糖心vlog

Key Points

Question听 What are the prevalence, strength, and patterns of seizure cycles in focal epilepsy?

Findings听 In this cohort study of 222 patients who received an implantable continuous electroencephalographic recording device, the prevalence was 12% of weak circannual seizure cycles, 60% of moderate multidien seizure cycles, and 89% of moderate circadian seizure cycles. Prevalent cycles were further classified into 5 multidien (7, 15, 20, and 30 days and irregular) and 5 circadian (morning, mid-afternoon, evening, early night, and late night) seizure chronotypes.

Meaning听 The findings establish the high prevalence of circadian and multidien seizure cycles and reveal the existence of distinct seizure chronotypes in focal epilepsy.

Importance听 Focal epilepsy is characterized by the cyclical recurrence of seizures, but, to our knowledge, the prevalence and patterns of seizure cycles are unknown.

Objective听 To establish the prevalence, strength, and temporal patterns of seizure cycles over timescales of hours to years.

Design, Setting, and Participants听 This retrospective cohort study analyzed data from continuous intracranial electroencephalography (cEEG) and seizure diaries collected between January 19, 2004, and May 18, 2018, with durations up to 10 years. A total of 222 adults with medically refractory focal epilepsy were selected from 256 total participants in a clinical trial of an implanted responsive neurostimulation device. Selection was based on availability of cEEG and/or self-reports of disabling seizures.

Exposures听 Antiseizure medications and responsive neurostimulation, based on clinical indications.

Main Outcomes and Measures听 Measures involved (1) self-reported daily seizure counts, (2) cEEG-based hourly counts of electrographic seizures, and (3) detections of interictal epileptiform activity (IEA), which fluctuates in daily (circadian) and multiday (multidien) cycles. Outcomes involved descriptive characteristics of cycles of IEA and seizures: (1) prevalence, defined as the percentage of patients with a given type of seizure cycle; (2) strength, defined as the degree of consistency with which seizures occur at certain phases of an underlying cycle, measured as the phase-locking value (PLV); and (3) seizure chronotypes, defined as patterns in seizure timing evident at the group level.

Results听 Of the 222 participants, 112 (50%) were male, and the median age was 35 years (range, 18-66 years). The prevalence of circannual (approximately 1 year) seizure cycles was 12% (24 of 194), the prevalence of multidien (approximately weekly to approximately monthly) seizure cycles was 60% (112 of 186), and the prevalence of circadian (approximately 24 hours) seizure cycles was 89% (76 of 85). Strengths of circadian (mean [SD] PLV, 0.34 [0.18]) and multidien (mean [SD] PLV, 0.34 [0.17]) seizure cycles were comparable, whereas circannual seizure cycles were weaker (mean [SD] PLV, 0.17 [0.10]). Across individuals, circadian seizure cycles showed 5 peaks: morning, mid-afternoon, evening, early night, and late night. Multidien cycles of IEA showed peak periodicities centered around 7, 15, 20, and 30 days. Independent of multidien period length, self-reported and electrographic seizures consistently occurred during the days-long rising phase of multidien cycles of IEA.

Conclusions and Relevance听 Findings in this large cohort establish the high prevalence of plural seizure cycles and help explain the natural variability in seizure timing. The results have the potential to inform the scheduling of diagnostic studies, the delivery of time-varying therapies, and the design of clinical trials in epilepsy.

Over the centuries, scholars and laymen have sought to explain the periodicity of epileptic seizures by invoking divine¹ or demonic² interventions, celestial motions,³ and hormonal cycles.⁴ In 1881, Gowers⁵ described daily seizure patterns, which he termed nocturnal, diurnal, and diffused. In 1938, Griffiths and Fox⁶ described longer rhythms, ranging from weeks to months, in institutionalized children, men, and women with epilepsy. Today, biological rhythms are known to be ubiquitous,⁷ with myriad manifestations in medicine. Practicing neurologists commonly encounter patients reporting seizures occurring at the same time of the day^6,8,9 or month,^6,10 or even during a particular season,⁶ although such anecdotal clinical observations are seldom quantified.

