AppetiteCheck: Feasibility of Momentary Vagus Nerve Stimulation as an Implicit Intervention for Eating Behavior

1 INTRODUCTION

• Overeating and emotional eating affect people without diagnosed eating disorders and are often exacerbated by distracted or otherwise unmindful eating.These behaviors are associated with increased risk of several health conditions.Source: §1 · p. 2
• Existing eating interventions include tracking and reflection, but active tracking can impose cognitive burden and harmful nudges.Other systems automatically detect eating-related behaviors or support diet management and eating pace.Source: §1 · p. 2
• AppetiteCheck explores physiological modulation as a complement to interventions that ask users to notice, interpret, or respond to cues.The paper frames this as a low-attention approach for habitual, distracted, and affectively driven eating episodes.Source: §1 · p. 2
• The study evaluates momentary tcVNS during eating as a potential intervention modality for future closed-loop eating systems.The feasibility evaluation used an experimenter-operated handheld device in a constrained laboratory setting.Source: §1 · p. 3
• Participants ate 9.6% less and 23.6% more slowly during tcVNS than during sham stimulation.They still felt satiated after snacking; heart rate was lower, and participants described stimulation as subtle and barely noticeable.Source: §1 · p. 3
• The paper reports a first evaluation of tcVNS during eating and frames physiological intervention as an alternative to sensory cues in a controlled lab setting.It also outlines a path toward closed-loop ubiquitous devices that pair eating-related sensing with low-attention intervention.Source: §1 · p. 3

2.1 Vagus Nerve Stimulation

• Momentary tcVNS applies stimulation in parallel with eating, extending VNS beyond chronic and episodic paradigms.Prior work examined implanted VNS and vagal blockade for appetite and weight management; a nearby taVNS study found no food-intake effect but stimulated before eating episodically.Source: §2.1 · p. 4

2.2 Implicit Health Interventions

• AppetiteCheck extends implicit health interventions from changing perception to physiological actuation during an ongoing activity.Prior examples use sensory or perceptual routes such as altered voice, tactile feedback, scent, or breathing guidance.Source: §2.2 · p. 4

2.3 Eating Management & Eating Interventions

• Eating interventions can act through sensory, cognitive, or physiological mechanisms, with different demands on user attention and effort.Sensory approaches depend on users perceiving cues; cognitive approaches require effort and compliance; physiological approaches target appetite or satiety pathways and carry tradeoffs.Source: §2.3 · p. 5
• Momentary, non-invasive physiological actuation during eating remains comparatively underexplored relative to implanted VNS and vagal blockade.AppetiteCheck measures food intake, eating rate, appetite, and physiological signals without aiming to alter food perception or require behavior monitoring.Source: §2.3 · p. 5
• Shared behavioral targets do not make clinical contexts interchangeable, since obesity, eating disorders, and general dietary support raise different safety concerns.Source: §2.3 · p. 5

3.1 System

• AppetiteCheck used an off-the-shelf, FDA-cleared handheld stimulator, chosen for replicability, safety, and institutional IRB compliance.The authors prioritized the cleared device over an initially designed wearable prototype for feasibility testing.Source: §3.1 · p. 5
• Table 1 compares eating interventions by modality, mechanism, user engagement, sample size, and reported food-intake and eating-rate changes.Its caption reports that AppetiteCheck's food-intake reduction was comparable to several non-pharmacological interventions and a low-dose pharmacological intervention in the controlled snacking task.Source: Table 1 · p. 6

3.2 Intervention Approach

• Stimulation alternated two minutes on and two minutes off for up to 20 minutes, beginning concurrently with eating.Participants were not required to handle the device or monitor timing; calibration increased strength only to the level at which they reported a skin sensation.Source: §3.2 · p. 6
• Active tcVNS was delivered on the right neck, while sham stimulation was delivered on the shoulder to control for device contact and handling.The figure also illustrates the cervical course and branches of the targeted vagus nerve.Source: Figure 2 · p. 7, §3.2 · p. 7

4 METHODS

• The evaluation used a controlled within-subject study design.Source: §4 · p. 7
• The intervention was evaluated during distracted snacking.Source: §4 · p. 7
• The study received approval from the Cornell University Institutional Review Board (IRB0148698).Source: §4 · p. 7

4.1 Participants

• 24 participants completed both sessions and were included; 3 of the 27 recruited participants were excluded after one session.Two could not schedule the second session, and one ate no snacks.Source: §4.1 · p. 7
• Participants with complete data were aged 22 to 53 (M=28.3, SD=6.9); 12 were male, 10 female, and 2 non-binary.The sample also included participants with varied ethnic backgrounds and prior experience with electrical stimulation.Source: §4.1 · p. 7
• The achieved sample of N = 24 exceeded the a priori minimum of N = 21 for 80% power at α = 0.05 and medium effect size 0.25.The calculation was for a repeated-measures ANOVA with within-subject factors.Source: §4.1 · pp. 7–8

