Learning and memory of school children with epilepsy: A prospective controlled longitudinal study

Learning and memory of school children with epilepsy: A prospective controlled longitudinal study

Schouten, A

The medical prognosis of idiopathic and cryptogenic epilepsy in children with normal intelligence is held to be favourable. Seventy percent of these children become seizure-free when treated with antiepileptic drugs (AEDs; Berg et al. 1995), and another 70% remain so after discontinuation of AEDs (Berg and Shinnar 1994). Throughout this paper, this subset of children with epilepsy is referred to as children with ‘epilepsy only’ (Sillanpaa et al. 1992).

The school career of children with ‘epilepsy only’ appears to be at risk. Several research groups have established that repeating a year at school is more frequent among children with ‘epilepsy only’ than among children with chronic diseases (e.g. asthma or migraine) or among healthy classmates (Austin et al. 1999, Bailet and Turk 2000, Schouten et al. 2001). Significantly more children with ‘epilepsy only’ than their healthy classmates require special educational assistance (Schouten et al. 2001). School-administered group tests and other assessments of academic proficiency have revealed underachievement at school in children with ‘epilepsy only’ (Huberty et al. 1992, Sturniolo and Galletti 1994, Austin et al. 1999, Metz-Lutz et al. 1999, Bailet and Turk 2000). The cause of the underachievement at school is unclear.

Neuropsychological studies report worse results in children with epilepsy when compared with controls, but there is no unanimity with respect to the pattern of impaired and intact domains of cognition. Impaired learning and memory in the presence of normal attention (Forsythe et al. 1991) and attention deficit in the presence of normal learning and memory (Stores et al. 1992) have both been described. Other studies have stressed transient effects on memory (Metz-Lutz et al. 1999, Bailet and Turk 2000) or both attention and memory (Deonna et al. 2000). Vulnerability of working memory was found in children with temporal lobe epilepsy, all of whom had normal intelligence. Their sub-normal performance was only detected when demand on working memory was increased, that is, when the interval between stimulus and recall was prolonged and when rehearsal was interfered with (Hershey et al. 1998). Unfortunately, the authors did not separately analyze the effect of abnormalities on MRI, present in seven of 28 children, nor did they examine the relevance of the finding for the school career of the children. This issue has not yet been addressed in any study

Ongoing seizures are generally considered to underlie both disappointing school progress and problems in learning and memory of children with ‘epilepsy only’ (Metz-Lutz et al. 1999, Schoenfeld et al. 1999, Deonna et al. 2000). Variables other than epilepsy-related ones may have an impact as well. Austin and colleagues (1999) suggested that contextual factors, such as attitudes of parents and of children or behavioural problems, should be among the objects of future investigations. A few case-by-case studies found memory deficits to be transient (Metz-Lutz et al. 1999, Deonna et al. 2000), whereas others reported persistent underachievement at school (Austin et al. 1999).

The primary aim of the present study was to establish whether children with recently diagnosed ‘epilepsy only’ who receive mainstream education, perform worse than healthy controls in various aspects of learning and memory. We intended to study the effects of both epilepsy variables and contextual variables (contextual variables were derived from semi-structured interviews at diagnosis). Finally, we queried whether poor performance in the learning and memory tasks was related to underachievement at school, that is, repeating a year and requiring special educational assistance. We studied the children group-wise as well as case-by-case, at diagnosis and three and 12 months later. The study was part of the multicentre comprehensive neuropsychological assessment of cognition and behaviour, a project of the Dutch Study Group of Epilepsy in Childhood (DuSECh; Arts et al. 1999), and was approved by the ethics and research committees of the ten participating hospitals. To investigate the main research question – Do children with ‘epilepsy only’ have problems of learning and/or behaviour? – we calculated the minimally required number of children in each group to be 62 if: (1) alpha and beta (chance of a false positive/false negative decision) were 5%, (2) the chance of learning problems in the population under study (‘epilepsy only’) was estimated at 30% (for the total population of childhood epilepsy the literature yields prevalence estimations varying between 50% and 5%), and (3) the prevalence of learning problems in the population of children in mainstream schools was 5%. This paper reports on a section of the entire study.