In the century since initial phenomenological descriptions of seizure cycles, few studies^10-12 have applied rigorous statistics to analyze the cyclical organization of seizure timing, to our knowledge. The advent of inpatient seizure monitoring enabled studies of seizure cycles over several days, necessarily with cross-sectional designs,^13-15 but short recording durations precluded identification of longer cycles. Recent studies avoided this limitation by using online diaries of self-reported seizures collected longitudinally outside the hospital.^16,17 Despite the limited accuracy of such diaries,^18,19 these studies confirmed circadian seizure cycling in many individuals and suggested the existence of longer periodicities,^16,17 albeit with weaker effects.¹⁷ However, critical questions about the prevalence, strength, and patterns of seizure cycles remain unanswered, in part because diary studies could not correlate self-reported seizures with concurrent recordings of brain activity.

Quiz Ref IDIn the last decade, the availability of continuous electroencephalography (cEEG) provided a unique opportunity to study seizure cycles at long timescales with objective data and suitable statistics.^20,21 Using the 2 existing cEEG data sets, NeuroVista²² and NeuroPace,²³ studies^20,24,25 uncovered the striking cyclical organization of interictal epileptiform activity (IEA) and electrographic seizures (ie, those apparent on electroencephalograms) in a limited number of patients. However, without larger, longer studies incorporating patient-reported seizures, the generalizability and clinical relevance of these findings remain unclear. Here, we uncover seizure chronotypes鈥攄iscrete, robust, and prevalent patterns of seizure timing鈥攊n a retrospective data set including up to 10 years of seizure diary data and concurrent cEEG data from individuals participating in a clinical trial of an implanted neurostimulation device for epilepsy.²³

We analyzed retrospective data from a total of 222 out of 256 adults with medically refractory focal epilepsy who participated in clinical trials of the RNS System (NeuroPace Inc), an implanted device that treats seizures with responsive neurostimulation.^23,26 The RNS System provides a limited form of cEEG by continuously monitoring intracranial EEG from the seizure focus or foci and quantifying detections of epileptiform activity (eFigures 1 and 2 in the Supplement). In addition, participants in the clinical trials kept diaries of seizures classified as simple motor, simple other, complex partial, and generalized tonic-clonic, without knowledge of the recorded epileptiform activity. According to the 2017 International League Against Epilepsy classification,²⁷ we considered complex partial and generalized tonic-clonic seizures as the disabling seizures with impaired awareness studied here and excluded patients without disabling seizures. For analyses of seizure cycles at 3 timescales, we used the following selection criteria (eFigure 3 in the Supplement): (1) For circannual seizure cycles, we included 194 patients who kept continuous diaries for 2 years or more showing 24 days or more with disabling seizures, based on the rationale that a uniform distribution over the months of the year would result in 2 seizures per month. (2) For multidien seizure cycles, we included 186 patients with 24 or more self-reported days with disabling seizures over a period with 6 months or more of continuous cEEG count data after discarding the initial segment (median, 131 days) Quiz Ref IDto avoid the 鈥渋mplant effect鈥� (the changes in EEG signal and reduction in seizure frequency that can occur after placement of intracranial electrodes).^22,28,29 (3) For circadian seizure cycles, we included 85 patients with 48 hours or more with nonzero counts of electrographic seizures in 6 months or more of continuous cEEG, based on the rationale that a uniform distribution over the 24-hour clock would result in 2 seizures per hour. Retrospective data analysis was approved by the institutional review boards of participating centers,^23,30,31 and written informed consent was obtained from the participants.

We analyzed the following 3 distinct types of data (eFigure 1 in the Supplement): (1) Self-reported disabling seizures were recorded in participants鈥� diaries as counts per calendar day; timestamps were not recorded, so self-reported seizures were analyzed only at multidien and circannual timescales. (2) Electrographic seizures were recorded by the RNS System with timestamps; owing to the limited storage of electrocorticograms by the device, electrographic seizures were identified through sustained detections of epileptiform activity (termed long episodes) exceeding a clinically prespecified duration, typically 15 to 40 seconds (eFigure 2 in the Supplement). The reliability of long episodes as a proxy for electrographic seizures depends on device detection settings, which are periodically adjusted for clinical purposes. For each patient and each epoch of stable detection settings, a board-certified epileptologist (V.R.R.) evaluated by visual inspection whether 90% or more of long episodes corresponded to electrographic seizures (ie, clear organization and spatiotemporal evolution of epileptiform activity) in the subset of long episodes that had stored electrocorticograms (median, 274 reviewed per patient), as described in detail previously.^20,32 A software tool that uses deep learning to group electrocorticograms based on electrographic feature similarity facilitated this visual analysis.³³ For patients with at least 1 epoch meeting this criterion (162 patients had none), electrographic seizures from a varying number of epochs were aggregated for analysis at circadian and multidien timescales. The frequency of changes in detection settings precluded analysis of electrographic seizures at the circannual timescale. (3) Interictal epileptiform activity was available for all patients and defined as hourly counts of detections of short-lived epileptiform discharges (eFigure 2 in the Supplement).^20,32

