Responsive deep brain stimulation guided by ventral striatal electrophysiology of obsession durably ameliorates compulsion

Abstract

Treatment-resistant obsessive-compulsive disorder (OCD) occurs in approximately one-third of OCD patients. Obsessions may fluctuate over time but often occur or worsen in the presence of internal (emotional state and thoughts) and external (visual and tactile) triggering stimuli. Obsessive thoughts and related compulsive urges fluctuate (are episodic) and so may respond well to a time-locked brain stimulation strategy sensitive and responsive to these symptom fluctuations. Early evidence suggests that neural activity can be captured from ventral striatal regions implicated in OCD to guide such a closed-loop approach. Here, we report on a first-in-human application of responsive deep brain stimulation (rDBS) of the ventral striatum for a treatment-refractory OCD individual who also had comorbid epilepsy. Self-reported obsessive symptoms and provoked OCD-related distress correlated with ventral striatal electrophysiology. rDBS detected the time-domain area-based feature from invasive electroencephalography low-frequency oscillatory power fluctuations that triggered bursts of stimulation to ameliorate OCD symptoms in a closed-loop fashion. rDBS provided rapid, robust, and durable improvement in obsessions and compulsions. These results provide proof of concept for a personalized, physiologically guided DBS strategy for OCD.

Keywords: NAc-VeP; OCD; RNS; Y-BOCS; closed-loop fashion; detection; rDBS; time-domain feature; ventral striatum.

Figure 1.. NAc-VeP electrophysiology derived from ambulatory task.

a, CT and MR imaging were co-registered to localize the rDBS electrode contacts: one electrode contact was located in the posterior border of the right VeP and the other electrode contact was in the ventral anterior border of the right NAc.This lead was off-label in an attempt to treat OCD in a patient with comorbid partial-onset epilepsy. Another lead targeting the right temporal lobe is not depicted in this figure but delivered stimulation when a seizure was detected. b, Top: iEEG trace from bipolar re-referencing between NAc and VeP. Bottom: 2s-AUC short-term trend (black line) was extracted from iEEG, and relative change (red line) of AUC from the average (black dotted line) and peaks (blue circle). The peaks were detected and extracted using the MATLAB function to find local maxima (blue circle). The relative changes at these peaks were quantified in our analyses. The blue vertical lines are instances when the threshold detection settings, % change in AUC, detected a peak in the AUC, i.e., when responsive stimulation would be triggered. c, Pearson correlation between 2s-AUC short-term trend across the spectral power. Low-frequency ranges had a significant correlation with 2s-AUC in 1/f corrected power (<15Hz) (one-sample two-sided sign test, and FDR-adjusted *p<0.001). d, Sensitivity and specificity for various threshold settings from ambulatory task. We can examine how accurate the rDBS is with the threshold changes. e, The number of peaks when relative changes exceed threshold derived from ambulatory task (mean±s.e.m.). Only the obsessive state iEEG snapshots had peaks over the 75% increase in AUC. f, Comparison of spectral power at AUC peaks with amplitude below vs exceeding 75% of the long-term trend in both obsessive state and control (left), and below vs exceeding 75% in the obsessive states (right). Peaks exceeding the 75% threshold (only observed in obsessive iEEG snapshots) had greater low-frequency power (<15Hz) compared to peaks below the 75% threshold observed in both obsessive and control iEEG snapshots (FDR-adjusted *p<0.001, one-sided student’s t-test).

Figure 2.. NAc-VeP electrophysiology derived from laboratory-based provocation of obsession.

a, Top row: naturalistic provocation task / Bottom row: items in VR provocation task. b, Sensitivity and specificity for various threshold settings from the naturalistic provocation task (top) and the VR provocation task (bottom). c, The number of peaks when relative changes exceed the designated threshold (mean±s.e.m.) from the naturalistic provocation task (top) and the VR provocation task (bottom). Only obsessive trials were observed with higher thresholds.

Figure 3.. In-lab provocation tasks and symptom-locked ambulatory electrophysiologic data all found low-frequency oscillations to be elevated during obsessive states.

a, Spectrograms after touch interaction (t=0), top row: naturalistic provocation task / bottom row: VR provocation task. b, Baseline normalized delta-frequency range (1–4 Hz) power in each condition (top: naturalistic / bottom: VR provocation tasks). There were significant power changes after touch interaction (*p<0.05). c, Absolute powers in the dash lines of a. 1–4 Hz (naturalistic) and 1–2 Hz (VR) absolute power of interaction with obsessive food are significantly higher than interaction with neutral items (FDR-adjusted *p<0.01). d, We quantified the power from 1–120Hz for ambulatory (magnet-swipe stored) snapshots of data (60 to 2 seconds preceding the magnet-swipe), comparing self-reported control and obsessive states. We did not find a significant difference in power between these two conditions for the ambulatory stored data at any specific frequency (FDR-adjusted p>=0.6553). e, As 2s-AUC was found to specifically correlate with power <15 Hz, we averaged ambulatory power from 1–15 Hz and compared between conditions. Average ambulatory power from 1–15 Hz in obsessive state was significantly higher than the power of control (*p<0.05, one-sided Student’s t-test).

Figure 4.. RDBS therapeutic outcomes.

a, Y-BOCS, Obsessive, and Compulsive scores over time. Y-BOCS scores of the baseline, and 24 hours and 1 year 18 weeks post NAc-VeP-rDBS activation were assessed by the clinician. Y-BOCS-SR was reported over the 2 years 22 weeks after surgery (biweekly, first month, intermittently afterward). Y-BOCS decreased rapidly and durably after rDBS was enabled. b, Relative changes of the absolute power at long episode detections (AUC peaks) from the average across the month (mean±s.e.m.). The relative change is calculated for each detection as short-term trend/long-term trend of absolute power. Two different detection settings are depicted by yellow and purple, blanks are no data with those detection settings. The frequency band with the most increased power was delta in 2021–2022 (purple). c, Relative changes at long episode detections in b across the frequency. All the relative power as gray dots across the frequency when rDBS detection was activated. The median of the relative power within the same frequency is depicted with the black solid line. The median of the low-frequency (<11Hz) power is significantly greater than zero in both detection settings depicted with blue solid line (one-sample two-sided sign test, and FDR-adjusted *p<0.000001). When rDBS detection is enabled by AUC detector, the median of delta frequency power was the most increased compared to its average.