Electroencephalographic gamma correlates with pain intensity attenuation in a prospective longitudinal study using an AI-powered digital device for pain self-management

Chronic low back or neck pain is one of the most common neurological disorders in the US. Electroencephalogram (EEG) is a leading noninvasive approach to identify biomarkers for chronic pain, but its validation in clinical settings remains limited by methodological challenges, including variable feature selections, subjective removal of EEG artifacts, and reliance on population-level analyses. In this single center, prospective, longitudinal study, we investigated the relationships among EEG features, patient-reported clinical outcomes via validated Patient-Reported Outcome Measures (PROMS) questionnaires, and daily self-management of pain-associated physical activity via a newly developed AI-powered mobile application Paindrainer PDCC (Paindrainer AB, Lund, Sweden) in patients with chronic pain undergoing standard care. In total, twenty patients with chronic lower back or neck pain were enrolled and evaluated during 3 visits over a 12-week period. Results show significant improvements in pain score, physical function, and pain interference scores over time, whereas mean EEG power spectral density across all electrodes and frequency bands was statistically insignificant at the group level between the three visits. However, logistic regression analysis revealed significant nominal inverse correlations between pain score reduction and gamma power changes (averaged globally and localized to the right temporal and left frontal regions), though these failed to retain significance after multiple comparison correction. Logistic regression showed a statistically significant correlation between increased EEG gamma oscillations (1 Hz bins) and decreased pain intensity in the right temporal EEG electrodes. These exploratory analyses suggest that improvements in self-reported pain outcomes in chronic pain patients undergoing standard treatments and complemented with digital-based behavioral interventions correlate negatively with measurable changes in EEG spectral features, which need further validation in future studies with larger cohorts.