Slower alpha rhythm associates with poorer seizure control in epilepsy

Abstract

Objective: Slowing and frontal spread of the alpha rhythm have been reported in multiple epilepsy syndromes. We investigated whether these phenomena are associated with seizure control.

Methods: We prospectively acquired resting-state electroencephalogram (EEG) in 63 patients with focal and idiopathic generalized epilepsy (FE and IGE) and 39 age- and gender-matched healthy subjects (HS). Patients were divided into good and poor (≥4 seizures/12 months) seizure control groups based on self-reports and clinical records. We computed spectral power from 20-sec EEG segments during eyes-closed wakefulness, free of interictal abnormalities, and quantified power in high- and low-alpha bands. Analysis of covariance and post hoc t-tests were used to assess group differences in alpha-power shift across all EEG channels. Permutation-based statistics were used to assess the topography of this shift across the whole scalp.

Results: Compared to HS, patients showed a statistically significant shift of spectral power from high- to low-alpha frequencies (effect size g = 0.78 [95% confidence interval 0.43, 1.20]). This alpha-power shift was driven by patients with poor seizure control in both FE and IGE (g = 1.14, [0.65, 1.74]), and occurred over midline frontal and bilateral occipital regions. IGE exhibited less alpha power shift compared to FE over bilateral frontal regions (g = -1.16 [-0.68, -1.74]). There was no interaction between syndrome and seizure control. Effects were independent of antiepileptic drug load, time of day, or subgroup definitions.

Interpretation: Alpha slowing and anteriorization are a robust finding in patients with epilepsy and might represent a generic indicator of seizure liability.

Figure 1

Alpha power is shifted towards lower frequencies in patients with poor seizure control across clinical syndromes. The three plots in the upper row (A–C) compare the following pairs of power spectra (from left to right): patients versus healthy subjects, patients with poor versus good seizure control (GSC), and patients with focal versus idiopathic generalized epilepsy. Lines indicate group averages and shaded areas 95% confidence intervals (CI). Tick marks and black lines above the x‐axis show frequencies at which power spectra diverge: a shift towards lower alpha frequencies can be appreciated for the whole patient cohort, for poor seizure control, and for focal epilepsy subgroups. Plots (D and E) show the statistical assessment of this observations in terms of the (log‐transformed) alpha‐power shift, that is the ratio of average low‐ to average high alpha‐power. Higher values indicate more low‐alpha power. Dots and error bars represent means ± 95% CI. P‐values are derived from pair‐wise contrasts of two analyses of covariance (see Methods for details). There was a significant difference between healthy subjects and patients, which was driven by poor seizure control patients (P < 0.001, panel D). The power spectrum of focal epilepsy patients was more shifted than the power spectrum of idiopathic generalized epilepsy patients (P = 0.007, panel E, horizontal line). However, there was no syndrome‐by‐seizure control interaction: alpha power in poor seizure control patients was always more shifted than alpha power in GSC patients (P = 0.001), and this occurred in equal measure in both syndrome categories (panel E, vertical line).

Figure 2

Alpha‐power shifts are topographically extended, indicating a forward spread of low‐alpha power. This figure shows topographical maps for three pair‐wise comparisons: panel (A), healthy subjects versus all epilepsy patients, panel (B), good‐seizure control versus poor‐seizure control patients, and panel (C), idiopathic generalized epilepsy versus focal epilepsy patients. The first two columns on the left show raw data: positive values (red) indicate a shift of electroencephalogram (EEG) power towards the low‐alpha band, negative (blue) values a shift towards the high‐alpha band. The third and fourth column show statistical maps, that is P‐values and effect sizes, respectively. Maps of family‐wise error (FWE) corrected P‐values were derived using permutation tests. White areas did not reach the significance threshold. The last column shows effect size maps (Hedges’ g), where g = 0.2 represents a weak, and g = 1.3 a very large effect. Note that all patients present significant shifts of alpha power compared to healthy controls over the entire scalp (panel A), with large effects occipitally, and that this effect is particularly pronounced in poor seizure control patients: very large effects are seen here over bilateral and midline frontal regions (panel B). Finally, t focal epilepsy patients had more frontal alpha‐power shift, that is more low‐alpha power, over frontal regions compared to idiopathic generalized epilepsy patients, (panel C).