Cognitive‐behavioural therapy (CBT) interventions for young people aged 10 to 18 with harmful sexual behaviour

Abstract

Background

Around 1 in 1000 adolescents aged 12 to 17 years old display problematic or harmful sexual behaviour (HSB). Examples include behaviours occurring more frequently than would be considered developmentally appropriate; accompanied by coercion; involving children of different ages or stages of development; or associated with emotional distress. Some, but not all, young people engaging in HSB come to the attention of authorities for investigation, prosecution or treatment. Depending on policy context, young people with HSB are those whose behaviour has resulted in a formal reprimand or warning, conviction for a sexual offence, or civil measures. Cognitive‐behavioural therapy (CBT) interventions are based on the idea that by changing the way a person thinks, and helping them to develop new coping skills, it is possible to change behaviour.

Objectives

To evaluate the effects of CBT for young people aged 10 to 18 years who have exhibited HSB.

Search methods

In June 2019, we searched CENTRAL, MEDLINE, Embase, 12 other databases and three trials registers. We also examined relevant websites, checked reference lists and contacted authors of relevant articles.

Selection criteria

We included all relevant randomised controlled trials (RCTs) using parallel groups. We evaluated CBT treatments compared with no treatment, waiting list or standard care, irrespective of mode of delivery or setting, given to young people aged 10 to 18 years, who had been convicted of a sexual offence or who exhibited HSB.

Data collection and analysis

We used standard methodological procedures expected by Cochrane.

Main results

We found four eligible RCTs (115 participants). Participants in two studies were adolescent males aged 12 to 18 years old. In two studies participants were males simply described as "adolescents."

Three studies took place in the USA and one in South Africa. The four studies were of short duration: one lasted two months; two lasted three months; and one lasted six months. No information was available on funding sources.

Two studies compared group‐based CBT respectively to no treatment (18 participants) or treatment as usual (21 participants). The third compared CBT with sexual education (16 participants). The fourth compared CBT (19 participants) with mode‐deactivation therapy (21 participants) and social skills training (20 participants). Three interventions delivered treatment in a residential setting by someone working there, and one in a community setting by licensed therapist undertaking a PhD.

CBT compared with no treatment or treatment as usual

Primary outcomes

No study in this comparison reported the impact of CBT on any measure of primary outcomes (recidivism, and adverse events such as self‐harm or suicidal behaviour).

Secondary outcomes

There was little to no difference between CBT and treatment as usual on cognitive distortions in general (mean difference (MD) 1.56, 95% confidence interval (CI) ‐11.54 to 14.66, 1 study, 18 participants; very low‐certainty evidence), assessed with Abel and Becker Cognition Scale (higher scores indicate more problematic distortions); and specific cognitive distortions about rape (MD 8.75, 95% CI 2.83 to 14.67, 1 study, 21 participants; very low‐certainty evidence), measured with the Bumby Cardsort Rape Scale (higher scores indicate more justifications, minimisations, rationalisations and excuses for HSB).

One study (18 participants) reported very low‐certainty evidence that CBT may result in greater improvements in victim empathy (MD 5.56, 95% CI 0.94 to 10.18), measured with the Attitudes Towards Women Scale, compared with no treatment. One additional study also measured this, but provided no usable data.

CBT compared with alternative interventions

Primary outcomes

One study (59 participants) found little to no difference between CBT and alternative treatments on post‐treatment sexual aggression scores (MD 0.09, 95% CI ‐0.18 to 0.37, very low‐certainty evidence), assessed using Daily Behaviour Reports and Behaviour Incidence Report Forms. No study in this comparison reported the impact of CBT on any measure of our remaining primary outcomes.

Secondary outcomes

One study (16 participants) provided very low‐certainty evidence that, compared to sexual education, mean cognitive distortions pertaining to justification or taking responsibility for actions (MD 3.27, 95% CI −4.77 to −1.77) and apprehension confidence (MD 2.47 95% CI −3.85 to −1.09) may be lower in the CBT group. The same study indicated that mean cognitive distortions pertaining to social‐sexual desirability may be lower in the CBT group, and there may be little to no difference between the groups for cognitive distortions pertaining to inappropriate sexual fantasies measured with the Multiphasic Sex Inventory.

Authors' conclusions

It is uncertain whether CBT reduces HSB in male adolescents compared to other treatments. All studies had insufficient detail in what they reported to allow for full assessment of risk of bias. 'Risk of bias' judgements were predominantly rated as unclear or high. Sample sizes were very small, and the imprecision of results was significant. There is very low‐certainty evidence that group‐based CBT may improve victim empathy when compared to no treatment, and may improve cognitive distortions when compared to sexual education, but not treatment as usual. Further research is likely to change the estimate. More robust evaluations of both individual and group‐based CBT are required, particularly outside North America, and which look at the effects of CBT on diverse participants.

Plain language summary

Can cognitive behavioural therapy reduce harmful sexual behaviour in adolescents?

Background

About one out of every 1000 young people aged 12 to 17 years old engages in harmful sexual behaviour like making other children engage in sexual activities. Some are convicted of a sexual offence. Many treatment programmes include cognitive‐behavioural techniques, tailored to individual needs. Cognitive‐behavioural therapy (CBT) is based on the theory that changing the way people think helps change behaviour. It has been used for adults, but it is not known if it works for adolescents with harmful sexual behaviour.

Review question

Is CBT better at reducing adolescent harmful sexual behaviour than no treatment or alternative treatment? We looked at evidence about the effect of CBT on offending rates and adverse events such as self‐harm. We also examined participants' emotional and psychological well‐being, as well as their sexual attitudes and behaviour.

Search date

In June 2019, we searched many databases for randomised controlled trials comparing CBT to no treatment or other treatments. Randomised controlled trials use a random method (like tossing a coin) to decide whether people get different treatments or no treatment.

Study characteristics

We found four small studies including a total of 115 young men exhibiting harmful sexual behaviour. In two studies, participants were aged 12 to 18 years old. In the other two studies the male participants were simply described as "adolescents".

Three studies were conducted in the USA and one in South Africa. Studies were short. One lasted two months; two lasted three months; one lasted six months. We do not know who funded these studies.

Two studies (39 participants) compared CBT to no treatment or treatment as usual. One study (16 participants) compared CBT to a sexual education programme. One study compared CBT (19 participants) with mode deactivation therapy (explicit, systematic, goal‐oriented approach to address problematic emotions, behaviours and thoughts) (21 participants) and social skills training (social skills development and role playing) (20 participants). In three studies, CBT was delivered in a residential setting by someone working there. In the fourth, it was provided in community by a licensed therapist studying for a PhD.

Key results

One study (59 participants) examined whether CBT reduced harmful sexual behaviour, or made participants less likely to offend. It provided very low‐certainty evidence showing CBT reduced sexual aggression at post‐intervention. This was similar to other treatments of mode deactivation therapy and social skills training. No studies examined whether it had unintended consequences such as self‐harm.

One study (59 participants) found little to no difference in how CBT improved psychological well‐being compared with other treatments (very low‐certainty evidence). One study (18 participants) showed CBT meant the young men understood how their behaviour had effected their victims, compared to the no‐treatment group (very low‐certainty evidence). One study (21 participants) measured this, but reported no usable data.

Two studies examined whether CBT improved the kind of thinking associated with harmful sexual behaviour (sexual attitudes and behaviour). One of these (21 participants) compared CBT with treatment as usual. It found no evidence that it made a difference. Another study (16 participants) compared CBT with sexual education. It found CBT improved some types of cognitive distortions. One study (18 participants) reported no significant difference between CBT and receiving no treatment on general cognitive distortions about sexual behaviour (very low‐certainty evidence).

Certainty of the evidence

We cannot tell whether CBT reduces harmful sexual behaviour in male adolescents. The four studies had very small sample sizes. Overall, there is very low‐certainty evidence that group‐based CBT may improve victim empathy compared to no treatment, and may improve cognitive distortions compared to sexual education, but not treatment as usual. The very low‐certainty of this evidence means that the results are likely to change when further studies are carried out. No studies looked at the impact of CBT on girls with harmful sexual behaviour. It was difficult to assess how well the studies were conducted. Available reports did not provide enough information or were rated at high risk of bias in some sections. More, better quality randomised controlled trials of individual and group‐based CBT are needed, particularly outside North America. Evaluations need to also include more diverse participants.

Summary of findings

Summary of findings 1. Cognitive‐behavioural therapy versus no treatment or treatment as usual for harmful sexual behaviour.

Cognitive‐behavioural therapy versus no treatment or treatment as usual for harmful sexual behaviour
Patient or population: young people (aged 10 to 18 years old) with harmful sexual behaviour Settings: community or secure settings Intervention: cognitive‐behavioural therapy (CBT) Comparison: no treatment or treatment as usual (TAU)
Outcomes		Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	Number of participants (studies)	Certainty of the evidence (GRADE)	Comments
		Assumed risk with no treatment or TAU	Corresponding risk with CBT
Recidivism: any sexual or nonsexual offence (no study reported data on this outcome)		‐	‐	‐	‐	‐	No study reported data on this outcome
Adverse events (no study reported data on this outcome)		‐	‐	‐	‐	‐	No study reported data on this outcome
Sexual attitudes and behaviour Follow‐up: end of treatment (last point of data collection)	Cognitive distortions about sexual behaviour ‐ general Measured by: Abel and Becker Cognition Scale (29 items, scores range from 29 to 145; higher scores indicate more problematic cognitive distortions)	The mean score for general cognitive distortions about sexual behaviour in the control group was 29.11	The mean score for general cognitive distortion about sexual behaviour in the intervention group was 1.56 higher (11.54 lower to 14.66 higher)	‐	18 (1 RCT)	⊕⊝⊝⊝ VeryLowa	‐
	Cognitive distortions about sexual behaviour ‐ pertaining to rape Measured by: Bumby Cognitive Cardsort Scale (36 items rated on 4‐point scale where 1 = strongly disagree to 4 = strongly agree; higher scores indicate more justifications, minimisations, rationalisations and excuses for HSB)	The mean score for cognitive distortion pertaining to rape in the control group was −2.33	The mean score for cognitive distortion pertaining to rape in the intervention group was 8.75 higher (2.83 to 14.67 higher)	‐	21 (1 RCT)	⊕⊝⊝⊝ VeryLowb	‐
Thinking patterns (no study reported data on this outcome)		‐	‐	‐	‐	‐	No study reported data on this outcome
Victim empathy: attitudes to women Measured by: Attitude Towards Women Scale (15 items; scores range from 0 to 45; higher scores indicate more egalitarian attitudes towards women) Follow‐up: end of treatment (last point of data collection)		The mean score for victim empathy in the control group was 0	The mean score for victim empathy in the intervention group was 5.56 points higher (0.94 to 10.18 higher)	‐	18 (1 RCT)	⊕⊝⊝⊝ VeryLowa	Karakosta 2015 also measured this outcome, but provided no usable data.
Social functioning (no study reported data on this outcome)		‐	‐	‐	‐	‐	No study reported data on this outcome
Emotional self‐regulation and impulse control (no study reported data on this outcome)		‐	‐	‐	‐	‐	No study reported data on this outcome
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% Cl) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; HBS: Harmful sexual behaviour; RCT: Randomised controlled trial
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect Moderate certainty: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low certainty: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect Very low certainty: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect

^aDowngraded two levels due to very serious imprecision (evidence based only on one RCT with 18 participants), and one level due to study limitations (risk of bias was predominantly high or unclear risk).
^bDowngraded two levels due to very serious imprecision (evidence based only on one RCT with 21 participants), and one level due to study limitations (risk of bias was predominantly high or unclear risk).

Summary of findings 2. Cognitive‐behavioural therapy versus alternative interventions for harmful sexual behaviour.

Cognitive‐behavioural therapy versus alternative interventions for harmful sexual behaviour
Patient or population: young people (aged 10 to 18 years old) with harmful sexual behaviour Settings: community or secure settings Intervention: cognitive‐behavioural therapy (CBT) Comparison: alternative treatment (sexual education programme, mode deactiviation therapy (MDT) and social skills training (SST))
Outcomes		Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	Number of participants (studies)	Certainty of the evidence (GRADE)	Comments
		Assumed risk with alternative treatment	Corresponding risk with CBT
Recidivism: behavioural reports of sexual aggression Measured by: Daily Behaviour Report cards and Behaviour Incidence Report forms completed by staff Follow up: post treatment		The mean score for sexual aggression in the control group was 0.38	The mean score for sexual aggression in the intervention group was 0.09 higher (0.18 lower to 0.37 higher)	‐	59 (1 RCT)	⊕⊝⊝⊝ VeryLowa	‐
Adverse events (no study reported data on this outcome)		‐	‐	‐	‐	‐	No study reported data on this outcome
Sexual attitudes and behaviour Measured by: post‐treatment structured interview (non‐validated) Follow‐up: end of treatment (last point of data collection)	Cognitive distortions about sexual behaviour ‐ pertaining to justification/taking responsibility for actions	The mean score for cognitive distortions pertaining to justification/taking responsibility for actions in the control group was 0.67	The mean score for cognitive distortions pertaining to justification/taking responsibility for actions in the intervention group was 3.27 lower (4.77 lower to 1.77 lower)	‐	16 (1 RCT)	⊕⊝⊝⊝ VeryLowb	‐
	Cognitive distortions about sexual behaviour ‐ pertaining to apprehension confidence	The mean score for cognitive distortions pertaining to apprehension confidence in the control group was 1.17	The mean score for cognitive distortions pertaining to apprehension confidence in the intervention group was 2.47 lower (3.85 lower to 1.09 lower)	‐	16 (1 RCT)	⊕⊝⊝⊝ VeryLowb	‐
	Cognitive distortions about sexual behaviour ‐ pertaining to inappropriate sexual fantasies	The mean score for cognitive distortions pertaining to inappropriate sexual fantasies in the control group was −0.33	The mean score for cognitive distortions pertaining to inappropriate sexual fantasies in the intervention group was 0.13 higher (1.52 lower to 1.78 higher)	‐	16 (1 RCT)	⊕⊝⊝⊝ VeryLowb	‐
	Cognitive distortions about sexual behaviour ‐ pertaining to social‐sexual desirability	The mean score for cognitive distortions pertaining to social‐sexual desirability in the control group was −4.83	The mean score for cognitive distortions pertaining to social‐sexual desirability in the intervention group was 8.53 higher (4.72 higher to 12.34 higher)	‐	16 (1 RCT)	⊕⊝⊝⊝ VeryLowb	‐
Thinking patterns (no study reported data on this outcome)		‐	‐	‐	‐	‐	No study reported data on this outcome
Victim empathy (no study reported data on this outcome)		‐	‐	‐	‐	‐	No study reported data on this outcome
Social functioning (no study reported data on this outcome)		‐	‐	‐	‐	‐	No study reported data on this outcome
Emotional self‐regulation and impulse control (no study reported data on this outcome)		‐	‐	‐	‐	‐	No study reported data on this outcome
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RCT: Randomised controlled trial
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect Moderate certainty: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low certainty: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect Very low certainty: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect

^aDowngraded two levels due to very serious imprecision (evidence based only on one RCT with 59 participants), and one level due to study limitations (risk of bias was predominantly high or unclear risk).
^bDowngraded two levels due to very serious imprecision (evidence based only on one RCTs with 16 participants), and one level due to study limitations (risk of bias was predominantly high or unclear risk).

