A Preoperative Electroencephalography Signature for Predicting Treatment Response to Deep Brain Stimulation in Obsessive-Compulsive Disorder

Deep brain stimulation (DBS) is an effective therapy for treatment-refractory obsessive-compulsive disorder (OCD), but outcomes are heterogeneous and patients who do not respond undergo significant financial and surgical burdens without clinical benefit. Here, we sought to identify and prospectively validate a scalable, non-invasive preoperative DBS treatment-responsive neurobiological signature for OCD. Using preoperative resting-state electroencephalography (EEG) data from a randomized, double-blind, sham-controlled trial of DBS targeting the bilateral nucleus accumbens and anterior limb of the internal capsule in patients with OCD (N = 24; ClinicalTrials.gov registration: NCT04967560), we identified relative delta-band power derived from a right fronto-temporal EEG electrode in the eyes-closed resting-state condition as a predictive signature of clinical outcomes. In particular, lower relative delta power emerged as a robust phenotype predictive of greater symptom reduction at six months, accounting for more than 40% variance in clinical outcomes and yielding over 20% improvement in the response rate relative to that of all-comers. This phenotype displayed aberrant EEG patterns relative to healthy controls but was not associated with differences in baseline symptoms. Moreover, its treatment-predictive relationship was specific to active stimulation and absent in the sham DBS group. The EEG signature demonstrated excellent short-term and long-term test-retest reliability, and its predictive utility was furthermore corroborated by source-space magnetoencephalography analysis. Notably, its out-of-sample predictive power was validated in an independent, prospective cohort (N = 8), where it successfully predicted clinical outcomes for 7 participants. Mechanistically, spatially-resolved transcriptomic analysis revealed that the treatment-predictive strength was strongest in regions with elevated expression of inhibitory-neuron markers, while event-related potential analysis demonstrated that DBS responders exhibited delayed N2 latency during response inhibition, indicating impaired inhibitory control. Additionally, longitudinal changes in the EEG signature were found to correlate with clinical outcomes, suggesting a potential readout for target engagement. Our findings establish a scalable, biologically grounded EEG signature that could enable patient selection and inform treatment optimization for DBS, thereby advancing a precision medicine approach for severe OCD.