Epilepsy and Sleep Characteristics Are Associated with Diminished 24-Hour Memory Retention in Older Adults with Epilepsy

Abstract

Objective:

Individuals with epilepsy often have memory difficulties, and older adults with epilepsy are especially vulnerable, due to the additive effect of aging. The goal of this study was to assess factors that are associated with 24-hour memory retention in older adults with epilepsy.

Methods:

55 adults with epilepsy, all over age 50, performed a declarative memory task involving the recall of the positions of 15 card pairs on a computer screen prior to a 24-hour ambulatory EEG. We assessed the percentage of encoded card pairs that were correctly recalled after 24 hours (24-hour retention rate). EEGs were evaluated for the presence and frequency of scalp interictal epileptiform activity (IEA) and scored for total sleep. Global slow wave activity (SWA) power during non-REM sleep was also calculated.

Results:

44 participants successfully completed the memory task. Two were subsequently excluded due to seizures on EEG. The final cohort (n=42) had a mean age of 64.3 ± 7.5 years, was 52% female, and had an average 24-hour retention rate of 70.9% ± 30.2%. Predictors of 24-hour retention based on multivariate regression analysis when controlling for age, sex, and education included: number of antiseizure medications (β =−0.20, p=0.013), IEA frequency (β =−0.08, p=0.0094), and SWA power (β =+0.002, p=0.02).

Significance:

In older adults with epilepsy, greater frequency of IEA, reduced SWA power, and higher burden of antiseizure medications correlated with worse 24-hour memory retention. These factors represent potential treatment targets to improve memory in older adults with epilepsy.

Introduction

Memory complaints are common in epilepsy patients and have been associated with a poorer quality of life.¹ Memories once encoded need to be consolidated, and sleep plays a central role in this process, especially slow wave activity (SWA) in sleep.^2,3 With aging, SWA tends to decrease, and this was found to correlate with worse long term memory retention.⁴

Accelerated long-term forgetting (ALF) is a phenomenon that has been described in epilepsy and other patient populations, whereby newly acquired information, despite apparently normal encoding and retrieval after short periods of time, decays at a faster than normal rate over days to weeks⁵, suggesting impaired memory consolidation. Higher rates of ALF have also been described in older adults without neurological disease.⁶ Older adults with epilepsy are thus an especially vulnerable population because of the adverse impact of both aging and epilepsy on memory consolidation. With regards to epilepsy-related factors, seizures, interictal epileptiform abnormalities (IEAs) and medications have been associated with worse cognitive performance.^7,8 This presents a significant clinical problem, as cognitive difficulties naturally increase the need for care. Cognitive assessments in epilepsy often rely on short intervals (typically 5–10 minutes, and almost always less than 1 hour) to evaluate memory and can miss patients whose memory impairments are related to impaired consolidation and may only manifest days later.⁹

In this study, we evaluated predictors of 24-hour memory retention in older patients with epilepsy, including epilepsy-related factors, EEG features, and sleep. We hypothesized that greater interictal epileptiform abnormalities and reduced SWA power in sleep would correlate with worse memory retention.

Methods

Study Population

All English-speaking patients aged 50 years and older, with a history of seizures who were referred for 24-hour ambulatory EEG at Brigham and Women’s Hospital (BWH) or South Shore Hospital (SSH) between 2016–2022 were approached to enroll in the study. Participants with known generalized epilepsy were excluded from the study.

An additional two participants referred for ambulatory EEG due to fluctuating cognition and an epileptiform routine EEG were included. Supplementary figure S1 shows the recruitment flow chart. The study was approved by the Institutional Review boards at BWH and South Shore Hospital SSH. Written informed consent was obtained from all participants before any study procedures were carried out.

Data extracted from the electronic medical record included demographics, epilepsy duration, epilepsy lateralization, presence and nature of MRI lesion, and medications. Medications were further divided into those associated with known negative cognitive effects (benzodiazepines, topiramate, phenytoin)^10,11 and those that were not. Epilepsy lateralization was determined based on prior and current EEG findings, and location of cortical MRI lesion; participants were then classified as: unknown, right, left, or bilateral. A mini mental status exam (MMSE) or Montreal cognitive assessment (MOCA) examination performed within 1 year of the study was also included. MOCA scores were then converted to MMSE scores¹² to allow for a correlation with retention scores.

