We agree that the conclusions from NICE and our review are surprisingly different, considering similar literature search periods and widely similar inclusion criteria for primary studies. However, our scope was to assess and evaluate ‘the mere evidence’ of clinical outcomes. The National Institute for Health and Clinical Excellence, in contrast, aims at the formulation of instructional recommendations for the British National Health Service. Thus, the steering group may have had a broader view and considered additional criteria such as cost-effectiveness within a complex system of healthcare. Differences may therefore, at least in part, stem from different perspectives and scopes: the assessment of the mere evidence and the formulation of instructional guidelines.

Indeed, there were and still are no RCTs on quetiapine available. We are aware of one RCT (the Verkes Borderline Study) that has not been published (yet). Thank you for the reference list. There are two more open-label trials of quetiapine in borderline personality disorder. ^{Reference Bellino, Paradiso and Bogetto1,Reference Roepke, Merkl, Dams, Ziegenhorn, Anghelescu and Heuser2} However, this list is not necessarily exhaustive.

We agree that forest plotting would have contributed to a more immediate understanding of the evidence. However, may we refer you to the full Cochrane review which is to be published soon in the Cochrane Library. Forest plots will be provided there whenever appropriate.

Finally, we thank you for indicating this passage which is indeed liable to misunderstanding. The American Psychiatric Association guidelines recommend low-dose antipsychotics in general, ³ whereas our findings indicate that second-generation antipsychotics are supported by the current RCT evidence in particular. This development of a shift towards second-generation antipsychotics has been foreshadowed by John M. Oldham in his guideline watch of 2005, ^{Reference Oldham4} but, to our knowledge, the original guideline recommendations have not been modified since.