Cognitive performance in distinct groups of children undergoing epilepsy surgery—a single-centre experience

Patient selection

Paediatric patients (≤19 years of age) investigated for and having undergone epilepsy surgery in Motol Epilepsy Centre between January 1st, 2000 and December 31st, 2017 with available data on (i) results of pre- and post-surgical neuropsychological evaluation, (ii) seizure outcome one year after epilepsy surgery in patients included in analysis of post-surgical IQ/DQ and on (iii) pre- and post-surgical epilepsy- and epilepsy surgery-related variables were included in the study. Patients who underwent multiple epilepsy surgeries were excluded. The dataset was completed to our best knowledge; however, due to the nature and length of the study, some data might have been missed. Since the study is observational in nature and no experimental procedures were performed, the approval of Motol University Hospital ethics committee was not required. Informed consent with the pre-surgical evaluation and epilepsy surgery was obtained prior to all procedures from patients or their legal representatives.

Study design

We retrospectively analysed pre-, post-surgical IQ/DQ and the change between post- and pre-surgical IQ/DQ in relation to multiple epilepsy- and epilepsy surgery-related variables. In addition, the study period was divided in two—pre-2011 and post-2011—to compare trends in cognitive outcomes between the periods of developing vs. established epilepsy surgery program, similar to methodology published elsewhere (Lamberink et al., 2015). The cut-off date was set to December 31st, 2010. The year 2011 was chosen as the dividing line for several reasons: it was marked by the introduction of stereo-EEG as a method of long-term invasive EEG monitoring in our centre, and the number of epilepsy surgery procedures stabilized after the previously observed increase every year (Belohlavkova et al., 2019).

By comparing two periods of epilepsy surgery program, we aimed to analyse whether changes in patient population (e.g., different distribution of aetiologies, epileptic syndrome, age at surgery, etc.) and novel diagnostic and treatment strategies (e.g., use of stereo-EEG and advances in neuroimaging) that we report elsewhere (Belohlavkova et al., 2019) were reflected in cognitive performance of patients included in the study. All studied epilepsy-related variables are listed in Supplementary materials. Solely for the purpose of this study we used the term “epileptic syndrome” to denote either temporal, or extratemporal or hemispheric electro-clinical epileptic syndrome without any relationship to the underlying structural or genetic aetiology; hypothalamic hamartomas were excluded for the purpose of analysis of epileptic syndromes. “Abnormal neurological finding” stands for any focal abnormality in standard neurological examination. Surgery type represents the type of epilepsy surgery procedure: hemispherectomy/hemispherotomy, tailored resection, lesionectomy, extended lesionectomy, standardized resection (in our series exclusively anterior mesial temporal resection); surgery extent denotes the extent of resection: hemispheric, focal, lobar or multilobar. Surgery location may be either left-, right-sided or midline (hypothalamic hamartomas were preserved for this analysis). Complications related to epilepsy surgery were classified as major or minor as published elsewhere (Bjellvi et al., 2015). Types of focal cortical dysplasia (FCD) were classified according to ILAE consensus classification (Blumcke et al., 2011). FCD type Ia, Ib and Ic are characterized by abnormalities in radial, tangential and radial and tangential migration. FCD type IIa and IIb display cytological abnormalities, including dysmorphic neurons (FCD IIa and IIb) and balloon cells (FCD IIb). FCD type III occurs adjacent to another pathology: hippocampal sclerosis (FCD IIIa), glial or glioneural tumour (FCD IIIb), vascular malformation (FCD IIIc) or lesion acquired early in life (FCD IIId) (Blumcke et al., 2011). In one type of analysis, histopathological lesions were divided in two groups: developmental vs. acquired. MCD, TSC and long-term epilepsy-associated tumours (LEAT) were considered developmental, and inflammatory, post-traumatic and glial scar lesions were considered acquired. Specifically for the purpose of this analysis, hippocampal sclerosis was excluded, as the proportion of its developmental or acquired origin has not yet been unequivocally established.

