The role of executive functioning in memory performance in pediatric focal epilepsy

Abstract

Objective

Learning and memory are essential for academic success and everyday functioning, but the pattern of memory skills and its relationship to executive functioning in children with focal epilepsy is not fully delineated. We address a gap in the literature by examining the relationship between memory and executive functioning in a pediatric focal epilepsy population.

Methods

Seventy children with focal epilepsy and 70 typically developing children matched on age, intellectual functioning, and gender underwent neuropsychological assessment, including measures of intelligence (WASI/DAS), as well as visual (CMS Dot Locations) and verbal episodic memory (WRAML Story Memory and CVLT-C). Executive functioning was measured directly (WISC-IV Digit Span Backward; CELF-IV Recalling Sentences) and by parent report (Behavior Rating Inventory of Executive Function (BRIEF)).

Results

Children with focal epilepsy had lower delayed free recall scores than controls across visual and verbal memory tasks (p = 0.02; partial η² = .12). In contrast, recognition memory performance was similar for patients and controls (p = 0.36; partial η² = .03). Children with focal epilepsy demonstrated difficulties in working memory (p = 0.02; partial η² = .08) and planning/organization (p = 0.02) compared to controls. Working memory predicted 9–19% of the variance in delayed free recall for verbal and visual memory; organization predicted 9–10% of the variance in verbal memory. Patients with both left and right focal epilepsy demonstrated more difficulty on verbal versus visual tasks (p = 0.002). Memory performance did not differ by location of seizure foci (temporal vs. extra-temporal, frontal vs. extra-frontal).

Significance

Children with focal epilepsy demonstrated memory ability within age-level expectations, but delayed free recall was inefficient compared to typically developing controls. Memory difficulties were not related to general cognitive impairment or seizure localization. Executive functioning accounted for significant variance in memory performance, suggesting that poor executive control negatively influences memory retrieval.

Children with epilepsy may exhibit deficits in learning and memory^1–3; however, the severity and pattern of impairment is not clear, with findings ranging from global memory disruption^4–6 to no deficits^7,8. A recent review observed that 78% of 88 studies showed children with epilepsy scored lower than controls on memory measures². By self-report, 70% of children and adolescents describe problems with learning and memory⁹, and both parents and patients indicate that cognitive effects of epilepsy are a primary concern¹⁰. Characterizing and understanding these impairments is important because learning and memory are essential skills for success in academic and everyday functioning¹¹.

Discrepant findings regarding memory profiles in pediatric epilepsy may be due to several issues in studies of memory functioning in this population. Most do not control for intellectual ability (IQ)², making it difficult to distinguish if memory difficulties are a specific impairment. One study that did control for IQ found that children with absence and focal epilepsy demonstrated worse performance on verbal memory compared to controls¹². Another issue is using epilepsy samples with heterogeneous patient characteristics (e.g., combining generalized and focal epilepsy).

Poor executive functioning (EF) skills are well documented in children with epilepsy^13–16. EF skills involve higher-level cognitive functions, including goal formation, planning, goal-directed behavior, and effective performance¹⁷. These skills include attention, working memory, and organization, and are essential to memory development¹⁸. Working memory in particular has been identified as one of the most frequent clinical elevations for children with epilepsy on parent ratings and executive dysfunction in epilepsy is related to quality of life¹⁹. Memory models purport the importance of attention and working memory in overseeing short term memory processing (e.g., phonological loop), organization of memory consolidation, and retrieval of information^20,21. Taken together, the executive components of memory may be essential for successful consolidation and retrieval and, when disrupted, may exert detrimental effects on memory performance.

Despite documented difficulties in children with epilepsy on both memory and executive functioning tasks, few studies have examined the relationship between these two domains. Research with other clinical populations, such as traumatic brain injury (TBI), has found a relationship between the two areas, with memory deficits in children with TBI related to working memory impairment²². In epilepsy, one study found that parent report of poor attention predicted self- and parent-reported everyday memory problems in children with intractable epilepsy¹¹. Another study in temporal lobe epilepsy (TLE) found that planning, abstraction, and mental tracking were correlated with memory performance via objective neuropsychological measures²³. After accounting for IQ, patients categorized as having more executive dysfunction performed worse on verbal and visual memory than patients with less executive dysfunction²³. Thus, initial studies support a role for executive functioning in poor memory performance in pediatric focal epilepsy.

Determining if these memory deficits are related to seizure localization or lateralization in pediatric epilepsy is also important. Material-specific memory deficits are found in adult TLE, with left TLE associated with lower verbal memory and right TLE with lower visual memory^24,25. Although a similar pattern has been found in pediatric TLE^6,26, the majority of studies do not show pre-surgical lateralizing memory impairments^24,27,28. Furthermore, regardless of side of focus, verbal memory may be most affected in pediatric TLE²⁹, and in pediatric epilepsy in general¹².

