# Which biomarkers are most commonly used for seizures associated with Lennox-Gastaut syndrome?

## The most frequent biomarkers for Lennox-Gastaut syndrome seizures are electroencephalographic measurements, specifically SSW and GPFA patterns.

# Abstract

Electroencephalography metrics appear most frequently for seizures in Lennox-Gastaut syndrome. Ten studies report slow spike‐wave (SSW) patterns, and seven document generalized paroxysmal fast activity (GPFA); other EEG features (such as gamma rhythms, paroxysmal fast activity, and network indices) also support diagnosis and treatment assessment. Standard, video, sleep, intracranial, and automated EEG methods serve as the primary detection techniques.

Imaging biomarkers (MRI, fMRI, PET), serum and cerebrospinal fluid markers, and measures of autonomic or cortical excitability are used less often. Studies employing these varied approaches link EEG findings to seizure type, network mapping, and treatment response, thereby establishing SSW and GPFA as the most common biomarkers in this context.

## Methods

We analyzed 34 sources from an initial pool of 998, using 6 screening criteria. Each paper was reviewed for 5 key aspects that mattered most to the research question.

Papers identified with Elicit search

n = 998

Papers screened using: Population - LGS Diagnosis, Biomarker Focus, Study Type and Measurements, Temporal Association, Sample Size, Study Population Type

n = 998

Papers screened out

n = 964

Papers included for extraction

n = 34

## Paper search

Using your research question “Which biomarkers are most commonly used for seizures associated with Lennox-Gastaut syndrome?”, we [searched](https://support.elicit.com/en/articles/553025) across over 126 million academic papers from the [Semantic Scholar](https://www.semanticscholar.org/) corpus. We retrieved the 998 papers most relevant to the query.

## Screening

We screened in sources based on their abstracts that met these criteria:

- **Population - LGS Diagnosis**: Does the study include patients with confirmed Lennox-Gastaut syndrome (LGS) who can be analyzed separately from patients with other epilepsy syndromes?
- **Biomarker Focus**: Does the study examine biological markers (molecular, cellular, imaging, or electrophysiological) in relation to seizures in LGS?
- **Study Type and Measurements**: Is this either a primary research study or a systematic review/meta-analysis that reports quantitative biomarker measurements?
- **Temporal Association**: Does the study demonstrate a clear temporal relationship between biomarker measurements and seizure activity?
- **Sample Size**: Does the study include 5 or more subjects?
- **Study Population Type**: Was the study conducted in human subjects?

We considered all screening questions together and made a holistic judgement about whether to screen in each paper.

## Data extraction

We asked a large language model to extract each data column below from each paper. We gave the model the extraction instructions shown below for each column.

- **Study Design**: Identify the type of study design used:
    - Specify whether it is an observational study (e.g., cross-sectional, cohort, case-control)
    - If an experimental study, note the specific type (e.g., randomized controlled trial, interventional study)
    - If multiple study designs are present, list all types
    - If the design is not clearly stated, write “Not clearly specified”
    
- **Biomarkers Investigated**: List ALL biomarkers examined in the study related to Lennox-Gastaut syndrome:
    - Include full names and abbreviations
    - Specify the biological source of the biomarker (e.g., serum, cerebrospinal fluid)
    - Note the measurement method used (e.g., ELISA, immunoassay)
    - If multiple biomarkers were studied, list them in order of appearance in the text
    
- **Participant Demographics**: Extract the following participant information:
    - Total number of participants with Lennox-Gastaut syndrome
    - Age range or mean age
    - Gender distribution (if reported)
    - Inclusion/exclusion criteria
    - Any specific subgroups within the LGS population
    
- **Biomarker Findings**: For each biomarker investigated:
    - Report statistical significance of findings
    - Note any correlations with clinical features (e.g., seizure severity, EEG characteristics)
    - Include specific numerical values (mean, standard deviation, p-values)
    - Highlight any comparative findings (e.g., differences from control groups)
    
- **Clinical Implications**: Extract the authors’ conclusions about:
    - Potential diagnostic or prognostic value of identified biomarkers
    - Suggested clinical significance of biomarker findings
    - Any recommendations for future research

