Fifteen studies were included (the number of patients was not reported). There were 3 RCTs, 3 randomised comparisons of active treatments, and 9 uncontrolled observational studies.
The included studies differed with respect to the patient characteristics, severity of illness, treatment modality and duration, study design, attrition rates (range: 0 to 51%) and assessment methods. The drop-out rates were higher for treatments of longer duration (greater than 16 weeks) than for those of shorter duration (16 weeks or less): 29.3% (9 studies) for longer duration versus 8.2% (5 studies) for shorter duration (p=0.004). Concurrent use of medication was rarely reported.
All the studies reported an improvement in personality disorder with psychotherapy.
The mean pre-treatment post-treatment effect sizes were significantly greater than 0. For active psychotherapies, the pooled self-rated effect size (12 studies) was 1.11 (95% CI: 0.88, 1.34; range: 0.54 to 1.74), and the pooled observer-rated effect size (12 studies) was 1.29 (95% CI: 0.78, 1.81; range: 0.34 to 2.51).
For control therapies, the pooled self-rated effect size (3 studies) was 0.25 (95% CI: -0.35, +0.86; range: 0.13, 0.45), and the pooled observer-rated effect size (2 studies) was 0.50 (95% CI: -2.29, +3.29; range: 0.19, 0.81).
Psychotherapy was more effective than no treatment in the RCTs. The mean difference in self-rated effect size weighted by sample size (3 RCTs) was 0.78 (p=0.002). The mean difference in observer-rated effect size weighted by sample size (2 RCTs) was 0.57 (p=0.085).
The correlation (r) between the treatment duration and effect size was negative for self-reported outcomes (N=12; r = -0.46, p=0.13) but positive for observer-related outcomes (N=12, r=0.14, p=0.66). The mean self-reported effect size was greater for shorter-term than for longer-term studies; the mean effect sizes were 1.38 and 0.92, respectively (p=0.02).
Publication bias: the investigation showed that the addition of one study with a zero difference in effect size between the active treatment and the control would diminish the self-rated findings to a trend (p=0.06).
The proportion of patients no longer meeting the criteria for a personality disorder ranged from 30 to 69% in 4 studies using medium- to long-term dynamic or interpersonal therapies. The mean proportion of patients was 51.8% (95% CI: 13.1, 90.4) after a mean of 78 sessions over a mean of 67 weeks. There was no statistically- significant difference between the percentage recovered and the treatment length or the number of therapy sessions.