We address gaps in understanding the memory profile in pediatric focal epilepsy by examining if 1) memory deficits are present in this population, 2) executive functioning can explain those deficits, and 3) memory difficulties are related to seizure lateralization or localization. Our study examined visual and verbal memory in relation to executive functioning in children with focal epilepsy compared to typically developing children matched on age, gender, and IQ. By matching our samples on IQ at the group-level, we controlled for global cognitive effects. To study a more homogenous population, we limited our sample to patients with focal epilepsy who had generally normal MRIs. We hypothesized that memory difficulties in children with focal epilepsy would be related to coexisting executive functioning impairment and not related to lateralization or location of seizure focus.

Methods

Participants

One hundred and forty children between the ages of 5 and 17 (16 years, 11 months) participated in this study; 70 children with focal epilepsy (42 male; 62 right-handed; mean age =10.2) were matched based on age to 70 typically developing (TD) controls (38 male; all right-handed; mean age =10.3). Following matching on age, intellectual functioning and gender were used to optimize matching patients to all eligible controls, allowing the samples to be matched on these variables at the group level. Participants were recruited through multiple clinical research protocols that included common measures. Children were excluded if IQ < 70 to rule out global impairment. Exclusion criteria for the TD controls included known history of a medical disorder or central nervous system (CNS) injury, learning disabilities, Attention Deficit/Hyperactivity Disorder (ADHD) (Combined presentation, Predominantly inattentive presentation, and Predominantly hyperactive-impulsive presentation), or a significant ongoing medical condition. All subjects were recruited from Children’s National Health System’s (Children’s National) neurology clinics and the Washington, D.C. metropolitan community via flyers, pamphlets and lectures. The study was approved by, and performed according to, the policies of the Children’s National Institutional Review Board. Informed consent and child assent was obtained prior to the evaluation.

Patients had focal epilepsy as determined by seizure characteristics, EEG and/or video EEG; 47 had a left-hemisphere focus and 23 had right focus. MRI was globally normal for all patients; incidental findings (e.g., vascular variant, pineal cyst, Chiari I malformation) were considered normal and found in 17 patients. Eight patients had ADHD; two had a history of depression; one had a history of ADHD and anxiety; one had a history of ADHD, anxiety, and depression; and one had a history of anxiety, tics, and Obsessive Compulsive Disorder (OCD). Maternal education was measured via the Hollingshead scale and divided into three categories (12 years, 14/15 years, 16+ years). Detailed antiepileptic medication (AED) information, seizure focus, seizure onset/duration, and clinical characteristics are provided in Table 1.

Table 1.

Patient Seizure Characteristics

EPILEPSY VARIABLES	n	Mean (range)
Age of Seizure Onset (years)	70	6.00 (0.5–15.0)
Epilepsy Duration (years)	70	4.19 (0.13–13.9)
Seizure Lobe Focus	Left (n=47)	Right (n=23)
Frontal	10	4
Temporal	12	3
Fronto-temporal	7	3
Parietal	2	0
Occipital	1	0
Multifocal (other than fronto-temporal)	4	5
Undetermined	11	8
AEDs
Total Participants on AEDs*	65
Oxcarbazine	30
Carbamazpine	12
Lamotrigene	13
Diastat/Diazepam (intermittent)	4
Levetiracetam	11
Valproate	7
Lacosamide	2
Clonazepam	2
Felbamate	1
Lorazepam	1
Phenobarbital	1
Phenytoin	3
Zonisamide	3
Topiramate	6
Number of AEDs at time of study*
0	4 (6%)
1	40 (58%)
2	17 (24%)
3	6 (9%)
4	2 (3%)

Patient seizure characteristics. AED, Antiepileptic drug.

^*Information was missing for one patient.

Neuropsychological Testing

Each neuropsychological measure has a specified age range. All study participants (5–17 years) received an age-appropriate measure within the domain, as highlighted below.

Intellectual Functioning

The Wechsler Abbreviated Scale of Intelligence (WASI) was administered to assess intellectual ability³⁰. The WASI contains two verbal (Vocabulary and Similarities) and two nonverbal (Matrix Reasoning and Block Design) subtests. The WASI provided a Full-Scale IQ (FSIQ) score used to match participants. The lower limit of the WASI age range is 6; therefore, the 5-year-old study participants received the Differential Ability Scales (DAS) to assess intellectual ability, with the DAS General Conceptual Ability reported for the FSIQ³¹.

Memory

Verbal memory was evaluated using the Story Memory subtest of the Wide Range Assessment of Memory and Learning (WRAML)³² and the California Verbal Learning Test for Children (CVLT-C)³³. Story memory assesses memory for a large amount of contextual information. It consists of immediate and delayed (30 minutes) free recall trials of two stories and a recognition trial. We used only delayed free recall and recognition in analyses to be statistically conservative.

