Authors' reply

We are pleased that Gupta found our study ¹ of interest. As he says, the findings provide ‘further evidence that psychosocial treatments can supplement pharmacological treatments and improve the course and outcome of bipolar disorder’. ²

Regarding the three issues he raises: first, we agree that given the rigorous randomisation process the reporting of P values for baseline differences between ‘treatment’ and ‘control’ groups is not essential. However, many studies do report these P values and we have followed this convention in our article. Reporting P values also allows a ‘check’ on the randomisation. Here we posit that the idea of randomisation is to obtain ‘comparable’ comparison groups. If randomisation ‘fails’ in any particular way, this seems important to be aware of in both the conduct of the analyses and the interpretation of the results.

Second, regarding the outcomes, we were particularly interested to see whether we could deliver an intervention that addressed both poles of the illness, hence our strategy to define the primary outcomes in the way we did. With regard to the pooling of mixed and manic episodes, that is the most commonly used convention in clinical trial reporting. ³

Lastly, the rate of relapse in itself is not as important as the difference in rates between the intervention and the control group. Relapse is the most commonly used primary outcome measure in comparable trials. ⁴ Our definition of relapse as the primary outcome was dictated by the fact that in bipolar disorder it is relapse that matters most for clinical care rather than symptomatic differences at any single point in time. We should also point out that the lower relapse rate in the intervention group was mirrored by a reduction in emergency psychiatric contacts and hospital admissions, lending support to the clinical utility of the intervention.