Biswas et al are correct that our study was a pragmatic trial, but beyond that there seem to be many misunderstandings and we are happy to respond to the points they raise.
First, we did not compare family psychoeducation with treatment as usual (TAU). The comparison was between psychoeducation plus TAU v. TAU alone. We asked the pragmatic question whether adding psychoeducation to TAU alone was any better than TAU and were able to answer it positively. The strengths and weaknesses of this type of comparison are fully discussed in our paper.
Second, we did not exclude previous non-responders. We did focus on responders to pharmacotherapy in the index episode because this was a trial of maintenance treatment, and it is very hard for us to logically imagine such a trial without focusing on responders. In addition, it appears meaningless to us that Biswas et al would like to assess bipolarity in a trial of major depression.
Third, Biswas et al seem to insinuate that we ignored ‘differences in treatment structure (e.g. number of antidepressants, doses and length of treatment/follow-up sessions)’. Our Table 1 shows that they were comparable between the two arms, where the doctors in charge of TAU were kept unaware whether their patients had their family participating in family psychoeducation or not. We strictly abided by the principle of ceteris paribus.
Fourth, we agree that adherence and allegiance are important but often ignored aspects in clinical trials. Adherence to the family psychoeducation by the family members was maximised because there was no missed session. Adherence to TAU by the patients may have been optimal or suboptimal but this is not a valid concern in our context because we minimised performance bias (i.e. differential TAU intensity between the two arms) by masking the doctors. Adherence to family psychoeducation by staff was ensured through videotaping and supervision. All these are explained in the paper. On the other hand, we admit we failed to mention our allegiance to psychoeducation as researchers and therapists. We tried to minimise its influence by masking both the doctors in charge of TAU and the outcome assessors.
Fifth, Biswas et al advise that we examine effect modifiers and moderators. In our paper we explain that we did examine one strong empirically supported candidate variable in this regard, namely expressed emotion. 1 And we failed to confirm its role as effect modifier or moderator.
Last but not least, unfortunately we must confess that we do not fully understand how the authors’ proposed ‘two-level randomisation’ or psychoeducation to ‘enhance the maintenance of treatment effects following drug withdrawal’ might work. We are more than willing to continue this discussion in order to ‘creatively tackle design issues when conducting drug–behaviour trials’.