I am writing to reply to Turkington et al (Reference Turkington, Kingdon and Turner2002: p. 525), who claim in their interesting and recently published paper on cognitive—behavioural therapy (CBT) for psychosis, that ‘The NNT [numbers needed to treat] of 13 for improvement in overall symptoms was compatible with the results achieved when CBT was delivered by expert therapists (Reference Kuipers, Garety and FowlerKuipers et al, 1997)’. We do not think this claim is justified.

First, in our study 64% of the CBT group achieved clinical improvement compared with 47% of the controls (Reference Kuipers, Garety and FowlerKuipers et al, 1997). We did not present the NNT but they are 6 at the end of treatment and 3 at the end of follow-up (Reference Kuipers, Fowler and GaretyKuipers et al, 1998).

Second, the two studies address different questions in different samples. Our study tested whether CBT for psychosis could improve outcome compared with treatment as usual, in a sample comprising subjects deliberately chosen to have at least one distressing, positive, medication-resistant symptom of psychosis (not from ‘lists of patients with schizophrenia receiving treatment’; Reference Turkington, Kingdon and TurnerTurkington et al, 2002: p. 523). We were aiming at a treatment-resistant group, a rather different sample from that recruited by Turkington and colleagues. Neither study compared 9 months of CBT with a briefer intervention. Nor did they test the efficacy of two different kinds of CBT.

We believe that it is misleading to claim comparability of trials between ‘expert’ and ‘non-expert’ therapists, and between results from 6 sessions and 20 sessions. Evidence for the efficacy of CBT for psychosis is at an early and promising stage; we think it is unhelpful to make unsubstantiated comparisons across trials, and hope that these comments provide some clarification.