# Results

## Characteristics of Included Studies

| Study                   | Study Design                          | Population Size | Biomarker Types Studied                       | Primary Outcomes                                          | Full text retrieved |
|-------------------------|---------------------------------------|-----------------|------------------------------------------------|----------------------------------------------------------|--------------------|
| Romero Milà et al., 2025| Observational cohort study            | 15              | Electroencephalography (EEG) functional connectivity | Evolution from infantile epileptic spasms syndrome (IESS) to Lennox-Gastaut syndrome (LGS), treatment response | No                 |
| Kobayashi et al., 2009  | Observational (cross-sectional)      | 20              | EEG gamma rhythms                               | Pathophysiology of tonic seizures                          | No                 |
| Bare et al., 1998      | Observational (cross-sectional)      | 38              | EEG (video monitoring)                          | Identification of atypical absence seizures                 | No                 |
| Hödl et al., 2021      | Observational (cross-sectional)      | 7               | Heart rate variability (HRV)                    | Autonomic imbalance, vagus nerve stimulation (VNS) response| No                 |
| Deering et al., 2024   | Retrospective observational cohort    | 33              | EEG slow spike-wave (SSW), clinical markers   | Seizure-free periods, SSW as early biomarker               | No                 |
| Dalic et al., 2022a    | Randomized controlled trial          | 17              | EEG generalized paroxysmal fast activity (GPFA), electrographic seizures | Deep brain stimulation (DBS) response, seizure frequency    | No                 |
| Cho et al., 2025       | Observational (cohort)                | 14              | Stereoelectroencephalography (SEEG) indices (epileptogenicity index [EI], connectivity entropy index [CEI], spikes, high-frequency oscillations [HFOs]) | Epileptogenic zone network (EZN) mapping, surgical outcome | No                 |
| Zhou et al., 2025      | Observational (retrospective)        | Not reported in abstract | EEG GPFA, mutual information (MI)             | Steroid response, network topology                          | No                 |
| Bullinger et al., 2025  | Experimental (clinical trial)       | Not reported in abstract | Thalamic EEG (ictal/interictal)                 | Ictal/interictal patterns, closed-loop stimulation          | No                 |
| Yagi, 2015             | Observational (retrospective cohort) | Not reported in abstract | EEG (polysomnography)                           | Seizure types, sleep effects, pathophysiology               | No                 |

### Summary of Study Characteristics:
- **Study design**:
  - 25 studies were observational (cohort, cross-sectional, case-control, retrospective, longitudinal)
  - 4 studies were interventional (randomized controlled trial, clinical trial, DBS trial, interventional case series)
  - 2 studies were systematic reviews
  - 3 studies did not clearly specify study design  
- **Biomarker types studied (some studies used more than one type)**:
  - 26 studies: EEG-based biomarkers (including scalp EEG, SEEG, intracranial EEG, EEG-fMRI, and related indices)
  - 8 studies: Imaging-based biomarkers (MRI, fMRI, PET, CT, US)
  - 3 studies: Blood/serum/CSF-based biomarkers (serum glial markers, P-glycoprotein, CSF IgG)
  - 1 study: Genetic biomarkers (exome sequencing)
  - 1 study: Heart rate variability/autonomic biomarkers
  - 1 study: TMS (cortical excitability) as a biomarker
- **Primary outcomes (some studies had multiple outcomes)**:
  - 12 studies: Treatment response (surgical, DBS, VNS, steroid, IVIG, drug resistance)
  - 12 studies: Network mapping or EEG-imaging correlation
  - 11 studies: Seizure types, features, or phenotyping
  - 6 studies: Biomarker validation or method development
  - 5 studies: Pathophysiology or mechanism
  - 2 studies: Prognosis or cluster analysis
  - 2 studies: Neuroinflammation or genotype-phenotype relationships

We did not find mention of missing biomarker type or primary outcome information in the available abstracts or full texts.