Background

Description of the condition

Definitions of harmful sexual behaviour

Defining harmful sexual behaviour (HSB) in young people is problematic because of changes across time and culture regarding what is considered 'normal' sexual behaviour, as well as how a 'child' is defined (Fowler 2016; McNeish 2018; Veneziano 2002), and because many do not come to the attention of authorities for investigation, prosecution or treatment (Shlonsky 2017). The ratio of self‐reported to adjudicated sexual offences by young people is estimated to be between 12:1 (Lee 2012) and 25:1 (Elliott 1995) respectively. This may be due to non‐disclosure by victims, unwillingness on the part of victims to prosecute or an inability to provide sufficient evidence. Victims of sexual offences may not disclose for many reasons, including fear, not knowing who to tell, believing what is happening is normal, or communication problems, particularly for children with learning disabilities. In the UK, a prevalence study showed that 83% of young people aged 11 to 17 years old who had been sexually assaulted by a peer had not told anyone about the assault. This is considerably higher than the 34% non‐disclosure rate of those sexually assaulted by an adult (Radford 2011). Under‐reporting may also be, in part, due to a reluctance by communities, particularly in certain cultures, to discuss sexual issues, deviant or otherwise (Grant 2000). Some harmful behaviour may be excused as 'exploratory' in nature and something that the young person will 'grow out of' (Barbaree 2006).

Sexual offences committed by youths encompass a wide spectrum of behaviours, in a variety of situations, with many types of victims (Hackett 2019; Kemper 2010; Righthand 2004; Ryan 1997; Scottish Government 2020). What is defined as a sexual offence differs across jurisdictions, as does the age of minimum criminal responsibility. In most jurisdictions, young people aged 12 years or older are considered by law to be old enough to be held criminally responsible, but not sufficiently old to be subject to the full range of adult criminal sanctions (Barbaree 2006; Hoghughi 1997). Civil measures of welfare, care, assistance, diversion programmes and protection may be triggered when children younger than the age of criminal responsibility are suspected of illegal acts. On attaining the age of criminal responsibility, the possibility exists of adult penal procedures and sanctions, although these are not always used and 'age threshold offenders' may not come to the attention of the criminal justice system (Cipriani 2009; Department of Health and Home Office 2006). Table 3 contains details of the minimum age of responsibility for criminal activities in various countries (Cipriani 2009). For the purposes of this review, a young person with HSB is defined as a person aged between 10 and 18 years old who has been reprimanded, warned or convicted of a sexual offence, or who has received civil measures for their sexual offending.

1. Minimum age of responsibility for criminal activities in various jurisdictions (Cipriani 2009).

Country	Age in years of criminal responsibility
USA	6‐12 (10 for federal crimes)
Egypt, Estonia, India, Mayanmar, Singapore, Thailand	7
Scotland, Indonesia	8
Bangladesh, Ethopia	9
Iran	9 for girls, 15 for boys
Austria, Australia, England, Northern Ireland, South Africa, Wales, Switzerland	10
Japan	11
Belgium, Brazil, Canada, Netherlands, Mexico, Morocco, Ireland, Israel, Portugal, Turkey, Uganda	12
Algeria, Greece	13
Bulgaria, China, Columbia, Germany, Hungary, Italy, Peru, Romania, Russian Federation, Slovenia, Spain, Ukraine, Veit Nam	14
New Zealand	14 (children can be charged with murder, manslaughter or minor traffic offences from 10 years of age; all other offences cannot be charged under 14 years of age)
Czech Republic, Denmark, Finland, Iceland, Norway, Philippines, Sweden	15
Argentina	16

Prevalence

There are few studies upon which to base population estimates of the prevalence of HSB and they all vary considerably in their estimates (NSPCC 2019; Warner 2015). In the UK, estimates of official cases over a year suggest that about one in 1000 young people aged 12 to 17 years old is identified as displaying HSB (Department of Health and Home Office 2006). Around a third of child sexual abuse is by other children or young people, with estimates ranging from one fifth to two thirds (Hackett 2014; NSPCC 2019; Radford 2011). Young people comprise a large proportion of those reprimanded, warned or convicted of sexual offences. In England and Wales, the number of police recorded sexual offences against under‐18 year olds, by under 18‐year olds, rose from 5215 in 2013 to 9290 in 2016 (NSPCC 2019). Children and young people aged 10 to 19 years old were the alleged offenders in 26% of sexual offences committed in Australia between 2015 and 2016 (Australian Bureau of Statistics 2017). In Germany, data from the Police Criminal Statistics Bureau indicates that juveniles aged between 14 and 20 years old are over‐represented in the category of 'sexual offences' (equivalent to one young person convicted of HSB per 1000 young people aged 14 to 20 years old in the general population) (Bullens 2004). Many young people displaying HSB have more than one victim: a study in The Netherlands found that the victim(s) were known to the offender in two thirds of the cases; half of the youths had victimised one person, 15% had two victims and 35% had victimised more than two people (Bruinsma 1996). To date, there has been little research into technology‐assisted HSB, although it is a potentially growing population (Hollis 2017; Lewis 2018). There is some cross‐over between online and offline HSB and between child sexual exploitation and HSB (Belton 2016; NSPCC 2019).

Profile

Young people engaging in HSB are not a homogeneous group. They show great variation in terms of their backgrounds, presenting problems and types of HSB, including those categorised as offences (Balfe 2020; Letourneau 2014; Malvaso 2020; Murphy 2017; Rich 2009; Veneziano 2002). Assessment tools may screen for whether behaviour is developmentally normal or HSB, or assess HSB behaviours and risk of recidivism. The evidence around these tools is very limited (Campbell 2016b; Schwartz‐Mette 2019; Worling 2017).

Risk factors for adolescent sexual offences are very similar to those observed for other forms of serious antisocial behaviour (Letourneau 2009; Van Wijk 2005). There are currently no validated classifications of young people with HSB. Some attempts to define subtypes have been offence‐driven (for example, rapists, child molesters), whereas others have been based on personality (for example, disturbed impulsive, pseudo‐socialised) (Veneziano 2002). Others have suggested categorisation by victim age (offending against children versus adolescents or adults), co‐offender status (offending as individuals or in groups) and crime history (with or without a previous history of crime) (Aebi 2012). Some have been described as having poor interpersonal skills, being socially isolated, lacking empathy and evidencing distinctive patterns of making claims on others (Epps Fisher 2004; Lane 1997). There may be evidence of psychiatric disorder (for example, conduct disorder, attention deficit disorder and adjustment disorder), low self‐esteem, depressive or anxious symptoms, antisocial behaviour (especially amongst juvenile rapists) and neuropsychological impairment (for example, impulse control problems, poor planning, lifestyle impulsivity and deficits in verbal cognitive functioning) (Sheerin 2004).

Some have a history of maltreatment or victimisation (physical, sexual or emotional abuse, neglect, witness to family violence) (Barra 2018; Bruinsma 1996). Being sexually abused as a child has been strongly linked to young people displaying HSB (Shlonsky 2017). They may exhibit cognitive distortions relating to the negative effects of the offence, blaming the victim and sexual attitudes or knowledge. Deviant sexual arousal may be reported (Seto 2010). Although not all families are dysfunctional, some are characterised by poor relationships, negative communication styles, instability, disorganisation and violence. The young person may have experienced a lack of adequate support and supervision, parent‐child attachment problems and physical or emotional separation from one or both parents (Bartosz 2016). Their parents may also have experienced mental health problems (Duane Morrison 2004; Långström 1999). Whilst some young people experience academic and behavioural problems in school, as well as learning difficulties, others show above average cognitive performance (Veneziano 2002).

Significant differences have been reported between younger perpetrators and older children. Pre‐adolescent children are likely to show less problematic or severe HSB, whilst the early teens are the peak time for HSB, most of which is displayed by boys (McNeish 2018). Studies show between 92% and 97% of young people showing HSB are male and 3% to 8% are female (NSPCC 2019).

Young people with learning disabilities are over‐represented in treatment services, although their offending behaviour is not thought to be linked to their learning disability per se (Department of Health and Home Office 2006). Whilst there are many similarities between developmentally‐disabled and non‐disabled youth in the range, types and elements of HSB, there are some differences in the associated cognitive process, the context of the behaviours and the level of sophistication (Lane 1997). One study found that these young people are more likely to suffer impulse control; less likely to be redirected by adults when displaying concerning behaviours; and less likely to receive sex education (Evertsz 2012). They are also at a greater risk of experiencing abuse and neglect, which can put them at an increased risk of trauma and sexualization (Evertsz 2012). Although there are few follow‐up studies, a recidivism rate of almost 31% has been reported for sexual offences among offenders with an intellectual disability, with the vast majority of re‐offences (84%) occurring in the first 12 months (O'Callaghan 2004). With respect to gender differences, young female perpetrators may be less likely than males to use violence, and there can be differences in perceptions, affective reactions and internal experience (Lane 1997). Juvenile females with HSB may be more likely than males to be slightly younger, more likely to be white and more likely to have a co‐offender. Females are less likely than males to commit rape and less likely to be processed formally by law enforcement (Vandiver 2010). Some studies suggest that females are also more likely to have been sexually victimised themselves (Bumby 2004). Young people who show violence as well as HSB manifest extreme offences, attitudes, dynamics and behaviours reflecting a disregard for other people's safety and welfare. They exhibit considerable psychopathy and their offences may include using weapons to injure their victims, sadism, ritual abuse or murder. They are unlikely to be treated in the community and overall the prognosis for change is thought to be poor (Lane 1997).

Most young people do not continue HSB into adulthood, although there is a subgroup at high risk for doing so (Cale 2016; Department of Health and Home Office 2006; Dopp 2017; Hackett 2013). Re‐arrest rates for registered juveniles with HSB in adulthood are as low as 5% (Vandiver 2006). Re‐offending rates for juveniles in England and Wales were 15.6% for sexual offences in 2016 (Ministry of Justice 2018). Predicting which adolescents are at greatest risk to sexually recidivate is very difficult. There is limited knowledge about which predictors are most accurately linked to sexual recidivism (Caldwell 2010), and uncertainty over how to best use instruments designed to predict re‐offending (Martinez 2007; Vitacco 2009), including use with particular groups such as individuals with intellectual disabilities (Griffin 2012) and non‐Western populations (Chu 2011). Young people with HSB are more likely to re‐offend with non‐sexual offences than sexual recidivism: observed recidivism rates for new sex offences tend to be quite low at around 10%. In a two‐year follow‐up study, young people convicted of a sexual offence were nearly 10 times more likely to have been charged with a non‐sexual offence than a sexual offence (Caldwell 2007). This highlights the need for interventions to focus on broad‐based behavioural and developmental goals and not just on preventing further sexual offending (Borduin 2009; Hackett 2004). Around half of all adults convicted of sexual offences report that their sexual deviance began during their childhood, and often their offences escalated in frequency and severity over time (Veneziano 2002; Zolondek 2001). It is important to identify those young people at risk of further offending and provide them with effective interventions and support. Services should avoid stigmatising young people as 'mini adult sex offenders' (McNeish 2018). Risk factors for both sexual and non‐sexual recidivism include unhealthy family environments, negative peer affiliations, social isolation and chronic or pervasive antisocial values and behaviours. Being highly impulsive, holding attitudes supportive of abusive behaviours and failing to complete treatment (or being terminated unsuccessfully from treatment) are also predictors. Risk factors specific to sexual recidivism include deviant sexual arousal, sexual compulsivity, sexual preoccupation, past sexual offences against two or more victims and the targeting of strangers as victims. These are similar to the risk factors of sexual recidivism amongst adults convicted of sexual offences (Worling 2003). An eight‐year follow‐up study of Canadian adolescents who had sexually offended found recidivism rates of 45% for a new criminal offence, 30% for a violent offence and 10% with a sexual offence. Paternal abandonment, childhood sexual victimisation, association with significantly younger children and having victimised a stranger were associated with a higher risk of sexual recidivism (Carpentier 2011).

