Effect of Levetiracetam on Cognition in Patients With Alzheimer Disease With and Without Epileptiform Activity: A Randomized Clinical Trial

Key Points

Question  What is the effect of 4 weeks of treatment with low-dose levetiracetam on cognitive function in patients with Alzheimer disease (AD)?

Findings  In this randomized clinical trial of 34 adults with AD, treatment with levetiracetam did not significantly modify cognitive function. However, the treatment did improve executive function and spatial memory among participants with AD who had seizures or subclinical epileptiform activity that was identified through extended neurophysiological recordings.

Meaning  This study’s findings indicated that extended neurophysiological assessments are important to identify patients with AD who could benefit from antiseizure approaches and that levetiracetam treatment could improve cognitive symptoms in the estimated 60% of patients with AD who have seizures and subclinical epileptiform activity.

Importance  Network hyperexcitability may contribute to cognitive dysfunction in patients with Alzheimer disease (AD).

Objective  To determine the ability of the antiseizure drug levetiracetam to improve cognition in persons with AD.

Design, Setting, and Participants  The Levetiracetam for Alzheimer’s Disease–Associated Network Hyperexcitability (LEV-AD) study was a phase 2a randomized double-blinded placebo-controlled crossover clinical trial of 34 adults with AD that was conducted at the University of California, San Francisco, and the University of Minnesota, Twin Cities, between October 16, 2014, and July 21, 2020. Participants were adults 80 years and younger who had a Mini-Mental State Examination score of 18 points or higher and/or a Clinical Dementia Rating score of less than 2 points. Screening included overnight video electroencephalography and a 1-hour resting magnetoencephalography examination.

Interventions  Group A received placebo twice daily for 4 weeks followed by a 4-week washout period, then oral levetiracetam, 125 mg, twice daily for 4 weeks. Group B received treatment using the reverse sequence.

Main Outcomes and Measures  The primary outcome was the ability of levetiracetam treatment to improve executive function (measured by the National Institutes of Health Executive Abilities: Measures and Instruments for Neurobehavioral Evaluation and Research [NIH-EXAMINER] composite score). Secondary outcomes were cognition (measured by the Stroop Color and Word Test [Stroop] interference naming subscale and the Alzheimer’s Disease Assessment Scale–Cognitive Subscale) and disability. Exploratory outcomes included performance on a virtual route learning test and scores on cognitive and functional tests among participants with epileptiform activity.

Results  Of 54 adults assessed for eligibility, 11 did not meet study criteria, and 9 declined to participate. A total of 34 adults (21 women [61.8%]; mean [SD] age, 62.3 [7.7] years) with AD were enrolled and randomized (17 participants to group A and 17 participants to group B). Thirteen participants (38.2%) were categorized as having epileptiform activity. In total, 28 participants (82.4%) completed the study, 10 of whom (35.7%) had epileptiform activity. Overall, treatment with levetiracetam did not change NIH-EXAMINER composite scores (mean difference vs placebo, 0.07 points; 95% CI, −0.18 to 0.32 points; P = .55) or secondary measures. However, among participants with epileptiform activity, levetiracetam treatment improved performance on the Stroop interference naming subscale (net improvement vs placebo, 7.4 points; 95% CI, 0.2-14.7 points; P = .046) and the virtual route learning test (t = 2.36; Cohen f² = 0.11; P = .02). There were no treatment discontinuations because of adverse events.

Conclusions and Relevance  In this randomized clinical trial, levetiracetam was well tolerated and, although it did not improve the primary outcome, in prespecified analysis, levetiracetam improved performance on spatial memory and executive function tasks in patients with AD and epileptiform activity. These exploratory findings warrant further assessment of antiseizure approaches in AD.

Trial Registration  ClinicalTrials.gov Identifier: NCT02002819

Alzheimer disease (AD) is the leading cause of neurodegenerative dementia worldwide.¹ Early symptoms include short-term memory loss, reduction in executive function, word-finding difficulties, and visuospatial dysfunction. Seizures can also occur early in the course of AD and exacerbate cognitive symptoms.²Quiz Ref ID An estimated 10% to 22% of patients with AD develop unprovoked seizures, with higher rates in familial and early-onset cases.^3-5 In addition, 22% to 54% of patients with AD exhibit subclinical epileptiform activity when monitored using overnight electroencephalography (EEG) and 1-hour magnetoencephalography with simultaneous EEG (MEG-EEG) recordings, and these epileptiform discharges are much more common during sleep.^6-9 Patients with AD and seizures or subclinical epileptiform activity experience a faster decrease in cognitive function.^6,9-11 Whether epileptiform activity contributes to cognitive worsening could not be evaluated in these longitudinal observational studies, but they provided a rationale to examine whether antiseizure approaches could benefit patients with AD who have an epileptic phenotype. Placebo-controlled clinical trials to address this question are lacking.

