Artículo en PDF
How to cite
Complete issue
More information about this article
Journal's homepage in redalyc.org
Sistema de Información Científica
Red de Revistas Científicas de América Latina y el Caribe, España y Portugal
Sexual offending is a kind of crime about which the general
public is particularly concerned. Discussions in the media and in
the political arena are emotionally laden and even legislation often
seems to be driven by serious single cases of sexual offending.
Accordingly, in the last decade many Western countries have
revised their criminal justice reactions to sexual offending towards
harsher punishment and incarceration. Because most sexual
offenses are not so serious to justify lifetime sentences there are
also increased attempts to prevent reoffending through
correctional treatment. In Germany, for example, a penal law
reform in 1998 introduced mandatory treatment for sexual
offenders who received a prison sentence of more than two years.
However, the changes in policy have been primarily based on
good intentions and not on sound knowledge of the effectiveness
of sexual offender treatment (Lösel, 2002). As Garrido, Farrington
& Walsh (2006) pointed out in the previous Psicothema section
devoted to crime prevention, policy and evaluation are not always
walking the same path. Accordingly, in many countries there is an
ongoing discussion of the question: What works in the treatment
of sexual offenders?
Early reviews of this issue revealed considerable inconsistency.
For example Marshall et al. (1991) proposed the effectiveness of
treatment, whereas Quinsey et al. (1993) were critically opposed
to their point of view. They mainly based their sceptical stance on
the methodological shortcomings of the available evaluations.
This is in accordance with what Furby, Weinrott and Blackshaw
(1989) concluded in a first systematic meta-analysis of findings on
the effects of sexual offender treatment. Some years later, various
meta-analyses came to more positive conclusions (e.g., Alexander,
1999; Gallagher, Wilson, & MacKenzie, 2000; Hall, 1995a;
Does sexual offender treatment work?
A systematic review of outcome evaluations
Martin Schmucker and Friedrich Lösel*
University of Erlangen-Nuremberg (Germany) and * University of Cambridge (UK)
The article reports a systematic review of controlled outcome evaluations of psychosocial and organic
sexual offender treatment. A comprehensive search of the literature in five languages revealed 80
independent comparisons between treated and untreated groups of sexual offenders (
N
= 22,181). The
majority of studies confirmed a positive treatment effect. Overall, 11.1% of treated offenders and
17.5% of controls showed sexual recidivism (37% difference). Findings for violent and general
recidivism were similar. Studies on surgical castration showed the strongest effect; however, this was
confounded with methodological and offender characteristics. Hormonal medication, cognitive-
behavioural, and behavioural approaches also revealed a positive effect. Non-behavioural treatments
did not show a significant impact. Other moderators such as small sample size, authors’ affiliation with
the program, program completion versus dropout, or type of outcome measure had a significant impact.
Methodological study characteristics explained the largest proportion of effect size variance. Overall,
findings are promising but more differentiated evaluations of high quality are needed.
¿Es efectivo el tratamiento de los delincuentes sexuales? Una revisión sistemática de los resultados
evaluados.
Este artículo presenta una revisión sistemática de los resultados de evaluaciones controla-
das acerca del tratamiento psicosocial y orgánico de los delincuentes sexuales. Un estudio exhaustivo
de la literatura destacó 80 comparaciones independientes entre grupos de delincuentes sexuales trata-
dos y no tratados (
N
= 22.181). La mayoría de los estudios mostraron un efecto positivo. En conjunto,
los delincuentes tratados reincidieron el 11,1%, mientras que los no tratados llegaron al 17,5% (es de-
cir, un 37% de diferencia). Los hallazgos en reincidencia violenta y no violenta fueron parecidos. Los
efectos más grandes los obtuvo la castración quirúrgica, aunque ciertas variables metodológicas y de
los sujetos contaminan ese dato. Los tratamientos no conductuales no mostraron ser efectivos. Otras
variables moderadoras como el tamaño pequeño de la muestra, la relación de los autores de los infor-
mes con el programa, acabar o no el mismo, o el tipo de resultado analizado tuvieron un efecto signi-
ficativo. Las características metodológicas de los estudios explicaron la mayor parte de varianza de los
resultados. En conjunto, los resultados son prometedores, aunque se necesitan evaluaciones diferentes
de alta calidad.
Fecha recepción: 18-6-07 • Fecha aceptación: 11-9-07
Correspondencia: Martin Schmucker
Institute of Psychology
University of Erlangen-Nuremberg
D-91054 Erlangen (Germany)
E-mail: martin.schmucker@psy.phil.uni-erlangen.de
Psicothema 2008. Vol. 20, nº 1, pp. 10-19
ISSN 0214 - 9915 CODEN PSOTEG
www.psicothema.com
Copyright © 2008 Psicothema
While Furby et al. (1989) could find only 11 controlled
evaluations of sexual offender treatment, Hall (1995a) integrated
12 controlled studies that have been issued since Furby’s
publication. Gallagher et al. (2000) drew upon 26 comparisons
and Hanson et al. (2002) found 43 studies that applied minimum
quality standards in evaluation design. Although these and other
meta-analyses found significant effects, there are substantial
differences in the size of the mean effect reported (Lösel &
Schmucker, 2003). For example, Hanson et al. (2002) reported a
relatively small mean effect of
d
= .12 for sexual reoffending,
whereas Gallagher et al. (2000) found a much larger overall effect
of
d
= .47. Such differences seem to be related to specific selection
criteria regarding the methodological quality of the studies and the
kinds of treatment that have been analyzed. In addition, previous
meta-analyses only considered the English-language literature.
Against this background, we conducted a comprehensive
meta-analysis of the international literature on sexual offender
treatment that contained both psychosocial and biological modes
of treatment and studies that have been published in various
languages. The study was carried out within the framework of the
Campbell Collaboration Crime and Justice Group. In the
following, we present selected data on the characteristics of the
evaluation studies, their general findings on outcome, and on
moderators that account for differences in effectiveness.
Method
Eligibility criteria.
In order to use all relevant evaluations on
sexual offender treatment we did not set any historical or regional
limits regarding the treatments evaluated. However, studies had to
meet minimal methodological standards to be eligible:
a) Treatment of sexual offenders
. Treatment included any
measures that used therapeutic interventions and aimed at
reducing reoffending. The nature of therapy could both be
psychosocial or organic, i.e. pharmacological or surgical.
Purely deterrent measures or managerial practices that did
not include therapeutic elements were not eligible.
b) Evaluation by means of recidivism outcomes
. Recidivism
was defined broadly and included new convictions,
incarcerations, arrests or charges as well as lapse behavior.
Measures of personality, e.g. empathy, or therapists’ ratings
of treatment success were not sufficient. The results had to
be presented in a way that allowed a reasonably accurate
estimation of the treatment effect.
c) Inclusion of a comparison group that did not receive the
treatment in question
. This would either be no treatment,
«treatment as usual» or other less intensive or less specific
treatment. Treatment dropouts were not counted as an
appropriate comparison group.
For economic reasons we restricted our selection to studies in
English, German, Dutch, French, and Swedish. However, there
seemed to be not much more controlled research published in
other languages. Minimum sample size was set to
N
= 10. This is a
compromise between the small samples of many studies and the
problem that very small samples may lead to excessive effect sizes
based on minor differences in absolute numbers.
We used a wide variety of sources to identify
relevant literature. First, the references of existing reviews and
meta-analyses formed a basic study pool. In addition we
performed searches of relevant databases (e.g. PsycInfo, MedLine,
Dissertation Abstracts International). Handsearches of journals
pertinent to the topic were performed and the references of
identified studies were scanned for further trials. To be sensitive to
unpublished evaluations of sexual offender treatment an internet
search was conducted. Finally, researchers in the field of sexual
offender treatment were personally contacted and asked for further
studies.