Memory Task and questionnaires

The memory task was a 2-D object-location memory task that has previously been shown to exhibit sleep dependent consolidation in adults^13,14 and was administered on the day of the EEG setup. It consisted of 15 pairs of colored pictures showing different animals and everyday objects displayed in a checkerboard-like 5×6 matrix (supplementary figure S2). Each pair contained identical pictures of one item. All 30 possible spatial locations were shown as black squares (‘the “back sides” of the cards) on a 15-inch laptop screen. At learning, the first card of each card-pair was presented alone for one second followed by the presentation of both cards for three seconds. After an interstimulus interval of three seconds, the next card-pair was presented in the same way. The entire set of card pairs was presented twice in different orders. Immediately at the end of these two exposures, recall of the spatial locations was tested using a cued recall procedure, i.e., the first card of each pair was presented, and the participant was asked to indicate the location of the second card with a computer mouse. Visual feedback was given in each case by presenting the second card at the correct location for two seconds, independent of whether the response was correct or not, allowing re-encoding of the correct location of the matching card. After presenting a card-pair, both cards were replaced by black squares again, so that the probability of being correct if guessing remained the same throughout each run. The cued recall procedure was repeated until the participant reached a criterion of 33% (5) correct responses, compared to chance rate of 3%. If a participant failed to reach this criterion after 5 repetitions of the cued recall procedure, their participation in the study was terminated. Once participants achieved this criterion, they were tested one last time without feedback, and the number of card pairs recalled was considered their baseline. When the participants returned for EEG removal, they were tested one last time. 24-hour retention rate was calculated as (number of correct card pairs recalled after 24 hours) / (number of correct card pairs encoded at baseline).

Participants were asked to rate their sleepiness using the Stanford Sleepiness Scale (SSS)¹⁵ on the day of EEG setup and upon return, as well as to report their level of memory concern over the past 4 weeks (1-a great deal; 2-somewhat; 3-only a little; 4-not at all) based on the memory question from the Quality of Life in Epilepsy-10 questionnaire.¹⁶

EEG recordings and sleep scoring

Scalp EEG electrodes were placed using the International 10–20 system with additional anterior temporal electrodes (T1, T2). Twenty-four–hour ambulatory EEG recordings were acquired with XLTEK TREX hardware (Natus Medical Inc, Pleasanton, CA), sampling at 200 Hz or with Arc Apollo EEG hardware (Cadwell, Kennewick, WA), sampling at 250Hz.

EEGs were visually reviewed separately by 2 of 3 board certified epileptologists (MP, RAS, SP) who manually marked any interictal epileptiform abnormalities (IEAs), including spike waves, sharp waves, and temporal intermittent rhythmic delta activity. A consensus by the 2 reviewers was reached for each abnormality in case of disagreement on initial review.

Sleep staging of the whole recording was performed by an expert sleep technician, blinded to diagnosis, in non-overlapping 30 second epochs according to standards by the American Academy of Sleep Medicine (AASM), with five stages: wake (W), non-rapid eye movement (NREM) stage 1 (N1), NREM stage 2 (N2), NREM stage 3 (N3), and rapid eye movement (REM).¹⁷

We combined the IEA annotations with the sleep staging to extract the frequency of IEA (number of IEAs per hour) during the study (both wakefulness and sleep) and during Non-REM sleep.

Slow wave activity (SWA) measurements

EEGs were first low-pass filtered at 30Hz (Butterworth 4th order) and high-pass filtered at 0.1Hz (Butterworth 4th order). Multi-taper spectrograms were calculated using non-overlapping 6-second windows, with 3 tapers and a time-bandwidth product of 2, resulting in a spectral resolution of ± 0.33Hz. Artifactual windows for each EEG channel were identified using thresholds for excessively high amplitudes, EMG artifact, and flat channels, and were removed prior to analysis. On average, across participants, the artifactual epochs removed for any given channel was <5%.