Preclinical studies in transgenic mouse models of AD found that suppressing epileptiform activity with antiseizure drugs was associated with improvements in behavior as well as histopathological signs of chronic network hyperexcitability in the hippocampus.^12,13 Levetiracetam is a widely used antiseizure drug that has been reported to suppress epileptiform spikes and improve synaptic and cognitive function in mouse models of AD.^12-16 Levetiracetam treatment has been found to be well tolerated and successful at suppressing seizures among patients with AD and seizure disorders, often at low doses.^2,17

Based on these data, we had a clear rationale to evaluate the ability of levetiracetam to improve cognitive function among patients with AD. We sought to explore whether patients with AD and epileptiform activity responded better to levetiracetam treatment than those without this sign of network hypersynchrony. Patients with AD who did not have epileptiform activity were included because they could have had aberrant neuronal activity that was lower than the level of detection of EEG and MEG-EEG,¹⁸ and they could have also responded favorably to treatment with levetiracetam.

The Levetiracetam for Alzheimer’s Disease–Associated Network Hyperexcitability (LEV-AD) study was an investigator-initiated phase 2a randomized double-blinded placebo-controlled crossover clinical trial conducted at the University of California, San Francisco (UCSF), and the University of Minnesota, Twin Cities (UMN). Study visits occurred from October 16, 2014, to June 9, 2017, at UCSF and from August 3, 2018, to July 21, 2020, at UMN. The clinical trial protocol (Supplement 1) was approved by the institutional review boards at both sites. All participants or their legally authorized representatives provided written informed consent. This study followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline for randomized clinical trials and was conducted in accordance with the Declaration of Helsinki.¹⁹

Eligible participants were adults 80 years and younger with a diagnosis of probable AD, a Mini-Mental State Examination score of 18 points or higher, and/or a Clinical Dementia Rating (CDR) score of less than 2 points.²⁰ In the initial study design, enrollment was restricted to those who experienced symptom onset at younger than 70 years and had a history of 1 or more unprovoked seizures or subclinical epileptiform activity within 5 years of enrollment. However, to improve recruitment, enrollment was expanded on August 8, 2015, to include participants up to age 80 years and those without seizure or epileptiform activity. Additional inclusion criteria and complete exclusion criteria are provided in the eMethods in Supplement 2. Patients receiving antiseizure drugs for any reason were excluded. Participants self-reported their race to assess diversity in the sample. Further details on the protocol, eligibility criteria, and study design are available in the clinical trial protocol in Supplement 1.

Participants received the following screening procedures before randomization into the clinical trial: magnetic resonance imaging of the brain; the Mini-Mental State Examination²¹; the CDR²²; an optional lumbar puncture or amyloid positron emission tomography to measure biomarkers of AD in the cerebrospinal fluid or brain; an overnight EEG, including a polysomnographic examination at the UMN study site; and a 1-hour MEG-EEG. Additional screening procedures are provided in the eMethods in Supplement 2.

Participants who met eligibility criteria after screening examinations were randomized to receive an oral treatment sequence according to a prespecified process (eFigure 1 in Supplement 2). Quiz Ref IDParticipants randomized to group A received placebo twice daily for 4 weeks followed by a 4-week washout period, then levetiracetam, 125 mg, twice daily for 4 weeks. Participants randomized to group B received this treatment sequence in reverse order. The investigational site pharmacist generated and maintained the randomization schedule and assigned a randomization code to each participant. The site pharmacists were the UCSF Drug Product Services Laboratory (San Francisco, California) from October 16, 2014, to April 31, 2016, and the Safeway Compounding Pharmacy (Safeway, Inc; San Jose, California) from March 1, 2016, to July 21, 2020. The active study drug was prepared using pharmaceutical-grade levetiracetam tablets that were crushed and encapsulated in dark green capsules. Placebo was prepared using the same inactive ingredients as the active drug (at the UCSF site) or cornstarch (at the UMN site), which were encapsulated in identical dark green capsules. Participants were randomized 4 at a time to ensure equal distribution in groups A and B at all phases of the clinical trial. The randomization key was held by the pharmacists, and all investigators and participants were blinded to treatment sequence. Participants were evaluated over 4 visits, with visit 1 occurring on study day 1, visit 2 on study day 29, visit 3 on study day 57, and visit 4 on study day 85. Details of visits 1 through 4 are provided in the eMethods in Supplement 2.