Neuropsychological examination was performed before the (first) surgery and at one year follow-up after the resection using (according to patient’s age and cognitive level) Wechsler Adult Intelligence Scale in 3rd revision (Wechsler, 2010), Wechsler Intelligence Scale for Children in 3rd revision (Wechsler, 2002) for evaluating IQ and Bayley Scales of Infant Development in 2nd revision (Bayley, 1993) for developmental quotient assessment. Stanford-Binet Intelligence Scale in 4th revision (Thorndike, Hagen & Sattler, 1986) was selected for IQ testing in children with lower cognitive skills (expected IQ < 60), despite its limited reproducibility and reliability of IQ testing in this group of patients.

Statistical procedures

Pre-surgical and post-surgical IQ/DQ and IQ/DQ change scores were selected as dependent variables. We first performed regression analysis to describe the relationship between pre- and post-surgical IQ/DQ in the entire cohort and separately for children above and below 6 years of age and for children above and below 12 years of age. Then, we did univariate regression analysis for continuous observed (independent) variables and selected those with p-value below 0.05 for multivariate analysis. We performed one-way ANOVA for categorical variables and calculated median difference and its 95% confidence interval using Hodges-Lehman estimator. For calculation of effect sizes, omega-squared effect size was used for ANOVA tests and r-squared effect size for correlation tests and multivariant models.

In the multivariate analysis, we calculated a multiple regression based on the general linear model by stepwise regression algorithm with the variables that reached statistical significance in univariate testing. The beta coefficients with p-values of < 0.05 were considered statistically significant. For the calculations software MatLab version 2017b and its statistical computing toolbox was used.

Trends in cognitive performance in the periods of pre-2011 vs. post-2011

Using regression analysis, we have shown there exists strong and significant correlation between pre- and post-surgical IQ/DQ, and that patients undergoing epilepsy surgery tend to have higher post-surgical IQ/DQ than pre-surgical IQ/DQ (Rho = 0.888, p < 0.001). The gain in IQ/DQ is more prominent in those with lower pre-surgical IQ/DQ and the group of children older than 6 years of age (Fig. 1). We found no significant difference in pre-, post-surgical IQ/DQ and IQ/DQ change between the periods of pre-2011 and post-2011 (p = 0.62, p = 0.65, p = 0.77, respectively). When analysed in groups according to the epileptic syndrome, differences in cognitive performance (pre-, post-surgical IQ/DQ and IQ/DQ change) between the two periods (pre-2011 vs. post-2011) were not significant for groups of patients with temporal lobe epilepsy (TLE), extratemporal (XTLE) and with hemispheric (HEMI) epilepsy. The same is true in relation to underlying histopathology and developmental vs. acquired lesions (see Supplementary materials). In the analysis of differences between the two periods, we classified FCD in three classes (FCD type I, FCD type II, FCD type III) as the small numbers in respective subcategories (FCD type Ia-Ic, IIa and IIb and IIIa–IIId) would preclude statistical calculation. We found a significantly lower pre- and post-surgical IQ/DQ in the period post-2011 (F = 8.28, p = 0.013 and F = 4.76, p = 0.05, respectively) in the group of FCD type I. We did not identify significant difference in post-surgical IQ/DQ and IQ/DQ change between the two periods in either seizure-free or non-seizure-free patient cohort (seizure free: F = 0.48, p = 0.49 and F = 0.24, p = 0.62, respectively, non-seizure-free: F = 0.8, p = 0.38 and F = 0.26, p = 0.62). For statistical details, including measures of effect size and degrees of freedom, see Supplementary materials.

Pre- and post- surgical cognitive performance in distinct groups of patients according to their epilepsy-related features