Management of children and young people who present with HSB

Often the response to young people engaging in HSB is one that seeks to restrain or contain the young person by means of restrictions on their liberty, exclusions (from school, or social gatherings) or surveillance of some kind. Clinical treatments for juvenile HSB include behavioural conditioning, pharmacological responses, family systems interventions, rational‐emotive counselling, music and art therapy, 'cycle'‐based approaches, cognitive‐behavioural therapy (CBT), relapse prevention programmes and ecological multisystemic approaches (Borduin 2015; Dopp 2015). Treatment programmes are often tailored for individual needs and may combine several approaches, even when they are theoretically quite different (Allardyce 2018; Campbell 2016a; Chaffin 2002; NICE 2016). Few are manualised (Marsh 2019). Treatment may be delivered individually or in a peer‐group environment, although concern has been expressed that this can sometimes have negative effects through delinquent peer influences and socialisation into delinquent behaviour and belief patterns (Chaffin 2002). More programmes now involve the young person's family (Calvert 2019; Campbell 2020; Dopp 2017; Letourneau 2009; Thomas 2004).

Description of the intervention

Cognitive‐behavioural therapy (CBT) is a highly structured, psychological therapy. A key aspect of the therapy is an educative approach whereby, through collaboration and guided discovery, the person learns to recognise negative or undesirable thinking patterns, re‐evaluate these thoughts and practice new ways of thinking and behaving. This, in turn, leads to an increase in protective factors and changes in overt behaviour. The therapeutic relationship in CBT is also thought to be an important element (Easterbrook 2017). CBT aims to work from a strength‐based model to change a broad range of internal processes (for example, thoughts, beliefs, emotions, physiological arousal, correction of offender misperceptions and reasoning biases associated with the offending behaviour) as well as overt behaviours (for example, social skills or coping behaviours) (Bilby 2012). It may be used in conjunction with other approaches and may be delivered individually or in groups. Many treatment programmes for young people with HSB use cognitive‐behavioural techniques (Richardson 1997; Veneziano 2002). These are also the basis of treatment in prison settings and community programmes in England, Canada, New Zealand and the USA for adults convicted of sexual offences (Bilby 2012; Hanson 2009; Marques 2005; Mews 2017). There is concern that these dominant interventions (that is, cognitive‐behavioural group treatments with an emphasis on relapse prevention) may fail to address the multiple determinants of juvenile HSB and could inadvertently result in adverse (iatrogenic) outcomes (Letourneau 2008).

How the intervention might work

The majority of people who have committed a sexual offence, irrespective of age, appear to hold ideas and beliefs about sexuality and interpersonal relationships that condone using others for sexual gratification (Grant 2000; Schmucker 2015). Given the importance of adolescence for developing self‐identity and social roles, cognitive distortions that serve to justify offending and reduce the offender's acceptance of responsibility need to be addressed with effective and timely interventions. Cognitive‐behavioural interventions are based on the idea that by changing the way a young person thinks, and helping them to develop new coping skills, it is possible to change their behaviour. Cognitive‐behavioural interventions for harmful sexual behaviour consist of three core elements: intervention aimed at increasing the offender’s accountability for their offending; work on relapse prevention; and work to address criminal thinking and factors associated with the development and maintenance of all forms of criminal behaviour. Treatment goals typically include: increasing empathy; enhancing problem‐solving skills and self‐awareness; decreasing cognitive distortions and deviant sexual arousal; sex education; improving social skills; resolving trauma; and improving anger management (Veneziano 2002). Co‐occurring problems, such as substance abuse, may also be addressed.

Why it is important to do this review

Young people showing HSB are, themselves, individuals in need. HSB are potentially damaging for the children who display them because HSB in children challenges social norms, and consequently, some adults and children may respond by labelling, isolating or condemning the child (Barter 2011). If addressed early and effectively, however, youth with HSB have a high rate of recovery and may experience less social exclusion (O'Brien 2010). Early intervention leads to the best rehabilitative outcomes for the children and young people involved (O'Brien 2010). Research suggests that between 20% and 30% of adults who commit sexual offences begin their offending in adolescence, representing a further imperative to intervene early (Evertsz 2012).

It is important, therefore, to try and help young people showing HSB before their beliefs and behaviour become entrenched and difficult to change. Treatment should be age‐appropriate and based on age‐appropriate assessment (Calder 1997; Calder 1999; Department of Health 1999). Although often recommended, psychological treatment for young people with HSB is based largely on modified adult programmes, whose efficacy with adolescents remains largely unproven (Chaffin 2002). 

Although young people with HSB consume much of the resources of the criminal justice, educational and mental health systems, relative to their small numbers, few empirically supported interventions exist to treat these youths (Borduin 2009; Walker 2004). Specialised treatment programs used for young people with HSB within the juvenile justice system have not been evidenced as any more effective for reducing sexual recidivism than general treatment as usual (TAU) (Kettrey 2018). Given the prevalence, as well as the often devastating effects on victims, it is important that we continue to develop effective, cost‐beneficial methods of reducing future risk (Marshall 2000). Recidivism rates for untreated adolescents are 17.8%, compared to 5.17% of those treated (Worling 2000). A meta‐analysis of the effectiveness of HSB treatment for juveniles reported the following recidivism rates over an average 59‐month follow‐up period: 12.53% sexual crimes; 24.73% non‐sexual violent crimes; 28.51% non‐sexual non‐violent crimes; and 20.40% unspecified non‐sexual crimes (Reitzel 2006). There has been a shift in practice towards using earlier intervention, community‐based rehabilitation approaches rather than incarceration, and family‐based approaches that may involve CBT (Balsamo 2016). Research has only begun to evaluate these programs for youth with HSB. Although systematic reviews exist for adults convicted of sexual offences (Dennis 2012; Khan 2015), offenders with a learning disability who have committed sexual offences (Ashman 2008), and other treatments such as Multi‐Systemic therapy (MST) with youth (Littell 2005), no recent systematic review has focused on CBT even though it is used with adolescents.

This systematic review will provide policymakers with a synthesis of evidence about the effects of cognitive‐behavioural interventions with this vulnerable group (Chaffin 2008). Since treatment aims to reduce recidivism, the importance of examining the efficacy of treatment programmes is an integral part of child protection responses, as well as having implications for juvenile justice. 

Objectives

To evaluate the effects of CBT for young people aged 10 to 18 years who have exhibited HSB.

Methods

Criteria for considering studies for this review

Types of studies

Randomised controlled trials (RCTs) that used parallel groups.

Types of participants

Young people aged between 10 and 18 years old who have received treatment in any setting for sexual offences or HSB.

Types of interventions

CBT compared with no treatment, waiting list, or standard care (defined as the care a person would normally receive had they not been included in the research trial). Where CBT was a major rather than sole component of service, we only included studies comparing CBT plus an adjunctive treatment with that same adjunctive treatment alone. This is so that we could assess the efficacy of the CBT component alone.

CBT is defined as an intervention that involves: (i) working with recipients to establish the links between their thoughts, feelings and actions; (ii) correcting misperceptions, irrational beliefs and reasoning biases related to target symptom/s; and (iii) either or both of the following: (a) recipients monitoring their own thoughts, feelings and behaviours with respect to target symptom/s, and (b) promotion of alternative ways of coping with target symptom/s (Brooks‐Gordon 2006).

We included studies if they used CBT as defined here, solely or if they included CBT as a well‐defined, major part (≥ 50%) of a broader intervention, irrespective of mode of delivery.

Types of outcome measures

Primary outcomes

Primary outcomes of concern for young people with HSB are whether they are likely to repeat the HSB behaviour or commit some other unacceptable behaviour (recidivism), or whether they harm themselves (self‐harm or suicide):

• Recidivism

  • Any sexual offence

  • Behavioural reports of sexual aggression

  • Any nonsexual offence

  • Time before re‐offence

• Adverse events (e.g. Juvenile Risk Assessment Scale (Hiscox 2007))

  • Increase in sexual offending

  • Increased seriousness of sexual offending

  • Self‐harm

  • Suicide attempt

  • Suicide

Secondary outcomes

• Reactions to the offending behaviour (e.g. Juvenile Sex Offender Assessment Protocol‐II (J‐SOAP‐II; Prentky 2003); Juvenile Sexual Offence Risk Assessment Tool‐II (J‐SORRAT‐II; Epperson 2014); or Estimate of Risk of Adolescent Sexual Offence Recidivism (ERASOR; Worling 2004))

  • Offence accountability: accepting responsibility for actions

  • Denial/minimisation: acknowledging they engaged in the behaviour or some aspect of the offence.

• Coping skills (e.g. ERASOR (Worling 2004))

• Psychological well‐being (e.g. Children's Depression Scale (Kovacs 2001))

  • Self‐esteem

  • General mental state

• Sexual attitudes and behaviour (e.g. J‐SOAP‐II (Prentky 2003); J‐SORRAT‐II (Epperson 2014); ERASOR (Worling 2004); Hare Psychopathy Checklist: Youth Version (PCL:YV; Hempel 2013); Attitudes Towards Women Scale (Spence 1972); or Bumby Cardsort Rape Scale (Bumby 1996))

  • Cognitive distortions about sexual behaviour

  • Deviant sexual interests, preferences or arousal

• Thinking patterns (e.g. J‐SOAP‐II (Prentky 2003); ERASOR (Worling 2004))

• Victim empathy (e.g. Attitudes Towards Women Scale (Spence 1972); Measures for the Assessment of Dimensions of Violence Against Women: A Compendium (Flood 2008))

• Aggression (e.g. Structured Assessment of Violent Risk in Youth (Borum 2002))

• Social functioning (PCL‐YV (Hempel 2013))

• Emotional self‐regulation and impulse control (e.g. J‐SOAP‐II (Prentky 2003); ERASOR (Worling 2004))

• Substance use (e.g. CRAFFT (i.e. Car, Relax, Alone, Forget, Friends, Trouble) (Knight 2002))

• Programme engagement

  • Completion of treatment programme

  • Dropouts

  • Treatment refusers

• Economic outcomes

  • Direct costs

  • Indirect costs

These outcome measures could have been assessed through validated questionnaires, structured interviews or analysis of case‐history information. We extracted data from independently validated measures. If measures had not been independently validated (for example, tools developed by the research team), we noted these as being at higher risk of bias.

See also Sneddon 2012 and Table 4 for additional methods not used in this version of the review.

2. Additional Methods Table.

Intendedmethods specified in protocol (Sneddon 2012) but not used in thereview		Reason for non use
Types of outcome measures	We intended, if possible, to either code the follow‐up period and then treat it as a continuous variable, or else divide outcomes into immediate (within six months), short term (greater than (>) six to 24 months), medium term (> 24 months to five years) and long term (> five years). We intended to draw information from psychometric tests, as well as police or other official data.	Insufficient data were presented in the results sections of eligible studies to code the data in this way.
	Economic outcomes Direct costs Indirect costs	We intended to extract any economic information included in study descriptions but none was available.
Search methods	We intended to search Criminal Justice Abstracts EBSCOhost in June 2019.	We no longer had access to the database.
Measures of treatment effect	Dichotomous outcome data We aimed to calculate odds ratios and 95% confidence intervals for dichotomous outcomes. For meta‐analyses of dichotomous outcomes that would have been included in 'Summary of findings' tables, we aimed to express the results as absolute risks, using high and low observed risks among the control groups as reference points.	Insufficient data were available.
	Continuous outcome data We aimed to calculate mean differences if all studies use the same measurement scale, or standardised mean differences if studies use different measurements scales, and 95% confidence intervals for continuous outcome measures. If necessary, we aimed to compute effect estimates from P values, T statistics, ANOVA (analysis of variance) tables or other statistics as appropriate. We aimed to calculate standardised mean differences using Hedges g.
	Multiple outcomes Had a study provided multiple, interchangeable, measures of the same construct at the same point in time (for example, multiple measures of obsessive thoughts), we aimed to calculate the average standardised mean difference across these outcomes, and the average of their estimated variances. This strategy aims to avoid the need to select a single measure, and to avoid inflated precision in meta‐analyses (preventing studies which report on more outcome measures receiving more weight in the analysis than comparable studies that report on a single outcome measure).
Dealing with missing data	We aimed to assess the sensitivity of any primary meta‐analyses to missing data using the strategy recommended by Higgins 2008.
Assessment of heterogeneity	We aimed to describe statistical heterogeneity by computing the I2 statistic (Schünemann 2019), a quantity which describes approximately the proportion of variation in point estimates that is due to heterogeneity rather than sampling error. In addition, we aimed to employ a Chi2 test of homogeneity, to determine the strength of evidence that heterogeneity was genuine.
Assessment of reporting biases	We aimed to draw funnel plots (estimated differences in treatment effects against their standard error) if we had found sufficient studies. Asymmetry could have been due to publication bias, or due to a real relationship between trial size and effect size, such as when larger trials have lower compliance and compliance is positively related to effect size. In the event that we found such a relationship, we aimed to examine clinical variation of the studies (Schünemann 2019). As a direct test for publication bias, we aimed to compare results extracted from published journal reports with results obtained from other sources (including correspondence).
Data synthesis	Where the interventions, comparators, participants and outcomes were the same, we aimed to synthesise the results in a meta‐analysis. Unless the model was contra‐indicated (for example, if there was funnel plot asymmetry), we planned to present the results from the random‐effects model. In the presence of severe funnel plot asymmetry, we would have presented both fixed‐effect and random‐effects analyses, in the knowledge that neither model is appropriate. If both indicated a presence (or absence) of effect we would have been reassured; if they did not agree, we aimed to report this. We aimed to calculate all overall effects using inverse variance methods. If some primary studies reported an outcome as a dichotomous measure and others used a continuous measure of the same construct, we aimed to convert results for the former from an odds ratio to a standardised mean difference, provided that we could assume the underlying continuous measure had approximately a normal or logistic distribution (otherwise we would have carried out two separate analyses).
Subgroup analysis and investigation of heterogeneity	If sufficient studies were found, we aimed to undertake the following subgroup analysis. Age at time of treatment Gender Location of treatment (institutional, community) Modality of treatment (individual versus group, or combination) Participants with or without a learning disability Violent sexual offending
Sensitivity analysis	We aimed to conduct sensitivity analyses to examine the robustness of the findings. This would have been done by exploring whether findings were sensitive to restricting the analyses to studies judged to be at low risk of bias. In these analyses, we aimed to restrict the analysis to: (a) only studies with low risk of selection bias (associated with sequence generation or allocation concealment); (b) only studies with low risk of performance bias (associated with issues of blinding); (c) only studies with low risk of attrition bias (associated with completeness of data). In addition, we aimed to assess the sensitivity of findings to any imputed data.