Cognitive testing included, in order of administration, the National Institutes of Health Executive Abilities: Measures and Instruments for Neurobehavioral Evaluation and Research (NIH-EXAMINER),²³ the Stroop Color and Word Test (Stroop), the Alzheimer’s Disease Assessment Scale–Cognitive Subscale (ADAS-Cog),²⁴ and a virtual route learning test (eFigure 2 in Supplement 2).

The NIH-EXAMINER was used to measure executive function. This battery of tests comprises psychometrically robust scales that were generated using item response theory and provides reliable estimations of real-world executive function.^23,25 In this study, tasks evaluated on the NIH-EXAMINER included antisaccade, set shifting, flanker task, dot counting, spatial 1-back, category fluency, and letter fluency. Four alternate forms were used depending on study visit (eMethods in Supplement 2). Composite scores on the NIH-EXAMINER (scores have an indefinite range, with higher scores indicating better performance) were generated using item response theory, and scores with SEs greater than 0.55 were classified as unreliable and excluded from analysis.²³ Composite scores from 2 participants were excluded on this basis.

The Stroop test was used to assess executive functions, including selective attention, cognitive flexibility, and processing speed. Tasks included color naming (naming the ink colors of words with meanings that were congruent with their font colors; score range 0-126, with higher scores indicating better performance) and interference naming (naming the ink colors of words with meanings that were incongruent with their font colors; score range, 0-77, with higher scores indicating better performance), and each task was restricted to 1 minute. None of the participants were color-blind, and almost all were able to participate, with the exception of 2 individuals who did not comprehend the instructions.

The ADAS-Cog was used to evaluate global cognitive functioning.²⁴ The ADAS-Cog is a 70-point scale (with lower scores indicating better cognitive function) that includes an assessment of verbal memory, language, orientation, reasoning, and praxis. Four alternate forms were used depending on study visit.

A virtual route learning test was used to assess navigation learning. Using a driving simulator displayed on a computer monitor, participants learned a predetermined route comprising 15 turns through a virtual neighborhood by trial and error over a maximum of 15 task repetitions. If a participant achieved 100% correct performance on 2 subsequent task repetitions, the test ended, and the maximum score was imputed for all remaining repetitions. Four alternate forms with different neighborhoods and routes were used in random sequence (eFigure 2 in Supplement 2). In each alternate form, a single building was repeated many times, but cues were different (eg, distinct buildings, a bush, grass, palm trees, a hot dog stand, or a school bus). Performance was measured by learning rates and the sum of accurate turns across task repetitions 2 through 15 or by the maximum number of repetitions completed consistently across all study visits.

Caregiver interviews were used for the following measures: the Clinician Assessment of Fluctuation (score range, 0-12 points, with higher scores indicating more fluctuations),²⁶ the One Day Fluctuation Assessment Scale (score range, 0-21 points, with higher scores indicating more fluctuations),²⁶ the CDR sum of boxes subscale (score range, 0-18 points, with higher scores indicating more impairment),²² the Alzheimer’s Disease Cooperative Study–Activities of Daily Living Inventory (score range, 0-78 points, with higher scores indicating less impairment),²⁷ the Alzheimer’s Disease Cooperative Study–Clinical Global Impression of Change (score range, 1-7 points, with lower scores indicating improvement and higher scores indicating worsening) subscale,²⁸ and the Neuropsychiatric Inventory (score range, 0-144 points, with higher scores indicating more behavioral symptoms).²⁹ Descriptions of caregiver-based measures are provided in the eMethods in Supplement 2.

At the end of each study visit, participants received a 1-hour resting MEG-EEG. We reviewed the initial overnight EEGs and 1-hour MEG-EEG recordings for sleep quality and determined that all participants reached all stages of sleep on the overnight EEGs and that all participants achieved stage 1 sleep (6 individuals [18.2%]) or stage 2 or higher sleep (28 individuals [81.8%]) on the initial MEG-EEG recording. In addition, participants who completed in-person visits achieved stage 1 sleep on 15 MEG-EEG recordings (15.6%) and stage 2 or higher sleep on 80 MEG-EEG recordings (83.3%) performed at study visits. Between March 13, 2020, and July 21, 2020, telephone-only visits were required because of the COVID-19 pandemic. Telephone visits were restricted to caregiver-based measures. Five participants had some or all visits conducted by telephone.