In total, over 2,000 citations were identified. Of these, 66
reports included 69 studies that met the inclusion criteria. When
studies reported results for more specific subgroups (e.g.
according to offense type) and reported characteristics of the
subgroups in as much detail as for the whole group, we chose
these subgroups as units of analysis. This allowed a higher
differentiation in moderator analysis while the samples remained
independent. Following this principle 80 comparisons were
available for the integration.
Coding
. Every study underwent a detailed coding procedure.
Variables included general study features, methodological
characteristics, treatment variables and characteristics of the
treated offenders. A detailed manual was used in order to ensure
objectivity. The first author coded all studies on the basis of a
detailed manual. To test interrater agreement, ten studies were
coded by another experienced rater. On average the interrater
agreement was at 91 % with no variable below 60 %. Core
variables such as design or principal treatment strategy were at
100 %.
To evaluate the overall methodological quality we adapted the
Maryland Scale of Scientific Rigor (Sherman et al., 1997) for our
purposes. This is a 5-point rating that integrates various
methodological study features regarding the validity of causal
interpretations. Level 1 of the Maryland Scale refers to
uncontrolled studies. These were not eligible for our meta-
analysis. In our adaptation the eligible studies were assigned to the
different levels according to the following definitions: (2) non-
equivalent control group (e.g. demonstrated differences in
psychosocial characteristics, treatment refusers); (3) group
assignment is incidental but equivalence can be assumed, i.e.
although assignment is not actively controlled by the researchers
the groups seem largely comparable (e.g. demonstrated
equivalence on important variables, assignment followed
principles unrelated to risk of recidivism); (4) systematic
procedures to ensure comparability of treatment and control group
(e.g. pairwise matching, statistical control); (5) uncompromised
random designs. If random assignment procedures were used but
flawed because of dropout or other reasons, studies were
downgraded.
Study sample characteristics.
Although we did not set
historical limits most of the studies were very recent. About three
quarters appeared since the 1990s and one third even in the new
millennium. The studies are mainly North-American (USA: 31;
Canada: 17). However, we also located eight studies from
German-speaking countries, eight from Great Britain, and five
from other countries. Unpublished evaluations comprised 36 % of
our study pool.
DOES SEXUAL OFFENDER TREATMENT WORK? A SYSTEMATIC REVIEW OF OUTCOME EVALUATIONS
11
than 15 % of the comparisons. This includes six comparisons on
hormonal medication and eight comparisons on surgical
castration. One third of the cases refer to outpatient treatment only.
Residential treatment was somewhat more frequent (prison:
k
= 25;
hospital:
k
= 14). Ten comparisons referred to a mixture of
outpatient and residential treatment settings. Seven comparisons
exclusively addressed adolescent offenders while the majority (
k
=
45) referred to adults only.
The design quality of the evaluations was generally poor. There
were only six uncompromised random designs on Level 5 of the
Maryland Scale. Seven comparisons corresponded to Level 4 and
19 to Level 3. The vast majority applied obviously non-equivalent
control groups (Level 2;
k
= 48). As far as risk variables were
reported, the majority of Level 2 studies contained treated groups
at higher risk (
k
= 14 vs.
k
= 5). For 29 comparisons the direction
was unclear or no data on relevant variables were reported. The
follow-up period ranged from 1 to 10 years and averaged 5.22
years (
SD
= 3.46). Small samples were common. 25 comparisons
referred to samples of 50 offenders or less. However, in one
comparison total sample size was as large as
N
= 2,557. The
median was at 118. Overall, the analysis refers to 22,181 offenders
of which 9,512 had been in the treatment groups.
Effect size computation and statistical integration.
Official
recidivism rate was the typical outcome criterion. As
recommended for dichotomous data (Fleiss, 1994; Lipsey &
Wilson, 2001) we used Odds Ratios (
OR
) as effect size measure
and conducted the statistical analyses on the natural log of the
Odds Ratio (
LOR
). Only some studies either reported the data in
other formats or presented more sophisticated analyses. In the
latter case we used the more sophisticated data if possible.
The unit of analysis was an individual study. If a study
contained more than one dependent (sub)sample we chose the
comparison that showed the best internal validity. Some studies
reported results for different independent subgroups (e.g.
according to offense type, age groups, or risk classification) but
did not qualify as independent comparisons as defined above. In
regard of the comparability of such subgroups we separately
calculated effect sizes for each subgroup and then used the
weighted average to obtain a study effect size.
We considered and integrated different domains of recidivism
(e.g. sex offenses, non-sexual violent offenses, any offenses)
separately. If more than one indicator of recidivism was reported
for an offense type (e.g. arrests and convictions) we averaged the
resulting effect sizes to one study effect size. This seemed to be
appropriate because comparisons between the different indices of
recidivism did not reveal systematic differences.
Effect size integration followed the procedure developed by
Hedges & Olkin (1985) which weighs the individual study effect
sizes according to their standard error. A
Q
test revealed that the
effect size distribution was significantly heterogeneous and we
thus applied the random effects model. Moderator analyses were
carried out using a mixed effects model accordingly (see also
Wilson, 2001).
Results
Overall effects
74 of the 80 comparisons reported sexual recidivism outcomes.
Data regarding (non-sexual) violent reoffending were provided for
20 comparisons and in 49 studies the authors presented data on
overall recidivism. The simplest approach to examine the overall
effect is to compare the direction of effects, irrespective of their
statistical significance. If there is no treatment effect, one would
expect as many comparisons indicating positive outcomes (i.e.
lower recidivism rates in the treated group) as negative outcomes
(i.e. the treated group recidivated at the same or an even higher rate
than the untreated group). Table 1 shows that positive outcomes
were significantly more frequent for all domains of recidivism.
Meta-analytic integration allows a more detailed look as it pays
attention to the size of the individual effects. The mean odds ratios
were quite consistent across the different domains of reoffending
ranging from 1.67 for any recidivism to 1.90 for violent
reoffenses. For sexual recidivism the average treatment effect was
OR
= 1.70 (each
p
<.001; see Table 1). Taking the
n
-weighted
average sexual recidivism rate of the treated groups (11.1 %) as a
base rate, this average effect translates to a 17.5 % average
recidivism rate in the comparison groups. Accordingly, the
recidivism rate of the treated offenders was 6.4 percentage points
or 37 % lower than in the control groups. Although the base rates
vary considerably for different domains of reoffending, in terms of
proportions the results are similar (violent recidivism: 44 %
lower; any recidivism: 31 % lower).
However, the average effects are only a rough indicator of the
treatment success in individual studies. Except for violent
MARTIN SCHMUCKER AND FRIEDRICH LÖSEL
12
Table 1
Total outcomes and mean effects
Outcome
Balance
Recidivism (%)
k
TG : CG
a
OR
CI
95%
QT
G
b
CG
c
Sexual recidivism
74
53 : 18***
1.70***
1.35 - 2.13
237.14***
11.1
17.5
Violent recidivism
20
18 : 2***
1.90***
1.49 - 2.33
0
19.68
***
0
6.6
11.8
Any recidivism
49
36 : 13**
1.67***
1.33 - 2.08
159.80***
22.4
32.5
Note: k
= number of comparisons; Balance TG : CG = number of comparisons favoring treated vs. untreated group, i.e. positive vs. negative treatment outcomes;
OR
= mean odds ratio;
CI
95%
=
95 % confidence interval;
Q
= test of homogeneity (
χ
2
,
df
=
k
- 1); TG= treated group; CG = comparison group.
a
χ
2
tests;
b
n
-weighted average;
c
estimated recidivism rate
***
p
<.01;
***
p
<.001
between the results of individual evaluations are clearly above
what would be expected by chance. We thus conducted moderator
analyses in order to isolate variables that might account for these
differences. We restricted these analyses to the domain of sexual
recidivism for reasons of space and also because this is the main
area of interest and provides a larger database.
