We did not set out to provide a definitive answer to a specific question; our objective was to provide a comprehensive synthesis of all available studies and, by using a systematic approach to data collection with limited exclusion criteria and a robust statistical analysis, we have produced the best summary available to date. Although data from RCTs are valuable, they are not the sole arbiters of association and so our information covers much more than the necessarily short-term span of an RCT. Even if we confine our analysis to the 14 RCTs in our review, we obtain an odds ratio of 1.60 (95% CI 1.25-2.06), indicating better resolution of a depressive episode without comorbid personality disorder. Both cohort studies and case series support this finding, with all groups identifying a poorer outcome in those with a personality disorder.

The overview by Kool et al (Reference Kool, Schoevers and de Maat2005) included just six RCTs, all of which involved drug treatment with antidepressants and none of which extended beyond 24 weeks. The judgement that these were the only trials of ‘high quality’ may be suspect, as it is difficult to assess quality from published papers (Reference Soares, Daniels and KumarSoares et al, 2004). In addition, despite their claim that studies were excluded when ‘they presented reanalyses of a study population that was already included’, we believe that their two largest studies (Reference Hirschfeld, Russell and DelgadoHirschfeld et al, 1998; Reference Russell, Kornstein and SheaRussell et al, 2003) both stem from the same trial (albeit with different outcomes) first reported by Rush et al (Reference Rush, Koran and Keller1998). Excluding Russell et al (Reference Russell, Kornstein and Shea2003), from their meta-analysis slightly widens the 95% CI for the reported (inverted) odds ratio of 1.14 from 0.93-1.39 to 0.88-1.45, neither of which are inconsistent with our own estimate above.

Our review also suggested that there may be a better response to the treatment of comorbid depression and personality disorder with antidepressant drugs than with other treatments, which is consistent with Kool et al (Reference Kool, Schoevers and de Maat2005). We remain optimistic about treating personality pathology successfully in this group, and think that newer treatments which focus on personality should be compared with aggressive pharmacotherapy for those who are regarded as having ‘resistant’ depression.