Newton-Howes et al (Reference Newton-Howes, Tyrer and Johnson2006) attempted to definitively answer the question of whether comorbid personality disorder affects outcome in people with major depression. Their search strategy, study selection, data summary and analysis are clearly described. However, the heterogeneous nature of their data does not allow such definitive answers as they claim. As has been noted previously (Reference Charney, Nelson and QuinlanCharney et al, 1981; Reference Black, Bell and HulbertBlack et al, 1988; Reference MulderMulder, 2002) people with depression and comorbid personality disorders are less likely to receive drugs or electroconvulsive therapy (ECT), precisely the treatments (as this meta-analysis reports) that they are more likely to respond to. Therefore, the only fair assessment of the effect of personality disorders on outcome is the randomised controlled trial (RCT). When the meta-analysis was confined to such trials the effect size was smaller but was still significant. A recent meta-analysis that restricted itself to RCTs of drug treatment reported no effect of recent comorbid personality disorder on outcome in people with depression (Reference Kool, Schoevers and de MaatKool et al, 2005). This suggests that better studies with more effective treatments will report less effect of comorbid personality disorder on outcome.

What does this mean clinically? Less than the authors claim, I would suggest. The sample size required to detect the difference between the outcome of patients with depression and personality disorders and similar patients but without personality disorder exceeds 1000 (and this by using all trials rather than just RCTs), suggesting minimal effect in normal clinical practice. Although it seems like a good idea, there is no evidence that targeting comorbid personality pathology is necessary and will result in better outcomes for those with depression. The numbers needed to show an effect of personality disorder on outcome suggest that a treatment trial specifically designed to look for a treatment effect would require such large numbers that it will never be performed.

What the meta-analysis suggests, along with many recent studies, is that good treatment of depression, particularly using drugs and ECT if indicated, will result for the most part in a similar outcome for people with and without personality disorders. Such treatments may in fact be effective for the comorbid personality disorder. Clinicians should be encouraged that aggressive treatment of mood disorder is likely to lead to a positive outcome in those with depression and comorbid personality disorder.