Moderator analyses for sexual recidivism outcomes
Content of treatment
. One set of comparisons clearly stood out
of the study pool, both in terms of treatment and in terms of effect
sizes. The eight comparisons on surgical castration showed an
average odds ratio of 15.03 (
z
= 9.03,
p
<.001). Moreover, the effect
size distribution is very homogenous with
Q
(7)= 1.76,
p
= .97. The
remaining 66 comparisons’ average effect is clearly lower but
remains significant (
OR
= 1.38,
z
= 3.16,
p
<.01). For various
reasons we decided to exclude the comparisons on surgical
castration from further analysis: First, this approach is rarely used
in contemporary practice. Second, it contains serious legal and
ethical problems. Third, because of the legal conditions for
surgical castration the respective studies do not contain equivalent
control groups. Fourth, the extremely high and homogenous
effects would have distorted any further analysis considerably.
Results of the moderator analyses of the more common
treatment programs are shown in Table 2. Even after the exclusion
of the surgical castration studies, the effect size distribution
remains highly heterogeneous,
Q
(65)= 163.92,
p
<.001, and the
main treatment approach used still exerts a significant influence
on the evaluation results. Hormonal medication shows the highest
mean effect. Of the psychosocial approaches only cognitive-
behavioral as well as classic behavioral approaches indicate
significant treatment effects. The odds ratios of the other
psychosocial approaches center around 1 and indicate no
difference in recidivism rates between treated and untreated
offenders.
The coding of general treatment approaches is not very subtle
because modern treatment programs do not strictly refer to one
therapeutic approach only but are more eclectic. For example,
hormonal treatment is often accompanied by psychotherapy and
psychosocial programs frequently contain various therapeutic
modules. Therefore, we additionally rated how far the evaluated
programs incorporated elements from different therapeutic
approaches. We used a 4-point scale ranging from 0 (not at all) to
3 (mainly). A multiple regression analysis revealed 21 % of
variance explained by these combinations (
p
= .05). Overall, the
initial analyses on the treatment approaches were confirmed in that
only the inclusion of cognitive and behavioral treatment elements
as well as hormonal medication revealed significant
ß
-weights
(each
p
<.05).
Other treatment characteristics.
Only outpatient treatment
showed a significant effect. The mean odds ratios for treatment in
institutional settings were considerably lower and not significant.
For mixed settings there was an intermediate effect. While these
differences did not reach significance, an ordering of the setting
variable from institutional to outpatient treatment revealed a
significant correlation of
r
= .27 (
p
= .02). However, this variable is
somewhat confounded with the treatment approach. If one
controls the latter by only analyzing comparisons on cognitive-
r
= .12,
p
=
.45).
What does make a difference, however, is whether the
treatment program was specifically designed for sexual offenders
or a general offender program that also incorporated sexual
offenders. This difference was also found in outcomes of general
recidivism (
p
= .05). Another finding related to the implementation
of the programs. Evaluations in which the author was in some way
involved in the program delivery showed clearly significant
effects but programs that were evaluated by independent
researchers did not. That model projects reveal better outcomes
than programs that ware implemented in everyday routine was a
related finding. However, the difference is not as clear-cut as for
author affiliation and both kinds of implementation revealed
significantly positive mean effects.
Offender characteristics.
Unfortunately, the description of
offender characteristics was often scarce. For example, only some
studies differentiated the results by the type of sexual offense. In
these, there were significant reductions in recidivism for any type
of offense except for incest offenders. The latter finding is mainly
due to the low recidivism base rate of incest offenders who are
often taken out of the respective family. The rather high effect
reported for rapists seems to contradict the common notion that
these offenders are particularly difficult to treat. However, this
finding is based on only five comparisons and should be regarded
with caution. Furthermore, rapists may be less sexually disordered
than, for example, child molesters with a preference for male
children.
Although programs that exclusively addressed juvenile
offenders show somewhat stronger effects than program for adult
offenders, this was not a significant difference. A related finding
refers to the age homogeneity of the treated offenders. It showed a
tendency of higher effects in more homogeneous age groups (
r
=
.23,
p
= .10).
As an indicator of therapy motivation, we compared programs
with voluntarily participating offenders with programs that
involved a more or less coerced treatment. Only the voluntary
treatment showed a significant mean effect. However, there was
considerable heterogeneity within any of the categories (each
p
<.02) and the between differences did not reach statistical
significance.
In contrast, whether offenders regularly terminated a program
or dropped out of treatment had a strong impact. Because the
control groups for these comparisons were partially dependent,
this difference could not be tested statistically but was rather
obvious. The odds to not recidivate for offenders who dropped out
of treatment was only half the odds for offenders who were not
treated at all. In contrast, for offenders who terminated treatment
according to the treatment plan the mean odds ratio was clearly in
the positive range.
Methodological characteristics.
Although the design quality
indicates how much one can trust a result, it did not have a
systematic impact on the effect sizes. At
p
= .06, comparisons
rated at Level 3 or higher on the Maryland Scale actually fared
somewhat better (
OR
= 1.69, CI
95%
: 1.26 - 2.28) than comparisons
on Level 2 (
OR
= 1.16; see Table 2). However, there was no linear
trend in the relation between effect size and design quality (
r
= .11,
DOES SEXUAL OFFENDER TREATMENT WORK? A SYSTEMATIC REVIEW OF OUTCOME EVALUATIONS
13
MARTIN SCHMUCKER AND FRIEDRICH LÖSEL
14
Table 2
Effects of selected moderator variables
CI
95%
Variables
k
OR
lower - upper
Q
bet
p
Total effect
66
1.38**
*
1.13
-
1.69
Moderator analyses
Treatment approach
12.60*
**
.027
Cognitive-behavioral
35
1.46**
*
1.12
-
1.89
Classical behavioral
0
7
2.18*
**
1.20
-
3.97
Insight oriented
0
5
1.01
***
0.52
-
1.96
Therapeutic community
0
8
0.87
***
0.54
-
1.39
Other psychosocial, unclear
0
5
0.94
***
0.52
-
1.68
Hormonal medication
0
6
3.11**
*
1.39
-
6.95
Specific treatment for sex offenders
4.70*
**
.030
Yes
56
1.56***
1.27
-
1.93
No
0
5
0.76
***
0.41
-
1.41
Setting of treatment
5.10
***
.165
Prison
21
1.16
***
0.84
-
1.60
Hospital
0
8
1.10
***
0.62
-
1.94
Outpatient
27
1.93***
1.35
-
2.77
Mixed
10
1.37
***
0.78
-
2.41
Status of treatment program
2.63
***
.105
Model project
0
9
2.41**
*
1.20
-
4.86
Routine practice
57
1.32**
*
1.07
-
1.63
Author affiliation to treatment program
10.95***
.001
Yes
32
1.92***
1.44
-
2.56
No
30
0.99
***
0.76
-
1.29
Age group
1.19
***
.275
Adolescents only
0
7
2.35*
**
1.01
-
5.43
Adults only
36
1.43*
**
1.08
- 1.90
Offense type
9.04*
**
.029
Rape
0
5
4.91**
*
1.64
-
14.68
Child molestation (not in family)
0
9
2.15*
**
1.11
-
4.16
Child molestation (incl. incest)
10
1.02
***
0.58
-
1.80
Exhibitionism
0
4
3.72*
**
1.27
-
10.93
Treatment participation
2.22
***
.329
Voluntary
28
1.45*
**
1.08
-
1.93
Non-voluntary
15
1.05
***
0.70
-
1.58
Mixed
0
7
1.01
***
0.57
-
1.77
Treatment termination
a
Treatment completed regularly
44
1.58***
1.23
-
2.05
Dropped out of treatment
14
0.51***
0.39
-
0.67
Sample size
31.43***
.000
10 to 50
18
4.03***
2.50
-
6.50
51-100
10
1.32
***
0.76
-
2.27
101-200
16
1.65**
*
1.13
-
2.41
201-500
12
1.00
***
0.72
-
1.38
More than 500
10
0.88
***
0.64
-
1.21
Design quality: Maryland Scale
6.13
***
.106
Level 2 (non-equivalent)
37
1.16
***
0.90
-
1.50
Level 3 (equivalence assumed)
17
2.08***
1.40
-
3.08
Level 4 (matching, statistical control)
0
6
1.19
***
0.67
-
2.12
Level 5 (randomization)
0
6
1.48
***
0.74
-
2.96
Control group formation
1.64
***
.200
Treatment refused
11
1.96**
*
1.20
-
3.20
Other
47
1.37*
**
1.07
-
1.75
Source of recidivism data
4.56*
**
.033
Criminal records only
57
1.28*
**
1.04
-
1.57
Also self report
0
5
3.32**
*
1.42
-
7.78
Publication status
2.91
***
.088
Published
40
1.62***
1.24
-
2.12
Unpublished
26
1.14
***
0.84
-
1.54
Note: k
= number of comparisons;
OR
= odds ratio; CI
95%
= 95 % confidence interval; CG= comparison group;
Q
bet
= test of between group differences (
χ
2
-distributed with
df=
number of
categories - 1).