Quiz Ref IDThe primary outcome was the ability of 4-week treatment with levetiracetam to improve executive function, as measured by the NIH-EXAMINER composite score, compared with 4 weeks of placebo. Secondary outcomes were (1) the ability of levetiracetam to improve executive function (measured by the Stroop interference naming subscale) and global cognitive function (measured by the ADAS-Cog), (2) its ability to improve the extent of disability and behavior (measured by the CDR sum of boxes subscale, the Alzheimer’s Disease Cooperative Study–Activities of Daily Living Inventory, the Alzheimer’s Disease Cooperative Study–Clinical Global Impression of Change, and the Neuropsychiatric Inventory), and (3) the efficacy of levetiracetam in suppressing epileptiform activity among those with epileptiform activity. Exploratory outcomes included performance on the virtual route learning test, fluctuations on the Clinician Assessment of Fluctuation and the One Day Fluctuation Assessment Scale, serum prolactin levels, and all measures in subgroups with and without epileptiform activity. Participants were included in the subgroup with epileptiform activity if they had a history of 1 or more unprovoked seizures or subclinical epileptiform activity within 5 years of study enrollment or subclinical epileptiform activity that was detected during the study.

We based sample size estimates on the hypothesis that treatment with levetiracetam would improve cognitive performance to an extent that was clinically meaningful compared with placebo. Standard sample size calculations for repeated-measures analysis of variance showed that F tests could detect a Cohen f effect size of greater than 0.35 with power greater than 0.80 using a sample of 36 participants and allowing for a 10% dropout rate.³⁰

For each of the scores, a 1-way within-participant repeated-measures analysis of variance was performed comparing the scores before and after receipt of placebo vs before and after receipt of levetiracetam. Analyses were 2-tailed, with the exception of the analysis of epileptiform activity, which was 1-tailed based on the expectation that levetiracetam treatment would reduce epileptiform activity. A linear mixed-effects model was used to compare the differences in learning rates on the virtual route learning test after treatment with placebo vs levetiracetam. Details of this model are provided in the eMethods in Supplement 2. For all analyses, the null hypothesis was rejected at P < .05. Analyses were performed using SPSS Statistics software, version 25 (IBM SPSS Statistics), and SAS software, version 9.4 (SAS Institute Inc).

Fifty-four adults were screened; of those, 11 individuals did not meet study criteria, and 9 individuals declined participation. A total of 34 individuals (mean [SD] age, 62.3 [7.7] years; 21 women [61.8%] and 13 men [38.2%]; 33 White individuals [97.1%] and 1 Black individual [2.9%]) were enrolled, with 17 participants randomized to group A and 17 randomized to group B (Figure 1). Eighteen participants were randomized at UCSF, and 16 participants were randomized at UMN. Thirteen participants (38.2%) were categorized as having epileptiform activity; of those, 11 participants had subclinical epileptiform activity that was detected during the study, 1 participant had a recent history of subclinical epileptiform activity, and 1 participant had a recent unprovoked seizure. The participant who experienced a seizure had 2 complex partial seizures within 1 year of enrolling in the study and declined receipt of antiseizure treatment because of the infrequency of their seizures and the availability of close supervision at home. Subclinical epileptiform activity was detected by EEG in 9 participants and by MEG-EEG in 5 participants. Two participants had epileptiform activity on both types of recordings. Epileptiform activity detected by EEG was predominantly found in the left temporal (n = 7), left frontotemporal (n = 1), and right frontal or bifrontal (n = 1) channels. Epileptiform activity detected by MEG-EEG was predominantly found in the left temporal (n = 1), left frontotemporal (n = 1), bitemporal (n = 1), right temporal (n = 1), and right temporoparietal (n = 1) channels. Spikes and sharp waves were sporadic, and their frequency ranged from rare to occasional, which was similar to their occurrence in a previous study.⁶

Baseline characteristics were well balanced between groups A and B (mean [SD] age, 62.9 [7.5] years vs 61.8 [8.1] years, respectively; 9 women [52.9%] vs 12 women [70.6%]; mean [SD] disease duration, 5.5 [3.4] years vs 4.3 [1.7] years) (Table 1). Compared with the UMN cohort, the UCSF cohort had an earlier age at onset of cognitive worsening (mean [SD], 61.0 [8.5] years vs 54.3 [5.3] years, respectively; P = .009 derived by t test), a longer disease duration (mean [SD], 3.8 [1.6] years vs 5.8 [5.3] years; P = .04 derived by Mann-Whitney U test), more years of education (mean [SD], 14.8 [1.9] years vs 16.8 [3.1] years; P = .04 derived by Mann-Whitney U test), and a higher proportion of participants categorized as having epileptiform activity (3 of 16 individuals [18.8%] vs 10 of 18 individuals [55.6%]; P = .04 derived by Fisher exact test) (eTable 1 in Supplement 2).

