Cognitive behaviour therapy to prevent harmful compliance with command hallucinations (COMMAND): a randomised controlled trial

Research output: Contribution to journalArticlepeer-review


  • Maximillian Birchwood
  • Alan Meaden
  • Nicholas Tarrier
  • shon lewis
  • til wykes
  • Lynda Davies
  • Emmanuelle Peters

Colleges, School and Institutes

External organisations

  • University of Nottingham
  • University of Warwick
  • Birmingham and Solihull Mental Health NHS Foundation Trust
  • King's College London
  • University of Manchester


Acting on command hallucinations in psychosis can have serious consequences for the individual and for other people and is a major cause of clinical and public concern. No evidence-based treatments are available to reduce this risk behaviour. We therefore tested our new cognitive therapy to challenge the perceived power of voices to inflict harm on the voice hearer if commands are not followed, thereby reducing the hearer's motivation to comply.

In COMMAND, a single-blind, randomised controlled trial, eligible participants from three centres in the UK who had command hallucinations for at least 6 months leading to major episodes of harm to themselves or other people were assigned in a 1: 1 ratio to cognitive therapy for command hallucinations + treatment as usual versus just treatment as usual for 9 months. Only the raters were masked to treatment assignment. The primary outcome was harmful compliance. Analysis was by intention to treat. The trial is registered, number ISRCTN62304114.

98 (50%) of 197 participants were assigned to cognitive therapy for command hallucinations + treatment as usual and 99 (50%) to treatment as usual. At 18 months, 39 (46%) of 85 participants in the treatment as usual group fully complied with the voices compared with 22 (28%) of 79 in the cognitive therapy for command hallucinations + treatment as usual group (odds ratio 0·45, 95% CI 0·23–0·88, p=0·021). At 9 months the treatment effect was not significant (0·74, 0·40–1·39, p=0·353). However, the treatment by follow-up interaction was not significant and the treatment effect common to both follow-up points was 0·57 (0·33–0·98, p=0·042).

This is the first trial to show a clinically meaningful reduction in risk behaviour associated with commanding voices. We will next determine if change in power was the mediator of change. Further more complex trials are needed to identify the most influential components of the treatment in reducing power and compliance.


Original languageEnglish
Pages (from-to)23-33
JournalThe Lancet Psychiatry
Issue number1
Publication statusPublished - 1 Jun 2014