Kleindienst and colleagues argue that our interpretation of the treatment outcome with regard to suicide attempts might lead to misunderstandings. They are right that a χ²-test comparing the absolute number of suicide attempters in both groups, transference-focused psychotherapy and treatment by experienced community psychotherapists, is not significant. However, this test does not seem appropriate in the present context, since the baseline, that is the number of suicide attempters during the year before treatment, was not equal in both groups (18 in the transference-focused psychotherapy group v.12 in the community psychotherapists group). Thus, a statistical approach had to be employed that controls for baseline data. Since no χ²-test exists that controls for baseline values, we defined change scores that allow for baseline control within a Mann–Whitney U-test. This test generated the P = 0.009 that, in our opinion, depicts the real changes in suicide attempters in both groups. A between-group effect size of 0.55 for the time × group interaction in suicide attempts was calculated from the χ²-statistics of the change scores (χ² = 7.126, d.f. = 2, P<0.028). Table DS2 of our paper only reports within-group effect sizes; between-group effect sizes were not calculated.

The issue of treatment drop-out is a limitation of this study, which has been thoroughly controlled for and discussed in our paper. After the decision to use treatment drop-out as a primary outcome criterion, we preferred to keep strictly to the intent-to-treat algorithm that demands every randomised patient to be part of the outcome analysis. Nevertheless, to address the understandable criticism raised by Kleindienst and colleagues, we repeated the drop-out analysis after excluding from it patients who did not begin therapy after randomisation. This analysis still revealed a significantly lower number dropping out of the transference-focused psychotherapy group (15 v. 23; χ² = 5.750, d.f. = 1, P = 0.016).

The changes in the primary outcome criteria had been made following the impression of an ongoing discussion in the literature addressing the adequacy of DSM–IV diagnostic criteria as outcome criteria in treatment studies. ^{Reference Skodol, Pagano, Bender, Shea, Gunderson and Yen1,Reference Zanarini, Frankenburg, Reich, Silk, Hudson and McSweeney2} Since our initial outcome criteria ‘number of DSM–IV borderline criteria’ and ‘GAF score’ revealed an even stronger superiority of transference-focused psychotherapy, we did not report this post-hoc change, because a bias in our decision was not suspected.

We thank Kleindienst and colleagues for their criticism and the Editor for giving us the opportunity to clarify important issues regarding our study. We hope that our comments will eliminate doubts concerning the fact that our study documents the efficacy of transference-focused psychotherapy for the treatment of borderline personality disorder.