SWA was measured during N2 and N3 sleep epochs. We calculated SWA on each electrode as the absolute power in the 0.6–1 Hz, and 0.6–4 Hz frequency bands. For each participant, and for each channel, we calculated the median SWA across all artifact-free, 6-second windows occurring during N2 and N3 sleep. Final Global SWA measures were calculated by averaging across the following electrodes: Fp1, F7, F3, Fp2, F8, F4.

Participants with prior intracranial surgeries (n=3) were excluded from the SWA analysis.

Statistics

To determine covariates of interest for 24-hour memory retention, a student’s t test was performed for the following binary variables: sex (female/male), education (higher education degree: yes/no), presence of MRI lesion (yes/no), daytime nap(yes/no), medications : selective serotonin reuptake inhibitor or selective serotonin and norepinephrine reuptake inhibitors (yes/no), benzodiazepines(yes/no), EEG findings: presence of focal slowing (yes/no), presence of generalized slowing (yes/no) with 24 hour memory retention as the outcome of interest. An ANOVA was performed for epilepsy localization and lateralization, and a Kruskal-Wallis test for the ordinal variables of memory complaints, and pre and post- SSS. For the continuous variables: age, epilepsy duration, number of anti-seizure medications, sleep stages, nap time duration, number of IEAs per hour, number of IEAs per hour NREM sleep, slow wave activity power, Pearson’s correlation or Spearman’s correlation was performed.

Variables that were identified as being significantly correlated with the retention rates on univariate analysis with a p≤0.05 were then included in a multiple linear regression model with memory retention as the outcome of interest. For variables which were found to be collinear, the variable with the higher correlation was favored.

Statistical analysis was performed using JMP version 16.2.

Results

Study demographics and epilepsy characteristics

A total of 55 participants were enrolled in the study (Supplementary figure S1). Two participants were excluded from analysis due to the presence of seizures on the ambulatory EEG. Eleven participants were excluded from analysis because they were unable to complete the baseline encoding task.

Forty-two participants were included in the analysis. The 11 participants who were unable to complete the task were more likely to be female (91% v.s. 52%, p=0.03) and be on a higher median number of anti-seizure medications (2 v.s. 1, p=0.02).

With regards to epilepsy localization, 26 had temporal lobe epilepsy, 3 had frontal lobe epilepsy, and 1 participant had a parietal epilepsy. Twelve participants did not have a clear epilepsy localization or lateralization.

MRI abnormalities included: left mesial temporal sclerosis (n=3), encephalomalacia (n=2) due to meningioma surgery in left temporal and right frontal lobes, parietal cavernoma, encephalomalacia due to right frontal trauma, and hyperintense left temporal lobe. Two participants used their CPAP machine during the study.

Figures 1a and 1b summarize the location of the IEAs and the percentage of IEAs in each sleep state per participant. Only one participant in the cohort had temporal rhythmic delta activity (5 left temporal runs), this participant did not have epileptiform discharges in the recording.

Figure 1a.

Number of IEAs per participant and their laterality

Figure 1b.

Fraction of IEAs per participant occurring in each state. Figure highlights that the majority of IEAs in sleep occurred in N2.

Memory performance

Forty-two participants successfully completed the encoding task (i.e., encoded at least 5 card pairs). The median number of trials needed for successful encoding was 2 (range:1–5), and the median number of card pairs encoded was 6 (5–11). The average 24-hour retention rate was 70.9% ± 30.2%. 24-hour retention rates were not associated with sex, level of education, subjective memory complaints or pre/post levels of subjective sleepiness (table 1). There was a negative correlation between the number of antiseizure medications and 24-hour memory retention rates (ρ= −0.30, p=0.05). The most common anti-seizure medications were levetiracetam (57%) and lamotrigine (19%). Retention scores were not different between those with (n=5) or without antiseizure medications with negative cognitive side effects (60.1% v.s. 72,3% p=0.43). Cognitive screening test data were available for 27 participants, 20 had MMSE scores (range : 25–30), 7 had MOCA scores (range: 19–28). There was a positive correlation between MMSE score or the MOCA converted to MMSE score and retention rate, but it did not reach statistical significance (ρ= +0.38, p=0.09).

