Promising results from a new digital therapy could help tackle one of psychiatry’s most intractable problems — hearing voices.

Auditory verbal hallucinations are one of the hallmarks of psychosis, particularly in people with schizophrenia, which affects 24 million people globally. These voices often bully or abuse the voice hearer, commenting incessantly on the person’s thoughts and behavior or even urging them to harm themselves.

A new multicenter phase 2/3 trial in the United Kingdom published in Nature Medicine found that distress from these voices and their severity and frequency lessened after 16 weeks of digital avatar therapy. In this behavioral therapy, participants talk with an animated head and voice that resembles the chatter in their head. The study authors suggest that externalizing the voices in a digital avatar might allow the person to gain power over the hallucinations.

“To our knowledge, this is the first therapeutic intervention that has a direct and sustained impact on the frequency with which people hear voices,” said Philippa Garety, psychology professor emeritus at King’s College London and the study’s lead author.

Decades of use drugs hours cognitive behavioral therapy to provide relief from these voices have produced lackluster results, spurring researchers to look for alternatives in recent years. It’s also part of a larger shift in psychiatry to take these voices more seriously and understand their content rather than dismiss them as meaningless. More than a decade ago, the late psychiatrist Julian Leff invented avatar therapy with this in mind. But this current study’s larger sample size and geographic spread have experts more excited about these results.

“It’s not like the voice necessarily disappears, although it decreases in frequency, but it’s not just there less frequently. It becomes less powerful, less potent,” said Alex Leff, a University College London neurologist who worked with his dad on the original 2013 pilot study of avatar therapy.

To start the therapy, the voice hearer works with their therapist to create the avatar — tweaking both the voice quality and the image to align with their internal babble. While up to 80% of people with schizophrenia have auditory hallucinations, this babble varies in number and type. Even within a single person, the voices can originate from nowhere, inanimate objects, the devil or even in the hearer’s own voice.

Acting as both themselves and the avatar, the therapist guides the patient in a three-way conversation in which they can safely confront the avatar in a ten to fifteen minute segment as part of a longer hour-long session.

“When the thing on the screen says, ‘you’re a piece of shit, you’re worthless, and you’ll never amount to anything,’ it’s really affecting. It’s really, really an emotional experience.” said Mark Huckvale, who created the original avatar software with Julian Leff and is one of the current study’s authors. “You’re trying to get people to build up their own self-esteem, and to do that they have to argue back against this persecution.”

Using data from nearly 300 participants, the study tested two different types of avatar therapy: a brief version where the participant abruptly confronted the avatar over six sessions and an extended version that was more tailored to their life history, and lasted 12 sessions. Every participant had to hear voices for at least six months, at least once a week, and find them at least somewhat distressing. And nearly every patient was also prescribed antipsychotic medicine.

After 16 weeks of total observation, both groups reported improvements across several measures, including voice-related distress, severity, empowerment, mood and well-being compared to the control group. When researchers followed back up three months later, the changes held but were less robust compared to the control group. The extended group demonstrated stronger and more lasting effects than their briefer counterpart.

Huckvale says these results demonstrate that the therapy is ready for clinical practice, especially after the lack of adverse events or side effects. While two members of the extended group died, an independent committee determined that neither were due to the therapy.

The regulatory body that provides clinical guidelines for health care in England and Wales recommended that avatar therapy can move forward to be tested out in NHS settings. If the real world evidence corroborates the effects shown here, they can roll out the therapy more broadly in three years. The research team also plans to test the therapy in Ethiopia and India and are also developing avatars that employ AI as a way to facilitate the conversation.

For people in the United States, the wait for avatar therapy will be longer. Many novel behavioral therapies for psychosis and schizophrenia are not available in the US, said Nev Jones, a University of Pittsburgh social work professor who has advocated for integrating lived experience into psychosis research. Jones said this has left clinicians with few good options for treating this symptom.

“We have such a gap between us and them that literally anything that US clinicians could be taught about evidence-supported therapies from the UK would be a good thing,” she said.

While she welcomed avatar therapy for US clinicians, Jones also expressed more skepticism about the result’s replicability, as she did after the therapy’s first randomized control trial. She respects the recent study, but she worries that media coverage will hype up digital behavioral therapies beyond what their results warrant.

“You see this again and again: you add AI, it’s sexy. You add something digital, you add avatars, it’s sexy. And it’s as if people believe that there’s gonna be something achieved that is, you know, quasi-miraculous. And that’s just never the case,” she said.

Jones said the heterogeneity of voices that people hear complicates the conclusions about the mechanisms at play in avatar therapy. The therapy’s higher-than-normal drop out rate further complicates the issue — a rate replicated for all three published studies. In the most recent study, nearly 20% of the brief group did not complete their sessions and over 40% of the extended group failed to complete their sessions.

Leff agreed that this higher dropout rate means that avatar therapy is not for everyone. But he also believes this means the therapy is more effective for the people who stick with it. Many of the therapy’s backers echoed this point, and emphasized the importance of developing a therapy that restores the voice hearer’s agency.

“The power, I think, comes from that interaction, that suddenly the person says something, and the voice changes appropriately in response to it,” said Tom Craig, a psychiatrist at King’s College London and one of the study’s authors. “I think that’s immensely empowering.”

STAT’s coverage of disability issues is supported by grants from the Robert Wood Johnson Foundation and The Commonwealth Fund. Our financial supporters is not involved in any decisions about our journalism.