Using ANOVA statistics we observed that patients with TLE differ from those with XTLE and those with HEMI epilepsy in their pre-surgical IQ/DQ (F = 6.83, p = 0.001) and in their IQ/DQ change (F = 6.63, p = 0.002); post-hoc tests showed that patients with HEMI score significantly lower in their IQ/DQ compared to those with XTLE and TLE (Fig. 2). Depending on their underlying histopathology, the analyses have identified group differences in pre- and post-surgical IQ/DQ (F = 6.56, p < 0.001 and F = 7.44, p < 0.001, respectively). Post-hoc analyses found significantly higher pre-surgical IQ/DQ in patients with tumour (LEAT) compared to those with MCD (p = 0.002), TSC (p < 0, 001) and glial scar (p = 0.005). Significant differences in post-surgical IQ/DQ were also observed between the aetiological sub-groups; patients with LEAT achieve significantly higher IQ/DQ scores than those with e.g., MCD, TSC, and those with hippocampal sclerosis score higher than patients with TSC. When comparing developmental vs. acquired histopathological lesions we found no significant difference in their pre-, post-surgical IQ/DQ and IQ/DQ change (data not shown). Furthermore, we analysed the differences in cognitive performance relative to FCD classes (Blumcke et al., 2011), and found significant difference between the classes in the pre- and post-surgical IQ/DQ (F = 4.89, p < 0.001 and F = 4.73, p < 0.001, respectively); the results remained significant for post-surgical IQ/DQ for the classes of FCD type I, II and III (F = 6.32, p = 0.003).We observed a trend towards higher pre-surgical IQ/DQ and IQ/DQ change in FCD type III vs. FCD type II and type I (Fig. 1).

Predictors of pre- and post-surgical IQ/DQ and its change over the entire period

The results of univariate testing are listed in Table 2 for categorical variables and in Table 3 for continuous variables associated with pre-surgical IQ/DQ, post-surgical IQ/DQ and IQ/DQ change; only the variables significant in univariate testing are shown here, all studied variables are listed in Supplementary materials. The used stepwise regression algorithm excludes variables that might skew results in the calculation process of the multiple regression based on the general linear model; therefore, some variables, despite being significant in univariate testing, do not appear in the results of multivariate testing, listed in Table 4. Some variables associated with severe epilepsy, e.g., the presence of infantile spasms affecting pre-surgical IQ/DQ remained significant in univariate testing only. The factors found significant in multiple regression model were (i) the presence of abnormal neurological finding and FCD type IIb and IIIb and younger age at first seizure affecting pre-surgical IQ/DQ; (ii) age at first seizure, status epilepticus, multilobar resection, unilobar resection, complete resection and TSC affecting post-surgical IQ/DQ; (iii) age at surgery and temporal or extratemporal epileptic syndrome affecting IQ/DQ change (see Table 4). The date of surgery, reflecting change in epilepsy surgery practice, was not rendered significant in the multivariate regression analysis model and was therefore excluded from the model by the used stepwise algorithm.

For the summary of statistically significant results of all tests, including post-hoc analyses and the respective p values, F values, effect sizes and degrees of freedom, see Tables 1–5.

Cognitive performance in children undergoing epilepsy surgery did not change between the periods of pre-2011 and post-2011

Comparing cognitive performance of all epilepsy surgery patients from the period of pre-2011 vs. post-2011 their pre-, post-surgical IQ/DQ and IQ/DQ change scores did not differ significantly. These findings might imply that the spectrum of epilepsy surgery patients has not changed significantly with respect to their cognitive performance. We further analysed whether distinct groups of patients in terms of epileptic syndrome and underlying aetiology differed in their cognitive performance between the two studied periods and again found no significant difference. These results reflect in part the development of paediatric epilepsy surgery practice in the region and in Europe (Baud et al., 2018); we are seeing increased referral of more complex patients and a changing population of paediatric epilepsy surgery candidates. While in the early period we encountered either very severe cases (epilepsy surgery as “the last resort”) or very straightforward ones (e.g., hippocampal sclerosis) with a very broad span of cognitive performance, in the later period we observed a decline in the straightforward cases (such as hippocampal sclerosis) and an increase in timely referral of complex cases (such as TSC, see Supplementary materials). The shift towards more complex patients, including extratemporal and MRI negative cases, has been observed in centres around Europe (Baud et al., 2018), and this might have been reflected in the cognitive performance of epilepsy surgery patients. Naturally, we cannot exclude the effect of small sample size in certain aetiological groups with less frequent pathologies. To summarize, the aetiological spectrum of epilepsy surgery candidates has changed significantly in terms of the underlying aetiology (see Supplementary materials and Belohlavkova et al., 2019) but the overall IQ/DQ scores remained similar.