Data were preprocessed as described previously,²⁰ with minor adjustments (eMethods in the Supplement). To resolve periodicity in epileptic activity at 3 timescales, a Morlet wavelet transform^20,21 was applied to (1) hourly counts of IEA (circadian: 0.8-1.2 days), (2) daily counts of IEA calculated by summing the hourly counts over a calendar day (multidien: broadband 4-45 days), and (3) detrended timeseries of self-reported seizures (circannual: broadband 90-450 days; eMethods in the Supplement).

Sample size was evaluated by the availability of data for each timescale. Individuals in the cohort can be viewed as independent replications of the statistical tests described here. Circular statistics (CircStat toolbox, Matlab; MathWorks) allow equivalence of 0掳 and 360掳 and can elucidate phase associations between a recurrent event and an underlying cycle. Quiz Ref ID鈥淧hase clustering鈥� reflects skewness of a circular distribution and is quantified by the phase-locking value³⁴ (PLV), where PLV tends to 0 for a uniform distribution of events over phases of a cycle, and the PLV is 1 when all events occur exactly at the same phase (eMethods in the Supplement). This quantification of effect sizes in circular data revealed the strength of diverse seizure cycles; that is, the tendency for seizures to cluster into phases of cycles of hourly or daily IEA, lunar phases, calendar months, weekdays, and clock times. At the individual level, the statistical significance of PLVs was tested against 1000 surrogate time series,³⁵ built by repeated random shuffling of seizure counts and intervals between seizure days.^36,37 This procedure serves as a control, testing the null hypothesis that a single seizure or a group of seizures occurs without association with underlying cycles as it destroys the original sequence of seizure counts while conserving its main internal statistics, such as mean, variance, and tendency for temporal clustering (eMethods in the Supplement).^21,36 To estimate prevalence, we adjusted the P value of significance such that the false discovery rate was 0.1 or less for each tested cycle. P values were from 1-sided tests and results were deemed statistically significant at P鈥�<鈥�.05, adjusted for the false discovery rate.

For individual analyses at the circannual timescale, self-reported seizure time series were cropped to encompass a multiple of 12 months of data, to avoid overrepresentation of certain months when calculating the distribution of seizures in monthly bins (eMethods in the Supplement).

To find recurring cyclical patterns across individuals at the multidien timescale, we used nonnegative matrix factorization,³⁸ an unsupervised soft clustering algorithm (eMethods in the Supplement). We grouped patients into 5 clusters based on individual characteristic periodograms derived from fluctuations in IEA.

For individual analyses at the circadian level, we clustered individual circadian peak times for electrographic seizures using a K-means algorithm and a rank of 5 (eMethods in the Supplement).⁸

At the population level, we estimated 99% CIs around population mean PLVs using bootstrapping.³⁹ The 鈥減eakedness鈥� of the circular distribution of angles of individual mean resultant vectors was measured using a calculation of circular kurtosis,³⁴ which has a value of 1 if all the individual angles coincide with the mean angle and decreases as individual angles diverge (eMethods in the Supplement).

We analyzed cEEG and seizure diary data from a total of 222 participants in the RNS System clinical trials (median age, 35 years [range, 18-66 years]; 112 men [50%] and 110 women [50%]) who had diverse focal epilepsies: 127 mesio-temporal (57.2%), 31 frontal (14.0%), 22 neocortical-temporal (9.9%), 9 parietal (4.1%), 3 occipital (1.4%), and 30 multifocal (13.5%) (eResults and eFigure 3 in the Supplement). The total data amounted to 1118 patient-years of cEEG (median, 5.9 years [range, 6 months to 9.5 years]), 754鈥�108 electrographic seizures (median, 1436 [range, 67-148鈥�073]), and 313鈥�995 self-reported seizures (median, 422 [range, 26-64鈥�882]).

