Cognitive Skills in Children with Intractable Epilepsy: Comparison of Surgical and Nonsurgical Candidates


Address correspondence and reprint requests to Dr. M.L. Smith at Department of Psychology, University of Toronto at Mississauga, 3359 Mississauga Road North, Mississauga, ON L5L 1C6, Canada. E-mail:


Summary:  Purpose: To compare neuropsychological performance of two groups of children with intractable epilepsy: those who are surgical candidates, and those who are not.

Methods: Intelligence, verbal memory, visual memory, academic skills, and sustained attention were measured in children aged 6–18 years. The effects of number of antiepileptic drugs (AEDs), seizure frequency, age at seizure onset, and duration of seizure disorder were examined.

Results: Both groups had high rates of impairment. Group differences were found only on the verbal memory task. Children who experienced seizures in clusters had higher IQ, reading comprehension, and arithmetic scores. Age at seizure onset and proportion of life with seizures were related to IQ. Performance did not vary with AED monotherapy versus polytherapy.

Conclusions: Few differences exist in cognitive performance between children with intractable seizures who are and those who are not surgical candidates. These findings suggest that children who are not surgical candidates can serve as good controls in studies on cognitive outcome of surgery.

Epilepsy in childhood results in an increased risk for cognitive, behavioral, emotional, psychiatric, and social impairments (see 1 for a recent review). These impairments occur with higher frequency in epilepsy than in other chronic medical illnesses, highlighting the importance of the underlying neurologic disorder in the etiology of the deficits (2–4). In the past several decades, epilepsy surgery has increased in prominence as a treatment for medically intractable seizures in children and adolescents (hereafter collectively referred to as children for the sake of simplicity). In carefully selected candidates, epilepsy surgery can be of benefit by ameliorating or significantly decreasing seizures in 50 to 90% of children (5–11).

Although the outcome of surgery has been examined with respect to its impact on seizure frequency and other neurologic variables, less is known about the effects on other aspects of the child and his or her development. In a review of the literature on the effects of surgery on psychosocial function in individuals younger than 16 years, Hermann (12) concluded that the few studies that existed had major methodologic flaws, including the lack of objective standardized measures, poor operational definitions, and anecdotal evidence. He pointed out the need for prospective studies testing consistent clinical hypotheses through rigorous methods. In contrast to the studies of psychosocial function, studies on cognitive function have used objective standardized measures, but they also have been limited by their failure to use a nonsurgical control group tested at comparable points in time (13–21). Ten years after Hermann wrote his critique of the outcome literature, a review by Strauss and Westerveld (22) again concluded that methodologically sound studies on psychosocial and cognitive function are lacking, and they pointed out the need for comparisons with unoperated-on patient control groups.

The choice of a control group for an evaluation of the effect of surgery is critical but is constrained by a number of practical factors. Although undoubtedly the best control group would consist of children who are determined to be good surgical candidates but who do not undergo surgery, it is very difficult to recruit such a group of patients. In our experience, very few parents and/or children refuse the surgical option once it has been decided that the child is a good candidate for surgery. Another option might be a wait-list control group, but this group would not be suitable for longitudinal studies. In the work described here, we have included, as a comparison group, children with medically intractable seizures who were evaluated in a tertiary-care epilepsy-monitoring unit, but who were deemed not to be surgical candidates.

The data reported here were collected in the baseline phase of an ongoing prospective longitudinal study of the outcome of pediatric epilepsy surgery. There are three main purposes of the initial phase of this study: (a) to describe the nature and extent of impairment across a range of cognitive variables in children with intractable epilepsy; (b) to determine whether there are similarities and differences between children who are deemed to be appropriate surgical candidates and those who are not. This objective will help establish, for future studies in other centers, whether children with nonfocal epilepsies can serve as appropriate comparison subjects for evaluating the short- and long-term effects of surgery. If differences are found, potential factors related to these differences (e.g., laterality of focus, site of focus) will be explored; and (c) to gather data with which to compare the effects of surgery and the effects of potential other medical variables (such as the effects of the eventual discontinuation of medication in those children who do obtain complete seizure relief after surgery).