Of the 34 randomized participants, 33 had biomarkers identified through cerebrospinal fluid or amyloid positron emission tomography that were consistent with an AD diagnosis; 1 participant with cerebrospinal fluid biomarkers that were inconsistent with an AD diagnosis was excluded from the efficacy analysis in accordance with the prespecified study protocol. Overall, 5 participants (14.7%) discontinued intervention, and 28 participants (82.4%) completed the study, 10 of whom (35.7%) had epileptiform activity. Four participants (11.8%) who completed the study had telephone-only visits because of restrictions during the COVID-19 pandemic. Additional characteristics of participants included in the efficacy analysis are provided in eTable 2 in Supplement 2.

The primary and secondary outcomes in the combined analysis of both sites are summarized in Table 2. Among those receiving levetiracetam treatment vs placebo, overall cognitive performance did not differ on the NIH-EXAMINER (mean score difference, 0.07 points; 95% CI, −0.18 to 0.32 points; P = .55), the ADAS-Cog (mean score difference, −1.0 points; 95% CI, −3.4 to 1.5 points; P = .43), or the Stroop interference naming subscale (mean score difference, 2.9 points; 95% CI, −1.0 to 6.8 points; P = .14). The Cohen f effect sizes for the potential benefit of levetiracetam treatment were 0.13 for the NIH-EXAMINER, 0.33 for the Stroop interference naming subscale, and 0.17 for the ADAS-Cog. Functional outcomes (eg, mean score difference on the CDR sum of boxes subscale, 0.1 points; 95% CI, −0.2 to 0.4 points; P = .63) and the frequency of epileptiform events (mean difference, 0.1 events; 95% CI, −0.9 to 1.1 events; P = .40) did not change among those receiving levetiracetam compared with placebo. No seizures were recorded on EEG during the study. The following exploratory measures did not change with levetiracetam treatment: Clinician Assessment of Fluctuation (difference vs placebo among 28 participants: 0.7 points; 95% CI, −1.0 to 2.4 points; P = .40), One Day Fluctuation Assessment Scale (difference vs placebo among 28 participants: 0.8 points; 95% CI, −0.3 to 1.8 points; P = .15), prolactin level (difference vs placebo among 24 patients: −0.2 ng/mL; 95% CI, −1.1 to 0.7 ng/mL; P = .63), and accuracy on the virtual route learning test (difference vs placebo among 12 participants: 3.8 correct turns; 95% CI, −8.5 to 16.0 correct turns; P = .52). The mean (SD) serum levetiracetam level at the end of active drug treatment was 3.8 (1.7) μg/mL.

In the prespecified exploratory analysis, we tested the hypothesis that participants with epileptiform activity would respond better to levetiracetam treatment than participants without epileptiform activity. We assessed cognitive outcomes in these 2 subgroups, which did not differ in most baseline characteristics (eTable 3 in Supplement 2). Among 9 participants with epileptiform activity, levetiracetam treatment improved performance on the Stroop interference naming subscale (net improvement vs placebo, 7.4 points; 95% CI, 0.2-14.7 points; F = 5.54; Cohen f = 0.83; P = .046), whereas among 13 participants without epileptiform activity levetiracetam treatment did not change performance on the Stroop interference naming subscale (net difference vs placebo, −0.3 points; 95% CI, −4.5 to 3.9 points; P = .88 [P = .04 vs participants with epileptiform activity]) (Figure 2A). Participants with epileptiform activity also showed improvement in learning rates on the virtual route learning test after receipt of levetiracetam treatment vs placebo (among 5 participants: t = 2.36; Cohen f² = 0.11; P = .02), whereas participants without epileptiform activity did not (among 7 participants: t = 0.04; P = .97) (Figure 2B). Compared with participants without epileptiform activity, those with epileptiform activity had net improvements in accuracy (sum of correct turns) on the virtual route learning test after receipt of levetiracetam treatment (improvement vs placebo, 17.4 correct turns; 95% CI, −0.6 to 35.4 correct turns; P = .03 for comparison between participants with vs without epileptiform activity, derived from t test) (eTable 4 in Supplement 2). Most participants with epileptiform activity had improved scores on the NIH-EXAMINER (6 of 8 individuals [75.0%]; Cohen f = 0.72) and the ADAS-Cog (7 of 9 individuals [77.8%]; Cohen f = 0.39) after receipt of levetiracetam treatment (eFigure 3 in Supplement 2), although the P values for group comparisons were not significant (eTable 4 in Supplement 2). Scores on ADAS-Cog components did not change after receipt of levetiracetam treatment among participants with or without epileptiform activity, although there was a slight improvement in orientation among those with epileptiform activity (mean score difference, −0.6; 95% CI, −1.1 to 0.0; P = .051) (eTable 5 in Supplement 2). In addition, functional outcomes did not change after receipt of levetiracetam treatment among 10 participants without epileptiform activity or 18 participants with epileptiform activity (eTable 6 in Supplement 2).

