Spectral EEG distributions in generalized anxiety disorder using a large hospital-based clinical EEG database

INTRODUCTION: Electroencephalography in clinical practice allows for single-participant interpretation of dynamic brain function in the context of neuropathology. Although EEG features might vary within psychiatric conditions, laboratory-like population thresholds describing distributions in clinical populations are lacking, and would vary with acquisition, preprocessing, artifact handling, state, montage, and context of use. Larger EEG datasets and biobanks are now improving the ability to separate biological signal from technical and systematic variability, creating the possibility of validated, context-specific EEG biomarkers rather than universal normality thresholds. GAD is a suitable test case because anxiety-related EEG findings are reported in the literature, but their population behavior in real-world clinical EEG data remains unclear. Rather than comparing GAD with healthy controls, the present analysis asks whether GAD is associated with selective shifts in spectral features relative to a broad hospital-based clinical EEG population without recorded GAD.

AIMS: To characterize population distributions of EEG spectral parameters in participants with generalized anxiety disorder compared with a clinical reference population.

METHODS: Participants were selected from a broad clinical population. Cases were a total of 859 participants with an EEG collected within -3 months to +12 months from the date of a primary diagnosis of GAD, and comparators were N=102,908 participants from a broader clinical EEG population without a recorded GAD diagnosis. Six groupings of EEG spectral parameters were calculated: theta (4–8 Hz) and beta-2 (20–30 Hz) power, alpha asymmetry indices, theta coherence, midline theta connectivity, and theta clustering coefficients. Cohen’s d was computed as the standardized mean difference between the GAD group and the clinical reference population using pooled standard deviation.

RESULTS: Effect sizes within the beta-2 power grouping were substantially greater than those observed across the five non-beta-2 groupings. The GAD-vs-non-GAD effect sizes within the beta-2 power grouping were 0.59 and 0.50 for relative and absolute power in the central region, and were 0.37 and 0.33 in the frontal region. Across the non-beta-2 groupings, the largest absolute standardized difference was ≤0.15. Notably, the theta and asymmetry features — despite being frequently cited in GAD literature — show negligible effect sizes in this real-world clinical sample.

CONCLUSIONS: In a large real-world clinical EEG dataset, beta-2 power showed selective distributional separation in GAD, whereas several commonly discussed theta and asymmetry features did not. Increased beta-2 power is consistent with the established neuroscience literature linking cortical hyper-arousal and elevated high-beta-2 oscillations to anxious states. If the GAD-vs-non-GAD contrast were driven mainly by nonspecific acquisition, preprocessing, referral, or database artifacts, one might expect broad displacement across many spectral features. Findings should be interpreted descriptively given hospital-based referral, diagnostic coding, medication exposure, and acquisition heterogeneity. The selective nature of the beta-2 shift, in contrast to negligible shifts across alpha, theta, asymmetry, connectivity and clustering measures, is not readily explained by a purely global database artifact and supports further evaluation of beta-2 power as a candidate GAD-associated EEG feature.

References

No References