## Primary Biomarker Categories

### EEG-Based Biomarkers

| Study                   | Biomarker Pattern                       | Detection Method                       | Clinical Validation Status     | Frequency of Use |
|-------------------------|----------------------------------------|---------------------------------------|-------------------------------|------------------|
| Romero Milà et al., 2025| Functional connectivity                 | Cross-correlation EEG                 | Correlates with LGS evolution, treatment response | Moderate         |
| Kobayashi et al., 2009  | Gamma rhythms                          | Spectral EEG                          | Pathophysiology of tonic seizures | Moderate         |
| Bare et al., 1998      | Atypical absence, SSW                  | Video EEG                             | Diagnostic accuracy         | High             |
| Deering et al., 2024   | SSW                                    | EEG                                   | Early biomarker, predictive | High             |
| Dalic et al., 2022a    | GPFA, electrographic seizures           | Sleep EEG                             | DBS response prediction    | High             |
| Cho et al., 2025       | EI, CEI, spikes, HFOs                  | SEEG                                  | EZN mapping, surgical outcome | Moderate        |
| Zhou et al., 2025      | GPFA, MI                               | Scalp EEG                            | Steroid response prediction  | Moderate         |
| Bullinger et al., 2025  | Ictal/interictal patterns              | Thalamic EEG                         | Closed-loop stimulation      | Moderate         |
| Yagi, 2015             | SSW, fast rhythms                      | Polysomnography                      | Seizure type, sleep effects | High             |
| Dalic et al., 2022b    | GPFA                                   | EEG                                   | DBS response                  | High             |

### Key findings from included studies:
- **Most frequently reported patterns**: Among the included studies, slow spike-wave (SSW) was the most frequently reported biomarker pattern (10 studies). Generalized paroxysmal fast activity (GPFA) was reported in 7 studies, and paroxysmal fast activity (PFA) in 4 studies. Other patterns (spike-wave, fast rhythms, polyspike, atypical absence, network or waveform features) were each reported in 1–2 studies.
- **Detection methods**: Standard EEG was the most common detection method (9 studies). Other methods included video EEG (2), sleep EEG (1), scalp EEG (1), SEEG (1), thalamic EEG (1), polysomnography (1), automated EEG (1), time-frequency EEG (1), intracranial EEG (1), s-EPIC (automated, 1), EEG-fMRI (1), cross-correlation EEG (1), and spectral EEG (1).
- **Frequency of use**: 11 studies were categorized as “High”, 10 as “Moderate”, and 2 as “Low” for frequency of use.

### Neurophysiological Biomarkers

| Study                   | Biomarker                               | Detection Method                       | Clinical Validation Status     | Frequency of Use |
|-------------------------|----------------------------------------|---------------------------------------|-------------------------------|------------------|
| Hödl et al., 2021      | Heart rate variability (high-frequency power) | Video-EEG, HRV analysis               | Autonomic imbalance, sudden unexpected death in epilepsy (SUDEP) risk | Low               |
| Bullinger et al., 2025  | Thalamic EEG patterns                  | Intracranial EEG                     | Closed-loop stimulation      | Moderate         |
| Badawy et al., 2012    | Cortical excitability                   | Transcranial magnetic stimulation (TMS) | Pathophysiology            | Low               |
| Warren et al., 2024a   | Thalamocortical activity              | Intracranial EEG                     | Network mapping              | Moderate         |
| Warren et al., 2024b   | Thalamocortical activity              | Intracranial EEG                     | Network mapping              | Moderate         |

### Key findings from included studies:
- **Most frequently studied biomarker**: Thalamocortical activity (2 studies).
- **Other biomarkers**: Heart rate variability (high-frequency power), thalamic EEG patterns, and cortical excitability were each studied in 1 study.
- **Detection methods**: Intracranial EEG was the most common (3 studies); video-EEG with HRV analysis and TMS were each used in 1 study.
- **Clinical context**: Network mapping was the most common clinical context (2 studies); autonomic imbalance/SUDEP risk, closed-loop stimulation, and pathophysiology were each reported in 1 study.
- **Frequency of use**: Moderate in 3 studies; low in 2 studies.

