Authors' reply

We thank Kripalani & Suleman for their critical remarks. Before addressing them point by point, a general remark is required. Our trial was an effectiveness, not an efficacy, trial. We compared a new treatment previously tested elsewhere (behavioural activation) with treatment as usual (TAU) (antidepressant medication) in Iran. An effectiveness trial aims to assess outcomes in usual care, not to test specific mechanisms, which affects the type of control condition(s). Some criticisms make sense from an efficacy study point of view, not from an effectiveness study point of view. Also of note is that the initial response to TAU was quite good, and that the longer-term response of behavioural activation accounted for its superiority.

We do not see how a placebo arm could have assessed cultural influences on TAU. To study this interesting topic, both a placebo and a natural course condition are needed to see whether placebo in Iran does worse than in other cultures compared with doing nothing.

Second, several sources state that 100 mg sertraline is a sufficient dose.^1,2 Moreover, the dose is a valid representation of usual practice in Iran, as there is reluctance to increase the dose given findings that ‘often, adequate clinical activity, and saturation of the 5-HT transporters, are achieved at starting dosages. As a rule, higher dosages do not increase antidepressant efficacy, but may increase the risk of adverse effects’.²

Third, the difference in the amount of attention given is an inherent aspect of comparing behavioural activation and TAU in routine practice. Adjusting for this difference would lead to an invalid comparison in an effectiveness study. The question whether extra attention given to the TAU group would reduce the difference between behavioural activation and TAU is a legitimate one, but goes beyond the scope of this study.

Fourth, last observation carried forward was not used - this is a misinterpretation of the paper; intention-to-treat analysis was used, as it is the gold standard. Analysing only completers leads to biased conclusions. We used mixed regression analyses that use all available data and yield valid estimates under certain assumptions in the light of missing data.³ The suggestion is that a therapy-completers analysis would yield different conclusions. However, the effects are quite similar when only treatment-completers are analysed - Hamilton Rating Scale for Depression: time×condition, F(1,78.02) = 10.05, P = 0.002; time squared× condition, F(1,78.40) = 7.94, P= 0.006; Beck Depression Inventory: time×condition, F(1,78.02) = 6.84, P = 0.011; time squared× condition, F(1,78.35) = 5.37, P = 0.023.

Fifth, the influence of referral type was analysed, and tables with statistics are available online.⁴ It is difficult to understand that this was missed (e.g. ‘referral did not change the condition× time and condition×time squared effects’, p. 207). Moreover, if anything, the differences between conditions were stronger in participants who were referred by healthcare professionals.

Finally, all patients were capable of understanding the information about the offered treatments and making the necessary decisions. All individuals were seen by a psychiatrist to check eligibility, including capacity to consent to participate in the study, as part of the good clinical practice guidelines we applied.