The CVLT-C consists of a 15-word list presented over five learning trials with recall after each trial, followed by a distracter list with recall, and then free and cued immediate delay of the original list. After a long delay (20 minutes), there is free recall [long delay free recall (LDFR)], cued recall, and yes/no recognition comprised of target and distracter items. Similar to Story Memory, we used only LDFR and recognition in analyses. The CVLT-C also includes measures of strategies used to learn and recall the words, including the section of the list remembered and types of errors made. We included those measures that relate to attention and executive functioning as described below.

Visual memory was assessed with the Dot Locations subtest of the Children’s Memory Scale (CMS)³⁴, which assesses memory for a visual-spatial pattern. CMS Dot Locations consists of a visual array of dots presented over three trials, followed by a new (distracter) visual array with recall. Immediately after the distracter and again after a long delay (30 minutes), there is free recall of the target array. Similar to the other tests, we focused on long delay free recall.

Executive Functioning

The Digit Span Forward (DSF) and Backward (DSB) subtest is part of the Wechsler Intelligence Scale for Children, Fourth Edition (WISC-IV) Working Memory Index³⁵. DSF is a measure of simple auditory attention, while DSB subtest requires mental manipulation, as it requires repetition of numbers in reverse order from what is presented. The WISC-IV age range is 6 years, 0 months to 16 years, 11 months; therefore, 5-year-olds did not receive DSF or DSB. The Recalling Sentences subtest from the Clinical Evaluation of Language Fundamentals, Fourth Edition (CELF-IV), also assesses working memory³⁶. This subtest requires repetition of verbatim sentences of increasing length and complexity.

The Behavior Rating Inventory of Executive Function (BRIEF)³⁷ was used to assess parent rating of everyday executive functioning skills. The BRIEF is an 86-item parent-rating questionnaire consisting of eight subscales: Inhibit, Shift, Emotional Control, Initiate, Working Memory (WM), Plan/Organize (P/O), Organization of Materials, and Monitor. The BRIEF also provides validity scales; no ratings in the unacceptable range were included.

Variables from the CVLT-C are also measures of executive functioning³⁸. The attention factor comprises three CVLT-C variables: list A trial 1, list B, percent total recall-middle; the organization/learning efficiency factor comprises three CVLT-C variables: semantic cluster ratio, percent recall consistency, and list A trial 5.

The majority of participants completed all assessment measures, except for the WRAML Story Memory subtest (Table 2 for sample sizes).

Table 2.

Memory Performance

	EPILEPSY GROUP Mean (SD) [% < −1.5+ SD of controls]	CONTROL GROUP Mean (SD)
SAMPLE-BASED Z-SCORES
Delayed Free Recall*	(n=41)	(n=41)
CVLT-C LDFR*	−0.60 (1.19) [17%]	−0.03 (1.15)
CMS-Dots Location Delay*	−0.59 (1.03) [22%]	0.02 (1.06)
WRAML-Story Memory Delay**	−0.54 (0.75) [15%]	0.00 (1.00)
Recognition Memory	(n=44)	(n=41)
CVLT-C Recognition Correct Hits	−0.16 (0.85)	−0.06 (1.13)
WRAML-Story Memory Recognition	−0.32 (0.94)	0.00 (1.00)
NORMATIVE Z-SCORES
MEMORY
Delayed Free Recall	(n=41)	(n=41)
CVLT-C LDFR	−0.37 (1.33)	0.24 (1.10)
CMS-Dots Location Delay	−0.09 (1.00)	0.44 (.95)
WRAML-Story Memory Delay	N/A	N/A
Recognition Memory	(n=44)	(n=41)
CVLT-C Recognition Correct Hits	−0.28 (1.28)	−0.01 (1.68)
WRAML-Story Memory Recognition	N/A	N/A

Table 2b. Executive Functioning Performance
	EPILEPSY GROUP Mean (SD)	CONTROL GROUP Mean (SD)
NORMATIVE Z-SCORES
Working Memory*	(n=53)	(n=67)
WISC-IV DSB^	−0.22 (0.87) [11%]	0.08 (0.86)
CELF-IV Recalling Sentences	−0.05 (1.04) [9%]	0.07 (0.95)
BRIEF Working Memory*	−0.57 (1.23) [13%]	0.01 (1.17)
Simple Auditory Attention	(n=57)	(n=66)
WISC-IV DSF	−0.18 (1.07)	0.11 (0.81)
Attention	(n=70)	(n=70)
CVLT-C Trial 1words Recalled	−0.15 (.96)	0.02 (1.04)
CVLT-C List B words Recalled	−0.21 (1.05)	−0.21 (1.00)
CVLT-C % Middle recall	−0.04 (1.27)	0.14 (0.90)
Organization/Learning Efficiency	(n=70)	(n=70)
CVLT-C Semantic clustering	−0.31 (1.16)	−0.04 (1.10)
CVLT-C % recall consistency	−0.13 (1.35)	0.01 (1.06)
CVLT-C Trial 5 words Recalled	−0.18 (1.28)	0.07 (1.02)
Planning/Organization	(n=67)	(n=70)
BRIEF Planning/Organization*	−0.51 (1.28) [15%]	0.02 (1.22)