All children participated on the basis of written informed consent by their parents. Older children (age 12-16 years) also gave their own informed consent.



Seventy-two children were consecutively invited to participate (Table I). As the parents of one child with absence seizures refused participation because of time limitations, seventy-one children were included between January 1997 and October 1998. Inclusion criteria were two or more unprovoked seizures, an idiopathic or cryptogenic aetiology according to the epidemiological standards set by the International League Against Epilepsy (ILAE; 1989); age between 5 and 16 years, and normal education. Cryptogenic epilepsy was presumed to be symptomatic, but no aetiology could be identified (ILAE 1989). Exclusion criteria were any associated neurological disorder (found by history, physical examination, or MRI), chronic illness (e.g. diabetes mellitus or asthma), or previous use of AEDs. When followed up, two children with epilepsy had to be excluded, one because of simultaneous presence of an additional chronic illness of which no one was initially aware, the other because the diagnosis was revised to non-epileptic psychogenic seizures, thereby reducing the sample to 69 children with epilepsy. Aetiology epilepsy type, and seizure type were classified by a paediatric neurologist in attention and doublechecked by two senior neurologists according to DuSEChprotocol based on the classifications of the ILAE (1981, 1989). Data on AED-treatment and on seizure remission when examined 12 months after diagnosis were provided according to a fixed protocol by children’s neurologists. Fifty children (72%) received an AED.

Healthy children who were matched to the children with epilepsy for age, sex, and educational level (classmates) provided data to control for effects of re-testing and of normal development. Children with epilepsy and their parents requested their cooperation. Three of the children with epilepsy could not find a suitable classmate within the short time span between diagnosis and the beginning of AED-treatment.

Cognition and school career

IQs (depending on age, Coloured or Standard Progressive Matrices; Raven et al. 1990; 1992) showed no differences between patients and controls. Before diagnosis, 15 children with epilepsy (22%) had repeated a year at school, compared with 10 control classmates (15%). Over the one-year followup, five more children with epilepsy (9%), including one boy who had already repeated a year before diagnosis, repeated a year of school, compared with two control children (3%; see Table I). In the general Dutch population 11% of the children repeat one year of school. Before diagnosis, special educational assistance had been requested for more children with epilepsy (54%) than for their healthy classmates (23%). Over the year, the percentages of children who had required special educational assistance did not increase in both groups (Schouten et al. 2001). There was no relationship between repeating a year nor requiring special educational assistance and epilepsy variables (idiopathic or cryptogenic aetiology, epilepsy type, either having or not having reached seizure remission of >6 months at 12 months after diagnosis) and whether or not the patients received AED treatment.

Contextual variables

At first assessment, semi-structured interviews were carried out with parents of the children with epilepsy. The interviews recorded prior adversity in the children’s lives (Oostrom et al. 2001). In Table II, parents’ adaptation refers to parents’ perceptions of changes in their parenting caused by the onset of epilepsy Children’s adaptation refers to parents’ perceptions of their children’s adaptation to their situation, changed by the onset of epilepsy. Long-standing behavioural problems are problems perceived as to have already been present before the first signs of epilepsy. Long-standing learning problems are problems perceived as already present before the first signs of epilepsy. Family problems refer to, for example, marital distress, divorce, or psychopathology in another member of the family (Table II).