In Figure 1, we show data from an exemplar patient who had cycles in electrographic and self-reported seizures at timescales ranging from hours to years. During 9 years of recording, this patient had a weak circannual cycle, with a peak around 365 days in the corresponding periodogram and a slight increase in seizure rate in the spring (PLV, 0.1; P鈥�=鈥�.002; Figure 1A-C). This patient also had coexisting approximately 7-day multidien cycles and approximately 30-day multidien cycles of IEA that did not consistently align with particular days of the week or month.⁴⁰ Electrographic and self-reported seizures showed preference for the rising phases of IEA cycles over days, with a PLV of 0.40 for electrographic seizures (P鈥�=鈥�.02) and 0.25 for self-reported seizures (P鈥�=鈥�.001; Figure 1D-F). At the circadian timescale, electrographic seizures showed preferential timing around 18:00 (PLV, 0.45; P鈥�=鈥�.001; Figure 1G-I), corresponding in this case to the rising phase of the circadian IEA cycle.

Among included patients, the prevalence of statistically significant seizure cycles was 12% (24 of 194) at the circannual timescale, 60% (112 of 186) at the multidien timescale, and 89% (76 of 85) at the circadian timescale (Figure 2). In addition to their individual statistical significance, we quantified the strength of seizure cycles (ie, circular effect size measured as the PLV; see Methods). At the population level, circannual seizure cycles were weaker (mean [SD] PLV, 0.17 [0.10]) than multidien seizure cycles (mean [SD] PLV, 0.34 [0.17]) and circadian seizure cycles (mean [SD] PLV, 0.34 [0.18]) (Figure 1), which did not differ significantly in strength (eFigure 4 in the Supplement). External cycles of fixed period length, such as calendar days¹⁷ and lunar phases,⁴¹ have been proposed to explain seizure periodicity. However, across patients, only 5% (10 of 186) had weak fluctuations of seizure rates according to days of the week (mean [SD] PLV, 0.19 [0.08]), and none had seizures modulated by days of the month or phases of the moon (Figure 2B-D). We next examined patterns of seizure timing at each timescale.

Among the 24 patients with a significant circannual cycle of self-reported seizures, individuals demonstrated weak to moderate seasonal seizure preference for the fall (Figure 3A), winter (Figure 3B), spring (Figure 3C), or summer (Figure 3D), without a group trend for seasonality (ie, circannual phase). In addition to this supervised approach, we found an approximately 365-day periodic fluctuation in seizure rate, using a wavelet transform on detrended seizure time series (eFigure 5 in the Supplement).

Across the 112 patients with significant multidien cycles of self-reported seizures, an unsupervised pattern recognition method extracted 5 distinct, recurrent features of IEA fluctuations, corresponding to peak periodicity at 7, 15, 20, and 30 days, as well as 1 feature accounting for more irregular periodicity (Figure 4A and B; eFigure 6 in the Supplement). These features represent shared periodicities in multidien cycles of IEA within the whole cohort, but a given patient could have 1 or more periodicities (Figure 4A), independent of sex (eFigure 7 in the Supplement) or seizure focus (eTable in the Supplement). For visualization, we show mean periodograms for patients clustered by dominant periodicity in Figure 4B. To characterize the association between IEA and seizures, we evaluated the tendency of seizures to recur at particular phases of multidien cycles of IEA (ie, phase clustering; see Methods). For most patients, electrographic (Figure 4C) or self-reported (Figure 4D) seizures occurred preferentially on the rising phase, close to the peak, highlighting that IEA increases during days around seizures, independent of a specific multidien period length. This consistent phase association between multidien IEA phases and both electrographic and self-reported seizures was apparent at the individual level (84 of 112 [75.0%] with PLV >0.2; Figure 4D) and across patients, with a high circular kurtosis of 0.58 (Figure 4D) indicating excellent correspondence between fluctuations in electrographic and clinical seizures over days (Figure 4C and D).

Across the 76 patients with significant circadian cycles of electrographic seizures at any localization, we observed 5 main seizure peak times, centered around 00:00 (N1), 03:00 (N2), 09:00 (D1), 14:00 (D2), and 18:00 (D3) (Figure 5A), corroborating historical observations.^6,8,9Quiz Ref ID Individuals with frontal lobe epilepsy tended to have nocturnal seizures (eTable in the Supplement).³² Regardless of peak seizure time, hourly IEA consistently peaked during the night (Figure 5B; eFigure 8 in the Supplement); thus, for a given patient, seizures could be in phase or out of phase with peak circadian IEA (eFigure 9 in the Supplement). Across patients, we found a slight preference of seizures for the rising phase of circadian cycles of IEA (eFigure 9 in the Supplement), although this phase association was less consistent (circular kurtosis鈥�=鈥�0.13) than for multidien cycles of IEA and seizures.