We report on the first two of these objectives. Intelligence, memory, attention, and academic skills were examined in surgical candidates and a comparison group of children with intractable epilepsy. The influence of seizure variables and medication on cognitive performance was also evaluated.



Subjects were 51 patients recruited through the Epilepsy Monitoring Unit at the Hospital for Sick Children in Toronto. All patients were in the Unit for prolonged video-EEG recordings lasting from 3 to 5 days for the purpose of determining the localization and types of seizures. In addition to the video-EEG recordings, all subjects had clinical neurologic examinations and magnetic resonance imaging (MRI) scans. Fourteen children (five surgical and nine nonsurgical) had normal MRI scans. Other investigations included positron emission tomography (PET), single-photon emission computed tomography (SPECT), subdural grids or strip electrodes, depth electrodes, and magnetoencephalography (MEG)/magnetic source imaging (MSI), although not all children underwent each of those procedures. Based on the results of these investigations, a determination was made, by a team of medical experts on epilepsy and epilepsy surgery, whether the child was a surgical candidate (n = 30) or not (n = 21). Three patients in the surgical group had had prior epilepsy surgery; one patient in this group had previously had a biopsy [revealing a dysembryoplastic neuroepithelial tumor (DNET)].

Table 1 presents the demographic and seizure-related characteristics of the two groups. Information on laterality and site of focus was based on the concordance of information across multiple tests. Further confirmation within the surgical group was obtained by intraoperative and, in the extratemporal cases, extraoperative electrocorticography. Without such information, there may be greater uncertainty with respect to the accuracy of the localizing features of the seizures in the control group (23,24); however, in all patients, these variables were classified according to the most current results with state-of-the art investigations. The control group did not include any children who had epilepsy syndromes that would not be treated with surgery (e.g., Lennox–Gastaut syndrome) or epilepsy associated with neurodegenerative disorders (e.g., progressive myoclonic epilepsy).

Table 1.  Demographic and medical characteristics of the two groups
Surgical (n = 30)Nonsurgical (n = 21)
  • AED, antiepileptic drug.

  • a

     Includes cases with diffuse bilateral abnormalities, generalized seizures, and independent bilateral focal electrographic abnormalities.

  • b

     Days to weeks without seizures, then several over 1–2 days.

Age (yr)  
 Mean (SD)13.25 (2.99)13.02 (3.21)
Full-scale IQ  
 Mean (SD)85.07 (17.92)82.67 (20.34)
Age at onset of seizures (yr)  
 Mean SD)6.67 (3.71)5.38 (4.70)
Number of AEDs  
Laterality of focus  
Site of focus  
Seizure frequency  

A review of the seizure history indicated that seizure frequency varied considerably among the children, and that, within individuals, seizure frequency varied over time. In addition, some children had nocturnal seizures, or seizures at school that were not always accurately reported. For these reasons, it was not possible to obtain an accurate lifetime estimate of seizure frequency. A straight count of seizure frequency, even over a short period, may also mask important effects; for example, it may be that having one seizure per day each week is not the same as having 6 days free of seizures followed by seven seizures on the next day, even though the seizure frequency would be the same in both cases. Therefore, as has been done in other studies (25,26), we developed a categoric classification of seizure frequency. Patients were placed in one of the following four categories: seizures occurring daily (n = 21); seizures not occurring daily but at least one per week (n = 11); seizures not occurring every week, but at least once per month (n = 11); and seizures occurring in clusters (the patient experienced more than two seizures over a 24- to 48-h period followed by intervals of 2 weeks to 3 months with no seizures (n = 8).