Search methods for identification of studies

Electronic searches

We searched the electronic databases and trial registers listed below in August 2014 and June 2019. Information was only available from Criminal Justice Abstracts until 1 August 2014, as we did not have access after that date.

• Cochrane Central Register of Controlled Trials (CENTRAL; 2019, Issue 6) in the Cochrane Library, which includes the Developmental, Psychosocial and Learning Problems Specialised Register (searched 26 June 2019).

• MEDLINE Ovid (1946 to 26 June 2019).

• Embase Ovid (1980 to 26 June 2019).

• PsycINFO Ovid (1967 to 26 June 2019).

• CINAHL EBSCOhost (Cumulative Index to Nursing and Allied Health Literature; 1937 to 26 June 2019).

• Conference Proceedings Citation Index ‐ Social Science & Humanities (CPCI‐SS&H; 1990 to 26 June 2019).

• Social Sciences Citation Index Web of Science (SSCI; 1970 to 26 June, 2019).

• Cochrane Database of Systematic Reviews (CDSR; 2019, Issue 6), part of the Cochrane Library (searched 26 June 2019).

• Database of Abstracts of Reviews of Effectiveness (DARE; 2015, Issue 2. Final Issue), part of the Cochrane Library (searched 26 June 2019).

• LILACS (Latin American and Caribbean Health Science Information database; lilacs.bvsalud.org/en; searched 26 June 2019).

• Criminal Justice Abstracts EBSCOhost (searched 1 August 2014. Not searched in June 2019 as we no longer had access to the database).

• Social Care Online (www.scie-socialcareonline.org.uk; searched 26 June 2019).

• ProQuest Dissertations & Theses: UK & Ireland (searched 26 June 2019).

• Networked Digital Library of Theses and Dissertations (NDLTD; www.ndltd.org/resources/find-etds; searched 26 June 2019).

• WorldCat (www.worldcat.org; searched 26 June 2019).

• ClinicalTrials.gov (clinicaltrials.gov; searched 26 June 2019).

• UK Clinical Research Network (UKCRN; www.ukcrc.org/research-infrastructure/clinical-research-networks/uk-clinical-research-network-ukcrn searched 26 June 2019).

• World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) (www.who.int/ictrp/en; searched 26 June 2019).

The exact search strategies are reported in Appendix 1. We did not apply any date or language restrictions.

Searching other resources

We used Google and Google Scholar to identify websites of relevant organisations to search for relevant studies. We searched the reference lists of included studies for additional trials, as well as the reference lists of relevant reviews found by searching the CDSR and DARE (see Electronic searches). In addition, we contacted the first author of each included study, as well as known experts in the field, for information regarding ongoing studies and unpublished data.

Data collection and analysis

In what follows, we detail only those methods that were deployed in this review. Other methods outlined in our published protocol, Sneddon 2012, but which were not required or possible to deploy in this first version of the review, are summarised in Table 4.

Selection of studies

Three review authors (HS, NL, DGG) independently assessed the titles and abstracts of all records retrieved from the searches and selected all that were potentially relevant. Working independently, HS and DGG obtained their full‐text reports and reviewed them against the inclusion criteria (Criteria for considering studies for this review). Review authors were not blinded to the names of the trial authors, institutions or journal of publication. Any disagreement was resolved by consensus following discussion with GM. We report the outcomes of our selection process in a PRISMA flow diagram (Moher 2009; Schünemann 2019).

Data extraction and management

For each included study, two review authors (HS, DGG) independently extracted and recorded the following data using a piloted data collection form, specifically designed for this review: study design and methods; sample characteristics; intervention characteristics (including theoretical underpinning of services, delivery, duration, outcomes and within‐intervention variability); outcomes; time points; and outcome measures. See Appendix 2.

In the event of disagreements, review authors first discussed these with reference to the study papers and, when necessary, sought clarification from the trial investigators until a consensus was reached ‐ in pairs (i.e. HS and DGG only). We collected information on study design and implementation in a format suited to completion of the 'Risk of bias' tables to appear in the completed review (Higgins 2017).

Assessment of risk of bias in included studies

We assessed the risk of bias of each included study using Cochrane's 'Risk of bias' tool (Higgins 2017). Two review authors (HS, DGG) independently assessed the risk of bias within each included study across the following, seven domains, and assigned ratings of low, high or unclear (uncertain) risk of bias: sequence generation; allocation concealment; blinding of participants and personnel; blinding of outcome assessment; incomplete outcome data; selective reporting; and other sources of bias (Appendix 3). When disagreements occurred between the judgements of the reviewers, they first sought to resolve the disagreements themselves. If needed, they consulted with GM.

Measures of treatment effect

We entered extracted data into Review Manager 5 (RevMan 5) (Review Manager 2014), and present these using the mean difference (MD) and its 95% confidence interval (CI). For additional methods archived for use in future updates of this review, please see our protocol, Sneddon 2012, and Table 4.

Unit of analysis issues

There were no cluster‐randomised or cross‐over trials in this area. See Sneddon 2012 and Table 4 for methods to manage these types of studies should we identify any in future updates of this review.

If a study had multiple treatment arms, we combined the relevant arms to make a single pair‐wise comparison, using RevMan 5 (Review Manager 2014)

Dealing with missing data

Where necessary, we attempted to contact the corresponding or other authors (or both) of the included studies to supply any unreported data (for example, group means and standard deviations (SDs), details of dropouts and details of interventions received by the control group). However, none responded to our requests. We described missing data and dropouts/attrition for each included study in the 'Risk of bias' tables (beneath the Characteristics of included studies tables), and discussed the extent to which the missing data could alter the results and conclusions of the review.

Assessment of heterogeneity

We assessed clinical variation across studies by comparing the distribution of important participant factors among trials (for example, age), and trial factors (randomisation concealment, blinding of outcome assessment, losses to follow‐up, treatment type, co‐interventions). See our protocol, Sneddon 2012, and Table 4 for how we will assess statistical heterogeneity in future updates of the review.

Assessment of reporting biases

We identified no unpublished data to compare with published journal reports. See our protocol, Sneddon 2012, and Table 4 for information on how we will investigate reporting biases should we identify sufficient studies in future updates of this review.

Data synthesis

There were insufficient comparable data to synthesise data in a meta‐analysis, or to undertake subgroup or sensitivity analyses. Please see our protocol, Sneddon 2012, and Table 4 for information on these methods, archived for use in future updates of this review.

We present a narrative synthesis of the data in the Effects of interventions section.

Summary of findings

Using GRADEprofiler (GRADEpro) software (GRADEpro GDT), we prepared 'Summary of findings' tables for the major comparisons of the review: CBT versus no treatment or TAU; and CBT versus alternative interventions.

We present the following outcomes assessed following treatment in the tables (see Types of outcome measures).

• Recidivism

• Adverse events

• Sexual attitudes and behaviour

• Thinking patterns

• Victim empathy

• Social functioning

• Emotional self‐regulation and impulse control

We provided a source and rationale for each assumed risk cited in the tables, and a rating of high, moderate, low or very low for the certainty of the evidence based on the presence of the GRADE criteria listed below and described in Chapter 11 of the Cochrane Handbook for Systematic Reviews of Interventions (Schünemann 2019).

• Risk of bias

• Imprecision

• Inconsistency

• Indirectness

• Publication bias

In terms of risk of bias, two review authors (HS with either DGG or NL) independently assessed the risk of bias within each included study across the following seven domains, and assigned ratings of low, high or unclear (uncertain) risk of bias: sequence generation; allocation concealment; blinding of participants and personnel; blinding of outcome assessment; incomplete outcome data; selective reporting; and other sources of bias (Appendix 2). When disagreements occurred between the judgements of the reviewers, they first sought to resolve the disagreements themselves; if needed, they consulted with GM.

Results

Description of studies

Results of the search

Our searches yielded 5527 records (5455 from database searching and 72 from other sources). Once duplicates had been removed, 3637 records remained. We excluded 3523 records based on their titles and abstracts. We then retrieved full texts of the remaining 114. We brought forward 30 records that were agreed from the title as relevant for full‐text review, and a further 84 records from title/abstract screening to check relevance. GM arbitrated the decision in some of these and we tried to secure the full texts for the rest. We extracted data from 114 of the full‐text reports and examined them.

Out of 114 reports, we included five reports of four studies in the review and excluded 109 reports. Of these, 18 studies (from 19 reports) were formally excluded, with reasons (see Characteristics of excluded studies tables). We identified no ongoing studies. (See Figure 1)

1.

Prisma Study flow diagram.

We contacted the author of one included study for further information, but without success (Mathѐ 2007). One study involved children aged between 6 and 12 years old, and we contacted the first author to obtain further information about the participants in our age range only (Bonner 1999), but received no response (see Characteristics of excluded studies tables).

Included studies

Four studies (115 participants) met the eligibility criteria for this review (Apsche 2005; Karakosta 2015; Mathѐ 2007; Piliero 1994).

Study design

All four studies were parallel‐group RCTs (Apsche 2005; Karakosta 2015; Mathѐ 2007; Piliero 1994).

Location

Three studies took place in the USA (Apsche 2005; Karakosta 2015; Piliero 1994), and one took place in South Africa (Mathѐ 2007).

In three studies the interventions were delivered in a residential setting by a person who worked there (Apsche 2005; Karakosta 2015; Mathѐ 2007), and in the other study, the intervention was delivered in a community setting by research students (Piliero 1994).

Participants

Age

Participants in the four included studies were under 18 years of age. Two studies stated that they included children older than 12 years of age (Karakosta 2015; Piliero 1994). Two studies listed participants as "adolescents" (quote) but gave no further age qualifiers (Apsche 2005; Mathѐ 2007).

Gender

All four studies included males only.

Offence profile

Participants in three studies were all sentenced for a sex offence, which they served in a residential setting (Apsche 2005; Karakosta 2015; Mathѐ 2007). The participants in Piliero 1994 were all still in the community but had been referred to the study for sexually inappropriate behaviour. In Piliero 1994, the participants were either on probation for a sex offence, or were volunteered by their parents because they had previously admitted to sexual contact with a sibling.

All four studies included participants with a history of violent sexual offending. Two studies included participants who, in addition to being offenders, were also victims of sexual abuse (Karakosta 2015; Piliero 1994).

Attrition

Two studies lost participants due to attrition. Piliero 1994 started with 20 participants but lost four due to a change in probationary status. Karakosta 2015 started with 25 participants, but also lost four, in this case, due to being moved to a different institution.

Other

Two studies included children who had a developmental delay or a learning difficulty (Karakosta 2015; Mathѐ 2007). The studies provided no other details or data for these young people as a subset of the samples.

Profile of therapists

In Apsche 2005, all staff and therapists were trained and supervised in Social Skills Training (SST) by a doctoral‐level psychologist. The expertise of the trainer in each treatment condition was noted as equal. The therapists shared the same professional degree and level of clinical experience in each of the three treatments and training was provided in each model prior to the study.

In Karakosta 2015, therapists for TAU were graduate‐prepared and were either licensed or license‐eligible by the state. Two Integrated Sex Offender Treatment Programme (ISOTP) therapists were licensed therapists but one resigned in the middle of the study period. The remaining therapist was thus required to train a new therapist about ISOTP services. The subsequent training may have affected the consistency and efficacy of the services provided by the new hire.

In Mathѐ 2007, the therapist was also the researcher. This approach, in itself, automatically removes any possibility for blinding, which has implications for the objectivity of the measurements and findings. There is no information on whether the therapist was licensed in CBT.

In Piliero 1994, one of the two therapists was the researcher herself who was undertaking her doctoral thesis. Both therapists had over seven years’ experience in sex education and CBT and were licensed therapists. There is no information on the therapists who delivered the alternative programme; presumably it was the same two therapists.

Interventions

Each of the four studies investigated the effects of a CBT treatment. Treatment protocols for three studies relied on standardised treatment curriculum, and offered self‐instructional training and social‐cognitive skills training through modules, which focused on relationships and sexuality, social and life skills and cognitive restructuring (Apsche 2005; Karakosta 2015; Mathѐ 2007).

All studies were short in duration: one lasted two months (Apsche 2005), two lasted three months (Karakosta 2015; Piliero 1994) and one lasted six months (Mathѐ 2007).

