The conclusion that couple therapy “is much more acceptable than antidepressant drugs” is not based on the findings of this study (Reference Leff, Vearnals and BrewinLeff et al, 2000). To generalise results of a trial on desipramine, a tricyclic antidepressant (followed by trazodone and fluvoxamine), to antidepressant drugs as a whole is misleading. Other antidepressant drugs may be more acceptable than the trial drugs and the conclusions should have been limited to the drugs used.

Intention-to-treat analysis can sometimes be taken to the extreme. In this study, the majority of subjects in the antidepressant group (56%) did not receive the antidepressant as randomised and therefore it is not surprising that the antidepressant group did poorly. It would have been interesting to see, in addition, the results of an explanatory analysis that would have shown the outcomes for those who actually received antidepressant drugs compared with those who actually received couple therapy. Although the study would, technically, no longer be a randomised trial, this would not be a problem because the baseline comparison shows that the two groups are comparable, the essence of randomisation. It is not always the case that intention-to-treat analysis is better than explanatory analysis.

Although comparatively new, the number needed to treat is now widely understood as a simple and meaningful analysis of trials. What is the improvement rate attributable to couple therapy in this trial and how many patients will need to be treated for one more patient to gain improvement with couple therapy over desipramine?

Finally, the effect of an important confounding variable has not been discussed. The results may have been confounded by marital discord that will respond better to couple therapy than to antidepressants.

These must be considered before recommending couple therapy over antidepressant drugs in people with depression living with partners.