Pilot RCT of self-guided smartphone-based VR exposure therapy for social anxiety in people who stutter (null result on primary outcomes)

Background

Following a scoping review that identified promising VR exposure designs for social anxiety, the same research group set out to test one such program directly with people who stutter. Social anxiety frequently accompanies stuttering and can be more distressing than the stuttering itself, yet few interventions specifically target this overlap.

What the researchers did

Twenty-five adults who stutter were recruited via online adverts on the British Stammering Association’s website (Stamma.org) and local stuttering groups across the United Kingdom between September 2021 and June 2022. Participants were randomly allocated to either three weekly sessions of smartphone-based VR exposure therapy (VRET; n = 13) or a three-week waitlist (n = 12). The waitlist group received the same VRET intervention after their second assessment.

VRET was delivered remotely using a Google Cardboard V2 headset and a smartphone app developed in Unity by the research group. Each session contained psychoeducation and three exposure exercises set in 360° video environments: ordering at a cafe, a telephone interaction, and a public-speaking task. A pre-recorded virtual therapist guided participants through each exercise using the inhibitory learning approach - helping them articulate expectations and safety behaviors before each exercise, then helping them evaluate outcomes after. Sessions were self-guided and could be completed at the participant’s own pace.

Outcomes were measured before treatment, after treatment (or after a three-week wait for the waitlist group), and at one-month follow-up for the VRET group. Primary outcomes were self-reported social anxiety (Social Phobia Scale), fear of negative evaluation (FNE-B), unhelpful thoughts about stuttering (UTBAS-6), and stuttering characteristics (WASSP). State-level distress was measured via a SUDS rating during an in-session behavioral assessment task. Two participants additionally provided physiological data (skin temperature, electrodermal activity, heart rate, heart rate variability) during the behavioral task at pre- and post-treatment.

What they found

The study’s primary finding was a null result: VRET did NOT significantly reduce social anxiety versus waitlist between pre- and post-treatment (b = -0.28, F(1, 19.56) = 3.10, p = 0.09; effect size d = -0.41). The corresponding equivalence test was inconclusive - the data were insufficient to either confirm or rule out a treatment effect of the size targeted. Secondary outcomes followed the same pattern: no significant effect of VRET on fear of negative evaluation (p = 0.19), unhelpful thoughts about stuttering (p = 0.59), stuttering characteristics, or in-session speech-related distress (p = 0.13).

A sensitivity analysis confirmed that the study was underpowered. For the social anxiety outcome, the interaction coefficient would have had to equal approximately -0.40 to achieve 80% power; the observed coefficient was -0.28. The authors concluded that the small sample size combined with substantial between-participant variability prevented the study from detecting the effect.

A more encouraging signal came from the one-month follow-up: among the four VRET participants who completed it, social anxiety (Hedge’s g = -0.34) and speech-related distress (g = -0.65) were lower than at post-treatment. This was not statistically tested due to the very small sample and should be treated as exploratory.

Practical feedback from participants was consistent: the cardboard headset was uncomfortable, the three pre-made scenarios did not match the situations they personally found most anxiety-provoking, and the pre-recorded virtual therapist did not recreate the supportive collaborative relationship that participants felt would have helped. Four of thirteen VRET participants did not complete post-treatment assessment; nine of the original thirteen completed all three sessions.

Why this matters

This is the first randomized controlled trial of VRET specifically designed for people who stutter, and it adopted a stronger methodology than the two prior exposure-therapy studies in this group (which had no control group or no statistical analysis). The null result is itself a useful contribution: it suggests that self-guided smartphone-based VRET with fixed pre-recorded scenarios may face engagement and personalisation barriers that a richer delivery model could potentially overcome. The participatory design process the authors used to build the intervention is a model future studies in this area could adopt.

For Therapy withVR specifically: this study did not use, evaluate, or compare against Therapy withVR. The intervention was a custom smartphone app on a Google Cardboard V2 headset using 360° video, developed by the Imperial College team. Its design choices (smartphone-based, fully self-guided, fixed scenarios, virtual therapist) differ from Therapy withVR (Meta Quest headset, clinician-controlled in real time, larger scenario library, live human clinician). The Chard et al. paper is included in this Evidence Hub because it adds to the broader evidence base on immersive VR for communication-related social anxiety, not because it relates to Therapy withVR.

Limitations

The authors flag the following in their own discussion:

• Underpowered. Sensitivity analyzes confirmed the observed effects would not reach significance at the achieved sample size; participant recruitment was challenging given stuttering affects only ~1% of the general population and not all PWS experience social anxiety.
• Group imbalance at baseline. Despite randomisation, the waitlist group had higher social anxiety and a larger proportion of women than the VRET group, potentially creating a floor effect that limited how much further the VRET group’s scores could fall.
• Allocation concealment. The primary researcher had knowledge of participant characteristics and conducted the randomisation, raising the possibility of selection bias.
• Disengagement and attrition. Four of thirteen VRET participants did not complete post-treatment assessment; only four completed the one-month follow-up. The authors attribute this partly to the remote delivery model, the cardboard headset discomfort, and the lack of personalised scenarios.
• Hardware and content limitations. Cardboard headsets offer limited immersion compared to higher-end consumer VR; 360° video offers limited interactivity; turn-taking in interactive scenarios required participants to click buttons to play the next video, which broke immersion.
• COVID-19 context. The remote-delivery design was a response to UK COVID-19 restrictions during the trial period and may have introduced additional engagement barriers compared to an in-person trial.

The authors are explicit that this pilot trial provides a basis for design improvements rather than a recommendation for clinical use of the current protocol.