Cognitive performance of children undergoing epilepsy surgery varies based on aetiology and epileptic syndrome

Given the broad spectrum of epileptic syndromes and underlying aetiologies in the studied population we hypothesised that these patients enter pre-surgical evaluation as distinct groups with distinct cognitive performances. According to our analyses, patients with hemispheric syndromes perform significantly lower in their IQ/DQ scores than their counterparts with temporal or extratemporal epilepsy; however, they also have the greatest potential for post-surgical increase in IQ/DQ scores, compared to patients with TLE or XTLE. These results are in line with the results from previous studies showing that patients with hemimegalencephaly profit from early epilepsy surgery (Bulteau, Otsuki & Delalande, 2013; Honda et al., 2013). Considering the aetiological spectrum of epilepsy surgery patients we hypothesised that patients with developmental structural lesions, such as TSC, MCD or long-term epilepsy-associated tumours differ from those with acquired lesions (e.g., inflammation, glial scar, etc.), but our data showed the opposite. It is rather the specific aetiology itself and the extent of involvement (temporal vs. extratemporal vs. hemispheric epilepsy) that distinguish these patients from each other. Patients with benign tumours achieve significantly better pre- and post-surgical IQ/DQ scores than those with other developmental abnormalities, e.g., TSC or MCD. Given their tendency to reach favourable seizure outcome, in addition to good cognitive outcome, they represent a group that might benefit from early referral to epilepsy surgery (Ramantani et al., 2014). Patients with TSC, on the other hand, tend to suffer from developmental delay spanning from mild cognitive impairment to profound intellectual disability in 14% to 31% respectively (Joinson et al., 2003). Despite on-going discussions about the pathogenesis of cognitive impairment in TSC (Curatolo et al., 2016; Chu-Shore et al., 2010), most authors agree that adequate seizure control, achieved by antiepileptic medication or epilepsy surgery, in TSC patients leads to improved cognitive outcomes (Arya et al., 2015; Chu-Shore et al., 2010).

In histopathological FCD subgroups, we observed a tendency towards better cognitive performance in FCD type III, with the most significant difference in relation to FCD type I. Lower IQ/DQ scores and worse outcomes of epilepsy surgery have been repeatedly observed in patients with FCD type I in contrast to FCD type II (Krsek et al., 2008; Krsek et al., 2009). FCD type III has only recently been distinguished as a separate category of FCD (Blumcke et al., 2011); therefore, it is difficult to compare the results of more and less recent studies. Nevertheless, FCD type III from its definition associates with other pathologies, e.g., hippocampal sclerosis (FCD IIIa) or brain tumour (FDC type IIIb), and patients with these pathologies achieved higher IQ/DQ scores which also explains the observed results. We observed lower pre- and post-surgical IQ/DQ in the period post-2011 in the group of FCD type I; this might be a result of the rather small sample size (n = 15). Another plausible explanation is that we tend to accept patients with more severe epilepsy and its related adverse cognitive sequelae in the epilepsy surgery program. It remains to be discovered whether and to what extent molecular genetic changes in patients with FCD influence their prognosis in terms of both seizure and cognitive outcome (Baldassari et al., 2018; Benova & Jacques, 2018).

Multiple factors influence cognitive performance in children undergoing epilepsy surgery

In the entire patient cohort we have identified multiple factors that might affect their cognitive performance. Factors associated with severe epilepsy, including early age at onset, frequent seizures, occurrence of infantile spasms and status epilepticus seemed to be associated with lower pre-surgical IQ/DQ scores. Given that patients suffering from infantile spasms tend to improve in their cognitive skills after successful epilepsy surgery (Asarnow et al., 1997) low IQ/DQ scores should not preclude timely referral for epilepsy surgery. In fact, patients with low pre-surgical IQ/DQ scores do benefit from epilepsy surgery even though their chance to achieve seizure freedom decreases with lower IQ/DQ scores (Malmgren et al., 2008). Overall, our data support the original concept of epileptic encephalopathy wherein epileptic activity itself contributes to cognitive decline (Berg et al., 2010). Therefore, in light of our findings supported by multiple studies advocating for epilepsy surgery as a treatment of choice in drug resistant focal epilepsy, we conclude that epilepsy surgery leads to favourable seizure and cognitive outcomes in carefully selected patients (Dwivedi et al., 2017; Harvey et al., 2008; Holthausen, Pieper & Kudernatsch, 2013; Ryvlin, Cross & Rheims, 2014; Van Schooneveld & Braun, 2013).