Memory (Table 2a) and executive functioning (Table 2b) performance for pediatric focal epilepsy versus typically developing children (z-scores). Sample-based z-scores are listed for measures when those scores were used in statistical models; normative z-scores are also listed for all measures when standard/normative (non-categorical) scores are available. For measures with significant group differences, we have added the frequency of the epilepsy group that show impairment (1.5+ SD below the TD mean) to Table 2. CVLT-C, California Verbal Learning Test for Children; CMS, Children’s Memory Scale; WRAML, Wide Range Assessment of Memory and Learning; BRIEF, Behavior Rating Inventory of Executive Function; LDFR, long delay free recall; DSB, Digit Span Backward; DSF, Digit Span Forward.

^*p < 0.05;

^**p < 0.01;

^***p < 0.001;

^{^}p < .08 (trend)

Analysis

Statistical analyses were conducted using SPSS version 23.0 (IBM Corp., Armonk, NY). The WRAML Story Memory Delay and Recognition provide categorical normative scores; to maintain scores as continuous variables, we created sample-based z-scores using the mean and standard deviation of TD control group’s raw scores. For analyses using the Story Memory subtest, we created and used sample-based z-scores for all memory measures in the models, then included age as a covariate. For all other analyses, we converted continuous normative scores to z-scores; BRIEF scores were inverted to be consistent with other measures (where lower z-scores reflected poorer functioning).

Multivariate analyses of (co)variance [MAN(C)OVAs] were conducted to test for group differences across groups of measures, with follow-up univariate ANOVAs to test for each dependent measure. Variables were examined for normality; when non-normality was found, post-hoc testing used bootstrapping for parameter estimates (based on 1,000 bootstrap samples)³⁹. Homogeneity of variance was examined using Levene’s Test of Equality of Error Variances.

Descriptive variables included age, gender, maternal education, and FSIQ. Continuous variables age and FSIQ were analyzed with ANOVAS, as described above. We used Pearson chi-square or likelihood ratios (based on cell size) to examine group differences for categorical variables, such as gender and maternal education.

Variables were grouped into factors according to neuropsychological aspects of memory and executive functioning (Table 2):

For memory analyses, two MANCOVAs examined group (patients vs. controls) differences with the following factors: 1) delayed free recall (3 variables): CVLT-C LDFR, WRAML Story Memory delayed free recall, and CMS Dot Locations delayed free recall; 2) recognition (2 variables): CVLT-C recognition correct hits and WRAML Story Memory recognition.

For executive functioning analyses, MANOVAs/ANOVAs examined group differences with the following factors: 1) working memory (3 variables): DSB, CELF-IV Recalling Sentences, and BRIEF WM; 2) simple auditory attention (1 variable): DSF; 3) attention (3 variables): CVLT-C list A trial 1, list B, percent total recall-middle³³; 4) 2 models for organization: 4a) organization/learning efficiency (3 variables): CVLT-C semantic cluster ratio, percent recall consistency, and list A trial 5; 4b) organization/planning (1 variable): BRIEF P/O.

Contribution of Executive Functioning to Memory

To examine the relationship between memory and EF, linear regression models were used to predict delayed free recall (Table 2) from EF measures. To limit the number of analyses, we included only memory and EF variables that distinguished between patients and controls (if they passed tests for multicollinearity). We also corrected for multiple comparisons (six), providing a new alpha of p < 0.008. In models involving WRAML Story Memory delayed free recall, where sample-based z-scores were used, age was entered first in the model (step 1), followed by EF measures (step 2).

Lateralization and Location of Seizure Focus

Variables were also grouped into factors according to memory modality (verbal versus visual memory) to test for seizure lateralization effects on memory. We used repeated measures ANOVA in SPSS MIXED (marginal/population average model with unstructured covariance matrix for the residuals) to examine the effects of memory modality (visual versus verbal) and hemisphere of seizure focus (left versus right) on delayed free recall (CVLT-C LDFR; CMS Dot Locations long delay free recall). The variables in the mixed model ANOVA were restricted to CVLT-C and CMS Dot Locations as they have parallel formats with learning trials of discrete units of information, followed by delayed recall. Follow-up analyses were computed within the repeated measures model.