a
Comparisons are based on identical CG in part; therefore, between-group differences could not be tested.
*
p
<.05; ***
p
<.01; ***
p
<.001
Whether the comparison group consisted of offenders who had
refused treatment or not had a small but insignificant influence.
Similar to design quality, the length of follow-up time did not
correlate with effect size (
r
= .00).
However, other methodological variables had a clear relation to
effect size. This was particularly the case for sample size (
r
= -.26;
p
= .03). This correlation resulted mainly from large effects in trials
with very small samples (
N
≤ 50; see Table 2). The recidivism base
rate was also significantly related to effect size (
r
= .30,
p
= .01).
Furthermore, studies using self reported recidivism alongside
criminal records showed higher effects than studies that relied on
official data only.
Another set of variables that can be regarded as methodological
in a broader sense refers to descriptive validity (see Lösel &
Köferl, 1989), i.e. the quality of study reporting. In particular, a
lack of reporting details on the treatment concept and on outcome
statistics correlated significantly with effect size (
r
= -.33,
p
<.001
and
r
= -.24,
p
= .03).
Although there was no significant difference between
published and unpublished studies, a significant mean effect
appeared in published studies only.
Moderator effects of treatment features after controlling for
confounding variables
Many of the moderator variables presented above are
confounded with each other. In order to test whether treatment-
specific variables remain significant after controlling for non-
specific variables, we conducted a hierarchical regression
analysis. In four steps we first entered unspecific and
methodological variables, then offender variables, general
treatment characteristics, and finally, variables referring to
treatment content. In a sample of only 66 studies it was not
reasonable to include all moderator variables in the model.
Therefore, on each hierarchical step we firstly included those
variables that seemed to be important from a theoretical point of
view or had proven to be empirically related in the bivariate
analyses (
r
≥.20). Following a stepwise procedure, we then
excluded any variable that did not significantly contribute to the
>.10). Only for the treatment content
cluster we added all types of concept and excluded them stepwise
as described above. In order to retain the full sample, missing
values were plugged by the sample mean (see Cohen & Cohen,
1983).
For the full model there was not much heterogeneity;
Q
(55,
k
=
66)= 65.40,
p
= .16. Therefore, the further calculations followed
the assumptions of a fixed effects model (see Lipsey & Wilson,
2001). Table 3 shows a summary of the analysis. With 60 %, the
full model explains a large proportion of effect size variance
(
p
<.001). The largest part of this was due to unspecific features
which accounted for almost half of the variance. Nevertheless, on
any consecutive step the added variables provided for a significant
increase in explained variance. Most notably, even after
controlling for an array of possible confounding variables, a
cognitive orientation of the treatment program still added
significantly to the explanation of effect size variance. Although
the increase in explained variance at this step is not large, one must
take the conservative nature of this estimate into account: Because
of confounding variables the shared variance needs always to be
regarded as originating from the lower step variable. This caveat
holds similarly for other treatment contents. Rather than
suggesting that other treatment methods do not work in sexual
offender treatment the present regression analysis gives further
evidence that the adoption of a cognitive framework serves to
enhance treatment effects independently of other factors.
Discussion
To our knowledge the present study is the most comprehensive
meta-analysis on the effects of sexual offender treatment. This is
in part due to our inclusion of the non-English literature as well as
unpublished trials. Another reason is the recent increase in the
evaluation of sexual offender treatment. Almost one third of the
integrated studies have been published since 2000. Although
Alexander’s (1999) study pool is larger than ours, our analysis is
restricted to controlled trials. Including a control group is a
minimum standard in order to come to valid conclusions regarding
the effectiveness of an intervention. However, we did not restrict
the analysis to randomized trials that can be considered as the
«gold standard» in program evaluation. In fact, most studies are
DOES SEXUAL OFFENDER TREATMENT WORK? A SYSTEMATIC REVIEW OF OUTCOME EVALUATIONS
15
Table 3
Hierarchical regression with effect size as dependent variable
Cluster of predictor variables
ΔR
2
ΔQ
p
Methodological characteristics
Variables: Quality of outcome reporting, Quality of treatment description, Small sample (
N
≤ 50), Treatment refusers as CG, No dropouts in TG
.45
73.21
.000
Offender characteristics
Variables: Age homogeneity of TG
.03
4.33
.037
General treatment characteristics
Variables: Involvement of authors, Group format, Specific for sexual offenders
.10
15.71
.001
Content of treatment
Variables: Cognitive orientation
.03
5.27
.022
Total
.60
98.52
.000
Note
: Variable changes in index direction correspond to higher effect sizes.
Δ
Q
= Change in the sum of squares with each hierarchical step (
χ
2
distributed with
df
= number of added variables)
There are several reasons why we also included non-randomized
studies. First of all, there are too few randomized trials on sexual
offender treatment with too heterogeneous modes of treatment as
to carry out a differentiated analysis. Secondly, not even
randomization can fully rule out relevant group differences; see
Marques, Wiederanders, Day, Nelson, & von Ommeren (2005) as
an example. Thirdly, the effect size of the randomized trials did
not differ significantly from the non-randomized evaluations. The
non-equivalent comparisons even indicated weaker effects.
Therefore, the overall effect seems not to be positively biased by
the inclusion of studies with poorer designs. Nevertheless, we
should remain critical about the overall positive effect. In order to
reach a more definitive answer on the question «Does sexual
offender treatment work?» we need more high quality studies.
The positive overall effect corresponds to a reduction in
recidivism of about a third. Results are similar for general and
violent reoffending. Many sexual offenders are not «specialized»
in sexual offending but have a broader range of offenses (Hanson
& Morton-Bourgon, 2005). Therefore, the finding that treatment
also reduces general and violent recidivism is of some importance.
