EEG Slow-Wave Patterns Track Oxcarbazepine Response in Focal Epilepsy

Roughly one in three people with focal epilepsy does not respond well to the first antiseizure medication they are prescribed, and clinicians currently have no objective way to know in advance who will benefit. A retrospective study published in Acta Epileptologica now reports that quantitative resting-state electroencephalography (EEG) features recorded during ongoing oxcarbazepine monotherapy differ between patients who became seizure-free and those who did not. In 27 drug-naïve patients with focal epilepsy, the group that continued to have seizures showed significantly higher theta-band relative power over frontal and central electrodes and stronger delta-band functional connectivity across frontal–central networks at follow-up, after correction for multiple comparisons. The findings suggest that persistent low-frequency synchronization may reflect ongoing network instability and reduced treatment efficacy, although the study design and sample size mean these are candidate response-associated features rather than validated predictors.

The research team, led by Shupeng Cheng, Wenkang Li and Jing Ning with senior authors Rong Rong and Xiaoshan Wang, set out to address a well-recognized gap in epilepsy care. Oxcarbazepine is a second-generation sodium-channel blocker widely used as first-line monotherapy for focal epilepsy, which accounts for approximately 60 percent of all patients with the disorder. Yet interindividual variability in treatment response is substantial: many patients fail to achieve complete seizure control on their initial regimen and may eventually develop drug-resistant epilepsy. Identifying biomarkers of oxcarbazepine efficacy is therefore considered clinically important for individualized treatment decisions. Because conventional EEG interpretation relies on visual inspection of epileptiform discharges, which cannot capture finer dynamic changes in brain network states, the authors turned to quantitative approaches—power spectral density analysis, which measures the distribution of oscillatory power across frequency bands, and functional connectivity analysis, which reveals coordinated activity between brain regions.

The study included 27 patients with focal epilepsy who had never previously taken antiseizure medications, started oxcarbazepine monotherapy at or after their first visit, continued it for at least six months, and completed long-term EEG recording before treatment. Each patient underwent resting-state EEG twice: once before oxcarbazepine initiation and again approximately one year into monotherapy, with the follow-up recording obtained while patients remained on the drug. Nineteen electrodes arranged according to the international 10–20 system captured data at sampling rates of 256 to 512 Hz. Preprocessing involved resampling, detrending, a 50 Hz notch filter to remove line noise, a 0.5–70 Hz band-pass filter, and independent component analysis to remove ocular, muscular and other artifacts. For each subject, the researchers selected a 60-second interictal resting segment free of epileptiform discharges, identified jointly by two experienced electrophysiologists with video recordings.

Spectral analysis used Welch’s method with 5-second windows and 50 percent overlap to estimate power spectral density, which was then converted to relative power by scaling each frequency band against total power across the full spectrum. Functional connectivity was quantified with amplitude envelope correlation with correction (AEC-C), a leakage-reduced measure chosen because conventional coherence- or phase-based metrics can spuriously inflate zero-lag coupling through volume conduction and field spread. The result was a 19-by-19 connectivity matrix for each canonical frequency band—delta (1–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), beta (12–30 Hz) and gamma (30–50 Hz)—for every patient at both time points. Group comparisons employed nonparametric statistics appropriate to the small sample and non-normal distributions: false discovery rate correction for power data and the network-based statistic (NBS) with 5,000 permutations for connectivity matrices, applied at two initial thresholds (t = 2.787, P < 0.01 and t = 3.725, P < 0.001) with cluster-level significance at P < 0.05.

Patients were classified as seizure-free (SF) or not seizure-free (NSF) based on electronic health records from follow-up visits occurring 6 to 24 months after oxcarbazepine initiation. The cohort comprised 15 NSF and 12 SF patients. Baseline characteristics were well matched: sex distributions, median ages (23.0 years in NSF versus 20.5 years in SF), the interval from seizure onset to treatment, the proportion with temporal lobe epilepsy (33 percent overall) and MRI abnormalities (26 percent), and follow-up duration all showed no significant between-group differences. This comparability strengthens the interpretation that the EEG differences observed later relate to treatment outcome rather than to demographic or clinical confounders measured at baseline.

The most robust spectral finding emerged in the theta band at the on-treatment follow-up recording. Patients in the NSF group exhibited significantly higher relative theta power than the SF group at ten frontal and central electrodes—Fp1, Fp2, F3, F4, C3, C4, F7, F8, Fz and Cz—and these differences survived false discovery rate correction. In other bands, uncorrected between-group differences appeared at scattered electrodes (P4 in delta, Cz and P4 in alpha, O2 in beta, and bilateral posterior temporal and occipital electrodes in gamma), but none remained statistically significant after correction. Notably, the interaction effect between treatment and clinical response—the formal test of whether the two groups changed differently over time—did not remain significant after multiple-comparison correction, a caveat the authors acknowledge. The cross-sectional pattern nevertheless indicates that seizure freedom on oxcarbazepine was associated with a relative attenuation of frontal–central slow-frequency oscillatory activity, whereas persistent elevation of theta power characterized the non-seizure-free group.

