + CPS1.1.a>1.5 yearsAttention, learning
2F13130GTC1.3.a>2 yearsLearning
3F1182GTC1.3.aOngoing seizuresBehaviour
4M8107Absence2.1.d>1.5 yearsAttention, academic skills
5M578CPS1.3.a>1.0 yearsLearning
6F592Absence2.1.d>1.5 yearsLearning, behaviour
7M698CPS1.3.a>1.5 yearsBehaviour
8F999CPS1.3.a>1.5 yearsBehaviour
  • * Engel; 2000

  • with AED.

  • GTC = generalized tonic clonic seizure(s); CPS = complex partial seizure(s).

χ2 tests of differences between these eight patients with persistent deficit(s) and the 34 patients without persistent deficit(s) showed overrepresentation of maladaptive parenting (χ2 = 4.72, P = 0.03), family problems (χ2 = 6.03, P = 0.01) and behavioural problems prior to the diagnosis (χ2 = 9.57, P = 0.01) in the eight patients. No statistically significant relationships were found between persistent deficits and epilepsy characteristics or school achievement.


Long-term psychosocial and social disadvantages of having had epilepsy in childhood have been convincingly described (Jalava et al., 1997; Morgan et al., 2000) but it remains unclear how these disadvantages arise. The present follow-up study was performed to prospectively assess cognitive and behavioural functioning in otherwise healthy schoolchildren with EO, in the first 3–4 years after diagnosis. This was done by comparing (subgroups of) children with EO with healthy age- and gender-matched classmates, by comparing within the epilepsy group those with idiopathic to those with cryptogenic aetiology, those on and off AED and those who had attained seizure freedom to those who had not, and by case-by-case analyses. Strengths of the study are the relative homogeneity of the epilepsy group as far as level of aetiological classification (e.g. idiopathic versus cryptogenic) and illness stage are concerned, developmentally sound assessment procedures, a control for response ambiguity in the behavioural questionnaires, and a combined qualitative and quantitative approach, the latter with appropriate processing of qualitative contextual information. Assets of the assessment van were that the children did not travel further than usual for hospital visits, thereby preventing travel-related fatigue of the children, and that the first assessment of patient and control child could be arranged flexibly within 48 h after diagnosis. Limitations of the study were, notwithstanding its multiple-centre approach and consecutive inclusion, the rather small number of patients and the sample attrition that resulted from the call-back procedure; however, epilepsy-related biases resulting from the dropping out of participants were not found. In the light of the pivotal role that contextual variables proved to play, we now regret our earlier decision to not interview the parents of the healthy control children.

Differences between groups are strongly influenced by school career

Throughout the follow-up, the group of children with EO lagged behind in both performance times and, particularly in the domains of learning and attention, content. They also exhibited more problematic behaviour. However, after controlling for the possible influence of repeating a grade at school, the only remaining disadvantage in the cognitive domains turned out to be increased susceptibility to proactive interference. Proactive interference is the well-known phenomenon that previous learning inhibits new learning of similar material. It is closely related to poor concentration and person-related variables, such as not being certain of oneself (Torgesen, 1994). Moreover, children with EO who had once repeated a grade continued to have poor scores in Colour Trails I and II and continued to make more writing errors, whereas repeaters without EO caught up with non-repeaters. In the comparisons with healthy classmates, especially children with cryptogenic epilepsy and those to whom AED was prescribed, scored more poorly. The latter finding cannot be explained as an adverse effect of the medication, as differences between children with and without AED already existed at the pretreatment baseline. Rather, specialists may be more apt to prescribe AED to children with EO and school or learning difficulties than to those whose parents do not have any complaints about their children's cognition and/or behaviour. Finally, the parents of children with ongoing seizures were found to report less behavioural problems than parents of children in seizure remission. This finding is at variance with earlier findings in children with new-onset epilepsy, which were obtained, however, in a less rigidly composed sample including children with symptomatic epilepsy (Austin et al., 2002). For the present results, one may speculate that being seizure-free may in some cases produce a ‘burden of normality’, as described after epilepsy surgery and other chronic conditions (Wilson et al., 2001).

Within the EO group, context rather than epilepsy variables are associated with cognitive and behavioural problems

Throughout the follow-up, neither idiopathic versus cryptogenic aetiology, seizure remission nor use of AED had clinically meaningful effects with respect to cognition or behaviour. On the other hand, contextual variables did turn out to be associated with these disadvantages, off-balance and maladaptive parenting in the earliest months after diagnosis being especially strongly correlated. Attention has previously been drawn to the predicament of caregivers. In particular, maternal adaptation to a child's epilepsy has been expounded to be a complex factor in the child's adaptation (Shore et al., 2004). The relationship between parental anxiety and the child's quality of life has also been described previously (Williams et al., 2003). Quite recently, research by Austin's group drew attention to moderating effects of family mastery on academic achievement in writing and possibly reading and on behavioural problems (Austin et al., 2004; Fastenau et al., 2004). We have reported previously that maladaptive parenting has traceable associations with the child's neuropsychological performance and school achievement. The present paper provides evidence that these associations are still present 3.5 years after diagnosis.

Persistently disadvantaged children

A few studies have addressed clinically significant cognitive disadvantages or deficits and reported these to be present in about 20% of the samples under study (Schoenfeld et al., 1999; Seidel and Mitchell, 1999), which is not dissimilar from the present finding. The case-by-case approach of the present study facilitated the evaluation of these disadvantages in respect of their distribution within the groups under study and their development over time. In more than 80% of children with EO cognitive and behavioural problems were absent or, interestingly, not persistent. Two issues are worthy of mention. In the almost 20% of children who were found to have a so-called deficit, poor parenting, unhappy family situations and prior existence of behavioural problems were overrepresented. This finding is at least partly in line with previous research, which found that children with epilepsy whose parents were perceived as having an overcontrolling approach had more problems than children with similar types of epilepsy who did not perceive their parents as overcontrolling (Carlton-Ford et al., 1997). Secondly, non-persistent cognitive and/or behavioural problems were demonstrated in nearly twice as many children with EO (38%) than in healthy children (20%).

These findings fit with and expand on our earlier work, in which we described multiconditional vulnerability to the cognitive and/or behavioural problems of children with EO in the first year after diagnosis (Oostrom et al., 2003). Although the predominant association of context variables with the vulnerability of children with EO to cognitive and behavioural problems may not be specific to children with epilepsy, it urges researchers and practitioners to go beyond the strictly medical variables and to explore parental and other contextual risks.


We are most grateful to all the children who participated in our study. We also thank their parents and teachers. We thank Mrs W. Meijer MA for her contribution in the data collection, Prof. Dr O. Brouwer for his critical comments on earlier versions of the manuscript and Mrs M. Schinkel MA for language editing. The study was supported by the Dutch Epilepsy Foundation (NEF), JANIVO Foundation, Peugeot Holland NV and Foundation for the Advancement of Neuropsychological Research in Children (BNOK). Participants of the Dutch Study of Epilepsy in Childhood (DuSECh) are: W. F. M. Arts MD PhD (UMC Rotterdam), A. C. B. Peters MD PhD (UMC Utrecht), O. F. Brouwer MD PhD (UMC Groningen), C. A. van Donselaar MD PhD (UMC Utrecht/St Clara Hospital Rotterdam), E. A. J. Peeters MD (Juliana Children's Hospital, The Hague), H. Stroink MD PhD (Hospital St. Elisabeth, Tilburg).


View Abstract