A Danish trial of virtual-reality therapy for voices found a small, real effect

Auditory verbal hallucinations are among the most frequent and debilitating symptoms of schizophrenia spectrum disorders, and current treatment options have suboptimal response rates — the stated premise of a trial published this month in The Lancet Psychiatry [s1]. Its proposal is one of the more unusual in the field: give the dominant voice a face in immersive virtual reality, and have the patient confront it.

The result is worth reporting precisely because it is modest. Challenge-VRT worked, on its primary endpoint, by a margin that is statistically significant and clinically small [s1].

The design

Challenge was a randomised, assessor-masked, parallel-group superiority trial run across outpatient psychiatric services in three Danish regions [s1]. Participants were 18 or older, had experienced auditory verbal hallucinations for at least three months, and had responded insufficiently to antipsychotic medication [s1]. People who could not identify a dominant voice, who had current substance dependence, or who had an organic brain disorder were excluded [s1].

Randomisation was 1:1 through a secure web-based system using permuted blocks of 12, stratified by site and by antipsychotic treatment, with allocation concealed from outcome assessors [s1]. The intervention was seven weekly immersive virtual reality sessions plus two booster sessions; the control arm received enhanced treatment as usual, matched to the intervention in frequency and duration [s1]. That matching matters — it means the comparison is not therapy versus nothing, but a specific therapy versus an equivalent dose of clinical contact.

The therapy itself was co-developed with people who have lived experience of auditory verbal hallucinations [s1].

Recruitment ran from 16 November 2020 to 14 June 2023 [s1]. Of 399 patients referred, 62 were excluded because they declined or could not be contacted, and 337 were assessed for eligibility; 66 did not meet inclusion criteria [s1]. That left 271 randomised — 140 to Challenge-VRT and 131 to enhanced treatment as usual — with one excluded after randomisation, giving 270 in the analysis [s1]. Mean age was 32.83 years (SD 11.9, range 18 to 72); 105 participants (39%) were male and 165 (61%) female [s1]. No data on ethnicity were collected [s1].

What it found

The primary outcome was clinician-rated severity of auditory verbal hallucinations on the Psychotic Symptoms Rating Scales-Auditory Hallucinations total score at 12 weeks after baseline [s1].

At 12 weeks, Challenge-VRT reduced that score relative to the control condition by an adjusted mean difference of 2.26 points (95% CI 0.25 to 4.26; p=0.027), with a Cohen's d of 0.27 [s1].

A Cohen's d of 0.27 is, by the conventional rough labels, a small effect. The confidence interval runs from 0.25 to 4.26 points, meaning the data are compatible with an effect close to nil at the lower bound. This is a positive trial, and it is a positive trial with a modest and imprecisely estimated effect. Both statements are true and both belong in any honest summary of it.

Among the secondary outcomes, only hallucination frequency moved significantly: an adjusted mean difference of 0.84 points at 12 weeks (95% CI 0.14 to 1.53; p=0.018) and 0.86 points at 24 weeks (95% CI 0.07 to 1.65; p=0.034) [s1]. The other pre-specified secondary outcomes — distress from voices, the Voice Power Differential Scale, the Voices Acceptance and Action Scale, assertive responding to voices, and the Personal and Social Performance Scale — showed no significant differences [s1].

That pattern deserves attention. The theory behind confronting a voice in virtual reality is that it shifts the power relationship between the person and the voice. The measure built to detect exactly that shift — the Voice Power Differential Scale — did not move [s1]. The intervention reduced how severe and how frequent the voices were without demonstrably changing the relational variable it was designed to change.

Safety

The intervention was generally well tolerated [s1]. Six serious adverse events occurred that might have been related to Challenge-VRT: five hospital admissions for exacerbation of auditory verbal hallucinations and one episode of self-harm [s1]. No deaths occurred during the trial [s1].

Those six events are the reason this is not a trivially safe intervention. Deliberately evoking a distressing hallucination in a controlled setting carries a risk of making it temporarily worse, and in this trial it appears to have done so in a handful of participants.

Limits

The primary endpoint was measured at 12 weeks [s1]. The authors describe the finding as short-term efficacy [s1], and only hallucination frequency was still separating the groups at 24 weeks [s1]. Whether the severity benefit persists beyond three months is not answered here.

Blinding was partial by necessity. Assessors were masked, but participants obviously knew whether they had been in a virtual reality headset [s1]. For a clinician-rated primary outcome that depends on patient report, participant expectancy cannot be ruled out as a contributor.

The trial ran in one country, in outpatient services, in a population that could identify a dominant voice — a requirement that excludes people whose hallucinations are diffuse or multiple [s1]. And no ethnicity data were collected, so nothing can be said about whether the effect generalises across populations [s1].

Why it still matters

The comparator was an active one. Participants in the control arm received enhanced treatment as usual matched in frequency and duration [s1], so the 2.26-point difference is not a measure of attention or contact — it is a measure of what the virtual reality component added on top of that.

For a symptom that persists after medication has done what it can, a small added effect from a nine-session psychological intervention is a real contribution, provided nobody oversells it. The authors' own conclusion is that the findings support further development and evaluation of immersive virtual reality-based therapies in this population [s1] — a claim appropriately sized to the evidence.

The trial is registered as NCT04661163 and is now complete [s1]. Its registry record was first posted on 10 December 2020, lists an actual enrolment of 270, and gives its official title as an evaluation of virtual reality-assisted exposure therapy for persistent auditory hallucinations versus supportive counselling in people with psychosis [s2] — a description of the comparator that is worth holding alongside the paper's "enhanced treatment as usual" [s1].

Sources

• [s1] Immersive virtual reality-assisted therapy targeting persistent auditory verbal hallucinations in patients diagnosed with schizophrenia spectrum disorders in Denmark: the Challenge assessor-masked, randomised clinical trial — The Lancet Psychiatry, Crossref record created 3 July 2025 (August 2025 issue). https://doi.org/10.1016/S2215-0366(25)00161-0
• [s2] CHALLENGE. A Randomised Clinical Trial Examining Virtual Reality Therapy, NCT04661163 — ClinicalTrials.gov registry record, first posted 10 December 2020. https://clinicaltrials.gov/study/NCT04661163