I read the special article about agomelatine with interest. ^{Reference Whiting and Cowen1} The authors state that controlled studies have suggested a favourable efficacy and tolerability profile of agomelatine in depression. This statement is not entirely accurate.

The article has missed the negative studies and is a glaring example of publication bias, issues that have been highlighted in a recent meta-analysis. ^{Reference Koesters, Guaiana, Cipriani, Becker and Barbui2} This meta-analysis of placebo-controlled trials of agomelatine in depression included unpublished trials and concluded that agomelatine is unlikely to be clinically superior to placebo. I am part of a group which has recently submitted a systematic review for the Cochrane Collaboration where we compared the efficacy of agomelatine with other antidepressant drugs in depression. Agomelatine did not seem to provide any significant advantage in efficacy. We also found evidence of publication bias. We contacted Servier, maker of agomelatine, for the unpublished trials, but did not receive any response. Furthermore, Servier has not provided data to the National Institute for Health and Care Excellence (NICE); hence NICE has not recommended agomelatine use. ³

The article states that more data are needed to assess the effectiveness of agomelatine in real-world conditions. However, the fact is that agomelatine's efficacy in controlled trials is not yet established. Almost all the studies have been sponsored by Servier or Novartis, the company which marketed it in the USA.

Reboxetine is a classic example of publication bias; in this case mostly positive studies were published. Many years after its introduction, in 2010, the unpublished data was accessed and a meta-analysis found reboxetine to be an ineffective and potentially harmful antidepressant drug. ^{Reference Eyding, Lelgemann, Grouven, Hrter, Kromp and Kaiser4} It is time that drug companies disclose all data from all trials irrespective of the outcome so the efficacy of a drug can be judged objectively.