Our study was designed in close cooperation with full-time clinicians and in both groups (CBT and BDT) application of manuals was highly flexible to be representative of the respective therapies as they are carried out in clinical practice and to enhance the external validity of the study. Sessions were audiotaped and scored using the Coding System of Therapeutic Focus on Action and Insight (CFAI; Reference Samoilov, Goldfried and ShapiroSamoilov et al, 2000) by two independent raters who were masked to the treatment group (interrater reliability (Kendall's W) ranged from 0.86 to 0.91). In general, results revealed that therapists adhered to the respective therapies (Reference Emmelkamp, Grauwelman, Rengers, Hoenink, Rexwinkel and RoelofsenEmmelkamp et al, 2004).

To the best of our knowledge there are no measures of ‘psychodynamic origin’ specifically related to avoidant personality disorder and hence we used the PDBQ. Furthermore, it was not feasible to keep the independent assessors who completed the SCID–II unaware of the treatment group in a number of instances.

Post-treatment CBT was significantly superior to BDT on all ‘primary’ outcome measures. A significance level of α=0.1 set rather than 0.01 as claimed by Leichsenring & Leibing. Even if we exclude the SPAI scores (P=0.09), this still leaves superior outcome for CBT on three out of four outcome variables. The lack of power to detect differences between the waiting-list control group and the active treatments is acknowledged as a limitation.

There are important differences between our study and that of Svartberg et al (2004). Svartberg et al included all types of cluster C and self-defeating personality disorders, rather than limiting their study to avoidant personality disorder. Two-fifths of their sample did not fulfil criteria for avoidant personality disorder treatment and treatment consisted of 40 rather than of 20 sessions. Furthermore, outcome with respect to personality disorders (sic) was only assessed with the Millon Clinical Multiaxial Inventory (Reference MillonMillon, 1994), rather than with the gold standard SCID–II. Finally, the lack of a control group in the study of Svartberg et al renders the results difficult to interpret.

In contrast to most other psychotherapy studies, we did our utmost to prevent an effect of investigator allegiance. The study was designed in close cooperation with two psychodynamic therapists (G.F. and H.K.) and two cognitive–behavioural therapists (A.B. and A.K.), who all fully participated in the design of the study, selection of measures, treatment manuals (including degree of flexibility) and therapists.