I would like to make some comments on the points raised by Kirov & Korszun and Anderson & Haddad. They both cite evidence from continuation and maintenance studies, but this is likely to be more flawed than evidence from acute treatment studies. In studies of long-term treatment, patients who have responded to acute treatment are randomised to continue active drugs or to be withdrawn to an inert placebo. However, it cannot be assumed that the state of having had treatment withdrawn is equivalent to never having had treatment in the first place. It is known that there is a discontinuation reaction with all classes of antidepressants (Reference Haddad, Lejoyeux and YoungHaddad et al, 1998). The symptoms of this reaction may themselves be mistaken for relapse, or they may unblind participants and predispose them to relapse because of fears of discontinuing treatment. This is likely to be a particular problem given that the initial sample of patients comprises people responsive to treatment who are therefore likely to have high expectations of the benefits of treatment.

In addition, the evidence on antidepressant effects and severity is complex. The majority of studies that show that increased efficacy correlates with increased severity are studies of out-patients. In in-patients, more-severe depression has been shown to respond less well to antidepressants than moderate depression does, independently of the presence of psychotic symptoms (Reference Kocsis, Croughan and KatzKocsis et al, 1990). In our meta-analysis we found no significant differences from placebo in in-patient studies (Reference Moncrieff, Wessely and HardyMoncrieff et al, 1998), which is in line with results from other large landmark in-patient studies such as the Medical Research Council study and the National Institute for Mental Health study described in my editorial (Reference MoncrieffMoncrieff, 2002).

Finally, if the benefits of antidepressants are so obvious, it seems surprising to me that we have little evidence that the burden of depressive illness is reducing in line with the vast expansion in antidepressant prescribing. In contrast, long-term incapacity related to depression has been rising rapidly both in absolute terms and in relation to other conditions (Reference Moncrieff and PommerleauMoncrieff & Pommerleau, 2000).