We also assessed cognitive measures in participants with early-onset AD (symptom onset at age <65 years) because a younger age of onset has been associated with a higher risk of seizures and subclinical epileptiform activity among those with AD.^4,31 Participants with early-onset AD improved on the Stroop interference naming subscale after receipt of levetiracetam treatment (net improvement among 20 participants: 4.0 points; 95% CI, 0-7.9 points; F = 4.41; Cohen f = 0.48; P = .049) but did not improve in other cognitive measures.

Because the UCSF cohort had a higher proportion of participants with epileptiform activity, we performed separate post hoc analyses of cognitive measures at the 2 study sites. After the receipt of levetiracetam treatment, participants at the UCSF site improved on the Stroop interference naming subscale (net improvement among 15 participants: 6.1 points; 95% CI, 1.4-10.7 points; F = 7.78; Cohen f = 0.75; P = .01) and the ADAS-Cog (net improvement among 15 participants: 3.1 points; 95% CI, 0-6.1 points; F = 4.74; Cohen f = 0.58; P = .047) but not on the NIH-EXAMINER or the virtual route learning test (eFigure 4 and eTable 7 in Supplement 2). Performance among participants at the UMN site did not significantly change in any of the cognitive measures (eFigure 4 and eTable 7 in Supplement 2).

No participants discontinued treatment in the clinical trial because of adverse events. Mild adverse effects were reported in 7 of 33 participants (21.2%) while receiving levetiracetam and 6 of 32 participants (18.8%) while receiving placebo (Table 3). Performance on the Neuropsychiatric Inventory did not change after receipt of levetiracetam treatment compared with placebo, indicating no measurable effects on mood.

Using data from the current study, we estimated sample sizes needed to show cognitive benefit after 4 weeks of treatment with levetiracetam for future clinical trials with a similar design using an α level of .05 for type I errors, a β level of .20 for type II errors, and effect sizes observed in Table 2 and eTable 4 in Supplement 2. A general study of AD would require 468 participants for the NIH-EXAMINER, 76 participants for the Stroop interference naming subscale, and 298 participants for the ADAS-Cog, whereas a study restricted to participants with epileptiform activity would require 20 participants for the NIH-EXAMINER, 15 participants for the Stroop interference naming subscale, and 60 participants for the ADAS-Cog.

Quiz Ref IDThis phase 2a randomized clinical trial showed that treatment with low-dose levetiracetam was well tolerated in patients with AD and, although levetiracetam treatment did not improve the primary outcome in our sample, it did improve cognitive function in those with epileptiform activity in the prespecified exploratory analysis. Levetiracetam treatment improved accuracy in spatial memory and executive functioning tasks among participants with epileptiform activity and had a good safety profile. Differences in response to levetiracetam treatment between participants with and without epileptiform activity justify further investigations of levetiracetam and other antiseizure medications among patients with AD who have epileptic phenotypes. The addition of extended neurophysiological examinations was important to identify patients who could benefit from levetiracetam treatment.

Antiseizure approaches for treating patients with AD are being increasingly investigated as evidence increases for the presence of network hyperexcitability in early disease stages that may contribute to disease progression.^{4,10,16,32-34} Sample size calculations based on data from the current clinical trial indicated that substantially fewer participants with AD would be needed to demonstrate cognitive benefits of levetiracetam treatment, particularly with regard to executive function, if the study preselected patients with seizures or subclinical epileptiform activity. The addition of neurophysiological examinations to identify the patients with AD who are most likely to benefit from an antiseizure strategy could lead to faster and more effective clinical trials. Preselecting patients with early-onset AD for clinical trials focused on network hyperexcitability could also be useful given that younger patients with AD could have network hyperexcitability without overt epileptiform activity, and this factor could account for why these patients responded better on the Stroop interference naming task after receipt of levetiracetam treatment in the current study. We cannot exclude other distinct factors in younger patients that could contribute to differences in response to levetiracetam treatment. Our study did not assess disease modification, and the findings still allow the possibility that levetiracetam treatment could generally slow disease progression in patients with AD when administered over longer courses. Numerous studies of levetiracetam for the treatment of AD are ongoing.^35-38

The low dosing of levetiracetam in the current study was based on optimal dosing reported in preclinical investigations as well as the findings of a clinical trial of amnestic mild cognitive impairment,^12,13,39 and our study achieved serum levetiracetam levels that were similar to the levels achieved in those studies. We used executive functioning measures because patients with AD who have subclinical epileptiform activity experience faster worsening of executive function and global measures than those without.⁶ Seizures in patients with AD are also associated with faster decreases in executive function.¹⁰

Both secondary cognitive outcomes significantly improved after receipt of levetiracetam treatment among those in the UCSF cohort but not the UMN cohort. Differences in the cohorts could account for this discrepancy. First, the UCSF cohort included a higher percentage of participants with epileptiform activity, partly because of the study’s initial focus on enrolling those with epileptiform activity. The UCSF cohort also had a younger age at symptom onset, longer disease duration, and more years of education, which have all been associated with a higher prevalence of seizures among those with AD.^2,5,7,11,40 Second, though unlikely to have a significant impact, there were minor differences in placebo composition between the UCSF and UMN sites.