Table 1.

Participant demographics and clinical variables.

	n= 42	p-value
Age in years (± SD)	64.3 ± 7.5	0.53
Sex		0.38
Female	52%
Male	48%
Higher Education Degree		0.35
Yes	64%
No	36%
Epilepsy Duration in years (± SD)	3.1 ± 7.4	0.33
Epilepsy Localization		0.15
Temporal	62%
Extra-temporal	10%
Unknown	29%
Epilepsy Lateralization		0.25
Right	17%
Left	38%
Bilateral	14%
Unknown	31%
MRI lesion		0.92
Yes	19%
No	81%
Medications
SSRI		0.90
Yes	19%
No	81%
Benzodiazepine		0.49
Yes	9%
No	91%
Median # of ASMs (range)	1 (1–3)	0.05
Median Pre-testing SSS (range)	2 (1–5)	0.46
Memory Complaints		0.78
A great deal	15%
Somewhat	26%
Only a little	46%
Not at all	13%
Post-testing SSS (range)	2(1–5)	0.95

ASM: Antiseizure medication, SSRI: Selective serotonin reuptake inhibitor, SSS: Stanford Sleepiness Scale. P-values represent associations with 24-hour memory retention rates.

Association of 24-hour memory retention with sleep features

Table 2 summarizes the duration of sleep and sleep stages for the cohort. None of the sleep macroarchitecture parameters, including duration of N1, N2, N3, and REM sleep correlated with 24-hour memory retention rates. There was a significant positive correlation between global SWA power in the 0.6–4Hz frequency range and 24-hour memory retention rates (ρ=+0.36, p=0.026) (figure 2).

Table 2.

Neurophysiologic Parameters

	n=42	p-value
Total Sleep time (minutes ± SD)	389.0 ± 80.5	0.46
N1	38.5 ± 16.9	0.81
N2	267.0 ± 66.6	0.50
N3	9.5 ± 18.4	0.14
REM	74.0 ± 39.4	0.17
Daytime Nap		0.73
Yes	43%
No	57%
Nap Time (minutes ± SD)	13.2 ± 26.8	0.89
EEG Findings
Generalized slowing on EEG		0.65
Yes	10%
No	90%
Focal Slowing on EEG		0.35
Yes	50%
No	50%
# of IEAs per hour (± SD)	0.74 ± 1.5	0.0004
# of IEAs per hour of NREM sleep (± SD)	3.4 ± 8.4	0.0015
Slow Wave Activity Power (μV2)	n=39
Global (0.6–1Hz)	10.8 ± 7.1	0.06
Global (0.6–4Hz)	40.7 ± 23.2	0.023

P-values represent associations with 24-hour memory retention rates.

Figure 2.

Retention rates after 24 hours (points) overlaid on regression model predicted retention rates (line) as a function of power of global slow wave activity (0.6–4Hz) in μV2, with antiseizure medication held constant at 1. Points are color coded based on frequency of IEAs per hour.

Association of 24-hour memory retention with IEAs

To determine the association of IEA burden on memory retention rates, the frequency of IEAs (number of IEAs per hour) over the entire 24-hour EEG recording as well as during only NREM sleep were calculated. Both measures were negatively correlated with 24-hour memory retention (ρ= −0.52, p=0.0004) and (ρ= −0.48, p=0.0015) respectively (table 2).

Multivariate Analysis

The following variables were found to be collinear with each other: SWA power (0.6–1Hz) and SWA power 0.6–4Hz (ρ=+0.97, p<0.0001); similarly, IEA frequency over 24 hours and IEA frequency in NREM sleep were also highly correlated (ρ=+0.95, p<0.0001). As a result, the variables of Global SWA in the 0.6–1Hz range and the IEA frequency in NREM sleep were excluded from the model. Alternative models including these variables are included as supplementary table S1.

After controlling for age, sex, and education, the final model included significant associations for SWA power (0.6–4Hz) (β =+0.002, 95%CI[+0.0008 to +0.0090], p=0.02), number of antiseizure medications (β =−0.20, 95% CI=[−0.35 to −0.05], p=0.013), and number of IEAs per hour (β =−0.08, 95%CI[−0.14 to −0.02], p=0.0094). For the model as a whole, the R²=0.45, i.e., 45% of the variance in memory retention rate was linearly accounted for by the 3 predictors. Adjusted R²=0.33, AICc=15.8.