To determine seizure localization effects on memory, we compared patients with temporal (n = 8) and extra-temporal foci (n = 33) using ANCOVA (with age as a covariate) for delayed free recall for CVLT-C, Story Memory, and CMS Dot Locations. We also examined the effects of frontal lobe epilepsy by using ANCOVA to compare patients with frontal (n = 8) and extra-frontal seizure foci (n = 33) for the same memory variables.

Results

Descriptive Analysis

Age, gender, maternal education, and FSIQ were not different between patient and control groups (p’s > 0.54). FSIQ fell within the Average range for controls (mean = 104, SD = 13) and patients (mean = 102, SD = 15).

Memory

The MANCOVA for delayed free recall revealed that patients performed worse overall than controls (across verbal and visual memory; F (3,77) = 3.51, p = 0.02, partial η² = .12). Variance was equivalent between groups for all analyses. As expected, the covariate of age was significant (i.e., regardless of group, delayed free recall improved with age) (F (3,77) = 11.24, p < 0.001, partial η² = .31). Post-hoc testing using bootstrapping for parameter estimates revealed that patients performed worse than controls on all three delayed free recall measures: CVLT-C (95% CI [−0.91, −0.11], p = 0.02), CMS Dot Locations (95% CI [−0.95, −0.19], p = 0.01), and WRAML Story Memory (95% CI [−0.81, −0.18], p = 0.007; Table 2, Figure 1). Despite group differences, mean standardized performance for delayed free recall was in the average range for patients and controls (Table 2).

Figure 1. Delayed free recall vs. Recognition.

Graphs of Delayed Free Recall and Recognition for typically developing controls (blue) and patients with focal epilepsy (green). Z-scores are sample-based for these graphs; dotted lines signify a z-score of 0. CVLT-C, California Verbal Learning Test for Children; CMS, Children’s Memory Scale; WRAML, Wide Range Assessment of Memory and Learning. *Indicates significant difference between groups.

The MANCOVA for recognition found no overall differences between patients and controls (F (2,81) = 1.05, p = 0.36, partial η² = .03; Table 2, Figure 1). Variance was equivalent between groups for all analyses. Mean standardized performance for recognition was within the average range for both groups.

Thirteen of the patients with epilepsy had histories of comorbidities, such as anxiety, depression, and ADHD, while these were screened out of the control group. We re-analyzed our main memory findings without these 13 and the results were stable: the MANCOVA for delayed free recall revealed that patients performed worse overall than controls (across verbal and visual memory; F (3,73) = 3.64, p = 0.02, partial η² = .13); the MANCOVA for recognition found no overall differences between patients and controls (F (2,77) = 1.08, p = 0.34, partial η² = .03).

Executive Functioning

The MANOVA for working memory revealed that patients performed worse than controls overall (across direct measures and parent ratings; F (3,116) = 3.47, p = 0.02, partial η² = .08). Post-hoc testing revealed that specific patients had more problems on BRIEF WM (95% CI [−1.02, −.15], p = 0.01) and demonstrated a trend to score lower on DSB (95% CI [−0.63, −0.004], p = 0.07), but no differences were found on CELF-IV Recalling Sentences (95% CI [−0.49, 0.27], p = 0.55). The univariate ANOVA for organization/planning (BRIEF P/O) demonstrated that parent ratings were lower for patients than controls (95% CI [−0.98, −0.13], p = 0.02). However, the ANOVA for simple auditory attention (95% CI [−0.61, 0.07], p = 0.10) and the other two MANOVAs for organization/learning efficiency and attention found no group differences [organization/learning efficiency: F (3,136) = 1.06, p = 0.37, partial η² = .02; attention: (F (3,136) = 0.64, p = 0.59, partial η² = .01)]. Variance across EF measures was equivalent between groups.

Association between Memory and Executive Functioning

Based on the above results, two working memory variables (BRIEF WM, DSB) and one organization variable (BRIEF P/O) were included in regression models. We ran separate regression models for working memory and organization variables because BRIEF WM and BRIEF P/O were highly correlated (r = 0.83, p < 0.001).

Working memory

In a combined group (patients and controls), working memory variables (BRIEF WM, DSB) significantly predicted delayed free recall across all measures (Table 3). Working memory accounted for 9% of the variance in CVLT-C delayed free recall (F (2,135) = 6.85, p = 0.001) and 9% of the variance in CMS Dot Locations delayed free recall (F (2,135) = 6.35, p = 0.002). For WRAML Story Memory delayed free recall, age accounted for 13% of the variance (F (1,79) = 11.43, p = 0.001). In the next step, working memory accounted for an additional 19% of the variance (F (3,77) = 12.07, p < 0.001; Figure 2).

Table 3.