This study has limitations. Quiz Ref IDFirst, a study of this nature is limited in sample size because of the extensive commitment required from participants and the low rate of study completion. Cognitive measures were also missing for 4 participants because of restrictions during the COVID-19 pandemic. Because of small samples and the lack of correction for multiple comparisons in the exploratory analysis, false-positive results and subgroup confounders are possible, and further evaluation of hypotheses generated from this study is required. Second, our study primarily assessed a population with early-onset AD and lacked participant diversity, and more inclusive recruitment will be important for future investigations. Third, the virtual route learning test was difficult for many participants to perform, limiting the results received from this test. Fourth, epileptiform events were infrequent, making it difficult to quantify the antiepileptic effects of levetiracetam treatment, and the study was not designed to assess correlations between the frequency of epileptiform activity and the response to levetiracetam treatment. Additional neurophysiological measures of network hyperexcitability are needed to determine the effects of antiseizure drugs on network dysfunction in patients with AD.^41-43

To our knowledge, this study was the first to include neurophysiological screening and treatment evaluations for levetiracetam treatment among patients with AD, which proved to be important for identifying estimations of response to the drug. These findings could lead to future personalized approaches to AD, in which patients with the epileptic variant of AD will receive different treatments than those without the epileptic variant. The implications are substantial when considering that an estimated 60% or more of patients with AD experience seizures and subclinical epileptiform activity.^6-9 Antiseizure approaches would complement and potentially enhance other strategies for treating AD, including those targeting disease protein aggregation as well as spreading and inflammation. Future AD clinical trials would benefit from including neurophysiological assessments whenever possible and adding antiseizure drugs when epileptic or epileptiform activity is detected.

Accepted for Publication: August 5, 2021.

Published Online: September 27, 2021. doi:10.1001/jamaneurol.2021.3310

Corresponding Author: Keith Vossel, MD, MSc, Mary S. Easton Center for Alzheimer’s Disease Research, Department of Neurology, David Geffen School of Medicine at UCLA, University of California, Los Angeles, 710 Westwood Plaza, RNRC C-224, Los Angeles, CA 90095 (kvossel@mednet.ucla.edu).

Author Contributions: Drs Vossel and Ranasinghe had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Vossel, Ranasinghe, Mucke, Possin, Boxer, Miller, Nagarajan, Kirsch.

Acquisition, analysis, or interpretation of data: Vossel, Ranasinghe, Beagle, La, Ah Pook, Castro, Mizuiri, Honma, Venkateswaran, Koestler, Zhang, Mucke, Howell, Possin, Kramer, Boxer, Nagarajan, Kirsch.

Drafting of the manuscript: Vossel, Ranasinghe, La, Ah Pook, Howell, Kramer.

Critical revision of the manuscript for important intellectual content: Vossel, Ranasinghe, Beagle, Castro, Mizuiri, Honma, Venkateswaran, Koestler, Zhang, Mucke, Possin, Boxer, Miller, Nagarajan, Kirsch.

Statistical analysis: Vossel, Ranasinghe, Venkateswaran, Mucke, Possin, Nagarajan.

Obtained funding: Vossel, Mucke.

Administrative, technical, or material support: Vossel, Beagle, La, Castro, Mizuiri, Honma, Venkateswaran, Koestler, Zhang, Howell, Boxer, Miller, Nagarajan.

Supervision: Vossel, Ranasinghe, Boxer, Miller, Nagarajan.

