How Brainify.AI provided a pharmaceutical partner with population-scale normative QEEG data to transform small-trial EEG observations into interpretable, population-referenced proof-of-mechanism evidence.
100,000+
Clinical EEG subjects
859
GAD patients characterized
14
Biomarkers analyzed
~4 Weeks
Engagement duration
A pharmaceutical company was running an approximately 30-subject Phase 1 study of a novel CNS compound for generalized anxiety disorder (GAD). The compound had a distinctive EEG pharmacodynamic signature, and the study faced a critical interpretive challenge common across early-stage CNS trials:
"Without a larger sample size or an external reference point, it was impossible to ascertain whether observed session-to-session EEG changes were genuine drug effects or merely natural variations between subjects."
Specifically, the client needed answers to questions that a small trial alone cannot address:
Brainify.AI deployed its proprietary Population Ruler Framework — a methodology that overlays clinical trial EEG data onto population-scale normative distributions, providing the statistical grounding needed to interpret small-trial results with the confidence of a 100,000+ subject reference group.
Robust reference distributions for 14 QEEG biomarkers generated from over 100,000 normative subjects and 859 GAD-diagnosed patients — spanning spectral power, hemispheric asymmetry, connectivity, and network topology.
Full summary statistics and probability density functions (KDE) for all features and both populations, establishing quantitative drug-change references including z-scores and Cohen's d effect sizes.
Standardized mean difference (Cohen's d) computed for all 14 features, identifying which biomarkers best distinguish the GAD brain profile from the broader population.
Quantitative inputs for a Batch Screening Method (BSM) to prospectively enrich future trials by selecting patients with the most pronounced QEEG abnormalities.
High-frequency beta (Beta-2, 20–30 Hz) power showed the largest and most consistent GAD vs. normative separation, with large effect sizes (d = 0.50–0.59) in central regions and medium effects (d = 0.33–0.37) frontally — consistent with cortical hyperarousal in anxiety.
Theta power, alpha asymmetry, connectivity, and network features showed negligible population-level GAD separation (d < 0.2). This likely reflects the heterogeneity of the broad clinical comparator — a finding to be tested against healthy controls in the next phase.
Population distributions combined with Phase 1 drug effect data create a quantitative proof-of-mechanism narrative — demonstrating that the compound shifts the GAD brain profile toward the normative range.
Population characterization provides the foundation for a Batch Screening Method, enabling enrichment of future trials by selecting patients with the most pronounced QEEG abnormalities.
The combined dataset supports a multi-pronged publication strategy covering mechanism of action, population-level QEEG characterization, and therapeutic effect magnitude.
Objective neurophysiological evidence anchored against well-characterized population distributions differentiates the clinical narrative from competitors relying solely on subjective symptom scales.
The analysis leveraged Brainify.AI's proprietary QEEG database. Data from 859 GAD-diagnosed patients (primary diagnosis, EEG within −3 to +12 months of diagnosis, 65% female) and approximately 103,000 normative subjects were processed through a rigorous pipeline:
Multiple EEG recordings averaged to reduce within-subject measurement variability
Gaussian Kernel Density Estimation (KDE) for full distribution characterization
Cohen's d computation for standardized population separation quantification
All recordings used 250 Hz sampling, standard 10–20 montage, average reference, with spectral analysis across 1–30 Hz. The 14 biomarkers were selected based on peer-reviewed literature spanning four domains: spectral power, hemispheric asymmetry, functional connectivity, and network topology.