The mean odds ratios for different areas of recidivism correspond
to a
d
-value of approximately 0.30 which is in the typical range of
the effects found for general offender treatment (Lösel, 1995; J.
McGuire, 2002). The mean effect reported by Hanson et al. (2002)
for sexual reoffending is lower than ours. One reason for this is
that Hanson and colleagues did not include organic treatment
which showed the highest effects in our analyses. This is
particularly true for surgical castration and it should be noted that
after removal of these studies the mean effect for sexual
reoffending decreases (
OR
= 1.38, equalling a 24 % reduction in
recidivism).
The interpretation of the very high effects of surgical castration
is not straightforward. On the one hand, the operated upon
offenders are hardly equivalent to the control groups. At least, they
have to be regarded as a highly selected and motivated group. This
sets them at a lower risk compared to the non-operated upon
offenders who often had refused the surgery. Quite probably then,
the reported effect is an overestimation. On the other hand, the
effects are very homogeneous across the studies and the
recidivism rate of the operated upon offenders is quite consistently
at about 5 % after relatively long follow-up periods of ten years
or more. Apart from the question of effectiveness, ethical
considerations have to be taken into account. Without going into
details, arguments can be raised for both pro and contra surgical
castration (Weinberger, Sreenivasan, Garrick, & Osran, 2005). As
for now, one can only say that abandoning this approach without
closer inspection may not be in the best interest of the society nor
the most serious sex offenders. However, it is necessary to collect
more solid knowledge on circumstances and modes that may
prove such a treatment to be reasonable.
This is also true for hormonal medication or the so-called
«chemical castration» by lowering the testosterone level. This
approach did show encouraging results in the present analysis.
One advantage over surgical castration is its reversible nature.
However, this is a disadvantage at the same time. In the light of
possibly serious side effects, non-compliance and treatment
dropout are common problems for hormonal treatments and after
cessation of the medication, not only do testosterone levels reach
initial levels but also recidivism rates show a marked increase (e.g.
Berlin & Meinecke, 1981; Meyer, Cole, & Emory, 1992). One
should also consider that hormonal treatment is usually
accompanied by psychotherapeutic measures. More over, in our
analysis, hormonal treatment was highly confounded with other
variables and did not show a significant impact in the hierarchical
regression.
Only a cognitive-behavioral treatment orientation consistently
demonstrated a positive impact. This effect is based on a relatively
solid number of 35 comparisons. For all other treatment
approaches the available number of independent findings was
much smaller. Still, the pattern of our findings on different modes
of treatment fits in well with what is known from general offender
therapy. The usually less clearly structured insight-oriented and
milieu-therapeutic approaches seem to be of little benefit while
highly structured cognitive-behavioral treatment shows good
effects (Lösel, 1995; J. McGuire, 2002). However, even in the
category of cognitive-behavioral treatment, there is considerable
variance in effects. For example, albeit applying this approach and
being one of the soundest studies so far, the recently finalized
California Sexual Offender Treatment and Evaluation Program
(Marques et al., 2005) did not show a positive overall effect.
Type of treatment is only one feature that moderated outcomes.
In particular, we found methodological factors to be strongly
related to effect size. This cluster of variables explained almost half
of the effect size variance. Saying this, a cautionary note seems to
be necessary. Applying a hierarchical model that introduced
methodological factors first, one may overestimate their influence.
In addition, the meaning of methodological variables needs also to
be regarded from a treatment perspective. For example, descriptive
validity was negatively related to effect size. This may indicate that
the respective evaluation referred to a less well-elaborated
treatment or to insufficient monitoring of implementation. The
latter may also relate to sample size. In larger samples it may be
more difficult to control and maintain a thorough treatment
implementation. Usually, smaller effects in larger samples are
discussed in terms of publication bias (e.g. Light & Pillemer,
1984). However, we found just the same relationship between
sample size and effect size in the unpublished studies. Therefore,
this relation may be linked to treatment integrity and this to author
involvement in implementation. If evaluators actively drive the
treatment under study they probably care that it is implemented
properly (Petrosino & Soydan, 2005). Actually, our data do
indicate that those studies in which measures to ensure treatment
integrity have been undertaken show better effects than studies in
which implementation problems were obvious. Unfortunately this
aspect could be rated only for a few primary studies and we cannot
rule out that the difference is due to sampling error. In the literature
on general offender treatment implementation issues have been
raised repeatedly (Bernfeld, Farrington, & Leschied, 2001; Lösel &
Wittmann, 1989; Weisburd, Lum, & Petrosino, 2001). Accordingly,
we need more information on treatment implementation and its
relationship to effect size in the field of sexual offender treatment.
However, the fact that treatments in bigger samples and without
author involvement show small or no effects does not allow us to
generalize
the
positive
overall
results
to
large-scale
implementations in routine practice.
Also other moderator analyses have to be interpreted
cautiously, both regarding the moderators we found as well as
those we failed to identify. There are several reasons for this:
MARTIN SCHMUCKER AND FRIEDRICH LÖSEL
16
DOES SEXUAL OFFENDER TREATMENT WORK? A SYSTEMATIC REVIEW OF OUTCOME EVALUATIONS
17
Firstly, from a statistical point of view, the random model is less
sensitive in revealing moderator effects (Overton, 1998). On the
other hand, multiple significance testing contains the danger of
inflated
α
-error. Secondly, variables are confounded and it is
virtually impossible to disentangle the multiple relationships with
regard to their individual causal impacts on effect size (Lipsey,
2003). Thirdly, sometimes the necessary information is not
provided in primary studies and analyses cannot be conducted or
are based on only a few studies. For example, offender
characteristics are probably more influential than our analyses can
demonstrate. But the lack of detailed information hampers more
differentiated analyses (Lösel, 2001). A clearly offender related
factor that shows worse outcomes is treatment dropout. However,
one should not reduce dropping out of treatment to a stable
offender characteristic but perceive it as a mismatch between
offender and therapy or a lack of motivational treatment elements
(Beyko & Wong, 2005; McMurran, 2002). This also seems to be
an important point with respect to whether offenders enter
treatment voluntarily or not. Effects tend to be better in voluntarily
participating offenders. But the difference to more or less coerced
treatment was not significant and effect sizes within the categories
were heterogeneous. This means that neither voluntary
participation is a sufficient condition for a positive development
nor does an external treatment enforcement preclude effective
therapy. Obviously, readiness for treatment is a more complex
construct and should be aimed at under both conditions (Ward,
Day, Howells, & Birgden, 2004).
In summary, our results indicate that sexual offender treatment
can significantly reduce recidivism rates. The size of the effect is
small to moderate but it is in accord with what we know from the
larger research literature on general offender treatment evaluation.
However, the evidence is based on studies that mostly apply a
weak methodological standard. Restricting the analysis to a few
randomized trials shows a comparable mean effect but it does not
render it statistically significant. As for now, our results indicate
that cognitive-behavioal treatment is promising. Also, hormonal
medication seems to improve outcomes. Although promising
findings have been reported on other pharmacological treatments
(Hill, Briken, Kraus, Strohm, & Berner, 2003), there are as yet no
controlled evaluations that show its usefulness with regard to
reoffending. Obviously, we need more high quality evaluations on
the whole range of sexual offender treatments to come to
unequivocal conclusions. This is also true with regard to the
descriptive validity of studies. More differentiated reports of the
treatment, its implementation, offender characteristics, and the
respective outcomes would clearly improve our ability to give a
more detailed answer to the question of «What works for whom
and under what conditions?».
References
(Studies included in the meta-analysis are indicated with an asterisk).
Alexander, M.A. (1999). Sexual offender treatment efficacy revisited.