Word Span

In school children, registration is commonly examined by means of digit repetition (De Bruyn et al. 1986, Baddeley 2000). Digits have poor relevance for younger school children (Kuhn 2000). We used nouns with a theme of which one could easily make a mental picture, as so-called ‘imageable’ words have shown to be better suited to this age group and to be appropriate also for older children (Klapwijk and de Jong 1995, Schouten et al. 2000a). Nouns denoting objects that could be used when travelling were selected from a list of words that are known to be acquired before the age of sixyears (Kohnstamm et al. 1981). The nouns were highly ‘imageable’ (scores 6 or 7 on the 7-point scale of ‘imageability’ norms for Dutch words; Van Loon-Vervoorn 1985) and had low association rates (De Groot 1980). As in the subtest ‘Digit Span’ of the Dutch version of the Wechsler Intelligence Scale for Children – Revised (De Bruyn et al. 1986), the nouns consisted of one or two syllables. Strings of nouns of increasing length were devised, each length was represented by two strings. The children were requested to repeat the orally presented strings of nouns until they failed on two strings of the same length.

Dependent variables were: (1) Span forwards, number of nouns in the longest string that the child repeated correctly in forward order; (2) Span backwards, number of nouns in the longest string that the child repeated correctly in backward order. This condition was more demanding than span forwards; (3) Number of intrusions, i.e., nouns not occurring in the strings.

Learning and retention: location learning

Learning and retention are usually examined by means of tasks requiring the memorization of lists of words. However, when rehearsing lists of words, young children (age 4 to 6 years) show little increment over trials (Goodman et al. 1999). We took advantage of the fact that children learn ‘where’ more easily than ‘what’ (Tipper and McLaren 1990, Schumann-Hengsteler 1996) and designed a computerized game that required the children to memorize the locations of 16 coloured pictures of natural objects, such as fruits, flowers, and butterflies. In children, spatial learning is known to be sensitive to dysfunction of the hippocampus (Hershey et al. 1998, Vargha-Khadem et al. 1997), which endorses the relevance of the material. The task was constructed similar to the much-used California Verbal Learning Test (Delis et al. 1986). As there were neither floor nor ceiling effects in any age group, the task was considered to be developmentally adequate (Schouten et al. 2000a). The relevance of learning tasks for scholastic progress depends not so much on retention within assessments as on the degree to which the learning result sinks in and is retained over longer time intervals. An index of long-term retention was acquired by calculating the savings at relearning after 3 and 12 months. Task demand was increased by the insertion of a trial with reordered locations of the pictures. Recall of the new locations yielded a measure of proactive interference (see below); after that, recall of the original locations yielded a measure of retroactive interference.

Dependent variables were: (1) Total immediate recall, summed number of correctly recalled locations in each of five successive, identical learning trials (maximum=80); (2) Within-assessment retention, number of correctly recalled locations after a delay of 30 minutes filled with other tasks, corrected for recall in trial 5. Calculated as (recall trial 5 minus delayed recall)/recall trial 5; (3) Recognition, number of correctly recognized locations (maximum= 16); (4) Long-term retention, savings when relearning after three and 9 months, calculated as (recall trial 1 in present assessment minus recall trial 1 in prior assessment)/recall trial 1 in prior assessment; (5) Proactive interference, number of correctly recalled reordered locations. We used raw scores, as the score in the proactive interference trial was independent of prior learning; (6) Retroactive interference, number of correctly recalled original locations after having recalled reordered locations, corrected for recall in trial 5. Calculated as (recall trial 5 minus recall retroactive interference trial)/recall trial 5; (7) Response times (ms), i.e., time elapsed between stimulus presentation and the child touching the location of choice.


The first assessment was within 48 hours of diagnosis and before AED-treatment. Re-assessments were 3 months and 12 months after diagnosis. The multicentre effort to obtain a sufficient number of children who fulfilled the inclusion criteria, and the requirement of assessing a patient and a matched control within 48 hours of diagnosis in standard conditions, necessitated the availability of transportable examination equipment. A van was converted into a laboratory with two examination rooms. Each room was provided with a table, two chairs, and a personal computer. The monitors had touch screens for responding, with the advantage that differences in familiarity with the use of a ‘mouse’ were avoided. The assessments were performed at the grounds of the hospitals where children with eplepsy were treated, thus preventing unwanted effects of differences in travelling time. The van was parked in a quiet area in the hospital grounds.


