I read the review article by Wijkstra et al (Reference Wijkstra, Lijmer and Balk2006) with much interest as I found the results in the abstract quite striking. However, I wish to raise a few points about the methodology and description of results.

First, the authors have concluded from two studies by Bruijn et al (Reference Bruijn, Moleman and Mulder1996) and van den Broek et al (Reference van den Broek, Birkenhäger and Mulder2004) that tricyclic antidepressants (TCAs) are more efficacious than non-TCAs in treating unipolar psychotic depression. They have also quoted these studies to state that antidepressant monotherapy is efficacious in treating this disorder. However, both these studies used haloperidol 1–15 mg/day as additional medication during the trial and this would affect the validity of these statements.

Second, the meta-analysis comparing TCAs with TCAs plus classical antipsychotics does not produce a statistically significant result because of the limited number (two randomised controlled trials) and few patients that have taken part in these trials. The overall effect calculated as the relative risk is 1.44 favouring the combination of tricyclics and classical antipsychotics with a confidence interval of 0.86–2.41. The wide confidence interval has affected the statistical significance of these results. However, I think that it is incorrect and possibly misleading for the authors to conclude that they found ‘no evidence that the combination of an antidepressant with an antipsychotic is more effective than an antidepressant alone’ as the result of the meta-analysis favours the combination and the reason for not obtaining a statistically significant result is the poor quality of constituent studies.