Sexual Abuse: A Journal of Research and Treatment, 11
(2), 101-116.
* Allam, J. (1998).
Effective practice in work with sex offenders: A recon-
viction study comparing treated and untreated offenders
. Birmingham:
West Midlands Probation Service Sex Offender Unit.
* Bakker, L., Hudson, S.M., Wales, D.S., & Riley, D. (1998).
«And there
was Light»: Evaluating the Kia Marama Treatment Programme for
New Zealand sex offenders against children
. Christchurch, NZ: Psy-
chological Service, Department of Corrections.
* Barbaree, H.E., & Seto, M.C. (1998).
The ongoing follow-up of sex of-
fenders treated at the Warkworth Sexual Behaviour Clinic
. Toronto:
Centre for Addiction and Mental Health.
* Barnes, J.M. (2000).
Recidivism in sex offenders: A follow-up compari-
son of treated and untreated sex offenders released to the community
in Kentucky.
Unpublished Dissertation, University of Louisville,
Louisville, KY.
Berlin, F.S., & Meinecke, C.F. (1981). Treatment of sex offenders with an-
tiandrogenic medication: Conceptualization, review of treatment
modalities, and preliminary findings.
American Journal of Psychiatry,
138
, 601-607.
* Berner, W., & Karlick-Bolten, E. (1986).
Verlaufsformen der
Sexualkriminalität. 5-Jahres-Katamnesen bei 326 Sexualdelinquenten
unter Berücksichtigung von Frühsozialisation, vorausgegangener
Delinquenz, psychiatrisch-psychologischer Diagnostik und Therapie
.
Stuttgart: Enke.
Bernfeld, G.A., Farrington, D.P., & Leschied, A.W. (Eds.). (2001).
Of-
fender rehabilitation in practice. Implementing and evaluating effec-
tive programs
. Chichester: Wiley.
Beyko, M.J., & Wong, S.C.P. (2005). Predictors of treatment attrition as
indicators for program improvement not offender shortcomings: A
study of sex offender treatment attrition.
Sexual Abuse, 17
(4), 375-389.
* Bluglass, R. (1980). Indecent exposure in the West Midlands. In D. West
(Ed.):
Sex offenders in the criminal justice system
(pp. 171-180). Cam-
bridge: Cambridge Institute of Criminology.
* Borduin, C.M., Henggeler, S.W., Blaske, D.M., & Stein, R.J. (1990). Mul-
tisystemic treatment of adolescent sexual offenders.
International Jour-
nal of Offender Therapy and Comparative Criminology, 34
, 105-113.
* Borduin, C.M., & Schaeffer, C.M. (2001). Multisystemic treatment of
juvenile sexual offenders: A progress report.
Journal of Psychology
and Human Sexuality, 13
(3/4), 25-42.
* Byrne, S.M. (1999).
Treatment efficacy of a juvenile sexual offender
treatment program.
Unpublished Dissertation, Memorial University of
Newfoundland, St. Johns, Newfoundland, Canada.
Cohen, J., & Cohen, P. (1983).
Applied multiple regression/correlation
analysis for the behavioral sciences
(2nd ed.). Hillsdale, NJ: Erlbaum.
* Cornu, F. (1973).
Katamnesen bei kastrierten Sittlichkeitsdelinquenten
aus forensisch-psychiatrischer Sicht
. Basel: Karger.
* Craissati, J., Falla, S., McClurg, G., & Beech, A. (2002). Risk, recon-
viction rates and pro-offending attitudes for child molesters in a com-
plete geographical area of London.
Journal of Sexual Aggression, 8
(1),
22-38.
* Di Fazio, R., Abracen, J., & Looman, J. (2001). Group versus individual
treatment of sex offenders: A comparison.
Forum on Corrections Re-
search, 13
(1), 56-59.
* Dünkel, F., & Geng, B. (1994). Rückfall und Bewährung von Karrieretätern
nach Entlassung aus dem sozialtherapeutischen Behandlungsvollzug und
dem Regelvollzug. In M. Steller, K.-P. Dahle & M. Basqué (Eds.):
Straftäterbehandlung: Argumente für eine Revitalisierung in Forschung
und Praxis
(pp. 35-74). Pfaffenweiler: Centaurus.
* Fedoroff, J.P., Wisner-Carlson, R., Dean, S., & Berlin, F.S. (1992).
Medroxy-progesterone acetate in the treatment of paraphilic sexual
disorders: Rate of relapse in paraphilic men treated in long-term group
psychotherapy with or without medroxy-progesterone acetate.
Journal
of Offender Rehabilitation, 18
, 109-123.
Fleiss, J.L. (1994). Measures of effect size for categorical data. In H.
Cooper & L.V. Hedges (Eds.):
The handbook of research synthesis
(pp.
245-260). New York: Russell Sage Foundation.
* Friendship, C., Mann, R.E., & Beech, A.R. (2003). Evaluation of a na-
tional prison-based treatment program for sexual offenders in England
and Wales.
Journal of Interpersonal Violence, 18
(7), 744-759.
MARTIN SCHMUCKER AND FRIEDRICH LÖSEL
18
* Frisbie, L.V. (1969).
Another look at sex offenders in California
(Mental
Health Research Monograph No. 12). Sacramento, CA: State of Cali-
fornia, Department of Mental Hygiene.
Furby, L., Weinrott, M.R., & Blackshaw, L. (1989). Sex offender recidi-
vism: A review.
Psychological Bulletin, 105
(1), 3-30.
Gallagher, C.A., Wilson, D.B., & MacKenzie, D.L. (2000). A meta-analysis of
the effectiveness of sex offender treatment programs.
Retrieved 5/23/2001,
from http://www
.wam.umd.edu/~wilsondb/papers/sexof
fender
.pdf
Garrido, V., Farrington, D.P., & Welsh, B.C. (2006). The importance of an
evidence-based approach in the current spanish policy for crime pre-
vention.
Psicothema, 18,
591-595.
* Groth, N.A. (1983). Treatment of the sexual offender in a correctional
institution. In J.G. Greer & I.R. Stewart (Eds.):
The sexual aggressor
(pp. 160-176). New York: Van Nostrand Reinbold.
* Guarino-Ghezzi, S., & Kimball, L.M. (1998). Juvenile sex offenders in
treatment.
Corrections Management Quarterly, 2
(1), 45-54.
Hall, G.C.N. (1995a). Sexual offender recidivism revisited: A meta-analy-
sis of recent treatment studies.
Journal of Consulting and Clinical Psy-
chology, 63
(5), 802-809.
* Hall, G.C.N. (1995b). The preliminary development of a theory-based
community treatment for sexual offenders.
Professional Psychology:
Research and Practice, 26
(5), 478-483.
* Hanson, R.K., Broom, I., & Stephenson, M. (2004). Evaluating commu-
nity sex offender treatment programs: A 12-year follow-up of 724 of-
fenders.
Canadian Journal of Behavioural Sciences, 36
(2), 87-96.
Hanson, R.K., Gordon, A., Harris, A.J.R., Marqués, J.K., Murphy, W.D.,
Quinsey, V.L., et al. (2002). First report of the collaborative outcome
data project on the effectiveness of psychological treatment for sex of-
fenders.
Sexual Abuse: A Journal of Research and Treatment, 14
(2),
169-194.
Hanson, R.K., & Morton-Bourgon, K.E. (2005). The characteristics of per-
sistent sexual offenders: A meta-analysis of recidivism studies.
Journal
of Consulting and Clinical Psychology, 73
(6), 1154-1163.
* Hanson, R.K., & Nicholaichuk, T.P. (2000). A cautionary note regarding
Nicholaichuk et al. (2000).
