Background
Sexual offender treatment programs to reduce reoffending have been implemented in many
countries as part of a strategy in managing this offender group. However, there are still
controversies regarding their effectiveness.
Objectives
A meta-analysis of relatively well-controlled outcome evaluations assessing the effects of treatment
for male sexual offenders to reduce recidivism is conducted. The aim is to provide robust estimates
of overall and differential treatment effects.
Search methods
We searched a broad range of literature databases, scanned previous reviews and primary studies
on the topic, hand-searched 16 relevant journals, carried out an internet search of pertinent
institutions, and personally contacted experts in the field of sex offender treatment. In total, we
identified more than 3,000 documents that were scanned for eligibility.
Selection criteria
Studies had to address male sexual offenders and contain an outcome evaluation with a treated
group (TG) and an equivalent control group (CG). Apart from randomized controlled trials (RCTs),
also quasi-experimental designs were eligible if they applied sound matching procedures,
statistically controlled for potential biases or the incidental assignment would not introduce bias.
The studies had to evaluate therapeutic measures aiming at reducing recidivism. Both,
psychosocial and organic treatment approaches were eligible. Case reports were not eligible and
sample size had to be at least n =10. To be eligible, studies had to report official recidivism data as
an outcome and provide sufficient information for effect size calculation. There were no
restrictions with regard to country of origin or language and both published and unpublished
documents were eligible.
Data collection and analysis
For each study/comparison we coded general features, characteristics of the sample, treatment
variables and methodological features. As most studies reported their results in terms of recidivism
rates, we chose the odds ratio (OR) as effect size measure. If results on treatment dropouts were
provided, we merged them with the treatment group results (“intent to treat” analysis). All
statistical analysis of effect sizes applied a random effects model.
Results
29 comparisons drawn from 27 studies met our inclusion criteria. This study pool comprised 4,939
treated and 5,448 untreated offenders. A quarter of the studies were retrieved from unpublished
sources. Most studies appeared since 2000 and more than half came from North America. The
evaluations mostly addressed cognitive-behavioral sex offender treatment. No study on hormonal
treatment met the inclusion criteria. Only about one fifth of the comparisons were RCTs and
matching designs were rare as well. The follow-up periods ranged from 1 to 19.5 years (M = 5.9
years). Most frequently recidivism was defined as a new conviction and with only one exception
studies presented data on sexual reoffending.
Overall, there was a positive, statistically significant effect of treatment on sexual reoffending (OR
= 1.41, 95% CI: 1.11 to 1.78, p < .01). The mean effect equates to 26.3% less recidivism after
treatment (sexual recidivism rate of 10.1% in treated sex offenders vs. 13.7 % in the control groups).
There was a comparable effect on general recidivism (26.4% less recidivism in treated groups; OR
= 1.45, 95% CI: 1.15 to 1.83, p < .01). The overall effects were robust against outliers, but contained
much heterogeneity.
Cognitive-behavioral programs showed a significant effect. Two RCTs on Multi-Systemic Therapy
(MST) which also contains many cognitive-behavioral elements revealed a particularly large effect.
Other intervention types showed weaker or no effects. There was a rather clear trend for better
treatment effects of more individualized programs. There was no significant difference between
various settings. We found significant effects for treatment in the community and in forensic
hospitals, but there is not yet sufficient evidence to draw conclusions regarding the effectiveness of
sex offender treatment in prisons.
The overall methodological quality of the studies was not significantly related to effect size. It
should be noted, though, that we could not demonstrate a significant effect on sexual reoffending
for the few RCTs in our study pool. Sample size was not linearly related to effect size but small
studies with fewer than 50 participants had larger effects. This may suggest publication selection
bias. However, studies from unpublished sources did not reveal weaker effects compared to
published studies. The strongest methodological moderator was descriptive validity. Most studies
lacked a detailed documentation of offender variables so that only few analyses could target this
factor. With regard to offender characteristics we found no significant treatment effect for low risk
participants. In contrast, medium and higher risk groups benefitted from treatment. Although the
treatment of adolescents fared somewhat better than for adults, this difference was not significant.
It made no difference whether offenders entered treatment voluntarily or on a mandatory basis.
Authors’ conclusions
Overall, the findings are promising, but there is too much heterogeneity between the results of
individual studies to draw a generally positive conclusion about the effectiveness of sex offender
treatment. However, the results reveal information that is practically relevant: For example, our
review confirms that cognitive-behavioral programs and multi-systemic approaches are more
effective than other types of psychosocial interventions. The findings also suggest various
conditions of success such as more individualization instead of fully standardized group programs,
an advantage of treatment in the community or therapeutic settings instead of prisons, a focus on
medium to high risk offenders, early treatment of young sexual offenders, and measures to ensure
quality of implementation.
Overall, and particularly with regard to moderators, the research base on sex offender treatment is still
not yet satisfactory. To enable more definite answers, more high-quality research is needed, particularly
outside North America. There is a clear need of more differentiated process and sound outcome
evaluations on various types of interventions (including pharmacological treatment), specific
characteristics of programs, implementation, settings and participants and research methods.