Conflict of Interest Disclosures: Dr Vossel reported receiving grants from the Alzheimer’s Association and the National Institute on Aging, National Institutes of Health and donations from the Fineberg Foundation, the N. Bud and Beverly Grossman Foundation, and the S. D. Bechtel, Jr. Foundation during the conduct of the study. Dr Mucke reported receiving grants from the S. D. Bechtel, Jr. Foundation during the conduct of the study and research funding for an independent project from Cure Network Dolby Acceleration Partners; personal fees from Eisai and Takeda Pharmaceutical; and serving on the scientific advisory boards of Acumen, Alkahest, Arvinas, Biogen, and Dolby Family Ventures outside the submitted work. Dr Possin reported receiving grants from the Global Brain Health Institute, the National Institutes of Health, the Rainwater Charitable Foundation, and Quest Diagnostics during the conduct of the study and personal fees from ClearView Healthcare Partners and Vanguard outside the submitted work. Dr Boxer reported receiving grants from the Alzheimer’s Association, the Alzheimer’s Drug Discovery Foundation, the Association for Frontotemporal Degeneration, Biogen, the Bluefield Project to Cure Frontotemporal Dementia, Eisai, Eli Lilly and Company, the National Institutes of Health, the Rainwater Charitable Foundation, Regeneron Pharmaceuticals, and the State of California; personal fees from Alector, Applied Genetic Technologies, Arkuda Therapeutics, Arvinas, AZTherapies, GlaxoSmithKline, Humana, Lundbeck, Oligomerix, Roche, the Roissy Foundation, Stealth BioTherapeutics, Transposon Therapeutics, TrueBinding, UCB, and Wave Pharma; and nonfinancial support from Eli Lilly and Company and Novartis; and serving as a consultant for Applied Genetic Technologies, Alector, Arkuda Therapeutics, Arvinas, Asceneuron, AZTherapies, BioAge Labs, GlaxoSmithKline, Humana, Oligomerix, Ono Pharmaceutical, Roche, Samumed, Sangamo Therapeutics, Stealth BioTherapeutics, Third Rock Ventures, Transposon Therapeutics, and Wave Pharma outside the submitted work. Dr Miller reported receiving grants from the Bluefield Project to Cure Frontotemporal Dementia and royalties from the Cambridge University Press, Guilford Press, John Hopkins Press, and Oxford University Press; being the editor-in-chief of Neurocase and a section editor of Frontiers in Neurology; serving as a scientific advisor, consultant, or board member for the Bluefield Project to Cure Frontotemporal Dementia, Biogen, the Buck Institute for Research on Aging, the National Institute for Health Research Cambridge Biomedical Research Centre and its Biomedical Research Unit in Dementia, the John Douglas French Alzheimer’s Foundation, the Larry L. Hillblom Foundation, the Massachusetts General Hospital/Harvard Medical School Alzheimer’s Disease Research Center, the Rainwater Charitable Foundation, SafelyYou, the Stanford Alzheimer’s Disease Research Center, The University of Texas at Dallas Center for BrainHealth, and the University of Washington Alzheimer’s Disease Research Center outside the submitted work. Dr Nagarajan reported receiving grants from RICOH MEG outside the submitted work. No other disclosures were reported.

Funding/Support: This study was supported by grant PCTRB-13-288476 from the Alzheimer’s Association made possible by Part the Cloud (Dr Vossel); grants K23 AG038357 (Dr Vossel), K08 AG058749 (Dr Ranasinghe), F32 AG050434 (Dr Ranasinghe), R01 NS100440 (Dr Nagarajan), RF1 AG062196 (Dr Nagarajan), R01 DC017696 (Dr Nagarajan), P50 AG023501 (Dr Miller), and P01 AG19724 (Dr Miller) from the National Institutes of Health; grant 2015-A-034-FEL from the Larry L. Hillblom Foundation (Dr Ranasinghe), and funding from the Fineberg Foundation (Dr Vossel), the N. Bud and Beverly Grossman Foundation (Dr Vossel), and the S. D. Bechtel, Jr. Foundation (Drs Vossel and Mucke).

Role of the Funder/Sponsor: The funding organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Data Sharing Statement: See Supplement 3.

Additional Contributions: We thank the participants and study partners for their generous time and commitment to this study. Marcos Contreras, PharmD, of the University of California, San Francisco, Drug Product Services Laboratory, and Karen Muir, RPh, of Safeway Compounding Pharmacy created and maintained the blinded randomization scheme. Deborah Barnes, PhD, of the University of California, San Francisco, provided advice on study design. Kirsten Peterson, RPSGT, and Amy Pearson, RPSGT, of the University of Minnesota provided technical assistance with sleep and electroencephalography studies. Sarah Ellis, CCRC, and Deanna Dickens, MD, of United Hospital assisted with setting up the magnetoencephalography plus electroencephalography research studies. Robert Rissman, PhD, of the University of California, San Diego, performed cerebrospinal fluid analysis on 3 samples. Melissa Terpstra, PhD, of the University of Minnesota assisted with setting up the magnetic resonance imaging research protocol. Zuzan Cayci, MD, of the University of Minnesota read the results of amyloid positron emission tomography for 1 participant. None of these individuals received financial compensation for their contributions.