The criterion for ‘under-performance’ was fixed at 2 standard deviations (SD) worse than the mean of age-matched controls. If under-performance persisted throughout the year it was called a ‘deficit’.

Analysis of the data was performed using SPSS (version 8.0). Longitudinal data on interval level were analyzed by repeated measures analysis of variance (full factorial), checked for homogeneity and normality of variance. A KolmogorovSmirnov difference of

Differences were considered to be statistically significant if their two-tailed p-values were 0.05 or lower. If ap-value was between 0.05 and 0.10, a tendency to significance was presumed. We report the actual mean difference as well as the relative difference expressed in percentages ([actual difference/grand mean]x100), the 95% confidence interval in case of statistical significance, and the 90% interval in case of a tendency to significance.

Analysis with inclusion and exclusion of the children with unclassifiable epilepsy (n= 3) made no difference. As these children were considered to have epilepsy they were included in all analyses, except for the analyses of epilepsy variables in which they were unclassifiable. Controls participated in all analyses, except for those concerning the variables derived from the interviews.



No statistically significant differences between the children with epilepsy and their healthy controls were found in span forwards, total immediate recall, retroactive interference, within-assessment retention, recognition, retroactive interference, long-term retention, errors in span forwards and backward, nor response times (Tables III and Ice. This means that registration, learning, and retention were basically normal. However, children with epilepsy performed worse than controls in conditions of increased demand. In span backwards the difference was statistically significant, in proactive interference the difference tended to be so. These effects remained present over the year, as indicated by the absence of any statistically significant interaction with time (Table III).


The small difference between patients and controls in span backwards was not related to aetiology of epilepsy. Children with ongoing epilepsy had a worse span backwards than controls throughout the year (mean difference 0.52 [17%], 95% CI 0.11.94). Children who had a six-months’ seizure remission after one year could not be distinguished from controls. Furthermore, untreated children had worse spans backwards than controls (mean difference 0.66 [21%], 95% CI 0.10-1.20), while the children who were treated could not be distinguished from controls. This difference already existed at diagnosis, when no child had been treated, and did not change over time. A similar pattern was seen in span forwards. From the start, untreated children had worse span forwards than controls (mean difference 1.00 [21%], 95% CI 0.26-1.74) and slightly worse than treated children (mean difference 0.13 [3%], 95% CI 0. 100-1.64), while treated children did not differ from controls.


The effect of increased demand present in location learning, was related to contextual variables. Children of maladapting parents tended to experience more proactive interference when learning locations than children of parents of well-adapting parents (mean difference 0.97 [18%], 90% CI (0.13-1.80). This effect was present in all three assessments. Maladaption of parents showed, furthermore, a relationship with total immediate recall (mean difference 5.76 [10%], 95% CI 1.04-10.49). The children of maladapting parents recalled locations in trial 1 worse than children of well-adapting parents, but the performance increase over the five trials was equal in the two groups. This effect remained stable during the year.

Two other contextual variables had interaction effects over time. Children’s perceived adaptation to the onset of epilepsy interacted with retroactive interference (p=

Long-standing behavioural problems (prior to onset of epilepsy) also interacted with time (p=


Patients who repeated a year had worse spans forwards and backwards than controls who progressed normally through school (mean difference 0.74 [18%], 95% CI 0.01-1.48, and mean difference 0.71 [22%], 95% CI 0.18-1.25 respectively). Controls who repeated a year did not differ from other controls, neither did patients who moved up normally. Location learning showed no relationship to having repeated a year at school.