Sexual Abuse: A Journal of Research and
Treatment, 12
(4), 289-293.
* Hanson, R.K., Steffy, R.A., & Gauthier, R. (1992). Long-term follow-up
of child molesters: Risk predictors and treatment outcome.
* Hedderman, C., & Sugg, D. (1996).
Does treating sex offenders reduce
offending?
(Research Findings No. 45). London: Home Office Re-
search and Statistics Directorate.
Hedges, L.V., & Olkin, I. (1985).
Statistical methods for meta-analysis
.
Orlando: Academic Press.
Hill, A., Briken, P., Kraus, C., Strohm, K., & Berner, W. (2003). Differen-
tial pharmacological treatment of paraphilias and sex offenders.
Inter-
national Journal of Offender Therapy and Comparative Criminology,
47
(4), 407-421.
* Huot, S.J. (2002, October).
Recidivism, recidivism, recidivism! An up-
date of several Minnesota recidivism studies.
Paper presented at the
21st Annual Research and Treatment Conference of the Association for
the Treatment of Sexual Abusers on ‘Best Practices: Clinical and Re-
search Collaborations’, Montréal, Québec, Canada.
* Janner, J. (1959). Über die Bedeutung psychiatrischer Betreuung verurteil-
ter Sexualdelinquenten.
Schweizerische Zeitschrift für Strafrecht, 74
,
310-319.
* Kansas Department of Corrections. (2000).
Offender programs evalua-
tion: Volume IV
. Topeka: Kansas Department of Corrections.
* Kaul, J., Huot, S.J., Epperson, D., & Dornfeld, M. (1994).
Sex offenders
released in 1988
(Unpublished report). St. Paul, MN: Minnesota De-
partment of Corrections.
* Lab, S.P., Shields, G., & Schondel, C. (1993). Research note: An
evaluation of juvenile sexual offender treatment.
Crime and
Delinquency, 39
(4), 543-553.
* Langevin, R., Paitich, D., Hucker, S.J., Newman, S., Ramsay, G., Pope,
S. et al. (1979). The effect of assertiveness training, Provera, and sex of
therapist in the treatment of genital exhibitionism.
Journal of Behavior
Therapy and Experimental Psychiatry, 10
, 275-282.
Light, R.J., & Pillemer, D.B. (1984).
Summing up: The science of
reviewing research
. Cambridge: Harvard University Press.
Lipsey, M.W. (2003). Those confounded moderators in meta-analysis:
Good, bad and ugly.
Annals of the American Academy of Political and
Social Science, 587
(May), 69-81.
Lipsey, M.W., & Wilson, D.B. (2001).
Practical meta-analysis
. Thousand
Oaks: Sage.
* Looman, J., Abracen, J., & Nicholaichuk, T.P. (2000). Recidivism among
treated sexual offenders and matched controls: Data from the Regional
Treatment Centre (Ontario).
Journal of Interpersonal Violence, 15
(3),
279-290.
Lösel, F. (1995). The efficacy of correctional treatment: A review and syn-
thesis of meta-evaluations. In J. McGuire (Ed.):
What works: Reducing
reoffending. Guidelines from practice and research
(pp. 79-111).
Chichester: Wiley.
Lösel, F. (2001). Evaluating the effectiveness of correctional programs:
Bridging the gap between research and practice. In G.A. Bernfeld, D.P.
Farrington & A.W. Leschied (Eds.):
Offender rehabilitation in practi-
ce. Implementing and evaluating effective programs
(pp. 67-92). Chi-
chester: Wiley.
Lösel, F. (2002). ¿Sirve el tratamiento para reducir la reincidencia de los
delincuentes sexuales? (Englisches Manuskript: Evaluation of sex of-
fender treatment). In S. Redondo (Ed.):
Delincuencia sexual y sociedad
(pp. 361-396). Barcelona: Ariel.
Lösel, F., & Köferl, P. (1989). Evaluation research on correctional treat-
ment in West Germany: A meta-analysis. In H. Wegener, F. Lösel & J.
Haisch (Eds.):
Criminal behavior and the justice system
(pp. 334-355).
New York: Springer.
Lösel, F., & Schmucker, M. (2003).
The efficacy of sex offender treatment:
A brief synthesis of meta-analyses.
Preliminary Report for the Camp-
bell Crime and Justice Group (Updated). University of Erlangen-
Nuremberg: Institute of Psychology.
Lösel, F., & Wittmann, W.W. (1989). The relationship of treatment in-
tegrity and intensity to outcome criteria.
New Directions for Program
Evaluation, 42
, 97-108.
* Lowden, K., Hetz, N., Harrison, L., Patrick, D., English, K., & Pasini-
Hill, D. (2003).
Evaluation of Colorado’s prison therapeutic communi-
ty for sex offenders: A report of findings
. Denver, CO: Office of Re-
search and Statistics, Division of Criminal Justice.
* Maletzky, B.M. (1991). The use of medroxyprogesterone acetate to as-
sist in the treatment of sexual offenders.
Annals of Sex Research, 4
(2),
117-129.
* Maletzky, B.M., & Field, G. (2003). The biological treatment of
dangerous sexual offenders. A review and preliminary report of the
Oregon pilot depo-Provera program.
Aggression and Violent Behavior,
8
(4), 391-412.
* Marqués, J. K. (1999). How to answer the question «Does sex offender
treatment work?».
Journal of Interpersonal Violence, 14
(4), 437-451.
Marqués, J.K., Wiederanders, M., Day, D.M., Nelson, C., & von Om-
meren, A. (2005). Effects of a relapse prevention program on sexual re-
cidivism: Final results from California’s Sex Offender Treatment and
Evaluation Project (SOTEP).
Sexual Abuse: A Journal of Research and
Treatment, 17
, 79-107.
* Marshall, W.L., & Barbaree, H.E. (1988). The long-term evaluation of a
behavioral treatment program for child molesters.
Behaviour Research
and Therapy, 26
, 499-511.
* Marshall, W.L., Eccles, A., & Barbaree, H.E. (1991). The treatment of
exhibitionists: A focus on sexual deviance versus cognitive and rela-
tionship features.
Behaviour Research and Therapy, 26
, 129-135.
Marshall, W.L., Jones, R.L., Ward, T., Johnston, P., & Barbaree, H.E.
(1991). Treatment outcomes with sex offenders.
Clinical Psychology
Review, 11
, 465-485.
* McConaghy, N., Blaszczynski, A., & Kidson, W. (1988). Treatment of
sex offenders with imaginal desensitization and/or medroxyprogestero-
ne.
Acta Psychiatrica Scandinavica, 77
, 199-206.
* McGrath, R.J., Cumming, G.F., Livingston, J.A., & Hoke, S.E. (2003).
Outcome of a treatment program for adult sex offenders. From prison
to community.
Journal of Interpersonal Violence, 18
(1), 3-17.
* McGrath, R.J., Hoke, S.E., & Vojtisek, J.E. (1998). Cognitive-behavioral
treatment of sex offenders. A treatment comparison and long-term fol-
low-up study.
Criminal Justice and Behavior, 25
, 203-225.
McGuire, J. (2002). Integrating findings from research reviews. In J.
McGuire (Ed.):
Offender rehabilitation and treatment: Effective pro-
grammes and policies to reduce re-offending
(pp. 3-38). Chichester,
UK: Wiley.
* McGuire, T.J. (2000). Correctional institution based sex offender treatment:
A lapse behavior study.
Behavioral Sciences and the Law, 18
(1), 57-71.