Discussion

In this study of retention rates of a visual declarative memory task in older adults with epilepsy, we found that greater frequency of interictal epileptiform abnormalities per hour, reduced slow wave activity (0.6–4Hz) power during NREM sleep, and higher number of antiseizure medications predict worse 24-hour memory recall.

Aging and memory consolidation

Declarative memories once acquired are initially hippocampally dependent and over time can be recalled with just the neocortex. For the brain to produce this shift, memories must be replayed during quiet wakefulness and NREM sleep.² Slow wave activity (SWA) in sleep is essential for this process of memory consolidation, and its enhancement is associated with improved memory retention.¹⁸ With aging, there is a change in sleep macroarchitecture with diminished N3 sleep and increased time in light sleep.¹⁹ There are also changes in the hippocampus and the frontal lobes, which are two key anatomical structures involved in the consolidation process.²⁰ Studies investigating long-term retention in older adults showed diminished SWA compared to younger controls, worse retention after a night of sleep⁴, and persistent deficits up to a week after the acquisition of new information.^4,6 In our study, we did not find age to be a predictor of memory retention rates, but all our participants were over age 50 and we did not have a younger control group for comparison. Some participants without IEAs recalled more correct card pairs at delayed 24-hour recall than at immediate recall, which suggests improved consolidation of card pairs that were not well encoded initially.

Impact of IEAs on Accelerated long-term forgetting

The phenomenon of accelerated long-term forgetting was first described in temporal lobe epilepsy patients.²¹ It can be seen within hours of encoding^22,23, and in both verbal and visual memory paradigms.^22–24 ALF has been noted across different patient populations and age ranges including children with epilepsy²⁵, patients with transient epileptic amnesia²⁶, and patients with generalized epilepsy.²⁷ Studies have also highlighted how epilepsy patients with subjective memory complaints may only exhibit deficits in long-term memory.²⁸

There have been few studies examining the effect of IEAs on retention rates. In one of the largest—a series of 70 patients with an age range of 18 to 60—the presence of IEAs on a routine EEG prior to testing was found to correlate with worse retention on verbal and visual memory tests four weeks later.²¹ In another study with a similar protocol, the presence of generalized discharges on routine EEG was associated with worse retention scores at 4 weeks on the Rey–Osterrieth Complex Figure Test (visual memory) only, but not on word list or story retention (verbal memory).²⁴ Another study comparing participants with temporal lobe epilepsy and extra-temporal lobe epilepsy, determined the presence of a hippocampal lesion including prior lobectomies was associated with worse memory retention rates at 24 hours; in the subset of 25 out of 32 patients who underwent a 4-day ambulatory EEG, the presence of IEAs was a significant predictor of memory retention rate only in the univariate analysis.⁹ In a 5-day ambulatory EEG study of epilepsy patients with generalized IEAs (n=5), focal IEAs (n=10), no IEAs (n=18), and controls (n=15), the participants with generalized IEAs showed worse retention of nonverbal memory and those with focal IEAs showed diminished retention of verbal memory between 30 minutes and 4 days.²⁹

Invasive EEG studies have highlighted the detrimental effects of hippocampal and extra-hippocampal IEAs, and IEA burden on encoding and recall of information.^30,31 The timing of the IEAs with regards to consolidation is important with IEAs occurring during wakefulness disrupting fast consolidation processes, and those occurring during sleep impacting the longer sleep-dependent consolidation processes.³¹ IEAs have also been linked to a transient decrease in functional connectivity, which suggests that this might be one mechanism contributing to their disruption of memory consolidation.³²

Whether IEAs are a treatment target to improve cognition remains a controversial issue.³³ One explanation for the association of IEAs and impaired long-term memory is that they represent a marker of a dysfunctional network although in our study, a similar correlation was not seen for focal slowing or MRI lesions. Since, in our study, memory retention worsened with the increased frequency of IEAs, this would indicate that the burden of does IEAs matter. We found that IEAs during sleep and wakefulness both negatively correlate with memory retention.