The need for special educational assistance was also related to word span. Patients for whom special educational assistance was required showed worse spans forwards than controls and than patients for whom this was not necessary (mean difference 0.75 [18%]; 95% CI 0.13-1.37, and mean difference 0.68 [17%], 95% CI 0.01-1.35 respectively) as well as worse spans backwards (mean difference 0.66 [21%], 95% CI 0.22-1.11, and mean difference 0.48 [16%], 90% CI 0.03-0.92 respectively). Controls who needed special educational assistance did not differ from other controls, nor did patients for whom special educational assistance was not necessary. Furthermore, patients who needed special educational assistance tended to have more proactive interference than controls who did not need special assistance (mean difference 1.16 [20%], 90% CI 0.03-2.31). Again, controls who did not require special educational assistance could not be distinguished from other controls, nor could patients who did not need special educational assistance.


Summated over the three assessments, 18 patients (26%) and 12 controls (18%) under-performed in one measure. Nineteen patients (28%) and two controls (3%) did so twice, whereas 10 patients (14%) and three controls (4%) under-performed at least three times in the same or in different assessments. The only feature shared by these 10 patients was the requirement of special educational assistance. The three under-performing controls were dissimilar in age, sex, level of intelligence, and achievement at school.

In every assessment, about the same proportion of children under-performed, but the composition of the group of underperformers changed. Only six children fulfilled the criteria for a deficit. In span forwards three younger patients (age below 7 years) persistently performed worse than their age-matched controls. One older patient (aged 11.7 years) and two controls (aged 6.2 and 8.9 years) under-performed persistently in total immediate recall and recognition. In other measures none of the children persistently under-performed.



When considered group-wise, normality of registration, learning and retention implies a more optimistic condition than has been suggested previously for children with ‘epilepsy only’ (D’Alessandro et al. 1990, Forsythe et al. 1991, Jambaque et al. 1993, Weglage et al. 1997, Croons et al. 1999, Metz-Lutz et al. 1999, Schoenfeld et al. 1999, Ballet and Turk 2000, Gulgonen et al. 2000, Pavone et al. 2001). Differences in selection of patients and controls may contribute to the disparity in findings. None of the previous studies features consecutive inclusion. Only three studies included children with newly diagnosed epilepsy (Forsythe et al. 1991, Metz-Lutz et al. 1999, Ballet and Turk 2000). Stores et al. (1992) also included only children with recently diagnosed epilepsy and reported normal memory performances. Hence, the stage of the disease may be relevant, as also suggested by reassuring findings in the end-stage of ‘epilepsy only’ (D’Alessandro et al. 1990).


When demand on working memory was increased, that is, when reproducing words in reversed order and when recalling previously learnt but randomly reordered material, we found the group of children with `epilepsy only’ performed worse than their matched controls. This finding extends those made by Hershey and colleagues (1998) in children with temporal lobe epilepsy, to the wider segment of children with ‘epilepsy only’. It shows, furthermore, that the effect of increasing demand on working memory pertains to a wider range of tasks than examined by Hershey and colleagues (1998). The group differences, although small, may reveal a vulnerability similar to ‘insufficiency of control processes’ (Jennekens-Schinkel et al. 1987), ‘lack of attention’ (Stores et al. 1992), and ‘instability of shifting set’ (Schouten et al. 2000b): features that previous studies found to characterize children with mild epilepsy. Authors who reported no deficits did not separately analyze spans forwards and backwards (Ballet and Turk 2000) nor did they introduce an interference condition in the learning task (Schoenfeld et al. 1999).