Using the largest existing cEEG data set, collected over years from ambulatory patients with epilepsy living in natural conditions,²³ we identified seizure chronotypes at multiple timescales. Circadian and multidien seizure cycles were equally strong and highly prevalent, whereas circannual seizure cycles were weaker and observed in a minority of patients. Thus, seizure cycles underlie the natural variability of seizure rates observed in clinical practice, and the proposed typology may help personalize care for individuals with focal epilepsy.

Circannual cycles of seizures were found in only 12% of patients, but, to our knowledge, this represents the first quantitative investigation of 鈥渟easonal epilepsy,鈥� a rare seizure chronotype. Circannual modulation of seizure timing might be superimposed on circadian effects and be associated with changes in photoperiod,^40,42 as observed in cluster headache.⁴³

Multidien cycles of seizures were found in 60% of patients, with periods centered at 7, 15, 20, and 30 days, corroborating the historical description of 鈥渞hythm in epilepsy鈥� and representing prevalent multidien seizure chronotypes.⁶ Despite greater awareness of catamenial and circadian influences on seizures in current practice,⁴⁴ we found that multidien cycles were as strong modulators of seizure timing and equally prevalent in men and women. Unlike the circadian period, multidien period length characteristically varies from cycle to cycle within individuals, and underlying periodicity was often apparent only in the alignment between seizures and IEA.²⁰ To account for this, we used spectral-domain methods that are more flexible than time-domain methods for elucidating cycles in epilepsy.^20,21 Here, seeking seizure cycles with fixed periodicity (days of the week,¹⁷ days of the month,¹⁷ and lunar phases⁴¹) failed to reveal the full extent of multidien cycles. As a result, and despite stringent statistical criteria, the prevalence of multidien cycles was here 2 to 4 times higher than reported (15%-30%) in calendar-based studies,^16,17 underscoring the advantage of tracking phases of multidien cycles of epileptic brain activity that are frequently nonstationary.²¹ This suggests that multidien cycles in epilepsy are free running; that is, they have an endogenous generator and are not paced by the environment.⁴⁰

Quiz Ref IDCircadian cycles of seizures were as prevalent in our study (89%) as previously reported.^16,17,32 Our description of circadian patterns provides contemporary analogs to Gowers鈥�⁵ seizure chronotypes and corroborates historical descriptions of 5 peak seizure times as follows^8,9: early and late peaks for nocturnal seizures and early morning, midafternoon, and evening peaks for diurnal seizures. In our cohort, an early morning peak occurred in only a few patients, possibly owing to the lack of included patients with generalized epilepsy.

Finally, our findings shed light on the association between IEA and seizures. At the timescale of hours, IEA consistently peaked during the night, but peak seizure times were dispersed around the clock, resulting in a loose circadian phase association between IEA and seizures.^20,24 In contrast, at the timescale of days, self-reported and electrographic seizures consistently occurred on the rising phase of multidien cycles of IEA, corroborating results from other human^20,25 and animal⁴⁵ studies. Thus, unique mechanisms may underlie multidien cycles in epilepsy. Sex hormones fluctuate with the menstrual cycle in women and possibly with a 20-day and 30-day periodicity in men,⁴⁶ and cortisol may have a 7-day cycle,⁴⁷ suggesting a possible hormonal basis for multidien cycles of seizures and IEA.¹³

This work has some strengths, complementing and extending previous studies^{16,17,20,24,25} in several ways. First, we analyzed separate data sets of electrographic and self-reported seizures, and the inclusion of the latter increases clinical relevance. Second, these well-curated data sets spanned an unprecedented number of individual cycles across many participants, enabling estimation of the prevalence of distinct seizure chronotypes. Third, we adopted a conservative statistical approach, based on surrogate time series testing,³⁵ that can elucidate cycles beyond the well-known clinical phenomena of within-day or within-hour temporal clustering,⁴⁸ as demonstrated by our recent computational study.²¹ Fourth, in addition to their statistical significance, we estimated the strength of individual seizure cycles, from weak modulation (slightly skewed circular distributions) to strong effects (seizures occurring over a narrow range of phases) that are likely to be clinically relevant.

