The review investigated the effect of psychotherapy for women who were sexually abused as children. It concluded that psychotherapy showed a moderate beneficial effect, which persisted at follow-up, in controlled studies. The authors' conclusions appear reliable but, since details about the interventions were missing, it was unclear which type of psychotherapy showed the best effects.

To review studies of psychotherapy with women who were sexually abused as children.

MEDLINE, EMBASE, PsycINFO, Sociological Abstracts and the Cochrane Controlled Trials Register were searched from 1966 to 2003. In addition, the reference lists of relevant papers and the last 10 volumes of the British Journal of Psychiatry, American Journal of Psychiatry, Journal of Consulting and Clinical Psychology, and Child Abuse and Neglect were searched. Only studies published in English, German and Scandinavian languages were eligible for inclusion.

Study designs of evaluations included in the review

The studies had to provide pre- and post-treatment or follow-up data.

Specific interventions included in the review

The review did not specify any inclusion criteria for the interventions. The included studies employed individual or group therapy or both; other than that, no information about the type of intervention in the individual studies was given.

Participants included in the review

Studies treating women with established childhood sexual abuse were eligible for inclusion in the review.

Outcomes assessed in the review

The studies had to report measurements on defined instruments; other than that, no inclusion criteria were specified. The included studies investigated depression, anxiety and trauma symptoms, Global Severity Index (GSI) and self-esteem.

How were decisions on the relevance of primary studies made?

Two reviewers independently selected papers for inclusion.

The authors rated randomisation, anonymity of judges, recruitment and selection, diagnostic system, control groups, attrition rates, quality of rating scales, description of treatment, supplementary treatment, side-effects and statistical procedures. The studies were given a quality score ranging from 1 (lowest quality) to 3. Two reviewers assessed quality independently, discussed the discrepancies and agreed a final score.

The reviewers did not say explicitly how the data were extracted for the review, or how many reviewers performed the data extraction. For the controlled studies, the post-treatment mean scores of the treatment and the control groups were compared. For uncontrolled studies, the pre-and post-treatment scores were compared using the pre-treatment standard deviation for pre-and post-treatment mean scores. In addition, the persistence of treatment effects was tested by comparing the post-treatment with the follow-up data. The analyses were based on completer rather than intention-to treat.

How were the studies combined?

For the controlled studies, the standardised and weighted Cohen's d (standardised mean difference) were calculated for five outcome groups (depression, anxiety, GSI, trauma, self-esteem), together with the 95% confidence intervals (CIs) and P-values. The main analysis pooled effects across outcomes and studies; several studies contributed to this analysis with more than one outcome. In addition, the mean effect sizes and 95% CIs across all outcomes were calculated for each study. For the uncontrolled studies, the effect sizes of each study were tabulated individually for the five outcome groups.

Publication bias was investigated by calculating a fail-safe N (number of unpublished negative studies that would be required to change the findings of the analysis) for the outcome depression.

How were differences between studies investigated?

The controlled and uncontrolled studies were analysed separately, and differences in the study design were discussed in the text. The reviewers reported that they tried to investigate patterns in the heterogeneity of the study results, and that effects of study quality and moderators such as level of education were followed up. The meta-analysis of controlled studies was performed with and without 2 studies that had exceptionally high mean effect sizes.

Twenty-four studies (n=1,087) met the inclusion criteria: 13 had control groups (n=598) and 11 were uncontrolled (n=489).

The mean quality score was 1.44 (range: 1.09 to 2.18) on a scale of 1 to 3. The fail-safe N method indicated a low risk of publication bias.

Thirteen controlled studies showed a pooled effect size of 0.63 (95% CI: 0.54, 0.72, P<0.01), indicating a statistically significant benefit of treatment. The omission of 2 studies with a high mean effect size changed the pooled value to 0.53 (95% CI: 0.43, 0.62). The effect sizes in the individual studies pooled across outcomes ranged from 0.19 to 1.99. The effect sizes for the outcome groups ranged from 0.44 for anxiety to 0.75 for self-esteem, all showing a statistically significant effect in favour of the treatment group.

The effect sizes in the 7 uncontrolled studies that enabled an analysis were slightly higher, ranging from 0.36 for self-esteem ratings in one study to 1.57 for depression symptoms in another.

Minimal changes from post-treatment to follow-up were observed in the 10 studies that presented these data.

Psychotherapy for women with childhood sexual abuse showed a moderate gain, which persisted at follow-up, in controlled studies. Multicentre studies with better designs are needed; these should precede a re-examination of the theoretical underpinnings for specific therapies in women with childhood sexual abuse.

The review appeared to have rather vague inclusion criteria and the interventions, especially, were not described in detail. The search was thorough but language restrictions were imposed, so the risk of language bias cannot be ruled out. An attempt to assess publication bias was made. The studies were assessed for quality and differences between the studies were apparently investigated where possible. The pooling of the studies was performed across several outcomes, with some studies contributing more than once to the analysis, but it was unclear whether the analysis was in any way adjusted for this fact. The reporting of the review was not always detailed, e.g. only point estimates without CIs were presented for some results and some abbreviations in the tables were not explained. Overall, the conclusions appear reliable but it was unclear what type of psychotherapy showed the best effects.

Practice: The authors did not state any implications for practice.

Research: The authors stated that studies with patients randomised to alternative treatment modalities should be carried out. Better study designs than the existing ones should be employed, with carefully defined childhood sexual abuse and adequate study power. More multicentre studies should be planned. Up-to-date statistical methodologies should be used to investigate changes over time.

H. Lundbeck (Norway) Ltd.

This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.