In Piliero 1994, one intervention involved sessions on victim empathy training, covert sensitization and masturbatory satiation. One study reported doing family interventions as part of the CBT protocol (Karakosta 2015).

Comparators and controls

In one study, Karakosta 2015, the control group was offered TAU. Another study had only a control group, which received no treatment (Mathѐ 2007). In the third study, Piliero 1994, the control group was offered a sexual education programme, which the authors described as "clinically inert" and which they did not expect to influence cognitive constructs. In the fourth study, Apsche 2005, the comparator groups were offered either mode deactivation therapy or social skills training.

Outcomes

Primary outcomes

One of the four studies, Apsche 2005, reported on the effects of CBT on recidivism: behavioural report of sexual aggression. The remaining three studies did not report on this, or any other measure of the outcome 'Recidivism'.

None of the four studies reported on the effects of CBT on adverse events (Apsche 2005; Karakosta 2015; Mathѐ 2007; Piliero 1994).

Secondary outcomes

The four studies measured the secondary outcomes listed below. Two studies used measures that were not standardised for children (Karakosta 2015; Mathѐ 2007): the Bumby Cognitive Cardsort Scale (Bumby 1996); Hostility Towards Women Scale, which was reported in a compendium of measures on violence against women (Flood 2008) that also included the Attitudes Towards Women Scale, (Spence 1972); and Abel and Becker Cognition Scale (Salter 1988).

Apsche 2005 measured the psychological well‐being before and after treatment using the following scales: internalising and externalising behaviour, including total scores, assessed with the Child Behavior Checklist (CBCL; Achenbach 1991) and the Devereux Scales of Mental Disorders (DSMD; Naglieri 1994). Standard deviations were not available for scores on the last measure. Therefore, we have provided a narrative description of those results in the Effects of interventions section.

Karakosta 2015 measured the following three outcomes before and after treatment.

• Psychological well‐being, assessed in terms of depression and anxiety using respectively, the Children’s Depression Inventory 2 (Kovacs 2001) and the Revised Children’s Manifest Anxiety Scale (Boehnke 1986).

• Sexual attitudes and behaviour, assessed using measures of sexual attitudes pertaining to rape and molestation; two of the scales included in the Bumby Cognitive Cardsort Scale (Bumby 1996).

• Victim empathy, assessed using scales from a compendium of measures on violence against women (Flood 2008). The study authors did not clarify the exact measures for this outcome used in the paper, and we were not able to contact the authors for clarification due to contact details being out of date.

Mathѐ 2007 assessed the following two outcomes, before, in the middle, and after treatment.

• Psychological well‐being, assessed as an improvement in self‐concept using the Self‐Concept Scale, which was developed by the researchers. According to the study authors, the scale is said to focus on anger and anxiety management, and developing insight and self‐awareness.

• Sexual attitudes and behaviour, measured with the Abel and Becker Cognition Scale (Salter 1988).

• Victim empathy, assessed using the Attitudes Towards Women Scale (Spence 1972).

Piliero 1994 assessed the impact of the interventions on sexual attitudes before and after treatment, using the juvenile form of the Multiphasic Sex Inventory and supplement scale — a measure of psychosexual characteristics (Nichols 1984). The study authors examined distortions pertaining to social‐sexual desirability, justification, apprehension confidence, and inappropriate sexual fantasies.

Funding sources

The studies did not state their funding sources (Apsche 2005; Karakosta 2015; Mathѐ 2007; Piliero 1994).

Excluded studies

We excluded 109 full‐text reports. Of these, we formally excluded 18 studies (from 19 reports): 14 studies because they were not RCTs (Dunham 2009; Feilzer 2004; Graham 1998; Gretton 2005; Hird 1996; Hout 2002; Jones 1998; Lab 1993; Marshall 2008; Pérez 2012; Thoder 2011; Viens 2012; Waite 2005; Worling 2000); two studies because they involved adult participants (Langdon 2007; Marshall 2008); and two studies because they involved either participants aged 6 to 12 years old (Bonner 1999) or did not meet the criteria for CBT used in this review (Weinrott 1997).

Risk of bias in included studies

We summarised our 'Risk of bias' judgements in Figure 2, and presented them as a graph in Figure 3 Each study carried substantial risks of bias, with a combination of elements assessed as unclear due to lack of information reported, or high risk of bias. Unless otherwise stated, our judgements of risk of bias affect all outcomes.

2.

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

3.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Allocation

Random sequence generation

We rated all four studies at unclear risk of bias as they did not provide sufficient information on the randomisation procedure (Apsche 2005; Karakosta 2015; Mathѐ 2007; Piliero 1994).

Allocation concealment

We rated all four studies at unclear risk of bias as they did not provide sufficient information on the randomisation procedure to be able to assess whether allocation was concealed from study personnel (Apsche 2005; Karakosta 2015; Mathѐ 2007; Piliero 1994).

Blinding

Blinding of participants and personnel (performance bias)

We assessed all four studies as being at high risk of performance bias because there was no blinding of participants and personnel in any study (Apsche 2005; Karakosta 2015; Mathѐ 2007; Piliero 1994).

Blinding of outcome assessment (detection bias)

Subjective outcomes

We assessed all studies at unclear risk of detection bias because blinding of outcome assessors was not discussed in the study reports (Apsche 2005; Karakosta 2015; Mathѐ 2007; Piliero 1994).

Objective outcomes

No study measured objective outcomes (e.g. recidivism, self‐harm).

Incomplete outcome data

We judged two studies to be at high risk of attrition bias due to high attrition rates (Karakosta 2015; Piliero 1994). We assessed two studies at low risk of attrition bias as there was no attrition (Apsche 2005; Mathѐ 2007).

Selective reporting

We considered one study to be at unclear risk of reporting bias, given that no comments were given on which findings were reported, the author did not report to the request for more information (whether only significant or all of them) and no protocol was available to allow for independent assessment (Mathѐ 2007).

We assessed two studies at high risk of bias because reporting of data was incomplete (Karakosta 2015; Piliero 1994). In Piliero 1994, there were data missing on both significant and non‐significant findings, which was not accounted for by the study authors. Karakosta 2015 only reported statistically significant data, according to the authors of the study.

One study, Apsche 2005, was assessed at low risk of bias as it appeared to present all outcomes.

Other potential sources of bias

We considered two studies to have an additional source of bias. In Mathѐ 2007 and Piliero 1994 the researcher was also the therapist delivering the intervention. Additionally, in Piliero 1994, testing instruments were administered by another practitioner who knew the participants personally. It is also concerning that two of the studies used measurement instruments not standardised for use with children.

We assessed one study, Karakosta 2015, at unclear risk of other bias due to a lack of information.

We assessed one study, Apsche 2005, at low risk of bias as it describes groups as similar at baseline for demographics and prior offence history.

Effects of interventions

See: Table 1; Table 2

Given that no study measured the same outcome using sufficiently similar measures, we were unable to combine the data in a meta‐analysis, and therefore are only able to report individual study results. Some studies gave only a description of the key findings, without presenting the data, and attempts to obtain them from study authors were unsuccessful. We present the findings below, organised by type of comparison (CBT versus no treatment or TAU, and CBT versus alternative treatment).

One study reported on the effects of CBT on the primary outcomes of recidivism (as listed in the Included studies section).

CBT versus no treatment or treatment as usual

Primary outcomes

The studies included in this comparison did not assess any of our primary outcomes.

Secondary outcomes

Psychological well‐being

Mathѐ 2007 (18 participants) reported that those in the CBT group scored higher on positive self‐concept (i.e. self‐esteem) (MD −4.78, 95% CI −7.86 to −1.70, Analysis 1.1), measured using the Self Concept Scale (developed by the researchers; scores range between 20 and 125; higher scores indicate poorer self‐concept) than those in the control group.

1.1. Analysis.

Comparison 1: CBT versus no intervention or TAU, Outcome 1: Secondary outcome: psychological well being ‐ self‐esteem (CBT vs no treatment)

Karakosta 2015 (21 participants) reported no significant reductions in depression for the CBT group compared to the treatment as usual group (MD 2.11, 95% CI −6.79 to 11.01, Analysis 1.4), measured using the Children's Depression Scale 2 (Kovacs 2001; higher scores indicate more severe depression, and higher levels of emotional or functional problems). The study author also reported using the Revised Children's Manifest Anxiety Scale (Boehnke 1986; higher scores on this measure represent higher levels of anxiety), but provided no data. Our interpretation of the summary of these findings suggests that there were no significant differences between the experimental and control group.

1.4. Analysis.

Comparison 1: CBT versus no intervention or TAU, Outcome 4: Secondary outcome: psychological well being ‐ depression (CBT vs TAU)

Sexual attitudes and behaviour

Mathѐ 2007 (18 participants) measured general cognitive distortions about sexual behaviour using the Abel and Becker Cognition Scale (29 items, scores range from 29 to 145; higher scores indicate more problematic cognitive distortions). The authors reported little or no difference between the groups (MD 1.56, 95% CI −11.54 to 14.66, Analysis 1.2; very low‐certainty evidence, Table 1).

1.2. Analysis.

Comparison 1: CBT versus no intervention or TAU, Outcome 2: Secondary outcome: sexual attitudes and behaviour ‐ cognitive distortions about sexual behaviour (general) (CBT vs no treatment)

Karakosta 2015 (21 participants) reported that those in the CBT group scored lower on cognitive distortions on rape (MD 8.75, 95% CI 2.83 to 14.67, Analysis 1.5; very low‐certainty evidence, Table 1), measured using Bumby Cognitive Cardsort Scale (Bumby 1996; 36 items rated on four‐point scale (where 1 = strongly disagree to 4 = strongly agree); higher scores indicate more justifications, minimisations, rationalisations and excuses for HSB), than those receiving TAU at the end of treatment. The study author also reported a significant reduction in cognitive distortions on molestation for those in the CBT arm, but provided no data to support this.

1.5. Analysis.

Comparison 1: CBT versus no intervention or TAU, Outcome 5: Secondary outcome: sexual attitudes and behaviour ‐ cognitive distortions about sexual behaviour pertaining to rape (CBT vs TAU)

Victim empathy

There is very low‐certainty evidence from Mathѐ 2007 (18 participants) that those in the CBT group scored higher in victim empathy (MD 5.56, 95% CI 0.94 to 10.18, Analysis 1.3; Table 1), measured using the Attitude Towards Women Scale (15 items; scores range from 0 to 45; higher scores indicate more egalitarian attitudes towards women), than those in the control group at the end of treatment.

1.3. Analysis.

Comparison 1: CBT versus no intervention or TAU, Outcome 3: Secondary outcome: victim empathy ‐ attitudes towards women (CBT vs no treatment)

Karakosta 2015 (21 participants) measured this outcome using measures from a compendium of scales (Flood 2008), but did not specify which measure and provided no usable data; therefore, no results could be included in the 'Summary of findings' tables.

Neither study assessed any of our other secondary outcomes: reactions to offending behaviour; coping skills; thinking patterns; aggression; social functioning; emotional self‐regulations and impulse control; substance use; programme engagement; and economic costs.

CBT versus alternative treatments

Primary outcome: recidivism

Apsche 2005 (59 participants) reported post‐treatment sexual aggression scores using Daily Behaviour Report cards and Behaviour Incidence Report forms completed by staff. The study reported very low‐certainty evidence showing little or no effect of CBT at post‐intervention on sexual aggression (MD 0.09, 95% CI −0.18 to 0.37, Analysis 2.1; Table 2) compared to alternative treatment.

2.1. Analysis.

Comparison 2: CBT versus alternative interventions, Outcome 1: Primary outcome: sexual aggression

Neither study included in this comparison assessed the primary outcome of adverse events.

Secondary outcomes

Psychological well‐being

Apsche 2005 (59 participants) reported psychological well‐being at post‐intervention using the Child Behaviour Checklist; higher scores on this measure represent increased severity of problems. Results (Analysis 2.2) indicated that there may be little to no difference between CBT and alternative treatments on internalising scores (MD 4.62, 95% CI −1.40 to 10.64), externalising scores (MD 5.79, 95% CI −0.73 to 12.32) or total scores (MD 4.51, 95% CI −1.12 to 10.14).

2.2. Analysis.

Comparison 2: CBT versus alternative interventions, Outcome 2: Secondary outcome: psychological well‐being ‐ CBCL

Apsche 2005 also reported psychological well‐being at post‐intervention using the Devereux Scales of Mental Disorders (DSMD); higher scores on this measure represent increased severity of problems (responses can be converted to T scores, with a mean of deviation of 10; a score of 60 or higher indicates an area of clinical concern). Mean DSMD scores for the internalising factor, externalising factor, critical pathology, and total score for the Mode Deactivation Therapy (MDT) group were at or near one standard deviation below the CBT group, suggesting that MDT improved psychological well‐being more than CBT.

Sexual attitudes and behaviour

Piliero 1994 (16 participants) reported on the effectiveness of CBT in changing young people's thinking about themselves and their relationships, compared with sexual education, using the juvenile form of the Multiphasic Sex Inventory (MSI) and supplement scale — a measure of psychosexual characteristics (Nichols 1984). The study reported very low‐certainty evidence that the CBT group showed improvements in cognitive distortions about sexual behaviour pertaining to justifications (MD −3.27, 95% CI −4.77 to −1.77, Analysis 2.3), apprehension‐confidence (MD −2.47, 95% CI −3.85 to −1.09, Analysis 2.4), inappropriate sexual fantasies (MD 0.13, 95% CI −1.52 to 1.78, Analysis 2.5), and social‐sexual desirability (MD 8.53, 95% CI 4.72 to 12.34, Analysis 2.6), at the end of treatment. See Table 2.