All patients in both groups had tried and found unsuccessful at least two antiepileptic drugs (AEDs). One child had tried all conventional AEDs and combinations of AEDs with no effect on seizure frequency, and therefore was not taking any medications at the time of the study. Of the remaining 50 participants, 12 children were taking one medication, 26 were taking two medications, 11 were taking three medications, and one child was taking four medications. There were five different single medications, 15 different two-medication combinations, and seven three-medication combinations represented in this sample, making it impossible to examine for specific drug effects. Instead, we investigated the potential influence of monotherapy versus polytherapy, given that polytherapy has been indicated as a risk factor for cognitive side effects (27–29).


This study was approved by the Research Ethics Board (REB) of the Hospital for Sick Children. Informed consent was obtained from the parents, and informed assent or consent from each of the children, in accordance with the REB guidelines. Each subject was tested individually by an experienced psychometrician. Measures of intelligence, attention, memory, and academic skills were administered (30–37). Table 2 provides details of the domains, tests, and scores obtained from these measures. The assessment was conducted across two or more sessions. Breaks were permitted when necessary to avoid fatigue and to accommodate the limited attention spans of some of the patients. Occasionally a child did not complete all tests because of time limitations or noncooperation.

Table 2.  Cognitive domains and measures
Cognitive domainCognitive testDependent measuresMetric
  • a

     Standard scores have a mean of 100 and SD of 15 in the general population.

  • b

     Scaled scores have a mean of 10 and a SD of 3 in the general population.

IntelligenceWISC-III (30); WAIS-3 (31)Full-scale IQ (FSIQ), Verbal IQ (VIQ), Performance IQ (PIQ)Standard scoresa
Academic skillsWechsler Individual Achievement Test (32)Reading decoding, reading comprehension, arithmetic, spellingStandard scoresa
Verbal memoryChildren's Memory Scale (33); Denman Neuropsychology Memory Scale (34)Delayed recall of storyScaled scoresb
Visual memoryRey-Osterrieth Complex Figure (35, 36)Delayed recall of geometric designScaled scoresb
Sustained visual
Vigilance and distractibility tasks from Gordon Diagnostic System (37)Number correct and errors of commissionPercentile scores


Mean differences were explored with independent group t tests or analyses of variance, and categoric data were examined with χ2 tests. When necessary, analyses were adjusted for heterogeneity of variance. Pearson product–moment correlations were used to investigate relations between medical and cognitive variables.

Demographic and epilepsy variables

There were no differences between the groups in age at the time of the assessment (t49 = 0.26; p < 0.79), Full-Scale IQ (FSIQ; t49 = 0.45; p < 0.66), age at seizure onset (t49 = 1.10; p < 0.27), sex distribution [χ2(1) = 0.03; p < 0.86], number of medications [χ2(4) = 3.28; p < 0.51], or category of seizure frequency [χ2(3) = 1.06; p < 0.78]. As would be expected, there was a greater proportion in the nonsurgical group who did not have lateralized seizures [χ2(2) = 24.39; p < 0.001]. Although there was a higher proportion in the surgical group with a temporal-lobe focus, the distribution of seizure foci did not differ significantly between the two groups [χ2(4) = 7.52; p < 0.11]. There were no differences in task performance between those children with normal and those with abnormal MRI findings.

Cognitive variables

The means for both groups on all tasks, as well as the number within each group completing each task, are presented in Table 3. Two dependent variables were analyzed for each of the sustained-attention tasks: number correct and errors of commission; the analyses showed the same pattern of results for both measures, and thus only the results for number correct are presented in the Results tables. Six children in each of the surgical and nonsurgical groups were unable to comply with the demands of the distractibility task, and the test had to be discontinued for them. Two different measures of delayed story recall were used, but there was no difference in performance on the two tasks (p > 0.98); therefore the results across the two tasks were pooled for the analysis to examine group differences. The surgical group performed worse than the nonsurgical group at recalling the details of a story after a delay. No other group comparisons were significant.

