The last Cochrane systematic review of early intervention for those with psychosis included cognitive–behavioural therapy (CBT), family therapy and medication, and reported no significant decrease in the development of psychosis at 12-month follow-up (Reference Marshal and LockwoodMarshal & Lockwood, 2004). The implications of the recent study of CBT for the prevention of psychosis (Reference Morrison, French and WalfordMorrison et al, 2004) need to be realistically interpreted with this background.

First, two people were excluded from the cognitive therapy arm after the trial had begun, which would have led to a non-significant result. This should have been acknowledged in the abstract, as an abstract has the most impact with service planners.

Second, after 6 months of cognitive therapy, there was a decrease in the development of psychosis compared with the control arm; however, there was similar distress for both groups. Cognitive therapy for psychosis has an aim of decreasing the distress of psychosis as well as the formulation of an explanatory model for that psychosis. It may be that a reframed and normalised explanatory language was taught to the individuals at high risk, and this led to the decreased identification of symptoms at 12 months and the masking of a psychotic episode. This would not ultimately lead to a decrease in distressing psychosis, but to a later identification of psychosis and a possible delay in pharmacological treatment.

The possible risk of harm or hazard was ignored, with a clear bias against the use of medication expressed by the authors in the discussion. Furthermore, the editorial comment alluded to the possibility of premature publication (Reference TyrerTyrer, 2004), but it is the implication of harm which needs to be explicitly stated.