### Molecular and Other Biomarkers

| Study                   | Biomarker                               | Detection Method                       | Clinical Validation Status     | Frequency of Use |
|-------------------------|----------------------------------------|---------------------------------------|-------------------------------|------------------|
| Cherkezzade et al., 2024| GFAP, HMGB1, CHI3L1, sCD163, TREM2    | Serum enzyme-linked immunosorbent assay (ELISA) | Neuroinflammation, prognosis | Low               |
| Kumar et al., 2015     | P-glycoprotein                        | PBMC assay                          | Drug resistance              | Low               |
| Espinosa Zacarías et al., 2002 | CSF immunoglobulin G (IgG)            | CSF analysis                         | IVIG response                | Low               |
| Widdess-Walsh et al., 2013 | De novo mutations, chromosome analysis | Blood, exome sequencing               | Genetic etiology             | Low               |
| Ennaim et al., 2013    | Brain atrophy (voxel-based morphometry)| MRI                                    | Network mapping              | Moderate         |
| Balfroid et al., 2023   | FDG-PET hypometabolism                  | PET                                 | Network mapping, cognitive status | Moderate         |
| Alanazi et al., 2022    | Centromedian nucleus (CMN) activity, MRI | Imaging, EEG                           | DBS targeting                | Moderate         |
| Micheu et al., 2012     | CT, MRI, US                            | Imaging                               | Early diagnosis               | Low               |
| Pedersen et al., 2015    | fMRI network properties                 | fMRI                                  | Network mapping               | Low               |

### Key findings from included studies:
- **Molecular biomarkers**: 4 studies (serum glial markers, P-glycoprotein, CSF IgG, genetic sequencing).
- **Imaging biomarkers**: 5 studies (MRI, PET, CT, US, fMRI).
- **Detection methods**: Blood/serum-based (3 studies), CSF-based (1), genetic sequencing (1), imaging-based (6), EEG-based (1).
- **Clinical purposes**: Network mapping (3 studies); neuroinflammation/prognosis, drug resistance, IVIG response, genetic etiology, cognitive status, DBS targeting, and early diagnosis (each in 1 study).
- **Frequency of use**: Low in 6 studies; moderate in 3 studies.

## Clinical Applications and Validation

### Diagnostic Applications
- EEG-based biomarkers: Slow spike-wave (SSW) and generalized paroxysmal fast activity (GPFA) were reported as central to the diagnosis of Lennox-Gastaut syndrome in several included studies, with high sensitivity and specificity reported in those studies, though most had small sample sizes or limited data.
- Video-EEG monitoring: Reported as essential for distinguishing seizure types, especially atypical absence seizures.
- Imaging biomarkers: MRI and PET were used as adjuncts for network mapping and identifying structural etiologies.

### Treatment Monitoring
- EEG changes and treatment response: Changes in GPFA and EEG functional connectivity were associated with treatment response, particularly in the context of neuromodulation (deep brain stimulation, vagus nerve stimulation) and steroid therapy.
- Automated EEG analysis: Automated EEG analysis and network biomarkers were reported to allow for more rapid and objective assessment of treatment efficacy.
- Heart rate variability and thalamic electrophysiology: These were explored for monitoring autonomic and network responses to therapy, but clinical utility was limited in the included studies.

### Prognostic Value
- EEG features and outcomes: Longer SSW discharges and disorganized EEG background were associated with poorer outcomes in the included studies.
- Serum glial markers and P-glycoprotein: These may have prognostic value for disease severity and drug resistance, though evidence was preliminary and based on small samples.
- Imaging biomarkers: Frontoparietal hypometabolism and brain atrophy were reported as potential predictors of cognitive outcomes and for guiding intervention planning.

## Summary
- **EEG-based biomarkers**: Slow spike-wave (SSW) and generalized paroxysmal fast activity (GPFA) were the most commonly reported and validated biomarkers for seizures associated with Lennox-Gastaut syndrome in the included studies. These biomarkers formed the foundation for diagnosis and monitoring.
- **Imaging and molecular biomarkers**: These are emerging as important adjuncts, especially for network mapping and understanding disease mechanisms, but require further validation for routine clinical use.
- **Trends**: The field is moving toward quantitative, automated, and network-based approaches, but the evidence base is limited by small sample sizes, heterogeneity, and incomplete reporting in the included studies.
- **Evidence limitations**: Most studies had small sample sizes, and many relied on retrospective or cross-sectional designs. Only a few systematic reviews or larger cohorts were included.