DOES SEXUAL OFFENDER TREATMENT WORK? A SYSTEMATIC REVIEW OF OUTCOME EVALUATIONS
19
McMurran, M. (Ed.). (2002).
Motivating offenders to change: A guide to
enhancing engagement in therapy
. Chichester, UK: Wiley.
* Meyer, W.J., Cole, C., & Emory, E. (1992). Depo provera treatment for
sex offending behavior: An evaluation of outcome.
Bulletin of the
American Academy of Psychiatry and the Law, 20
(3), 249-259.
* Mulloy, R., & Smiley, W.C. (1996, August).
Recidivism and treated
sexual offenders.
Paper presented at the International Congress of
Psychology, August 16-21, Montreal, Canada.
* Nicholaichuk, T.P. (1996). Sex offender treatment priority: An illustra-
tion of the risk/need principle.
Forum on Corrections Research, 8
(2),
38-41.
* Nutbrown, V., & Stasiak, E. (1987). Research monograph: A retrospec-
tive analysis of O.C.I. cost effectiveness 1977-1981.
Ontario Correc-
tional Institute, 2
, 1-16.
* Ohio Department of Rehabilitation and Correction. (1996).
Five year re-
cidivism follow-up of sex offender releases
(Unpublished Report).
Columbus, OH: Ohio Department of Rehabilitation and Correction.
* Oregon Department of Corrections. (1997). Outcome evaluation of the
Jackson County Sex Offender Supervision and Treatment Program.
Re-
* Ortmann, R. (2002).
Sozialtherapie im Strafvollzug: Eine experimentelle
Längsschnittstudie zu den Wirkungen von Strafvollzugsmaßnahmen auf
Legal- und Sozialbewährung
. Freiburg: Edition iuscrim.
Overton, R.C. (1998). A comprison of fixed-effects and mixed (random-
effects) models for meta-analysis tests of moderator variable effects.
Psychological Methods, 3
(3), 354-379.
* Perkins, D. (1987). A psychological treatment programme for sex offenders.
In B.J. MacGurk & D.M. Thornton (Eds.):
Applying psychology to
imprisonment: Theory and practice
(pp. 191-217). London: Her Majesty’s
Stationery Office Books.
* Peters, J.J., Pedigo, J., Steg, J., & McKenna Jr., J. (1968). Group psy-
chotherapy of the sex offender.
Federal Probation, 32
, 41-45.
Petrosino, A., & Soydan, H. (2005). The impact of program developers as
evaluators on criminal recidivism: Results from meta-analyses of ex-
perimental and quasi-experimental research.
Journal of Experimental
Criminology, 1
(4), 435-450.
* Procter, E. (1996). A five-year outcome evaluation of a community-
based treatment program for convicted sexual offenders run by the pro-
bation service.
Journal of Sexual Aggression, 2
, 3-16.
Quinsey, V.L., Harris, G.T., Rice, M.E., & Lalumière, M.L. (1993). As-
sessing treatment efficacy in outcome studies of sex offenders.
Journal
of Interpersonal Violence, 8
, 512-523.
* Rasmussen, L.A. (1995).
Factors related to recidivism among juvenile
sexual offenders.
Unpublished Dissertation, The University of Utah.
* Rattenbury, F.R. (1985).
The outcomes of hospitalized and incarcerated
sex offenders: A study of offender types, recidivism rates, and identifying
characteristics of the repeat offender.
Unpublished Dissertation, Loyola
University of Chicago, Chicago, IL.
* Rice, M.E., Harris, G.T., & Quinsey, V.L. (1993). Evaluating treatment
programs for child molesters. In J. Hudson & J.V. Roberts (Eds.):
Evaluating justice: Canadian policies and programs
(pp. 189-203).
Toronto: Thompson.
* Robinson, D. (1995).
The impact of cognitive skills training on post-re-
lease recidivism among Canadian federal offenders
(Research Report
No. R-41). Ottawa, Ontario, Canada: Correctional Service of Canada.
* Romero, J.J., & Williams, L.M. (1983). Group psychotherapy and inten-
sive probation supervision with sex offenders: A comparative study.
Federal Probation, 47
(4), 36-42.
* Ruddijs, F., & Timmerman, H. (2000). The Stichting ambulante
preventie projecten method: A comparative study of recidivism in first
offenders in a Dutch outpatient setting.
International Journal of
Offender Therapy and Comparative Criminology, 44
(6), 725-739.
* Schmid, P. (1988).
Was geschieht mit den Sexualstraftätern in der Psy-
chiatrie? Darstellung und Bewertung der psychiatrischen Behandlung
von Sexualstraftätern im Psychiatrischen Landeskrankenhaus Bad
Schussenried in den Jahren 1978-1987.
Unpublished Dissertation, Uni-
versität Tübingen, Tübingen.
Sherman, L., Gottfredson, D., MacKenzie, D., Eck, J., Reuter, P., & Bush-
way, S. (1997).
Preventing crime: What works, what doesn
t, what
s
promising. Report to the U.S. Congress
. Washington, DC.
* Smith, M.A. (2000).
The impact of prison-based drug treatment on the
timing and risk of reincarceration of sexual offenders.
Unpublished
Dissertation, Florida State University, Tallahassee, FL.
* Song, L., & Lieb, R. (1995).
Washington State sex offenders: Overview
of recidivism studies
(No. 95-02-1101). Olympia, WA: Washington
State Institute for Public Policy.
* Stalans, L.J., Seng, M., Yarnold, P., Lavery, T., & Swartz, J. (2001).
Process and initial impact evaluation of the Cook County Adult Proba-
tion Department
s Sex Offender Program: Final and summary report
for the period of June, 1997 to June, 2000
. Chicago, IL: Illinois De-
partment of Corrections.
* Stürup, G.K. (1953). Sexual offenders and their treatment in Denmark
and other Scandinavian countries.
International Review of Criminal
Policy, 4
, 1-19.
* Stürup, G.K. (1968). Treatment of sexual offenders in Herstedvester
Denmark: The rapists.
Acta Psychiatrica Scandinavica Supplement,
204
, 1-62.
* Taylor, R. (2000).
A seven-year reconviction study of HMP Grendon
Therapeutic Community
(Research Findings No. 115). London: Home
Office Research and Statistics Directorate.
Ward, T., Day, A., Howells, K., & Birgden, A. (2004). The multifactor of-
fender readiness model.
Aggression and Violent Behavior, 9
(6), 645-673.
Weinberger, L.E., Sreenivasan, S., Garrick, T., & Osran, H. (2005). The
impact of surgical castration on sexual recidivism risk among sexually
violent predatory offenders.
Journal of the American Academy of Psy-
chiatry and the Law, 33
(1), 16-36.
* Wille, R., & Beier, K.M. (1989). Castration in Germany.
Annals of Sex
Research, 2
, 103-133.
Weisburd, D., Lum, C.M., & Petrosino, A. (2001). Does research design
affect study outcomes in criminal justice?
Annals of the American
Academy of Political and Social Science, 578
, 50-70.
Wilson, D.B. (2001). Meta-analytic methods for criminology.
Annals of
the American Academy of Political and Social Science, 578
, 71-89.
* Worling, J.R., & Curwen, T. (2000). Adolescent sexual offender recidi-
vism: Success of specialized treatment and implications for risk pre-
diction.
Child Abuse & Neglect, 24
(7), 965-982.
* Ziethen, F. (2002).
Rückfallpräventive Effizienz der sozialtherapeutis-
chen Behandlung von Sexualstraftätern. Evaluation der Sozialtherapie
in der JVA Berlin-Tegel.
Unpublished Diplomarbeit, Freie Universität
Berlin.
logo_pie_uaemex.mx