| ||Literature Coverage Dates||Number of Studies||Number of Study Participants|
|Meta-Analysis 1||1984 - 2003||22||5536|
|Meta-Analysis 2||1996 - 2007||6||2343|
Babcock and colleagues (2004) examined the effect of cognitive behavioral therapy (CBT) interventions for male perpetrators of domestic violence. Their literature search identified studies that examined the effectiveness of three different types of interventions: Duluth Model, CBT, and others.
Their review focused on evaluations that measured violent recidivism of the perpetrator. A search of PsycINFO was conducted to identify all studies retrieved by entering the keywords ‘‘batterers’’ and ‘‘domestic violence,’’ and cross- referenced with terms, including ‘‘treatment’’ and ‘‘intervention.’’ This strategy yielded 68 empirical studies on the efficacy of batterers’ treatment programs. This number was further refined by using inclusion criteria that included 1) the presence of some form of comparison group of batterers, and 2) reliance on a victim report or police record as the index of recidivism. These criteria resulted in the identification of 22 studies (5 experimental and 17 quasi-experimental). Seven of the studies were unpublished, and 15 were published in peer-reviewed journals. These 22 studies provided 44 effect sizes in which various or multiple treatments were compared with each other and with comparison groups. There were 5,536 participants across the various treatment conditions in the reviewed studies. The authors did not present a demographic breakdown.
The outcomes of interest in this review were recidivism and victimization. Recidivism was considered any report of physical violence reported by the victims and/or domestic violence incidents reported by the police during the follow-up period. For the recidivism outcomes, the meta-analysis reported an effect size for five quasi-experimental CBT interventions (k
= 5). For the victimization outcomes, the meta-analysis reported on three quasi-experimental effect sizes for CBT interventions (k
The authors used a hierarchical fixed effects analysis model, because of a non-significant overall Q statistic, with inverse variance weights. Moderator analyses were performed by quality of methods (experimental and quasi-experimental) and by treatment type (CBT interventions, Duluth Model, and other interventions). Additionally, effect sizes were broken down by police report (recidivism) and partner report (victimization) data. Effect sizes were calculated as Cohen’s d
One potential limitation of the results from this meta-analysis is that 13 quasi-experimental studies used dropouts as a comparison group, which may have confounded the analysis by comparing the most motivated participants with the least. Therefore, some of the results of this meta-analysis should be interpreted with caution, particularly when they are dependent on the effects of quasi-experimental studies.Meta-Analysis 2
Smedslund and colleagues (2011) examined the effects of cognitive behavioral therapy (CBT) on men who physically abused their female partners. Their literature search concentrated on interventions that have a clear CBT element.
To be included in the meta-analysis, the studies needed to meet the following criteria: 1) used randomized controlled trials, 2) used male participants who were physically violent toward a female partner, 3) used interventions that included a form of CBT recognizable by its description, and 4) measured outcomes on violent behavior toward the female partner. The literature search targeted randomized trials in eight different databases and repositories. The study authors also consulted field experts and authors and searched conference papers. Of the six randomized trials identified, four used a no-treatment control group, and the remaining two used “other treatment” as a comparison group.
The six trials used in the meta-analysis had 12 citations: five trials had been published in at least one peer-reviewed journal. The six trials accounted for 2,343 of the participants, and the four trials with no-treatment control groups comprised 1,771 participants. The authors did not provide a demographic breakdown of the total sample. For the purpose of this review, only the controlled trials with a no-treatment comparison group (k
= 4) were considered, and the outcome of interest was violent behavior.
The authors reported effect sizes as relative risk scores with 95-percent confidence intervals. For the analysis of the trials with a no-treatment condition, a random effects model was used, due to statistically significant heterogeneity in the studies’ effect sizes.