This study also has some limitations. First, although our cohort included diverse focal epilepsies, all participants were treated with an implanted neurostimulation device, which limits generalizability of the results. However, although precise timestamps of delivered stimulations were not available for analysis in our data set, neurostimulation does not appear to cause the observed cycles, which persist when the stimulator is turned off²⁰; in addition, similar phenomena are found in rats⁴⁵ and dogs with epilepsy,⁴⁹ as well as in humans who did not receive neurostimulation.^6,8,9,24,25 Second, we included long-term self-reports of seizures, which remain the standard criterion in epilepsy trials but may be inaccurate.^18,19 This potential inaccuracy and the fact that our methods may not fully resolve some bimodal circular distributions or times when multidien IEA cycles vanish could have been associated with the underestimation of the prevalence and strength of seizure cycles. However, we found excellent correlation between IEA, electrographic seizures, and self-reported seizures over long timescales. Additional factors, such as sleep-wake cycles and sleep homeostasis, may play a role in seizure timing but were not examined here.

Our findings have several clinical implications. First, in a given patient, apparent patterns of seizure timing often defy explanation, leading to spurious correlations with behavioral or environmental factors.⁴⁰ Our results help demystify fluctuations in seizure rates and refocus the dialogue between clinicians and patients on objective metrics associated with long-term disease course. Second, diagnostic studies intended to capture seizures, such as inpatient video EEG, might have a higher yield if they are scheduled based on times of highest seizure likelihood.⁵⁰ Similarly, knowledge of seizure cycles could directly inform chronotherapy, the time-varying adjustment of medications, and other interventions to reduce seizures.⁵¹ Third, given the variability of seizure rates across and within patients, short-term clinical trials in epilepsy encounter sampling issues that limit interpretation of the results.^52,53 The design of future clinical trials in epilepsy and efforts toward seizure forecasting⁵⁴ will benefit from accounting for seizure cycles.

The chronobiology of epilepsy is undergoing a renaissance fueled by recent technological developments. Broader use of minimally invasive devices⁵⁵ capable of monitoring cycles in epileptic brain activity could help personalize diagnosis and treatment in epilepsy, and long-term recordings from such devices may help elucidate the mechanistic basis of seizure chronotypes.

Accepted for Publication: December 11, 2020.

Published Online: February 8, 2021. doi:10.1001/jamaneurol.2020.5370

Corresponding Author: Maxime O. Baud, MD, PhD, Sleep-Wake-Epilepsy Center, NeuroTec, Center for Experimental Neurology, Department of Neurology, Inselspital Bern, University Hospital, University of Bern, Bern, Switzerland (maxime.baud.neuro@gmail.com); Vikram R. Rao, MD, PhD, Department of Neurology, University of California San Francisco Epilepsy Center, 400 Parnassus Ave, 8th Floor, San Francisco, CA 94143 (vikram.rao@ucsf.edu).

Author Contributions: Drs Rao and Baud 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. Drs Rao and Baud are co鈥搒enior authors and contributed equally to this work.

Concept and design: Leguia, Andrzejak, Tcheng, Rao, Baud.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Leguia, Andrzejak, Tcheng, Rao, Baud.

Critical revision of the manuscript for important intellectual content: Andrzejak, Rummel, Fan, Mirro, Tcheng, Rao, Baud.

Statistical analysis: Leguia, Andrzejak, Rummel, Fan, Tcheng, Baud.

Obtained funding: Baud.

Administrative, technical, or material support: Mirro, Rao, Baud.

Supervision: Andrzejak, Rao, Baud.

Conflict of Interest Disclosures: Ms Mirro and Dr Tcheng are employees of NeuroPace Inc and report having equity ownership and receiving stock options. Dr Rao reported receiving grants and personal fees from NeuroPace Inc during the conduct of the study. Dr Baud reported receiving grants from Swiss National Foundation for Science, Wyss Center for Bio and Neuroengineering, and Velux Stiftung; personal fees from Wyss Center for Bio and Neuroengineering as a part-time employee during the conduct of the study; and having a patent for a brain monitoring device pending under PCT 62665486. No other disclosures were reported.

Funding/Support: Dr Leguia is supported through a grant for a postdoctoral fellowship from the Wyss Center for Bio and Neuroengineering. Dr Rao is supported by the Ernest Gallo Foundation Distinguished Professorship in Neurology at the University of California, San Francisco. Dr Baud is supported by an Ambizione grant (PZ00P3_179929) from the Swiss National Science Foundation and by a grant from the Velux Stiftung (1232).

Role of the Funder/Sponsor: The funding sources 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.

Disclaimer: The contents do not represent the views of the US Department of Veterans Affairs or the US government.