2.3. Analysis.

Comparison 2: CBT versus alternative interventions, Outcome 3: Secondary outcome: sexual attitudes and behaviour ‐ cognitive distortions about sexual behaviour pertaining to justifications (CBT vs sexual education programme)

2.4. Analysis.

Comparison 2: CBT versus alternative interventions, Outcome 4: Secondary outcome: sexual attitudes and behaviour ‐ cognitive distortions about sexual behaviours pertaining to apprehension confidence (CBT vs sexual education programme)

2.5. Analysis.

Comparison 2: CBT versus alternative interventions, Outcome 5: Secondary outcome: sexual attitudes and behaviour ‐ cognitive distortions about sexual behaviours pertaining to inappropriate sexual fantasies (CBT vs sexual education programme)

2.6. Analysis.

Comparison 2: CBT versus alternative interventions, Outcome 6: Secondary outcome: sexual attitudes and behaviour ‐ cognitive distortions about sexual behaviour pertaining to social‐sexual desirability (CBT vs sexual education programme)

Neither study included in this comparison assessed any of our other secondary outcomes: reactions to offending behaviour; coping skills; thinking patterns; victim empathy; aggression; social functioning; emotional self‐regulations and impulse control; substance use; programme engagement; and economic costs.

Discussion

Summary of main results

We identified four small studies that assessed the effectiveness of a group‐based CBT programme designed to address HSB in children and young people aged between 10 and 18 years old. Karakosta 2015 evaluated the ISOTP compared with TAU. Historically, the ISOTP has a high standing as a preferred treatment modality for young people with HSB. For description of the ISOTP, see Rehfuss 2013

One of the four studies reported uncertain evidence regarding the effect of CBT compared to alternative treatments on the primary outcome of recidivism in terms of behavioural reports of sexual aggression. None of the four studies assessed the impact of CBT on the outcomes of adverse events, thinking patterns, social functioning and emotional regulation or impulse control.

Of the secondary outcomes assessed, there is uncertain evidence regarding whether CBT is an effective intervention for young people engaged in HSB. One study (18 participants) reported that CBT may result in higher positive self‐concept and improvements in victim empathy compared with no treatment. One study (21 participants) found very low‐quality evidence that mean cognitive distortions about rape were lower in the CBT group than the TAU group. The same study found little to no difference in depression when CBT was compared with TAU. One study (16 participants) provided very low‐certainty evidence that, compared to sexual education, mean cognitive distortions pertaining to justification or taking responsibility for actions and apprehension confidence may be lower in the CBT group. The same study indicated that mean cognitive distortions pertaining to social‐sexual desirability may be lower in the CBT group, and there may be little to no difference between the groups for cognitive distortions pertaining to inappropriate sexual fantasies.

All four studies had significant methodological problems with regards to incomplete outcome data and absence of blinding, which makes it difficult to draw any firm conclusions. Also, it may be useful to acknowledge whether some harm was caused by these interventions, although none of the studies addressed this issue in detail. For instance, the negative effects of delving deep into the roots of one's distortions and the potential trauma that can result from this for the child or young person.

Overall completeness and applicability of evidence

As indicated above, only one of the included studies assessed the impact of CBT programmes on one of the primary outcomes of this review, namely recidivism. As a result, we cannot draw any conclusions about the effectiveness of CBT in influencing HSB.

The different local contexts of the studies pose a challenge for generalising, including where and how they were delivered, and the very definitions of HSB.

Three of the four studies were undertaken in the USA (Apsche 2005; Karakosta 2015; Piliero 1994), and the third study in South Africa (Mathѐ 2007), so the findings may not generalise to other countries.

There was variation across the studies as to how HSB was defined. Piliero 1994included all young men who were convicted of sexual misconduct, or voluntarily admitted to a history of criminal sexual behaviour. Mathѐ 2007 included young men who committed a sexual offence against a female that would warrant housing in a maximum security unit. Karakosta 2015 included adjudicated young males who committed any kind of sexual offence. This means that some studies included participants who had been imprisoned following rape, and others who had shown less violent, non‐penetrative sexual offending or inappropriate sexual behaviour. Apsche 2005 included adolescent males with documented incidents of physical and sexual aggression. All had been diagnosed with a conduct or personality disorder, or both.

Finally, all four studies recruited only boys. Given that this is a global problem that affects boys and girls, and one which encompasses a wide range of problems, the available evidence and its applicability is, therefore, extremely limited. Lack of follow‐up, small and potentially biased samples, coupled with unclear protocols further compound the problem.

Given that there has been more than four decades of specialist treatment for young people with HSB, it is surprising that there is little good‐quality evidence for whether CBT helps these young people. This may be due to a lack of resources or different priorities. Much of the research in this areas focuses on whether the HSB is repeated, rather than the safety, well‐being and development of children identified with HSB. Insufficient attention is paid to the impact of underlying factors such as past trauma, poverty, stigma, gender, co‐occurring diagnoses (such as mental health problems, developmental or learning difficulties), family disruption or living in out‐of‐home care (Shlonsky 2017). Although multi‐component treatment is often used in practice with components tailored to respond to individual need, the effectiveness of CBT within these types of treatment has not been evaluated.

Quality of the evidence

The objective of this review was to evaluate the effects of CBT for young people aged 10 to 18 years, who have exhibited HSB. Using the GRADE approach, we considered the certainty of the evidence included in this review to be very low (see Table 1; Table 2). We found no protocols for three of the included studies (Karakosta 2015; Mathѐ 2007; Piliero 1994), all of which had serious methodological weaknesses. The failure to find protocols is highly damaging to the evidence in the studies. The fourth study, Apsche 2005, had stronger methodology with only some weaknesses.

We downgraded the certainty of the evidence by one level due to study limitations, as the risk of bias in the included studies was predominantly high in relation to procedures for random sequence generation, incomplete outcome data and lack of blinding procedures. We also downgraded the certainty of evidence by two levels due to imprecision, as findings for all outcomes were based on evidence from RCTs only, with sample sizes ranging from 16 to 60 participants. It was not possible to assess formally statistical heterogeneity or the likelihood of publication bias due to the small number of included studies. We did not downgrade the certainty of the evidence for indirectness. Overall, this very low‐certainty evidence means that it is difficult to draw conclusions about the effectiveness of CBT as a treatment option for young people with HSB. No evidence from RCTs was available for individually‐delivered CBT (ISOTP in Karakosta 2015 was only done in groups). it is not clear whether Apsche 2005 delivered CBT individually or in groups.

Potential biases in the review process

We followed standard Cochrane procedures when developing the protocol for this review (Sneddon 2012), and in conducting the review. We searched extensively for relevant studies; and screening and data extraction were undertaken by three and two reviewers respectively, who acted independently. Two reviewers also checked the data entered into RevMan 5 for accuracy (Review Manager 2014). None of the review authors have any known conflicts of interest. The decision was made by review authors to not combine data for the different types of 'CBT versus alternative' comparisons in the same pooled analysis due to the potential differences between these types of comparators. If the decision had been made to pool these data together, the evidence may have been more precise, and thus more certain. However, the review authors agreed not to take this approach to limit the heterogeneity in this result.

Agreements and disagreements with other studies or reviews

There are no other systematic reviews of RCTs specifically focused on the effects of CBT on improving outcomes for young people with HSB.

Reitzel 2006's meta‐analysis of the effect of treatment on recidivism for juveniles included no studies that assessed the impact of CBT on recidivism. Similarly, the systematic review and meta‐analysis conducted by Kettrey 2018 only examined the impact on general and sexual recidivism of "psychosocial, therapeutically oriented treatments" provided solely to juveniles who had been convicted of sexual offences. The authors noted that seven of the eight included studies were quasi‐experimental, and commented on the paucity of accumulated evidence between 1950 and 2015 given the prevalence of the types of programmes included in their review, largely comprising variants of individual or group counselling.

In keeping with the findings of this review, an evidence review of treatments for youths who had engaged in 'illegal sexual behaviours' by Dopp 2017 classified the treatments according to five criteria often used to categorise the evidential status of interventions, and concluded that CBT fell into the category of 'Experimental' (i.e. having limited research support).

A rapid evidence assessment of approaches for treating HSB concluded that multi‐systemic therapy (MST) is amongst the most promising approaches to HSB (Shlonsky 2017). The authors suggest CBT likely contributes to the efficacy of MST, but the evidence they reviewed did not allow its unique contribution to this broad‐based family and community based treatment to be unpicked. Shlonsky 2017 also highlighted the need to improve the quality of evidence available with regard to interventions provided to this group of young people.

Authors' conclusions

Implications for practice.

It is unclear from the evidence available whether CBT improves outcomes for young people with HSB. There is some very low‐certainty evidence that group‐based CBT may improve some aspects of sexual attitudes and behaviour, compared to alternative interventions, as well as victim empathy and psychological well‐being of young people compared to no treatment. The findings, however, are inconsistent and the risk of bias in these studies varied between unclear and high. All of the studies examined only males, so we have no information on the effects of group‐based CBT on females. We suggest that all children presenting with HSB should be assessed and offered support to help them avoid escalation or prevent them harming other children, or both. Services should be delivered early and, even when surveillance, restrictions and exclusions are thought to be necessary, these should not be in place of support and therapeutic assistance.

Implications for research.

The questions of whether CBT is an effective intervention for young people with HSB, or indeed whether it may cause harm, have yet to be answered. The available evidence on CBT's effectiveness comes from four small studies, three of which were conducted between 13 and 26 years ago. Since then, CBT has developed considerably, and often forms part of broader‐based interventions targeting the range of factors that are thought to contribute to HSB or that can help to address HSB. Further, robust evaluations of both individual and group‐based CBT are required, both as a distinct intervention and as a component of multi‐component approaches to the management of HSB. Whatever their focus, future trials need to be rigorous in design and delivery, with larger sample sizes and better reporting to enable appraisal and interpretation of results. Evaluators should be independent of the service being examined. Studies should examine children and young people of all ages, both males and females, and explore effects in different contexts. Contact and non‐contact HSB (including internet‐based behaviours) should be examined.

As with the evidence base for treatment of adults convicted of sexual offences, more RCTs are needed, particularly outside of North America. Researchers should document and report differences in delivery e.g. frequency of sessions, number of sessions, number of therapists involved (within and across sessions), therapists' experience, professional affiliation; whether the therapist was also the researcher. In addition, there is a clear need for a more differentiated process and outcome evaluations that address the questions of what works with whom, in what contexts, under what conditions, with respect to what outcomes and why. There also needs to be more dialogue among programme commissioners, practitioners and researchers about organising and funding research on assessment, intervention and outcome, which will enable improved testing of interventions in specific contexts and with diverse participants.

It is the review authors' opinion that a deeper inquiry is necessary into how services are defining their objectives and whether there are practical incentives and pathways for rigorous evaluation in order to deliver benefits and reduce risks. It is clear that rigorous evaluation is not receiving due attention, and it would be useful to know why. It is possible that the nature of the HSB is such that it does not lend itself well to RCTs. This may be because follow‐up of recidivism and adverse events would include long‐term surveillance, restrictions and exclusions, which should not be legitimised in the name of research. The preference should be for treatment and support. To support this hypothesis, a cursory search for non‐controlled studies on the impact of CBT on HSB returned a substantial number of papers, many of which advocate for, at least, some positive impact of CBT, but without the restrictions needed to follow up recidivism. One of the suggestions for future reviews, may be to broaden the scope of this review, to include other research. It would also be useful for future research to describe differences in delivery (e.g. frequency of sessions, number of sessions, regular vs. random therapist, number of therapists present, therapists' experience, professional affiliation; whether the therapist was also the researcher).

What's new

Date	Event	Description
16 July 2020	Amended	Correcting error in Plain language summary: when describing the sex of the sample, one study had been erroneously omitted.

History

Protocol first published: Issue 5, 2012
Review first published: Issue 6, 2020

Appendices

Appendix 1. Search strategies

The search strategies draw on the sex offending terms identified by Kenworthy and colleagues (Kenworthy 2003), and have been expanded.

