Virtual Reality–Based Avatar Intervention for Eating Disorders: Mixed Methods #feasibility Study
Background: There is a growing interest in developing novel psychological interventions for eating disorders, with an emphasis on targeting maintaining factors. One hypothesized mechanism underlying illness maintenance is the experience of an “inner eating disorder voice,” which reinforces maladaptive thoughts, emotions, and behaviors. Preliminary studies suggest that the eating disorder voice is common among patients and is linked to greater illness severity. Objective: This single-arm, mixed methods pilot #feasibility study aimed to evaluate a novel virtual reality (VR)–based therapy targeting the eating disorder voice. The intervention was adapted from AVATAR therapy for psychosis and was examined as an adjunct to treatment as usual in individuals with eating disorders. In this adaptation, participants engaged with a therapist-controlled avatar representing their inner eating disorder voice in VR. The primary objectives were to assess the #feasibility, acceptability, and safety of the intervention and to provide preliminary estimates of its clinical efficacy. Methods: Adults with anorexia nervosa (9/10, 90%) or bulimia nervosa (1/10, 10%) took part in a 7-session VR-based therapy course at the Mental Health Centre Copenhagen, Copenhagen University Hospital, Denmark, alongside their treatment as usual. Quantitative measures of #feasibility (recruitment, retention rates, and satisfaction scores), safety, and eating disorder–related outcomes were collected at baseline and after treatment between June 2023 and January 2024. Qualitative interviews conducted after the intervention (October 2023 to November 2023) explored participants’ experiences. Descriptive statistics, paired t tests, and thematic analysis were conducted, and the analyses were finalized in October 2025. Results: Recruitment targets were met: 14 individuals were referred, and 11 provided consent well within the prespecified time frame. Treatment completion rate was 80% (8/10; 95% CI 44%-97%), and no serious adverse events occurred. Participants reported high satisfaction (7/10, 70%; mean 9, SD 1.15 on a 10-point Likert scale; median 9, IQR 8.5-10.0; range 7-10), and qualitative data (8/10, 80%) suggested that they valued the immersive virtual representation of their eating disorder voice. Exploratory analyses indicated improvements in eating disorder symptoms (Hedges g=−0.99, 95% CI −1.74 to −0.24; P=.01), power dynamics associated with the eating disorder voice (Hedges g=−1.63, 95% CI −2.59 to −0.67; P=.002), and emotion regulation via cognitive reappraisal (Hedges g=0.87, 95% CI 0.08-1.66; P=.04). Conclusions: The VR-based avatar intervention for eating disorders was feasible, acceptable, and safe, with preliminary signals of clinical improvement. These findings support further development and evaluation of the intervention in a randomized clinical trial.