Regression Analyses Predicting Delayed Free Recall

Regression Models Predicting Delayed Free Recall from Working Memory Variables
	CVLT-C LDFR
Predictor	b	CI 95%	SE B	β	p	ΔF	ΔR2
Step 1						6.85**	0.09
DSB	0.16	−0.06, 0.43	0.12	.11	.15
BRIEF WM	0.27	0.10, 0.44	0.08	.27	.002**

	CMS Dot Locations Delayed Free Recall
Predictor	b	CI 95%	SE B	β	p	ΔF	ΔR2
Step 1						6.35**	0.09
DSB	0.32	0.15, 0.51	0.10	.29	.002**
BRIEF WM	0.02	−0.10, 0.14	0.06	.03	.71

	WRAML Story Memory Delayed Free Recall
Predictor	b	CI 95%	SE B	β	p	ΔF	ΔR2
Step 1						11.43**	0.13
Age	0.14	0.50, 0.23	0.04	.36	.002**
Step 2						10.95***	0.19
Age	0.16	0.06, 0.23	0.04	.39	.001***
DSB	0.26	0.10, 0.46	0.10	.25	.02*
BRIEF WM	0.25	0.08, 0.38	0.08	.32	.001**

Regression Models Predicting Delayed Free Recall from BRIEF P/O
	CVLT-C LDFR
Predictor	b	CI 95%	SE B	β	p	ΔF	ΔR2
Step 1						13.67***	0.09
BRIEF P/O	0.29	0.12, 0.46	0.08	.30	.002**

	CMS Dot Locations Delayed Free Recall
Predictor	b	CI 95%	SE B	β	p	ΔF	ΔR2
Step 1						0.62	0.05
BRIEF P/O	0.05	−0.09, 0.17	0.07	.07	0.46

	WRAML Story Memory Delayed Free Recall
Predictor	b	CI 95%	SE B	β	p	ΔF	ΔR2
Step 1						11.33**	0.13
Age	0.14	0.06, 0.22	0.04	.36	.002**
Step 2						9.36***	0.10
Age	0.13	0.04, 0.22	0.04	.33	.008**
BRIEF P/O	0.22	0.07, 0.35	0.07	.31	.006**

The first section is for results of the linear regression models used to predict delayed free recall (CVLT-C, CMS Dot Locations, WRAML Story Memory) from Working Memory (BRIEF Working Memory, Digit Span Backward). The second for results of the linear regression models used to predict delayed free recall from BRIEF P/O. In total, 6 models total were run. The models for CVLT-C and CMS Dot Locations used standard/normative z-scores, while WRAML Story Memory z-scores were sample-based. 95% Confidence Intervals (CI) are based on bootstrapping for parameter estimates (confidence intervals based on 1000 bootstrap samples). CVLT-C, California Verbal Learning Test for Children; CMS, Children’s Memory Scale; WRAML, Wide Range Assessment of Memory and Learning; BRIEF, Behavior Rating Inventory of Executive Function; LDFR, long delay free recall; DSB, Digit Span Backward; WM, Working Memory; P/O, Planning/Organization.

^*p < 0.05;

^**p < 0.01;

^***p < 0.001.

Figure 2. Material Specificity.

Graph of verbal and visual delayed free recall for left (blue) and right (green) focal epilepsy. The dotted line signifies a standard z-score of 0. CVLT-C, California Verbal Learning Test for Children; CMS, Children’s Memory Scale. *Indicates significant difference between groups.

Organization/planning

In a combined group, BRIEF P/O predicted delayed free recall for verbal measures (Table 3). BRIEF P/O accounted for 9% of the variance in CVLT-C delayed free recall (F (1,135) = 13.67, p < 0.001). For WRAML Story Memory delayed free recall, age accounted for 13% of the variance (F (1,78) = 11.33, p = 0.001), and an additional 10% was explained by BRIEF P/O (F (2,77) = 10.97, p < 0.001). BRIEF P/O did not predict CMS Dot Locations delayed free recall (F (1,135) = 0.62, p = 0.43).

Seizure Lateralization and Location

Although the repeated measures model revealed a main effect of memory modality (F (1,68) = 10.14, p = 0.002), this effect was driven by the two-way (hemisphere of focus x memory modality) interaction (F (1,68) = 5.22, p = 0.03; Figure 3). Follow-up comparisons revealed that memory modality was different within each group, such that the right focal group demonstrated a trend toward lower verbal delayed free recall scores than visual, and the left group did not (right: p = 0.001; left: p = 0.44). The focal groups (left versus right) were not different for visual memory, but the right focal group demonstrated a trend toward lower verbal delayed free recall scores than the left group (verbal: p = 0.05; visual: p = 0.58). There was no main effect of hemisphere of focus (F (1,68) = 1.40, p = 0.24). To investigate potential reasons for these seizure lateralization differences, we examined the differences in age of seizure onset and epilepsy duration via ANOVAs, and differences in the number of AEDs (0–1 versus 2+) between groups (left versus right) using Pearson chi-square. No significant group differences were found for epilepsy duration or age of seizure onset, but the right group demonstrated a trend toward an earlier age of seizure onset than the left (mean age of onset: right = 5.09, left = 6.43; 95% CI [−0.31, 2.89], p = 0.09). We found no relationship between number of AEDs and seizure lateralization.