Table 3.  Mean performance by both groups across tasks
MeasureSurgicalNonsurgicalt Value
Mean (SD)No.Mean (SD)No. 
Verbal IQ88.3 (17.9)3087.1 (18.6)210.23, NS
Performance IQ84.7 (17.5)3080.3 (21.7)210.80, NS
Reading decoding92.2 (17.2)2986.1 (18.5)201.20, NS
Reading comprehension90.6 (19.3)2691.2 (18.8)19−0.09, NS
Spelling91.5 (15.8)2986.1 (18.5)201.10, NS
Arithmetic83.3 (15.3)2979.8 (22.0)200.68, NS
Story recall6.9 (4.4)298.9 (2.4)21−2.03, p < 0.05
Design recall6.0 (2.9)286.7 (4.0)16−0.68, NS
Vigilance39.9 (41.4)2644.3 (36.3)19−0.38, NS
Distractibility32.9 (26.8)2438.5 (35.5)15−0.56, NS

This finding for story recall was explored further by examining the potential contributions of variables relating to the site and side of the seizure focus. These analyses were done first by pooling the surgical and nonsurgical patients, and second by examination within each of the surgical and nonsurgical groups. Analyses on subgroups formed by dividing the patients into those with left hemisphere, right hemisphere, and nonlateralized foci revealed no differences for either the combined or separate surgical and nonsurgical groups. When subgroups were compared on the basis of the site of dysfunction (temporal only, frontal only, other), no significant differences emerged for the combined patients or the separate surgical and nonsurgical groups.

Rates of impairment

Inspection of the results for the two groups indicated considerable variability in performance across all measures. For this reason, the data were examined further by establishing ranges in terms of clinical criteria for performance. Scores >−1 SD of the mean for the normal population were classified as “average or above”; scores between 1 and 2 SDs below the mean were classified as representing “mild–moderate impairments”; and scores >2 standard deviations below the mean were considered to represent “severe impairments.” The percentage of patients within each group in each of these performance ranges is presented in Table 4, and potential differences were explored by using χ2 analyses. The only variable on which the distribution was statistically different between the groups was that of story recall, in which more children who were surgical candidates showed severe impairments [χ2(2) = 8.21; p < 0.02].

Table 4.  Percentage of subjects in each group in each of the three performance ranges
MeasureRange of performanceχ2 Value
Average or aboveMild–moderate impairmentSevere impairment 
Verbal IQ63.357.123.328.613.314.30.22, NS
Performance IQ43.342.936.719., NS
Reading Decoding72.450.017.330.010.320.02.57, NS
Reading Comprehension61.557.923.131.615.410.50.52, NS
Spelling79.355.010.325.010.320.03.33, NS
Arithmetic44.830.037.945., NS
Story Recall37.971.424.123.837.94.88.72, p < 0.02
Design Recall35.750.042.925.021.425.01.48, NS
Vigilance53.868.430.815.815.415.81.39, NS
Distractibility72.785.718., NS

Seizure frequency

The next question that was addressed was whether seizure frequency had an impact on the cognitive performance of these children. The results of the surgical and nonsurgical groups were pooled, and an analysis of variance was conducted to compare each cognitive variable across the seizure-frequency categories. Because the Verbal and Performance IQ were highly correlated (r = 0.76; p < 0.001), the composite measure of intellectual functioning, FSIQ, was used in this and subsequent analyses. The results, shown in Table 5, indicate that seizure frequency had an impact on FSIQ, Reading Comprehension, and Arithmetic. Post hoc comparisons using Tukey's test indicated that FSIQ was higher in the patients who had clusters of seizures compared with those patients who had seizures on either a daily (p < 0.02) or a weekly basis (p < 0.006). Children with seizures in clusters also had an advantage over those who had seizures on a weekly basis in performance on Reading Comprehension (p < 0.04) and Arithmetic (p < 0.03).

