Dr Lam (Reference Lam2006) comments on our study (Reference Scott, Paykel and MorrissScott et al, 2006a ), the largest randomised controlled trial (RCT) of psychological treatment for bipolar disorder conducted so far. We respond as follows.

1. (a) Dr Lam seems to misinterpret the nature and purpose of pragmatic trials. It is not a matter, as he suggests, simply of appropriate outcome measures, which should be a feature of all trials. Pragmatic trials are intended to test therapies in the practical circumstances of everyday clinical settings, using large multicentre samples (Reference Hotopf, Churchill and LewisHotopf et al, 1999). Most previous trials of therapies for bipolar disorders were single-centre efficacy studies designed to try out new interventions in specialist services or where the originator worked.

2. (b) Dr Lam comments on the number of cognitive–behavioural therapy (CBT) sessions received. We believe that 20 sessions with 2 boosters is as many as is practical in most National Health Service (NHS) settings. That patients attended about 14 of the sessions offered is frustrating but reflects clinical reality and is remarkably similar to the attendance achieved in Dr Lam's own study (Reference Lam, Watkins and HaywardLam et al, 2003: average 13.9 sessions, s.d.=5.5).

3. (c) Our analysis strategy was determined before inspection of the data under the scrutiny of a trial steering committee appointed by the Medical Research Council. Dr Lam confounds several issues and recommends an actuarial analysis that is fundamentally incorrect in the context of an RCT (ICH Harmonised Tripartite Guideline, 1999). In an intention-to-treat analysis, the date of randomisation determines the start of the clock and everyone who is randomised is analysed; it is wrong to delay the inclusion of any participant in the analysis until they are asymptomatic. We also reported in the text on the issue he raised, namely that there was no difference in time to next bipolar episode or mean severity of symptoms in the sample who were in acute bipolar episode at baseline, not in acute bipolar episode at baseline or the whole sample.

4. (d) Dr Lam suggests it was inappropriate to include in our study individuals who were in a current episode or not on mood stabilisers. However, given that RCTs of therapy for mental disorders are usually undertaken with participants who are currently symptomatic, we believe it is important and informative to explore the potential effects of therapies commenced in the acute phase of bipolar disorder. Furthermore, in Judd et al's (Reference Judd, Akiskal and Schlettler2002) 12-year follow-up it was shown that individuals with bipolar disorder spend 50% of their time with syndromal or sub-syndromal symptoms of mood disorder, predominantly depression. Not receiving or not adhering to recognised mood stabilisers is a similar well-documented issue in 20–50% of individuals with bipolar disorders (Reference Scott and ColomScott & Colom, 2005). Our sample thus reflects the realities of clinical practice.

5. (e) It is standard practice in RCTs to control for design variables and also to pursue additional analyses that control for potential confounders (Reference Schulz and GrimesSchulz & Grimes, 2005). None of our analyses failed to converge, a common consequence of multi-collinearity.

6. (f) Dr Lam points out that a median split of a continuous variable can lose information. In fact, this was the reason why we looked for a trend across four groups as shown in Fig. 4 (p. 318).

Previous studies of psychological therapies have mostly involved more selected populations at relatively lower risk of relapse. In those circumstances CBT appears beneficial. Our study used a mixed patient sample; many were high-risk or currently symptomatic. We designed the trial in this way to address an issue not explored so far in any other psychological therapy study, namely whether the treatment would be effective in all patients who might be considered for it. Patients were only excluded if participation was unfeasible or unethical.

Our findings indicate that 22 sessions of CBT may not be effective for most people seen in NHS general adult psychiatry settings. In our lower-risk subgroup, similar in characteristics to Dr Lam's sample (Reference Lam, Watkins and HaywardLam et al, 2003), CBT may be very helpful. The clinical implications are that for a stable, lower-risk population, early in their history of bipolar recurrences, CBT should be considered as an adjunctive treatment option to further enhance their outcome. For high-risk, complex cases, other forms of therapy should be considered, such as those targeted at medication adherence or relapse prevention, before considering CBT. These recommendations are consistent with the results from published meta-analyses and other findings on psychological therapies in bipolar disorders (Reference Scott and ColomScott & Colom, 2005; Reference Scott, Colom and VietaScott et al, 2006b ).