Regarding the effect of seizure localization, we found no differences between patients with temporal and extra-temporal foci on delayed free recall (F (3,36) = 0.52, p = 0.67, partial η² = .04), or for patients with frontal compared to extra-frontal foci (F (3, 36) = 0.17, p = 0.92, partial η² = .01). Variance was equivalent between groups for the analyses. We were not able to examine side and lobe of focus in the same model due to small number of cases per cell.

Discussion

We found that children with focal epilepsy had lower scores on verbal and visual delayed free recall compared to TD children. Despite demonstrating difficulty with retrieval compared to controls, children with focal epilepsy performed comparably on recognition. These findings cannot be attributed to the general downward shift of intelligence found in pediatric epilepsy⁴⁰, as patients and controls had comparable IQs. Importantly, although delayed free recall performance differences existed between patients and controls, patients generally demonstrated memory skills within age-level expectations. This pattern of increased difficulty with retrieval has previously been found in epilepsy using the CVLT-C⁸ and may be related to the different EF demands required in each format. Free recall requires generation of a response, which relies on devising a strategy to execute that response; however, recognition requires a simple decision (yes/no or choosing from options), which reduces executive demands. The degree to which children with focal epilepsy differed from controls mirrors this hierarchy of executive support; free recall was significantly worse in patients, while recognition was not.

This pattern of memory performance may be related to reduced executive control in children with focal epilepsy. Patients performed worse than controls on several aspects of EF, including working memory (both via parent report and a trend for direct testing) and planning/organization (via parent report). Similar to previous research¹⁹, patients generally were not as good at keeping information in mind—working memory capacity was lower—which may have constrained the amount of material that could be recalled. This was supported by our finding that working memory ability positively predicted delayed free recall for verbal (list-learning and stories) and visual (visual array) information. Better organizational skills also positively predicted delayed free recall for verbal information. Therefore, executive functioning weaknesses likely contributed to difficulty with appropriately maintaining and organizing information during encoding and/or retrieval.

Notably, patients did not differ from controls across all aspects of EF. No group differences were found for simple auditory attention (WISC-IV DSF) or the CVLT-C factor of attention. Potential reasons for these findings are as follows, both DSF and the CVLT-C attention factor are based on the ability to remember items after a single exposure (e.g., CVLT-C Trial 1) and our recent work suggests that simple attention is not different in pediatric epilepsy compared to complex attention⁴¹. Similarly, no group differences were found for the CVLT-C factor of organization/learning efficiency. The discrepant organization findings (CVLT-C vs. BRIEF) may be related to the fairly rote approaches of serial and categorical organizational strategies in CVLT-C as compared to the BRIEF P/O which captures complex, future-oriented organization and planning. Thus, performance was not different on basic tasks of attention and organization in children with focal epilepsy and average IQ, but difficulties emerged as complexity increased.

Alternatively, this memory profile may be interpreted as a classic free retrieval deficit, with intact encoding and storage allowing for strong recognition ability. In the adult literature certain disorders, such as Alzheimer’s disease, show generalized encoding/storage deficits, which affect both retrieval and recognition, while patients with Huntington’s Disease demonstrate specific recall/retrieval deficits, with preserved recognition^42,43. This pattern is thought to be related to different neuroanatomical involvement in each disorder: predominantly mesial temporal (hippocampal) involvement in Alzheimer’s compared to frontal-subcortical (and caudate, in particular) in Huntington’s. This suggests a possible framework such that recall versus recognition differences relate to the neuroanatomic substrates that underlie those skills, and these brain regions may be differentially involved in pediatric focal epilepsy. In the current study, children without hippocampal sclerosis performed similarly to patients with Huntington’s Disease (retrieval differences with preserved recognition). This suggests that the free retrieval deficit may be related to frontal-subcortical dysfunction, rather than hippocampal impairment. Frontal-subcortical networks also underlie EF⁴⁴. As a result, the classic free retrieval deficit and EF explanation for these findings may be synergistic.

Functional neuroimaging is a potential tool for further investigating the neuroanatomical underpinnings of recognition/retrieval deficits and the relationship with EF. While we have discussed specific free recall/retrieval difficulties as frontal-subcortical, and generalized encoding/storage deficits as hippocampal, these are likely integrated networks. EF skills, similar to recognition skills, may involve functional connections from frontal regions to several areas, including the hippocampus⁴⁴. Thus, it is possible that regardless of location of seizure focus, disruption at various places in the network may result in similar functional impairment.