Table 5.  Cognitive performance in relation to category of seizure frequency
MeasureSeizure frequencyF Value
Daily (n = 21) Mean (SD)1–10/wk (n = 11) Mean (SD)1–10/mo (n = 11) Mean (SD)Clustering (n = 8) Mean (SD)
Full-scale IQ80.3 (17.2)75.0 (20.7)86.2 (10.6)102 (17.3)4.56, p < 0.007
Reading decoding91.5 (16.3)82.8 (23.3)90.3 (16.6)94.8 (15.7)1.01, NS
Reading comprehension93.1 (18.3)78.9 (19.0)90.6 (11.9)102.9 (22.5)2.83, p < 0.05
Spelling91.4 (14.0)82.4 (20.0)87.8 (13.3)95.1 (23.3)1.41, NS
Arithmetic80.1 (16.5)74.7 (21.0)84.8 (15.7)94.5 (18.3)3.02, p < 0.04
Story recall7.5 (3.9)6.1 (3.1)8.9 (3.8)8.9 (4.2)0.83, NS
Design recall6.9 (2.9)5.1 (3.1)4.8 (3.1)8.1 (3.5)2.36, NS
Vigilance correct40.3 (36.4)22.1 (38.5)54.2 (40.8)47.4 (41,9)1.12, NS
Distractibility correct33.1 (33.5)15.0 (8.4)39.7 (27.9)37.3 (31.9)1.06, NS

Antiepileptic drugs

Analyses of covariance, using seizure frequency as a covariate, were also conducted to evaluate the potential impact of monotherapy versus polytherapy on cognitive performance. The one child taking no medication was eliminated from these analyses. No significant differences were found related to number of AEDs.

Age and time variables

Correlation coefficients were used to investigate the effects of three other seizure-related variables, age at seizure onset, duration (in years) of epilepsy, and the proportion of the child's life for which seizures were present [calculated as (duration/chronologic age)]. We chose to examine proportion of life with epilepsy in addition to duration, as the latter can be confounded by the child's age. For example, there is potentially a larger difference in the impact of a seizure disorder of 3 years' duration on a child who is age 4 years compared with one of age 12 years (the proportion in the former being 0.75 and in the latter being 0.25). Age at seizure onset was positively correlated with FSIQ (r = 0.302; p < 0.03), whereas proportion of life with seizures was negatively correlated (r = –0.301; p < 0.03) with FSIQ. Neither age at seizure onset nor proportion of life with seizures was correlated with any other cognitive variable. Duration of the seizure disorder did not correlate with any of the measures.


In this study we captured a high rate of cognitive impairment among children with intractable epilepsy. Although it is often assumed that the degree of cognitive impairment in epilepsy is a consequence of a long-standing history of seizures, these results demonstrate that even children with a relatively short duration can show considerable impairment in neuropsychological function.

It also was remarkable that there was a wide range of functioning within the groups. For example, IQ ranged from the Intellectually Deficient (<1st percentile) to the Very Superior (>99th percentile) level. Although these results demonstrate that epilepsy does not necessarily affect the child's cognitive function, the risk is very high. Depending on the measure used, 14 to 70% of the children had an impairment of a mild degree or greater, and the average rate of impairment across tasks for both groups was 45%. This rate is considerably higher than that expected in the general population, where ∼16% of children would be expected to have impairments of the degree as defined in our study. All of the children are in school, and the majority require modifications to accommodate their cognitive impairments and enhance their learning potential. These modifications differ depending on the school or school district, but encompass a number of approaches such as special class placement, one-to-one help within the classroom from an educational assistant, and curriculum modifications.

Comparability of the groups

These data were collected in the baseline phase of a longitudinal study of children undergoing epilepsy surgery. At this stage, our purpose was to document the points of similarities and differences in the cognitive function of children who are surgical candidates and those who are not. Because of statistical restraints imposed by our relatively small samples, we had restricted the number of measures to include intelligence and those areas that have consistently been reported as problematic in children with epilepsy: attention, memory, and academic function. The only measure on which a difference between groups was obtained was for the test of delayed story recall, an index of verbal memory. Analyses were conducted across all patients and within each of the surgical and nonsurgical groups to attempt to identify the variables that might impinge on this aspect of verbal memory; however, no differences relating to site (temporal, frontal, other), or to laterality (left hemisphere, right hemisphere, or nonlateralized foci) were found.