4.2 Measures

• The measurement plan paired physical measures with questionnaires and open-ended feedback about the stimulation and participant experience.Heart-rate monitoring was used to assess physiological responses.Source: §4.2 · p. 8

4.2.1 Behavioral Measures.

• Food consumed was measured by weighing snack portions before and after eating, while eating duration and grams-per-minute eating rate captured pace.The offered amount varied by snack type, so analyses also used the percentage of the offered portion eaten.Source: §4.2.1 · p. 8

4.2.2 Self-Reported Measures.

• Self-reports assessed satiety, eating impulsivity, stimulus perception, agency, and embodiment using questionnaires and 7-point Likert scales.Eating impulsivity was categorized as Impulsive above an average rating of 4 and Not Impulsive below 4.Source: §4.2.2 · pp. 8–9

4.2.3 Physiological Measures.

• Heart rate and two heart-rate variability metrics, RMSSD and high-frequency band power, were derived from ECG recordings.Lower heart rate and higher RMSSD are associated with parasympathetic activity; HF power is associated with vagal activity but has a more complex relationship.Source: §4.2.3 · p. 9

4.2.4 Qualitative Feedback.

• Open-ended feedback was coded by sensation, intensity, timing, likeness, and secondary impacts; device form-factor comments and daily-life adoption were analyzed separately.One annotator with sensory-analysis experience performed the coding.Source: §4.2.4 · p. 9

4.3 Study Procedure

• The two approximately one-hour sessions were held on different days with order counterbalanced, and each included baseline recording, calibration, distracted eating, and a post-session questionnaire.Participants watched videos of their choice while eating in a quiet study room.Source: §4.3 · p. 9, Figure 3 · p. 10

4.3.1 Snack Selection and Random Assignment.

• Participants selected one of four snacks and ate the same snack in both sessions; portions were standardized to typical-snacking Mean+2SD amounts.Participants were randomly assigned to counterbalanced tcVNS/sham order, and were asked to fast from food and coffee for at least 3 hours beforehand.Source: §4.3.1 · pp. 9–10

4.3.2 Setup and Calibration.

• Each session began with consent and questionnaires as needed, heart-rate monitoring, at least 5 minutes of baseline recording, and stimulation calibration to the lowest perceptible level.The study was single-blind; the experimenter held the gel-coated stimulator in place.Source: §4.3.2 · p. 10

4.3.3 Distracted Eating & Stimulation.

• During up to 30 minutes of video-distracted snacking, participants received 2-minute stimulation periods alternating with 2 minutes off, for up to five periods.Participants were not asked to attend to the stimulation; sham stimulation controlled for device contact, handling, and tactile sensation away from the cervical vagus nerve.Source: §4.3.3 · p. 11

4.3.4 Post-Experiment.

• After eating, participants completed a post-study questionnaire with qualitative feedback, removed the heart-rate monitor, and were offered wipes to remove gel.The stimulator was sanitized and recharged between sessions.Source: §4.3.4 · p. 11

4.4 Statistical Analyses

• Hierarchical linear mixed-effects models analyzed quantitative measures, with post-hoc mean differences estimated using Bonferroni-corrected tests when multiple-comparison models showed significant effects.The models used lme4 and post-hoc tests used emmeans in R.Source: §4.4 · p. 11
• The study was powered for within-subject tcVNS-versus-sham comparisons; interaction effects, including condition by snack type, were exploratory.All models included participant as a random intercept.Source: §4.4 · p. 11
• Exploratory models compared nested condition-only, additive, and interaction models using AIC; video type was examined, but sparse video-platform data were excluded.Video type was coded as informational or entertainment.Source: §4.4 · p. 11

5.1 Behavioral Measures

• Across snacks, tcVNS yielded 9.6% less food, 23.6% slower eating, and episodes 11.5% longer than sham.Source: Figure 4 · p. 12, §5.1 · pp. 12–13

5.2 Self-Reported Measures

• Satiety rose after snacking in both conditions without a condition difference; impulsivity showed no significant association with eating rate or satiety change.Source: Figure 5 · p. 13, §5.2 · pp. 13–14
• Participants rated tcVNS as affecting eating more than sham despite similar perceived stimulus strength and frequency, while agency stayed high and embodiment low across conditions.Source: Figure 6 · p. 14, §5.2 · p. 14

5.3 Physiological Measures

• Heart rate was lower under tcVNS, but HRV measures did not differ by condition and RMSSD and HF decreased during eating.Source: §5.3 · p. 15, Figure 7 · p. 16