Impact of Anti-seizure medications

Anti-seizure medications are associated with adverse cognitive effects, especially at higher drug loads³⁴, and in the setting of polytherapy³⁵. In our study we found an association between higher antiseizure medication number and worse performance on the memory task, we also found that the subjects who failed the task were also on a higher drug load. This represents an appealing target to aid memory in persons with epilepsy in the clinic.³⁶ This correlation could also be attributed to the medications serving as a marker of a more severe epilepsy requiring more aggressive treatment, but none of the participants analyzed had a seizure during the study. Ultimately a balance would have to be achieved between the side effects of medications experienced by patients and trying to treat IEAs.

Sleep variables

Sleep is essential for the transformation of short-term memories into longer term and more stable memories. SWA has been found to correlate with declarative and procedural memory consolidation³⁷ in healthy individuals and is an essential component of the process of memory consolidation. In a pilot study of epilepsy patients in the epilepsy monitoring unit, retention rates were higher after 12 hours including sleep compared to 12 hours of wakefulness and correlated with duration of N3 sleep.¹¹ Attempts to improve consolidation in healthy participants have focused on cuing such as auditory³⁸ cuing with memory enhancement correlating with cue-induced increases in relative SWA power.

Studies looking at sleep variables in ALF have also been limited. In a cohort of 11 patients with transient epileptic amnesia tested on a word pair association task over 12-hour intervals with sleep or wakefulness, slow wave sleep (SWS) percentage did not correlate with retention in the epilepsy group, as compared to the control group. None of the epilepsy participants had IEAs. The authors concluded that, unlike their healthy counterparts, the epilepsy group did not benefit from SWS.³⁹ In an ambulatory EEG study of temporal and extratemporal lobe epilepsy patients, the participants who took a nap after training showed better memory retention scores than those who did not, while the other sleep variables investigated, such as total sleep, total REM, and first half SWS, were not significant predictors.²⁶ Our study found that slow wave activity power in NREM sleep was positively correlated with memory retention but no significant correlation between memory retention and sleep macro-architecture such as duration of individual sleep stages. Our findings are in line with prior studies in older cohorts without epilepsy, showing a positive correlation between slow wave activity and memory consolidation.⁴⁰ Thus, sleep microarchitecture may be more important for long term memory in epilepsy patients, and any treatments may be more successful if they aim to preserve or enhance SWA. Treatment of any sleep fragmenting disorders, and interventions ranging from meditation to medications, and neuromodulation are being explored as possible therapeutic avenues⁴¹, and trials in older adults have already investigated auditory deep sleep stimulation.⁴²

Limitations

Our study has several limitations; due to the sample size, an adequate correction for epilepsy lateralization and localization was difficult to perform. Several participants had epilepsy that was difficult to lateralize and localize, given the absence of IEAs on EEG and absence of MRI lesion. In addition, cognitive screening data were not available for all participants, and the cohort on average did not have significant subjective memory complaints (59% responded not at all or a little) which is more than one would expect for a population of older adults with epilepsy. We are also unable to assess the prevalence of sleep disordered breathing, given the absence of oximetry. Finally, our declarative memory task was visually based, and it is unclear if a language-based task would have yielded different results. Future studies examining list-learning or figure learning paradigms should be pursued.

Conclusions

In this study of long-term memory retention in older adults with epilepsy, we found greater frequency of interictal epileptiform abnormalities, reduced slow wave activity power, and greater number of antiseizure medications to correlate with worse overnight retention on a visual memory task. These findings may represent potential therapeutic targets for improving memory in future studies.

Key Points.

This was a prospective study of older adults with epilepsy who performed a memory task prior to and after a 24-hour EEG.

Duration of individual sleep stages was not associated with memory retention

Greater frequency of interictal epileptiform abnormalities correlated with lower 24-hour memory retention rates.

Other predictors of lower memory retention included reduced slow wave activity power during non-REM sleep, and higher number of antiseizure medications.

Data availability statement:

The data that support the findings of this study are available from the corresponding author upon reasonable request.