Emmelkamp et al (Reference Emmelkamp, Benner and Kuipers2006) reported that cognitive–behavioural therapy (CBT) was more effective than brief dynamic therapy (BDT) for the treatment of avoidant personality disorder. However, the study has several methodological shortcomings.

In the BDT group it is not clear whether and to what extent a manualised treatment was realised. The article includes non-specific references to several psychodynamic manuals and it is not clear what therapeutic procedures were actually carried out. Furthermore, no disorder-specific treatment manual was used. In contrast, in the CBT group the manual of Beck & Freeman (Reference Beck and Freeman1990) for avoidant personality disorder was applied. No data with regard to adherence and competence were reported and thus it is not clear whether both treatments were carried out with equal competence.

Besides the presence or absence of the diagnosis according to the Structural Clinical Interview for DSM–IV Axis II Disorders (SCID–II) several self-report measures were applied as ‘primary outcome measures’. However, the authors focus on a specific measure that they regarded as primary. In addition to other outcome measures, Emmelkamp et al used the Personality Disorder Belief Questionnaire (PDBQ; Reference Arntz, Dreessen and SchoutenArntz et al, 2004). Arntz et al (Reference Arntz, Dreessen and Schouten2004) explicitly included items from Beck & Freeman (Reference Beck and Freeman1990) and hence the PDBQ is specifically tailored to the effects of CBT. Possibly the most convincing difference between CBT and BDT was found with regard to the number of patients still fulfilling the SCID–II criteria at follow-up (9 v. 36%). However, it is not clear whether the ‘independent assessor’ was masked to the treatment group.

In two outcome measures that refer more specifically to the features of avoidant personality disorder, the Social Phobia Anxiety Inventory (SPAI) and the Avoidance Scale, another measure developed by the authors (Reference EmmelkampEmmelkamp, 1982), both CBT and BDT achieved large and nearly identical pre-/post-treatment effect sizes: 0.92 v. 0.82 (SPAI) and 1.88 v. 1.75 (Avoidance Scale). Emmelkamp et al reported that ‘CBT was significantly superior on all primary outcome measures.’ However, for the difference between the CBT and BDT groups in SPAI score the P was 0.09, which is not significant at the level of α=0.01 set by the authors. Furthermore, at follow-up, there were no differences between CBT and BDT groups in SPAI and Avoidance Scale scores. Differences were only reported for the PDBQ and for two scales that refer to other personality disorders. For BDT, ‘no significant difference was found between BDT and control’ but no data are reported. Compared with the waiting list control, CBT was only superior in two of six measures but the sample size of the waiting list control was small (n=15 v. 26 for CBT and 28 for BDT post-treatment). The fact that almost no differences were found between the waiting-list control and both BDT and CBT is (at least in part) a result of the insufficient sample size. Furthermore, at least in some measures, the waiting-list group achieved medium or even large effect sizes.

The results reported by Emmelkamp et al (Reference Emmelkamp, Benner and Kuipers2006) are at variance with those reported by Svartberg et al (Reference Svartberg, Stiles and Seltzer2004), who found BDT and CBT to be equally effective for cluster C personality disorders.

Overall, the design, statistical analyses and reporting of the results raise serious concerns about an investigator allegiance effect (Reference Luborsky, Diguer and SeligmanLuborsky et al, 1999).