Cochrane Central Register of Controlled Trials (CENTRAL; 2019, Issue 6) in the Cochrane Library (searched 26 June 2019)

#1MeSH descriptor: [Sex Offenses] this term only
#2MeSH descriptor: [Rape] this term only
#3MeSH descriptor: [Child Abuse, Sexual] this term only
#4MeSH descriptor: [Incest] this term only
#5MeSH descriptor: [Paraphilias] 1 tree(s) exploded
#6(incest* or paraphil*):ti,ab
#7(rape or rapist):ti,ab
#8p*edophil*:ti,ab
#9((child* or sibling* or inter next famil* or interfamil*) near/1 (sex* or molest*)):ti,ab
#10(sex* near/3 (abus* or aggress* or assault* or coerc* or delinquen* or devian* or exploit* or homicid* or masochis* or molest* or murder* or offen* or predator* or recidiv* or reoffend* or re‐offend* or torture* or violen*)):ti,ab
#11exhibitionis*:ti,ab
#12((lewd* or indecen* or obscen*) near/2 (behav* or act* or expos*)):ti,ab
#13(child* near/1 porn*):ti,ab
#14{or #1‐#13}
#15[mh ^"Behavior Therapy"]
#16[mh Psychotherapy]
#17[mh "Family Therapy"]
#18[mh meditation]
#19[mh relaxation]
#20[mh Imagery]
#21[mh "Biofeedback, Psychology"]
#22(cognitive next behavio*):ti,ab
#23CBT:ti,ab
#24((multi next systemic or multisystemic) near/2 therap*):ti,ab
#25(family next therap* or psychotherap* or psycho‐therap*):ti,ab
#26(social next skill* near/1 train*):ti,ab
#27((cognition or cognitive) near/5 (therap* or rehabilitat* or interven* or program* or treat* or approach* or technique*)):ti,ab
#28(behav* near/5 (modification* or therap* or rehabilitat* or interven* or program* or treat* or approach* or technique*)):ti,ab
#29(aversive or aversion or biofeedback or desensiti*ation or relaxation or meditat*):ti,ab
#30{or #15‐#29}
#31#14 and #30
#32(juvenil* or teen* or adolescen* or young person* or young next people or youth* or child* or minor*):ti,ab
#33[mh CHild] or [mh Adolescent] or [mh "adolescent behavior"] or [mh "juvenile delinquency"] or [mh Minors]
#34#32 or #33
#35#31 and #34

MEDLINE Ovid (1946 to 26 June 2019)

1 Sex Offenses/
2 Child Abuse, Sexual/
3 Rape/
4 Incest/
5 exp Paraphilias/
6 incest$.tw.
7 paraphil$.tw.
8 (rape or rapist).tw.
9 p?edophil$.tw.
10 ((child$ or sibling$ or inter famil$ or interfamil$) adj1 (sex$ or molest$)).tw.
11 (sex$ adj3 (abus$ or aggress$ or assault$ or coerc$ or delinquen$ or devian$ or exploit$ or homicid$ or masochis$ or molest$ or murder$ or offend$ or offen#e$ or predator$ or recidiv$ or reoffend$ or re‐offend$ or torture$ or violen$)).tw.
12 (sex$ adj3 harm$ adj1 behav$).tw.
13 ((lewd$ or indecen$ or obscen$) adj3 (behav$ or act$ or expos$)).tw.
14 exhibitionis$.tw.
15 (child$ adj1 porn$).tw.
16 or/1‐15
17 child/
18 Adolescent/
19 Minors/
20 Adolescent Behavior/
21 Juvenile Delinquency/
22 (juvenil$ or teen$ or adolescen$ or young person$ or young people or youth$ or child$ or minor$).tw. (1293009)
23 or/17‐22
24 exp Behavior Therapy/
25 Psychotherapy/
26 Family therapy/
27 cognitive behavio$.tw.
28 CBT.tw.
29 ((multi systemic or multisystemic) adj2 therap$).tw.
30 (family therap$ or psychotherap$ or psycho‐therap$).tw.
31 (social skill$ adj1 train$).tw.
32 ((cognition or cognitive) adj5 (therap$ or rehabilitat$ or interven$ or program$ or treat$ or approach$ or technique$)).tw. (25463)
33 (behavio?r$ adj5 (modification$ or therap$ or rehabilitat$ or interven$ or program$ or treat$ or approach$ or technique$)).tw. (62151)
34 meditation/
35 relaxation/
36 Biofeedback, Psychology/
37 Imagery/
38 (aversive or aversion or biofeedback or desensiti#ation or relaxation or meditat$).tw.
39 or/24‐38
40 16 and 23 and 39
41 randomized controlled trial.pt.
42 controlled clinical trial.pt.
43 randomi#ed.ab.
44 placebo$.ab.
45 drug therapy.fs.
46 randomly.ab.
47 trial.ab.
48 groups.ab.
49 treatment outcome/
50 Program Evaluation/
51 Comparative Effectiveness Research/
52 ((effectiveness or efficacy) adj1 (compare or comparision or study or studies or research)).tw.
53 or/41‐52
54 exp animals/ not humans.sh.
55 53 not 54
56 40 and 55

Embase Ovid (1980 to 26 June 2019)

1 sexual crime/
2 child sexual abuse/
3 exp sexual assault/
4 incest/
5 sexual deviation/
6 incest$.tw.
7 paraphil$.tw.
8 (rape or rapist).tw.
9 pedophilia/
10 p?edophil$.tw.
11 ((child$ or sibling$ or inter famil$ or interfamil$) adj1 (sex$ or molest$)).tw.
12 (sex$ adj3 (abus$ or aggress$ or assault$ or coerc$ or delinquen$ or devian$ or exploit$ or homicid$ or masochis$ or molest$ or murder$ or offend$ or offen#e$ or predator$ or recidiv$ or reoffend$ or re‐offend$ or torture$ or violen$)).tw.
13 (sex$ adj3 harm$ adj1 behav$).tw.
14 ((lewd$ or indecen$ or obscen$) adj3 (behav$ or act$ or expos$)).tw.
15 exhibitionis$.tw.
16 (child$ adj1 porn$).tw.
17 or/1‐16
18 child/
19 juvenile/
20 exp child/
21 exp adolescent/
22 (juvenil$ or teen$ or adolescen$ or young person$ or young people or youth$ or child$ or minor$).tw.
23 or/18‐22
24 exp cognitive therapy/
25 exp behavior therapy/
26 psychotherapy/
27 family therapy/
28 guided imagery/
29 psychophysiology/
30 meditation/
31 aversion therapy/
32 ((cognition or cognitive) adj5 (therap$ or rehabilitat$ or interven$ or program$ or treat$ or approach$ or technique$)).tw.
33 cognitive behavio$.tw.
34 CBT.tw.
35 (behavio?r$ adj5 (modification$ or therap$ or rehabilitat$ or interven$ or program$ or treat$ or approach$ or technique$)).tw.
36 (aversive or aversion or biofeedback or desensiti#ation or relaxation or meditat$).tw.
37 ((multi systemic or multisystemic) adj2 therap$).tw.
38 (family therap$ or psychotherap$ or psycho‐therap$).tw.
39 (social skill$ adj1 train$).tw.
40 or/24‐39
41 17 and 23 and 40
42 Randomized controlled trial/
43 controlled clinical trial/
44 Single blind procedure/
45 Double blind procedure/
46 triple blind procedure/
47 Crossover procedure/
48 (crossover or cross‐over).tw.
49 ((singl$ or doubl$ or tripl$ or trebl$) adj1 (blind$ or mask$)).tw.
50 Placebo/
51 placebo.tw.
52 prospective.tw.
53 factorial$.tw.
54 random$.tw.
55 assign$.ab.
56 allocat$.tw.
57 volunteer$.ab.
58 treatment outcome/
59 program evaluation/
60 exp comparative study/
61 ((effectiveness or efficacy) adj1 (compare or comparision or study or studies or research)).tw.
62 or/42‐61
63 41 and 62

PsycINFO Ovid (1967 to 26 June 2019)

1 exp Sex Offenses/
2 exp paraphilias/
3 pedophilia/
4 incest$.tw.
5 paraphil$.tw.
6 (rape or rapist).tw.
7 p?edophil$.tw.
8 ((child$ or sibling$ or inter famil$ or interfamil$) adj1 (sex$ or molest$)).tw.
9 (sex$ adj3 (abus$ or aggress$ or assault$ or coerc$ or delinquen$ or devian$ or exploit$ or homicid$ or masochis$ or molest$ or murder$ or offend$ or offen#e$ or predator$ or recidiv$ or reoffend$ or re‐offend$ or torture$ or violen$)).tw.
10 (sex$ adj3 harm$ adj1 behav$).tw.
11 ((lewd$ or indecen$ or obscen$) adj3 (behav$ or act$ or expos$)).tw.
12 exhibitionis$.tw.
13 (child$ adj1 porn$).tw.
14 or/1‐13
15 exp Cognitive Behavior Therapy/
16 cognitive therapy/
17 exp behavior modification/
18 psychotherapy/ )
19 child psychotherapy/
20 exp adolescent psychotherapy/
21 multisystemic therapy/
22 Family Therapy/
23 cognitive behavio$.tw.
24 CBT.tw.
25 ((multi systemic or multisystemic) adj2 therap$).tw.
26 (family therap$ or psychotherap$ or psycho‐therap$).tw.
27 (social skill$ adj1 train$).tw.
28 ((cognition or cognitive) adj5 (therap$ or rehabilitat$ or interven$ or program$ or treat$ or approach$ or technique$)).tw.
29 (behavio?r$ adj5 (modification$ or therap$ or rehabilitat$ or interven$ or program$ or treat$ or approach$ or technique$)).tw.
30 exp relaxation therapy/
31 meditation/
32 imagery/
33 (aversive or aversion or biofeedback or desensiti#ation or relaxation or meditat$).tw.
34 or/15‐33
35 14 and 34
36 clinical trials/
37 (randomis$ or randomiz$).tw.
38 (random$ adj3 (allocat$ or assign$)).tw.
39 ((clinic$ or control$) adj trial$).tw.
40 ((singl$ or doubl$ or trebl$ or tripl$) adj3 (blind$ or mask$)).tw.
41 (crossover$ or "cross over$").tw.
42 random sampling/
43 Experiment Controls/
44 Placebo/
45 placebo$.tw.
46 exp program evaluation/
47 treatment effectiveness evaluation/
48 ((effectiveness or evaluat$) adj3 (stud$ or research$)).tw.
49 exp experimental methods/
50 or/36‐49
51 35 and 50

CINAHL EBSCOhost (Cumulative Index to Nursing and Allied Health Literature; 1937 to 26 June 2019)

S63 S47 AND S62 Search modes ‐ Boolean/Phrase Interface ‐ EBSCOhost
Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S62 S48 OR S49 OR S50 OR S51 OR S52 OR S53 OR S54 OR S55 OR S56 OR S57
OR S58 OR S59 OR S60 OR S61 Search modes ‐ Boolean/Phrase Interface ‐
EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S61 (MH "Treatment Outcomes") Search modes ‐ Boolean/Phrase Interface
‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S60 (MH "Program Evaluation") Search modes ‐ Boolean/Phrase Interface
‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S59 TI ("prospective study" or "prospective research") or
AB("prospective study" or "prospective research") Search modes ‐
Boolean/Phrase Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S58 TI ("follow‐up study" or "follow‐up research") or AB ("follow‐up
study" or "follow‐up research") Search modes ‐ Boolean/Phrase
Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S57 AB((trebl* N1 mask*) or (trebl* N1 blind*)) Search modes ‐
Boolean/Phrase Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S56 AB("cross over") Search modes ‐ Boolean/Phrase Interface ‐
EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S55 (MH "Crossover Design") Search modes ‐ Boolean/Phrase Interface ‐
EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S54 AB((tripl* N1 mask*) or (tripl* N1 blind*)) Search modes ‐
Boolean/Phrase Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S53 AB ((doubl* N1 mask*) or (doubl* N1 blind*)) Search modes ‐
Boolean/Phrase Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S52 AB ((singl* N1 mask*) or(singl* N1 blind*)) Search modes ‐
Boolean/Phrase Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S51 AB(trial) Search modes ‐ Boolean/Phrase Interface ‐ EBSCOhost
Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S50 AB(random*) Search modes ‐ Boolean/Phrase Interface ‐ EBSCOhost
Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S49 (MH "Random Assignment") Search modes ‐ Boolean/Phrase Interface
‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S48 (MH "Clinical Trials+") Search modes ‐ Boolean/Phrase Interface ‐
EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S47 S31 AND S37 AND S46 Search modes ‐ Boolean/Phrase Interface ‐
EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus 1,431
S46 S38 OR S39 OR S40 OR S41 OR S42 OR S43 OR S44 OR S45 Search modes
‐ Boolean/Phrase Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S45 (aversive or aversion or biofeedback or desensiti* or imagery or
relaxation or meditat*) Search modes ‐ Boolean/Phrase Interface ‐
EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S44 (social skill* N1 train*) Search modes ‐ Boolean/Phrase Interface
‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S43 (family therap* or psychotherap* or psycho‐therap*) Search modes ‐
Boolean/Phrase Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S42 ((multi systemic or multisystemic) N2 therap*) Search modes ‐
Boolean/Phrase Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S41 (behav* N5 (modification* or therap* or rehabilitat* or interven*
or program* or treat* or approach* or technique*)) Search modes ‐
Boolean/Phrase Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S40 ((cognition or cognitive) N5 (therap* or rehabilitat* or interven*
or program* or treat* or approach* or technique*)) Search modes ‐
Boolean/Phrase Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S39 cognitive behavio* OR CBT Search modes ‐ Boolean/Phrase Interface
‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S38 (MH "Psychotherapy+") Search modes ‐ Boolean/Phrase Interface ‐
EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S37 S32 OR S33 OR S34 OR S35 OR S36 Search modes ‐ Boolean/Phrase
Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S36 (juvenil* or teen* or adolescen* or young person* or young people
or youth* or child* or minor*) Search modes ‐ Boolean/Phrase Interface
‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S35 AG adolescent Search modes ‐ Boolean/Phrase Interface ‐ EBSCOhost
Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S34 AG child:6 ‐12 years Search modes ‐ Boolean/Phrase Interface ‐
EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S33 (MH "Minors (Legal)") Search modes ‐ Boolean/Phrase Interface ‐
EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S32 (MH "Juvenile Offenders") OR (MH "Juvenile Delinquency") Search
modes ‐ Boolean/Phrase Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S31 S16 OR S17 OR S18 OR S19 OR S20 OR S21 OR S22 OR S23 OR S24 OR S25
OR S26 OR S27 OR S28 OR S29 OR S30 Search modes ‐ Boolean/Phrase
Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S30 (child* N1 porn*) Search modes ‐ Boolean/Phrase Interface ‐
EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S29 exhibitionis* Search modes ‐ Boolean/Phrase Interface ‐ EBSCOhost
Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S28 ((lewd* or indecen* or obscen*) N3 (behav* or act* or expos*))
Search modes ‐ Boolean/Phrase Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus 19
S27 (sex* N3 harm* N1 behav*) Search modes ‐ Boolean/Phrase Interface
‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S26 (sex* N3 (abus* or aggress* or assault* or coerc* or delinquen* or
devian* or exploit* or homicid* or masochis* or molest* or murder* or
offen* or predator* or recidiv* or reoffend* or re‐offend* or torture*
or violen*)) Search modes ‐ Boolean/Phrase Interface ‐ EBSCOhost
Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S25 ((child* or sibling* or inter famil* or interfamil*) N1(sex* or
molest*)) Search modes ‐ Boolean/Phrase Interface ‐ EBSCOhost Research
Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S24 pedophil* or paedophil* Search modes ‐ Boolean/Phrase Interface ‐
EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S23 (rape or rapist) Search modes ‐ Boolean/Phrase Interface ‐
EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S22 paraphil* Search modes ‐ Boolean/Phrase Interface ‐ EBSCOhost
Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S21 incest* Search modes ‐ Boolean/Phrase Interface ‐ EBSCOhost
Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S20 (MH "Child Abuse, Sexual") Search modes ‐ Boolean/Phrase
Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S19 (MH "Paraphilias+") Search modes ‐ Boolean/Phrase Interface ‐
EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S18 (MH "Rape") Search modes ‐ Boolean/Phrase Interface ‐ EBSCOhost
Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus
S17 (MH "Incest") Search modes ‐ Boolean/Phrase Interface ‐ EBSCOhost
Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus Display
S16 (MH "Sex Offenders") Search modes ‐ Boolean/Phrase Interface ‐
EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus Display
S15 S1 OR S2 OR S3 OR S4 OR S5 OR S6 OR S7 OR S8 OR S9 OR S10 OR S11 OR S12 OR S13 OR S14 Search modes ‐ Boolean/Phrase Interface ‐ EBSCOhost
Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus Display
S14 (MH "Treatment Outcomes") Search modes ‐ Boolean/Phrase Interface
‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus Display
S13 (MH "Program Evaluation") Search modes ‐ Boolean/Phrase Interface
‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus Display
S12 TI ("prospective study" or "prospective research") or
AB("prospective study" or "prospective research") Search modes ‐
Boolean/Phrase Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus Display
S11 TI ("follow‐up study" or "follow‐up research") or AB ("follow‐up
study" or "follow‐up research") Search modes ‐ Boolean/Phrase
Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus Display
S10 AB((trebl* N1 mask*) or (trebl* N1 blind*)) Search modes ‐
Boolean/Phrase Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus Display
S9 AB("cross over") Search modes ‐ Boolean/Phrase Interface ‐
EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus Display
S8 (MH "Crossover Design") Search modes ‐ Boolean/Phrase Interface ‐
EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus Display
S7 AB((tripl* N1 mask*) or (tripl* N1 blind*)) Search modes ‐
Boolean/Phrase Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus Display
S6 AB ((doubl* N1 mask*) or (doubl* N1 blind*)) Search modes ‐
Boolean/Phrase Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus Display
S5 AB ((singl* N1 mask*) or(singl* N1 blind*)) Search modes ‐
Boolean/Phrase Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus Display
S4 AB(trial) Search modes ‐ Boolean/Phrase Interface ‐ EBSCOhost
Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus Display
S3 AB(random*) Search modes ‐ Boolean/Phrase Interface ‐ EBSCOhost
Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus Display
S2 (MH "Random Assignment") Search modes ‐ Boolean/Phrase Interface ‐
EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus Display
S1 (MH "Clinical Trials+") Search modes ‐ Boolean/Phrase Interface ‐
EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ CINAHL Plus Display