Functional connectivity results pointed in the same slow-frequency direction. At the P < 0.01 threshold, NBS revealed significantly stronger delta-band connectivity in the NSF group across multiple frontal and central nodes, including Fp1, Fp2, F3, F4, C3, C4, F7, F8, Fz and Cz. Under the stricter P < 0.001 threshold, a smaller but more robust difference persisted: connectivity between the left parietal electrode P3 and the left anterior temporal electrode F7 was significantly stronger in the NSF group. No significant between-group connectivity differences were found in other frequency bands after correction, and the connectivity interaction effect also failed to reach statistical significance. The authors interpret the persistent delta hyperconnectivity in non-responders as evidence that oxcarbazepine failed to dismantle the abnormal network supporting seizure generation and propagation in those patients, whereas reduced slow-wave connectivity in responders suggests successful attenuation of pathological neuronal synchronization.

These results fit with prior pharmaco-EEG literature. Previous studies have reported enhanced low-frequency power in the background EEG of people with epilepsy and attenuation of such slowing under antiseizure medication treatment. Using standardized low-resolution brain electromagnetic tomography, one prior study found that oxcarbazepine significantly decreased background low-frequency activity in left frontotemporal and limbic regions in children with benign epilepsy with centrotemporal spikes after six months. In the present cohort, however, the researchers did not observe a statistically significant pre-to-post reduction in low-frequency power across the entire group after correction, which they attribute to the limited sample size and inter-individual variability. Theta-band activity is physiologically linked to cognitive functions, and its pathological elevation is commonly viewed as a marker of cortical dysfunction; in drug-resistant temporal lobe epilepsy, executive control deficits have been associated with altered theta-band connectivity.

The study carries important limitations that temper any clinical application. The sample of 27 patients is small, and although the main findings withstood multiple-comparison correction, they have not been validated in an independent cohort. Follow-up duration varied across patients, raising the possibility of misclassification in a small number of seizure-free patients with shorter observation periods. Although all follow-up EEGs were obtained during continuous monotherapy, differences in treatment duration, medication dose, treatment stage and natural disease course may still have influenced the EEG measures. The authors also note that they did not perform subgroup analyses by seizure focus origin, and that the single analysis pipeline—chosen for stability and reproducibility on 19-channel routine EEG—leaves multi-method and source-space comparisons for future work. The Group × Time interaction results, which did not survive correction, further underscore that the reported differences are cross-sectional associations at follow-up rather than demonstrated longitudinal treatment effects.

Within those constraints, the study offers a concrete electrophysiological picture of how oxcarbazepine response differs across patients. Seizure-free patients showed relatively attenuated slow-frequency power and connectivity, while non-responders exhibited persistent slow-wave synchronization and hyperconnectivity over frontal, central and left parieto-temporal pathways. The authors propose that follow-up theta power over frontal and central regions and delta-band connectivity monitoring could serve as objective EEG features associated with treatment response, potentially helping clinicians identify insufficient response and adjust regimens earlier. They emphasize, however, that these are promising candidate features requiring further investigation, ideally in larger, prospective cohorts with standardized follow-up and, potentially, multimodal imaging or machine learning approaches to improve outcome prediction. For now, the study adds to a growing body of pharmaco-EEG evidence suggesting that the brain’s slow-frequency networks provide a measurable window into whether an antiseizure medication is stabilizing pathological activity or failing to do so.

Subject of Research: Resting-state EEG power spectral and functional connectivity biomarkers associated with oxcarbazepine treatment response in focal epilepsy

Article Title: Efficacy of oxcarbazepine in treating focal epilepsy based on resting-state EEG functional connectivity and power spectral analyses

Article References: Cheng, S., Li, W., Ning, J., Wang, Y., Rong, R., & Wang, X. (2026). Efficacy of oxcarbazepine in treating focal epilepsy based on resting-state EEG functional connectivity and power spectral analyses. Acta Epileptologica, 8(1), Article 27. https://doi.org/10.1186/s42494-026-00270-6

Image Credits: AI Generated

DOI: 10.1186/s42494-026-00270-6

Keywords: focal epilepsy, oxcarbazepine, EEG, functional connectivity, power spectral density, theta band, delta band, biomarkers, antiseizure medication, pharmaco-EEG, network-based statistic, retrospective study