5.4.1 Sensory Analysis.

• Both conditions felt mostly low-intensity and vibration-like, but tcVNS produced broader sensations; some participants cited pressure discomfort and preferred wearable application.Source: §5.4.1 · pp. 15–16, §5.4.2 · p. 16

5.4.3 Adoption in Daily Life.

• Among 21 participants, 2 would adopt, 11 were conditionally willing—mainly pending efficacy—and 8 would not; six envisioned eating-regulation uses.Source: §5.4.3 · pp. 16–17

6.1 Feasibility and Effectiveness of Momentary tcVNS

• Participants ate 9.6% less and 23.6% more slowly during tcVNS than during sham stimulation in a single controlled snacking episode.These differences were numerically larger than SSpoon’s reported changes and similar to low-dose intravenous GLP-1 for intake, but smaller than the higher dose’s effect.Source: §6.1 · p. 18
• Post-snacking satiety did not differ significantly, heart rate was lower during tcVNS, and HRV measures did not differ by condition.The study does not establish longer-term effectiveness, effects during regular meals, or the physiological pathway linking stimulation to behavior change.Source: §6.1 · p. 18
• Pairing eating detection with momentary physiological intervention could extend implicit eating systems, but subtle stimulation raises questions about user awareness and control.The authors identify preserving user control and making activation accountable as design challenges.Source: §6.1 · pp. 18–19

6.2 Towards Closed-Loop, Wearable tcVNS

• Real-world use would require further evaluation of safety, stimulation parameters, and effectiveness beyond the controlled laboratory study.The discussion favors wearable deployment over a handheld device to better fit a low-effort, peripheral intervention.Source: §6.2 · p. 19

6.2.1 Power and Operational Lifetime.

• Across deployment scenarios, stimulation contributes little to the energy budget; eating-detection sensing is the main power constraint.Estimates range from 13 days for on-demand use to 4.2 hours for active acoustic sensing with a 110mAh battery.Source: §6.2.1 · p. 19

6.2.2 Device and Electrode Design.

• A necklace, choker, wrap, or shawl could house the electronics and maintain electrode contact, but electrode comfort and safety remain unresolved.Participants’ tingling discomfort resolved after more electrode gel was added; gel-free flexible electrodes need investigation for longitudinal use.Source: §6.2.2 · pp. 19–20, Figure 9 · p. 20

6.2.3 Real-World Deployment.

• Deployment could use user-initiated or eating-detection-triggered sessions, but closed-loop use needs conservative timing, intensity, and frequency limits and user override.Potential target populations require careful study; clinical efficacy and safeguards for people with obesity or other conditions remain unestablished.Source: §6.2.3 · pp. 20–21

6.3.1 Altered Behavior Due to Experimental Conditions.

• Video and snack choices did not always reflect participants’ usual snacking, handheld placement sometimes slowed or stopped eating, and the sham could produce unexpected tingling.Expectation or response bias cannot be ruled out because the sham location did not create an identical sensory experience.Source: §6.3.1 · p. 21

6.3.2 Snacking and Meal Pacing.

• The supplied snacks were often finished, and distracted video snacking covered a limited range of eating behaviors that may make stimulation effects easier to observe.The near-binary consumption pattern complicated analyses; eating rate was the only behavioral metric that accounted for both groups of participants.Source: §6.3.2 · pp. 21–22

6.3.3 Generalizing from Healthy Participants.

• Participants were healthy, had no eating disorders, and their BMI was not collected, so effectiveness in obesity or clinical populations remains unknown.Source: §6.3.3 · p. 22

• AppetiteCheck uses non-invasive tcVNS to provide a low-effort intervention rather than an overt eating prompt.Source: §7 · p. 22
• In a 24-participant in-lab study, food intake was 9.6% lower and eating rate was 23.6% slower during distracted snacking under tcVNS than sham stimulation.Participants reported the same post-snacking satiety; stimulation was subtle and barely noticeable, while heart rate was lower and HRV measures did not differ.Source: §7 · p. 22
• The findings motivate future wearable, self-administered, and closed-loop eating interventions that combine physiological actuation with eating-related sensing while preserving comfort, safety, and user control.Source: §7 · p. 22

• The appendix records the final qualitative codebook and normalization rules used to reduce lexical variability in free-text responses.Source: Appendix A · p. 28
• Sensation responses were grouped into categories such as vibration, tingling, buzzing, hum, pulsing, and soreness, with overlapping terms normalized.Source: Sensation · p. 28
• Low-magnitude sensations were coded as subtle and reports of no perceptual awareness as unnoticeable; temporal codes distinguished initial notice, intermittence, pulsing, fading, and smoothness.Pulsing remained distinct from intermittent because it denotes structured rhythmic modulation rather than variability in awareness.Source: Intensity · p. 28, Temporal Characteristics · p. 28