Consonant with this hypothesis, memory performance was not affected by specific seizure foci (temporal or frontal) in the current study. Patients with temporal compared to extratemporal foci did not differ in either memory modality (verbal or visual), and neither did those with frontal foci. This lack of localized pattern of memory weakness may be due to greater plasticity of the developing brain, compensating for disruption at various parts of the memory network. In our population this is particularly plausible, given that six was the mean age of seizure onset in our sample.

Memory difficulties depended on modality, with lower verbal delayed free recall than visual for children with focal epilepsy regardless of seizure lateralization. Similarly, several studies have found that verbal memory is most affected in children with epilepsy ^12,29. In the current study, although both groups had more difficulty with verbal memory, the right focal group alone performed significantly worse on verbal than visual memory. This finding was unexpected and limited by the small sample size (n = 23) and variability within the right focal group. Further, the right focal group demonstrated a trend toward an earlier age of seizure onset than the left, which may explain some of the differences in memory functioning. Lack of material specificity may be related to our sample of children with generally normal MRIs. In adults, when material specificity is observed, it is associated with hippocampal damage, while our patients did not have mesial temporal sclerosis (MTS).

A strength of the current study is the use of direct, objective measures of memory and the comparison of those scores with a control group, as well as with normative data. Although delayed free recall performance differences existed between patients and controls, patients generally demonstrated memory skills within age-level expectations. Consequently, comparison with normative data rather than a control group may have failed to detect these differences. For measures with significant group differences, we have added the frequency of the epilepsy group that show impairment (1.5+ SD below the TD mean) to Table 2. This may explain discrepant findings in the literature for memory in pediatric epilepsy. This is an important point for clinicians to consider when interpreting neuropsychological data in pediatric focal epilepsy. Important differences in the memory profile, particularly between free recall and recognition, may be missed when studies and clinicians are relying only on normative data instead of comparing to a control group (for studies) or looking at intra-individual differences (for clinicians).

Limitations

Despite studying a homogenous sample of focal epilepsy with normal MRI, there was still variability in lobe localization and some undetermined cases in our study population. Not all subjects had video EEG, which is considered the most rigorous tool for confirming localization. Future research will include more patients to be able to study both side and lobe of focus in the same model. We included a wide age range (5–17) in the current study and our primary analyses used measures that encapsulated this age range; however, the resultant population was heterogeneous regarding age, particularly given the developmental differences between the lower limit of the age range and the upper limit.

Using sample-based z-scores for WRAML Story Memory delayed recall and recognition was the best option because standardized scores for that measure are categorical. We felt that the benefit of including a measure of contextualized memory outweighed any limitation involved in not using standard z-scores. Further, it should be noted that although working memory and organization/planning significantly predicted delayed free recall, the EF variables accounted for a wide range of the variance (9–19%), with the most variance predicted by working memory for the story memory task. Thus, the link between executive functioning and memory is modest and should be investigated further in future studies. In addition, in the future more direct, objective measures of EF will be combined with parent ratings for a more thorough evaluation of EF.

Despite these limitations, our findings provide a strong framework for the investigation of memory performance in pediatric focal epilepsy. Future studies should include both free retrieval and recognition in order to capture the range of abilities in this population. Similarly, clinicians should consider including both free retrieval and recognition in the neuropsychological battery for children with epilepsy, as well as several executive functioning measures. Future studies should also employ a longitudinal design to examine the developmental trajectory of memory difficulties, which cannot be addressed with our cross-sectional design.

Our findings potentially have implications for how to intervene to improve memory in children with epilepsy. Our results indicate that these children likely would benefit from interventions employing strategies that lower executive control demands during learning; however, future studies should investigate the effectiveness of interventions targeting executive functioning in children with epilepsy, and specifically if they generalize by improving memory performance.

Conclusions

Children with left and right focal epilepsy demonstrated memory ability that is generally within age-level expectations; however, delayed free recall was inefficient compared to TD children, particularly for verbal information. In contrast, they performed comparably on recognition. Memory difficulties in pediatric focal epilepsy were not related to general cognitive impairment or seizure localization. We propose EF weaknesses exist in children with focal epilepsy, and may contribute to their reduced learning and memory performance. Generally, development of memory skills may be dependent on good executive control¹⁸.

KEY POINTS.

• Previous research shows that children with epilepsy may exhibit deficits in learning and memory; however, the severity and pattern of impairment is not clear.

• Our results suggest delayed free recall is inefficient in pediatric focal epilepsy, with preserved recognition memory.

• Executive functioning weaknesses may contribute to reduced learning and memory performance in pediatric focal epilepsy.

We confirm that we have read Epilepsia’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.