Under-sampling in epilepsy: Limitations of conventional EEG

Maxime O. Baud, Kaspar Schindler, Vikram R. RaoClinical 2020 · DOI

TL;DR

Conventional EEG, typically recorded for short durations, severely under-samples the complex, cyclical nature of epileptic brain activity. This leads to misinterpretations and suboptimal clinical decisions regarding medication, driving, and prognosis. The authors advocate for a paradigm shift towards chronic, continuous EEG monitoring to better understand and manage epilepsy.

What Problem Does This Paper Address?

Epilepsy is clinically defined by recurrent seizures. For decades, electroencephalography (EEG) has been a cornerstone in diagnosing and managing epilepsy, primarily through short-duration recordings (e.g., 20-minute routine EEGs or short inpatient stays). However, recent research, particularly using long-term intracranial EEG (icEEG) from implanted devices, has revealed that epileptic brain activity, including interictal epileptiform activity (IEA) and seizure occurrence, exhibits significant cyclical patterns over various timescales (circadian, multidien, and even circannual). Conventional EEG, due to its limited recording duration, fails to capture these crucial long-term dynamics. This "under-sampling" leads to several critical problems:

  • Misinterpretation of IEA: Short EEG recordings can present a misleading snapshot of IEA, leading clinicians to incorrectly infer worsening or improvement in seizure risk based on transient fluctuations.
  • Suboptimal Treatment Decisions: Decisions about titrating anti-seizure medications (ASMs) are often based on these misinterpreted IEA patterns, potentially leading to over- or under-treatment.
  • Inaccurate Prognostication: The long-term outlook for seizure freedom or relapse is difficult to assess accurately with limited data.
  • Driving Restrictions: The interpretation of EEG findings for driving eligibility can be inconsistent due to the cyclical nature of IEA.

The paper argues that these limitations hinder the effective management of epilepsy and proposes that a move towards continuous, long-term EEG monitoring is essential to unlock a more quantitative and personalized approach to epilepsy care.

Methods

This paper is a review article, synthesizing existing research and proposing a new perspective on epilepsy management. The authors' "method" involves:

  • Reviewing recent discoveries: They highlight findings from studies using months- to years-long intracranial EEG (icEEG) recordings, particularly those from implanted devices like the RNS® System. These studies have uncovered the cyclical nature of IEA and seizures.
  • Critiquing conventional EEG practices: The authors analyze how the limitations of short-duration EEG (routine and inpatient) lead to misinterpretations and flawed clinical decisions.
  • Proposing a paradigm shift: They advocate for the adoption of chronic, continuous EEG monitoring, drawing parallels to other chronic disease management strategies (e.g., continuous glucose monitoring in diabetes).
  • Discussing technological advancements: They touch upon the development of next-generation devices capable of long-term, potentially non-invasive, EEG monitoring.

The core "novelty" lies not in new experimental methods, but in the re-framing of existing data and the critical evaluation of current clinical practices through the lens of newly discovered long-term brain activity cycles.

Datasets

The paper does not introduce new datasets. Instead, it references and discusses findings from studies that utilized:

  • Years-long intracranial EEG (icEEG) recordings: Primarily from patients with drug-resistant focal epilepsy who have implanted devices (e.g., RNS® System). These datasets are crucial for uncovering the multidien and circannual cycles.
  • Routine outpatient EEG recordings: Standard short-duration EEGs used for diagnosis and monitoring.
  • Inpatient EEG recordings: Longer duration EEGs (days) typically performed in epilepsy monitoring units (EMUs).

The authors emphasize that the limitations of the latter two types of recordings stem from their short duration, preventing the observation of the cycles revealed by the icEEG data.

Key Insights and Ideas

  • Epilepsy is a "cyclical disease": Beyond known catamenial epilepsy, seizures and interictal epileptiform activity (IEA) exhibit robust circadian, multidien (e.g., 7-30 days), and even circannual rhythms.
  • Short EEG is a "snapshot" with severe limitations: Routine EEGs are like listening to only three notes of a song; they provide insufficient information to understand the full "musical structure" of epileptic brain activity.
  • IEA fluctuations are key, not absolute counts: The changes in IEA over time, particularly the rise preceding seizures (pro-ictal states), are more informative than the absolute number of discharges in a short recording.
  • Seizure risk is not constant: These cycles define periods of heightened seizure risk, which can be leveraged for better prediction and management.
  • "Subclinical seizures" are questionable: The term "electrographic seizure" is preferred, acknowledging that not all seizures are clinically apparent.
  • Chronic monitoring is the future: Just as continuous glucose monitoring transformed diabetes care, continuous EEG monitoring is proposed as the next frontier for epilepsy management.

Key Results

The paper is a review and perspective piece, so it doesn't present new quantitative results. However, it highlights findings from previous studies:

  • Multidien cycles of IEA exist in ~60% of patients: These cycles are robust and occur independently of circadian rhythms.
  • Seizures tend to occur when IEA is rising: This "pro-ictal" state, lasting 3-7 days, can increase seizure risk by up to 10-fold.
  • IEA vanishing can indicate seizure freedom: Sustained absence of multidien IEA cycles may be a biomarker for remission.
  • Scalp EEG is insensitive: It captures only a fraction of the epileptiform discharges seen with icEEG.
  • Inpatient EEGs are often too short: They typically capture circadian cycles but not the longer multidien cycles.

Limitations and Future Work

Limitations:

  • Data primarily from drug-resistant focal epilepsy: Most of the evidence for long-term cycles comes from patients with implanted devices for focal epilepsy. It is not formally established if generalized epilepsies exhibit the same trends.
  • Scalp EEG sensitivity: It remains unclear how reliably these long-term cycles and their fluctuations can be detected with less invasive scalp EEG.
  • Interpretation of "vanishing IEA": While suggested as a biomarker for remission, this needs further validation.

Future Work:

  • Develop minimally or non-invasive chronic EEG devices: This is crucial for widespread adoption.
  • Validate findings in generalized epilepsies: Extend the understanding of cyclical dynamics to other epilepsy types.
  • Leverage next-generation devices for chronic monitoring: Utilize these devices to capture quantitative biomarkers of disease state.
  • Integrate chronic EEG data with ASMs: Enable a "closed-loop" system for dynamic and personalized treatment titration.
  • Empower patients: Provide patients with direct access to their health information to take a more active role in their management.

Summary updated

Abstract

The cyclical structure of epilepsy was recently (re)-discovered through years-long intracranial electroencephalography (EEG) obtained with implanted devices. In this review, we discuss how new revelations from chronic EEG relate to the practice and interpretation of conventional EEG. We argue for an electrographic definition of seizures and highlight the caveats of counting epileptiform discharges in EEG recordings of short duration. Limitations of conventional EEG have practical implications with regard to titrating anti-seizure medications and allowing patients to drive, and we propose that chronic monitoring of brain activity could greatly improve epilepsy care. An impending paradigm shift in epilepsy will involve using next-generation devices for chronic EEG to leverage known biomarkers of disease state.
BibTeX
@inproceedings{baud2020under,
  author = {Maxime O. Baud and Kaspar Schindler and Vikram R. Rao},
  title = {{Under-sampling in epilepsy: Limitations of conventional EEG}},
  booktitle = {Clinical},
  year = {2020},
  doi = {10.1016/j.cnp.2020.12.002},
}
Figure