Whether these findings are causally related to the epilepsy (‘epilepsy specific’), or have to be conceived of as concomitants of having an illness, or more generally of having experienced a stressful event, needs further discussion. In favour of ‘epilepsy specificity’ is the relation with seizure remission. The children who had a favourable outcome could not be distinguished from controls, while the children who did not reach six months of seizure freedom had worse spans backwards than controls in all three assessments. This is in accordance with previous reports of the influence of active epilepsy (D’ Alessandro et al. 1990, Mandelbaum and Burack 1997, Weglage et al. 1997, Metz-Lutz et al. 1999, Schoenfeld et al. 1999, Deonna et al. 2000). Exploring the effect of epilepsy type on word span and location learning, we found that, groupwise, children with absence epilepsy had poor span backwards (data not shown) and those with temporal lobe epilepsy started poorly when learning locations. These findings need replication in larger epilepsy subgroups. However, this study was designed to investigate learning and memory in children with ‘epilepsy only’, and analyses of performances in subgroups await new studies. Furthermore, we found a vulnerability in conditions of increased demand only in patients with preexisting difficulties at school and not in controls with these difficulties. The fact that the school difficulties predate diagnosis leads us to suspect some underlying process which gives rise to both these difficulties as well as to the epilepsy. In a similar vein, Austin and colleagues (2001) report behavioural problems which predate diagnosis.

However, we cannot rule out that these children may face an effect of having experienced stressful events that have nothing to do with epilepsy. The significant association (113%) of the children’s adaptation to the onset of epilepsy on retroactive interference suggests that children who are thrown off balance by the onset of epilepsy are more susceptible to interference than well-adapting children. Similarly, but less pronounced, maladaptive parenting and long-standing behavioural problems had associations with proactive interference.

In other illnesses it becomes increasingly clear that illness factors and contextual factors have complex and interactive effects on performance and behaviour of children (Holmbeck 1997). That we found both epilepsy and contextual variables to be related to poor performance in conditions of increased demand, leads us to suspect such interaction effects in childhood epilepsy as well. These complex effects can, however, only be studied in large groups.


The group differences, although statistically significant, remained small and restricted to increased demand. This should not conceal that 54% of the children with epilepsy, as opposed to 26% of healthy classmates, under-performed in one or the other aspect of the tasks. Hence, although the major components of memory may be adequate, the individual risk of under-performing may be increased. However, at every assessment over the year after diagnosis, different children were responsible for the stable percentages of underperformance and this non-persistence is not compatible with the concepts of deficit, disorder, nor impairment. Therefore, we prefer the notion of vulnerability to define our major finding.

To conclude, children who are receiving mainstream education and who are diagnosed as having ‘epilepsy only’ are not compromised in the major components of learning and memory processing. Increase of task demands, e.g., by reordering the material or the required response, reveals a vulnerability that cannot be explained by epilepsy variables alone. Not reaching seizure freedom and particularly context variables, such as having difficulty adapting to the adversity of epilepsy onset, contribute to explain the vulnerability in children with ‘epilepsy only’. In individual cases, under-performance is neither consistent nor persistent. Yet, it is twice as frequent as among controls. For teachers the present findings imply that they should be alert to the occurrence of these inconsistent under-performances which may be remediated by providing more structure than is normally provided (Meltzer 1994). In the course of time small deviations can add up to inordinate ones.

Accepted for publication 11 March 2002

Participants of the Dutch Study of Epilepsy in Childhood: WFM Arts (Sophia Children’s Hospital, Rotterdam); JH Begeer (University Hospital, Groningen); OF Brouwer (at the time of study: Leiden University Medical Center); CA van Donselaar (University Hospital, Rotterdam); AT Geerts (University Hospital, Rotterdam); EAJ Peeters (Westeinde Hospital and Juliana Children’s Hospital, Den Haag); H Stroink (Sophia Children’s Hospital, Rotterdam); G Hageman (Medical Spectrum Twente, Enschede); R ten Houten (Medical Center, Alkmaar); JF de Rijk-van Andel (Ignatius Hospital, Breda); LME Smit (Free University Hospital, Amsterdam); MJ Wennekes (De Wever & Gregorius Hospital, Heerlen).


We thank the children, their parents, and teachers for their participation in the project, Prof FGI Jennekens for critical comments on the paper and Mrs G Griffiths for language editing. The study was supported by the National Epilepsy Foundation (Grant 97-04), the JANNO Foundation, and Peugeot Holland NV.

Copyright Mac Keith Press Dec 2002

Provided by ProQuest Information and Learning Company. All rights Reserved.