Pilot study of responsive nucleus accumbens deep brain stimulation for loss-of-control eating

Abstract

Cravings that precede loss of control (LOC) over food consumption present an opportunity for intervention in patients with the binge eating disorder (BED). In this pilot study, we used responsive deep brain stimulation (DBS) to record nucleus accumbens (NAc) electrophysiology during food cravings preceding LOC eating in two patients with BED and severe obesity (trial registration no. NCT03868670). Increased NAc low-frequency oscillations, prominent during food cravings, were used to guide DBS delivery. Over 6 months, we observed improved self-control of food intake and weight loss. These findings provide early support for restoring inhibitory control with electrophysiologically-guided NAc DBS. Further work with increased sample sizes is required to determine the scalability of this approach.

Fig. 1. Initial characterization of human accumbens electrophysiological signal provoked by exposure to preferred food stimuli.

a, Schematic of the multi-stage, staggered enrollment, early feasibility study design (NCT03868670). The current stage of each participant within the trial is highlighted (blue. subject 1; green, subject 2). b, Two quadripolar depth electrodes placed bilaterally in the NAc. Stereotactic coordinate with fGATIR and T1 images were used for direct targeting optimization of the electrode trajectory within the NAc. NAc-Red, anterior limb of internal capsule, orange; putamen, yellow; caudate, blue. Coordinates are detailed in Supplementary Table 2. c, The multi-item buffet of preferred foods allowing us to examine NAc electrophysiology during LOC eating as well as the preceding in-lab standardized meals. The 2 s preceding a bite was quantified during LOC versus standard meals. Both participants reported LOC after mood provocation, which correlated with significant low-frequency increases in power (2–8 Hz) during the 2 s preceding a bite when compared with standard meal bites (buffet (red): subject 1, 2.4 ± 1.5 dB, n = 16 bites; subject 2, 5.6 ± 3.1 dB, n = 12 bites; standard meal (black): subject 1, 0.6 ± 1.0 dB, n = 15 bites; subject 2, 0.3 ± 0.9 dB, n = 11 bites; Student’s t-test, ^*P < 0.05). BP, bandpower. Bar graphs with error bars are presented as mean bandpower ± s.e.m.

Fig. 2. Bilateral NAc low-frequency oscillation captured during self-reported, at-risk moments for LOC eating.

a, Both subjects reporting LOC eating events (high craving, low hunger) in the real world consistent with the diagnosis of BED, and kept a food diary describing the LOC, craving and hunger intensity. Analysis of the left and right ventral NAc showed an increase in low-frequency oscillations during LOC eating when subjects experienced craving, compared with hunger and control periods (blue arrow corresponds to frequency quantified in bandpower insert). This increase in low frequency was observed in both our subjects (craving-red trace: power (V² (volts squared) per Hz) (mean ± s.e.m.): subject 1, left NAc: 0.21 ± 0.11, right NAc: 0.16 ± 0.06, n = 10 events; subject 2, left NAc: 0.58 ± 0.14, right NAc: 0.21 ± 0.07, n = 71 events; control-black trace: subject 1, left NAc: 0.1 ± 0.04, right NAc: 0.04 ± 0.01, n = 9 events; subject 2, left NAc: 0.19 ± 0.04, right NAc: 0.09 ± 0.04, n = 80 events; hunger-blue trace: subject 1, left NAc: 0.06 ± 0.01; right NAc: 0.03 ± 0.01, n = 13 events; subject 2, left NAc: 0.27 ± 0.11, right NAc: 0.11 ± 0.03, n = 37 events; corrected for multiple comparisons with P value adjustment for the three conditions using Bonferroni’s corrections (ANOVA, followed by Student’s t-tests comparing each state): subject 1, left NAc: F = 3.50, P = 0.04; right NAc: F = 4.95, P = 0.03; subject 2, left NAc: F = 5.14, P = 0.02, right NAc: F = 0.07, P = 0.93, ^**P < 0.05). b, The detection algorithms programmed to detect low frequency in both left and right ventral NAc simultaneously. An analysis of the LFP time-locked to low frequency triggered detections captured over 8 weeks during scheduled awake and user-initiated LOC recordings and confirmed that our detection algorithms were indeed detecting changes in low-frequency power. Bar graphs with error bars are presented as mean bandpower ± s.e.m.

Extended Data Fig. 1. 6-Month rDBS outcomes.

Bilateral stimulation has been well tolerated by both subjects (~1.5μC/cm² charge density, 10 s duration). Both subjects report an increased sense of control specific to food cravings and eating-related behaviors. Here, behavior outcomes are reported at 6-months with stimulation (rDBS) compared to pre-surgery baseline. A) For both subjects, LOC frequency decreased with stimulation (Subject 1: 16-point decrease; Subject 2: 13-point decrease). B) With LOC episode severity, both subjects showed decrease in severity from baseline to 6-months with stimulation (Subject 1: Baseline = 1.84 + /−0.69 (n = 13 samples), 6-month = 0.69 + /−0.053 (n = 13 samples), student’s t-test, p = 0.05; Subject 2: Baseline = 0.84 + /−0.15 (n = 13 samples), 6-month = 0.15 + /−.09 (n = 13 samples), student’s t-test, p = 0.09). Further, both subjects also experienced decreases in C) For both subjects, Objective Binge Eating events (OBE) decreased with stimulation (Subject 1: 8-point decrease; Subject 2: 12-point decrease). D) Weight (kg, %) and E) BMI (Subject 1: Weight = −5.9 kg, 4.5%, BMI = −2.2; Subject 2: Weight = −8.2 kg, 5.8%, BMI = −2.9). Bar graphs with error bars are presented as mean band power + /- S.E.