Turkington & McKenna (Reference Turkington and McKenna2003) debate the disingenuous title of whether CBT is worthwhile for psychosis. It is clear that conflict of interest has led to a publication bias with an absence of negative clinical reviews of CBT in the literature. Even so, the current evidence base relies on studies that are based in experimental settings, address a broad spectrum of different diagnoses, and have problems with fidelity to a specific CBT treatment. The most recent meta-analysis (Reference Cormac, Jones and CampbellCormac et al, 2003) finds no convincing change on rating scales at long-term follow up of CBT treatment for schizophrenia, and Pilling et al (Reference Pilling, Bebbington and Kuipers2002) suggest that further research is needed to elucidate the therapeutic factors that mediate mental state changes in psychosis. In addition, the most recent published randomised controlled trial of CBT reported no significant differences in clinically significant outcome or even on a unique patient-rated scale (Reference Durham, Guthrie and MortonDurham et al, 2003).

The efficacy of these trials with reference to relapse rates or hospital usage is not proven when compared with standard treatment or generic supportive counselling, and the case for the effectiveness and cost-effectiveness of specific psychological treatment teams working with people with psychosis has not been made.

The aim of CBT for psychosis is to develop a collaborative explanation of symptoms and experiences, with a theoretical mechanism of effectiveness to increase control and decrease distress. This aim cannot be accomplished without an aetiological understanding based on hard evidence. The problem of diagnosis of psychosis needs to be addressed since the trials included subjects with a mixture of chronic and acute psychoses, and there was no attempt to assess the duration of untreated illness prior to intervention (independent of whether the index episode was a first or subsequent episode). Worse outcome has been associated with longer duration of untreated psychosis (Reference Johnstone, Frith and CrowJohnstone et al, 1992). Rather than abandon CBT for psychosis because of its unproven clinical effectiveness, the way ahead may be to focus on symptom profiles linked to an axis of duration. A theoretical stage-specific CBT model would not exclude the clear biological neurotoxic aetiology of non-affective psychoses, and allows this debate to move on.