At first glance it might seem that the lack of relation with laterality and site may result from small sample sizes once our group is broken down into these dimensions, but the finding is consistent with other reports of memory in children who undergo surgery. Studies of pediatric surgical candidates have shown no preoperative differences related to laterality on measures of verbal memory or visual memory (13,15,18,19,21,38). Furthermore, Mabbott and Smith (38) found no differences in story recall between children with temporal lobe seizure foci and those with foci in extratemporal sites. Thus intractable seizures in childhood appear not to have specific effects on memory that vary reliably with the laterality and site of the seizure focus. Such specific effects have been demonstrated, however, in children with medically controlled seizures (39,40), suggesting that severity of the seizure disorder may play a role in the expression of the memory deficits.

AED findings

The literature on the effects of AED treatment on cognitive function is characterized by contradictory findings and methodologic confounds (41–43). Although it has been suggested that polypharmacy may be most detrimental to cognitive performance (27–29), that relation was not observed here. However, a wide variety of single AEDs, combinations of AEDs, and of doses were represented within the sample, and any potential differences may have been obscured by these variations and the resulting small number of children taking any particular drug or combination (44). Neurologic factors, such as the presence of the underlying brain pathology and the relatively high frequency of seizures, may outweigh any additional contribution of taking more than one AED (1). It also is possible that aspects of neuropsychological function not examined in the present study may be differentially affected by number of AEDs. For example, Dodrill (42) suggested that the effects of AEDs are most likely to be seen on tasks of psychomotor speed.

Timing variables

The negative impact of seizures on cognitive development was seen in the correlation between age at seizure onset and IQ. The duration of the seizure disorder also was found to be negatively correlated with IQ, but only when measured by the proportion of life with seizures rather than by the total number of years. The former measure may take better account of the impact of recurring seizures on development.

Seizure frequency

The adverse effects of epilepsy on cognition in adults and children has also been documented in the association between seizure frequency and performance on cognitive tasks (26,45–49), although this relation is not always found (e.g., 25,50). In this study, we demonstrated an effect not reported before, that children who experienced their seizures in clusters with seizure-free intervals between have an advantage in IQ and certain academic skills (reading comprehension and arithmetic). There may be a protective effect of a more prolonged “rest” from seizures that translates into higher functioning in this group. Although other studies have used a categoric approach to classifying seizure frequency, none has used that of clustering as used here, so this finding awaits further verification in additional and larger samples of children with epilepsy. The results also emphasize that children with reliably chronic and frequent seizures have the most deleterious outcome with respect to cognition.

Implications of the findings

The two groups studied here were comparable on a number of important dimensions: age, sex, age at seizure onset, seizure frequency, and number of medications. The important finding in this study is the high degree of similarity between the surgical and nonsurgical groups in their cognitive performance. The Pediatric Surgery Commission of the International League Against Epilepsy recently emphasized the lack of outcome studies on surgery and cognition in children (51). Our experience demonstrates the feasibility of recruiting a comparison group of children with intractable epilepsy for the purpose of conducting prospective outcome research on the effects of epilepsy surgery.

Acknowledgment: This research was assisted by the Ontario Mental Health Foundation. We thank Dr. Mary Desrocher and Shameela Hoosen-Shakeel for their assistance in testing the patients, and Nicholas Blanchette and Darren Kadis for help with the data analysis. We are grateful to Drs. William Logan, J. Rutka, O. C. Snead III, and S. Weiss for the opportunity to study their patients. Finally, we thank all the children, teenagers, and their parents who generously gave of their time to participate in this study.