To summarize, patients with varying underlying aetiologies tend to differ in their pre- and post-surgical IQ/DQ scores but not in their IQ/DQ change, we therefore suggest that despite the differences in pre-surgical cognitive performance, patients with various underlying aetiologies may all benefit from epilepsy surgery. These findings are also supported by the regression curve equation that shows that patients with lower pre-surgical IQ/DQ tend to profit most from epilepsy surgery. Understandably, they would still achieve significantly lower post-surgical IQ/DQ scores as post-surgical IQ/DQ scores correlate strongly with pre-surgical IQ/DQ.

Admittedly, our study displays certain limitations, the greatest being the short follow-up period of one year. Indeed, some authors show that only longitudinal studies over lengthy study periods provide solid evidence of cognitive improvement after epilepsy surgery, associated with cessation of antiepileptic medication (Sibilia et al., 2017; Skirrow et al., 2011). On the other hand, the length of follow-up period in previously published studies spans from few months to many years (Moosa & Wyllie, 2017; Van Schooneveld & Braun, 2013). Most recently published studies do provide evidence of cognitive improvement or improved cognitive trajectory over longer time periods; however, these studies have other limitations, e.g., smaller sample size or they fail to discern between different aetiologies (Puka, Tavares & Smith, 2017; Sibilia et al., 2017). Most importantly however, in our study, we aimed primarily to compare trends in cognitive performance between the periods of developing vs. established epilepsy surgery centre and to analyse whether epilepsy surgery candidates differ in their cognitive performance based on underlying aetiology and epileptic syndrome. Further longitudinal studies are needed to provide unequivocal evidence that paediatric epilepsy surgery can reverse the course of epileptic encephalopathy and lead to cognitive recovery or at least prevent further cognitive decline.

Due to repeated neuropsychological testing, we could not have completely eliminated a possible effect of repeated testing. Although some authors claim that the practice effect in paediatric population is not significant (Westerveld et al., 2000), others believe it may interfere with the correct evaluation of cognitive outcomes in paediatric epilepsy surgery (Sherman et al., 2003). In addition, given the age diversity of our population, our study faced the issue of comparability of various age-adjusted versions of IQ/DQ tests. We believe both of these issued will be addressed in future studies in which we aim to validate our findings over a longer follow-up period with wider intervals between repeated testing on a more homogenous age-groups of children.

Despite rather large sample size, some numbers of patients in individual subgroups remain small and this precluded some further analyses, e.g., the effect of complications of epilepsy surgery on post-surgical IQ/DQ score. However this also shows that overall, major complications of epilepsy surgery occur rarely in our centre as well as others (Bjellvi et al., 2015). Our study also lacked data on socioeconomic status of patients’ families that might have affected IQ/DQ change; in fact, parents’ education seems to independently contribute to IQ/DQ increase after epilepsy surgery (Meekes et al., 2015), while income and residence do not have significant effect (Puka et al., 2016). Therefore, the effect of families’ socioeconomic status on cognitive performance in children undergoing epilepsy surgery warrants further consideration, both in research and in clinical practice.

Another significant limitation concerns the effect of antiepileptic drugs (AED) on post-surgical cognitive outcome. It has been shown that AED withdrawal leads to improved cognitive skills in children undergoing epilepsy surgery, and that IQ scores tend to increase with decreasing number of AED (Boshuisen et al., 2015b). Given the short period of post-operative follow-up period we were unable to assess the effect of AED withdrawal as the vast majority of patients remain on pre-surgical dose of AED at least one year after epilepsy surgery. Accumulating evidence advocates for early AED withdrawal (Boshuisen et al., 2012; Braun & Schmidt, 2014), and a randomized control trial on the effect of early vs. late AED withdrawal on cognitive outcomes is ongoing (Boshuisen et al., 2015a).