Conference Proceedings Citation Index ‐ Social Science & Humanities Web of Science (CPCI‐SS&H; 1990 to June 2019; searched 26 June 2019)

#20#19 AND #18
DocType=All document types; Language=All languages;
#19TS=(random* or RCT or control or trial or group* or effectiveness or efficacy or experiment* or evaluat* or outcome)
DocType=All document types; Language=All languages;
#18#17 AND #9
DocType=All document types; Language=All languages;
#17#16 OR #15 OR #14 OR #13 OR #12 OR #11 OR #10
DocType=All document types; Language=All languages;
#16TS=(aversive or aversion or biofeedback or desensit* or imagery or relaxation or meditat*)
DocType=All document types; Language=All languages;
#15TS=("cognitive behav*" OR CBT)
DocType=All document types; Language=All languages;
#14TS=("social skill*" NEAR/1 train*)
DocType=All document types; Language=All languages;
#13TS=("family therap*" or psychotherap* or "psycho‐therap*")
DocType=All document types; Language=All languages;
#12TS=(("multi systemic" or multisystemic) NEAR/2 therap*)
DocType=All document types; Language=All languages;
#11TS= (behav* NEAR/3 (modification* or therap* or rehabilitat* or interven* or program* or treat* or approach* or technique*))
DocType=All document types; Language=All languages;
#10TS=((cognition or cognitive) NEAR/3 (therap* or rehabilitat* or interven* or program* or treat* or approach* or technique*))
DocType=All document types; Language=All languages;
#9#8 AND #7
DocType=All document types; Language=All languages;
#8TS=(juvenil* or teen* or adolescen* or young person* or "young people" or youth* or child* or minor*)
DocType=All document types; Language=All languages;
#7#6 OR #5 OR #4 OR #3 OR #2 OR #1
DocType=All document types; Language=All languages;
#6TS=(incest* or paraphil* OR p*edophil* or rape or rapist or exibitionist* )
DocType=All document types; Language=All languages;
#5TS=(child* NEAR/1 porn*)
DocType=All document types; Language=All languages;
#4TS=((lewd* or indecen* or obscen*) NEAR/3 (behav* or act* or expos*))
DocType=All document types; Language=All languages;
#3TS=(sex* NEAR/3 harm* NEAR/1 behav*)
DocType=All document types; Language=All languages;
#2TS=((child* or sibling* or "inter famil*" or interfamil*) NEAR/1 (sex* or molest*))
DocType=All document types; Language=All languages;
#1TS=(sex* NEAR/3 (abus* or aggress* or assault* or coerc* or delinquen* or devian* or exploit* or homicid* or masochis* or molest* or murder* or offen* or predator* or recidiv* or reoffend* or re‐offend* or torture* or violen*))
DocType=All document types; Language=All languages

Social Sciences Citation Index Web Of Science (SSCI; 1970 to 26 June 2019)

#20#19 AND #18
DocType=All document types; Language=All languages;
#19TS=(random* or RCT or control or trial or group* or effectiveness or efficacy or experiment* or evaluat* or outcome)
DocType=All document types; Language=All languages;
#18#17 AND #9
DocType=All document types; Language=All languages;
#17#16 OR #15 OR #14 OR #13 OR #12 OR #11 OR #10
DocType=All document types; Language=All languages;
#16TS=(aversive or aversion or biofeedback or desensit* or imagery or relaxation or meditat*)
DocType=All document types; Language=All languages;
#15TS=("cognitive behav*" OR CBT)
DocType=All document types; Language=All languages;
#14TS=("social skill*" NEAR/1 train*)
DocType=All document types; Language=All languages;
#13TS=("family therap*" or psychotherap* or "psycho‐therap*")
DocType=All document types; Language=All languages;
#12TS=(("multi systemic" or multisystemic) NEAR/2 therap*)
DocType=All document types; Language=All languages;
#11TS= (behav* NEAR/3 (modification* or therap* or rehabilitat* or interven* or program* or treat* or approach* or technique*))
DocType=All document types; Language=All languages;
#10TS=((cognition or cognitive) NEAR/3 (therap* or rehabilitat* or interven* or program* or treat* or approach* or technique*))
DocType=All document types; Language=All languages;
#9#8 AND #7
DocType=All document types; Language=All languages;
#8TS=(juvenil* or teen* or adolescen* or young person* or "young people" or youth* or child* or minor*)
DocType=All document types; Language=All languages;
#7#6 OR #5 OR #4 OR #3 OR #2 OR #1
DocType=All document types; Language=All languages;
#6TS=(incest* or paraphil* OR p*edophil* or rape or rapist or exibitionist* )
DocType=All document types; Language=All languages;
#5TS=(child* NEAR/1 porn*)
DocType=All document types; Language=All languages;
#4TS=((lewd* or indecen* or obscen*) NEAR/3 (behav* or act* or expos*))
DocType=All document types; Language=All languages;
#3TS=(sex* NEAR/3 harm* NEAR/1 behav*)
DocType=All document types; Language=All languages;
#2TS=((child* or sibling* or "inter famil*" or interfamil*) NEAR/1 (sex* or molest*))
DocType=All document types; Language=All languages;
#1TS=(sex* NEAR/3 (abus* or aggress* or assault* or coerc* or delinquen* or devian* or exploit* or homicid* or masochis* or molest* or murder* or offen* or predator* or recidiv* or reoffend* or re‐offend* or torture* or violen*))
DocType=All document types; Language=All languages;

Cochrane Database of Systematic Reviews (CDSR; 2019, Issue 6), part of the Cochrane Library (searched 26 June 2019)

#1(juvenile* or adolescen* or teen* or young*) near/3 (sex* next offen*)

Database of Abstract of Reviews of Effectiveness (DARE; 2015, Issue 2. Final issue), part of the Cochrane Library (searched 26 June 2019)

#1(juvenile* or adolescen* or teen* or young*) near/3 (sex* next offen*)

LILACS (Latin American and Caribbean Health Science Information database; http://lilacs.bvsalud.org/en/; searched 26 June 2019)

(tw:((juvenile* or adolescen* or teen* or child or young) )) AND (tw:(sex* offen*))

Criminal Justice Abstracts EBSCOhost (searched 1 August 2014. Not searched in June 2019 as we no longer had access to the database)

# Query Limiters/Expanders Last Run Via Results Action
S23 S21 AND S22 Search modes ‐ Boolean/Phrase Interface ‐ EBSCOhost
Research Databases
Search Screen ‐ Advanced Search
Database ‐ Criminal Justice Abstracts
S22 random* or RCT or control or trial or group* or effectiveness or
efficacy or experiment* or evaluat* or outcome Search modes ‐
Boolean/Phrase Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ Criminal Justice Abstracts
S21 S11 AND S20 Search modes ‐ Boolean/Phrase Interface ‐ EBSCOhost
Research Databases
Search Screen ‐ Advanced Search
Database ‐ Criminal Justice Abstracts
S20 S12 OR S13 OR S14 OR S15 OR S16 OR S17 OR S18 OR S19 Search modes
‐ Boolean/Phrase Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ Criminal Justice Abstracts
S19 (aversive or aversion or biofeedback or desensit* or imagery or
relaxation or meditat*) Search modes ‐ Boolean/Phrase Interface ‐
EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ Criminal Justice Abstracts
S18 (behav* N5 (modification* or therap* or rehabilitat* or interven*
or program* or treat* or approach* or technique*)) Search modes ‐
Boolean/Phrase Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ Criminal Justice Abstracts
S17 ((cognition or cognitive) N5 (therap* or rehabilitat* or interven*
or program* or treat* or approach* or technique*)) Search modes ‐
Boolean/Phrase Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ Criminal Justice Abstracts
S16 (social skill* N1 train*) Search modes ‐ Boolean/Phrase Interface
‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ Criminal Justice Abstracts
S15 (family therap* or psychotherap* or psycho‐therap*) Search modes ‐
Boolean/Phrase Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ Criminal Justice Abstracts
S14 ((multi systemic or multisystemic) N2 therap*) Search modes ‐
Boolean/Phrase Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ Criminal Justice Abstracts
S13 CBT Search modes ‐ Boolean/Phrase Interface ‐ EBSCOhost Research
Databases
Search Screen ‐ Advanced Search
Database ‐ Criminal Justice Abstracts
S12 cognitive behavio* Search modes ‐ Boolean/Phrase Interface ‐
EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ Criminal Justice Abstracts
S11 S9 AND S10 Search modes ‐ Boolean/Phrase Interface ‐ EBSCOhost
Research Databases
Search Screen ‐ Advanced Search
Database ‐ Criminal Justice Abstracts
S10 (juvenil* or teen* or adolescen* or young person* or young people
or youth* or child* or minor*) Search modes ‐ Boolean/Phrase Interface
‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ Criminal Justice Abstracts
S9 S1 OR S2 OR S3 OR S4 OR S5 OR S6 OR S7 OR S8 Search modes ‐
Boolean/Phrase Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ Criminal Justice Abstracts
S8 (rape or rapist) Search modes ‐ Boolean/Phrase Interface ‐
EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ Criminal Justice Abstracts
S7 paraphil* Search modes ‐ Boolean/Phrase Interface ‐ EBSCOhost
Research Databases
Search Screen ‐ Advanced Search
Database ‐ Criminal Justice Abstracts
S6 incest* Search modes ‐ Boolean/Phrase Interface ‐ EBSCOhost
Research Databases
Search Screen ‐ Advanced Search
Database ‐ Criminal Justice Abstracts
S5 (child* N1 porn*) Search modes ‐ Boolean/Phrase Interface ‐
EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ Criminal Justice Abstracts
S4 ((lewd* or indecen* or obscen*) N3 (behav* or act* or expos*))
Search modes ‐ Boolean/Phrase Interface ‐ EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ Criminal Justice Abstracts
S3 ((child* or sibling* or inter famil* or interfamil*) N1 (sex* or
molest*)) Search modes ‐ Boolean/Phrase Interface ‐ EBSCOhost Research
Databases
Search Screen ‐ Advanced Search
Database ‐ Criminal Justice Abstracts
S2 sex* N3 harm* N1 behav*) Search modes ‐ Boolean/Phrase Interface ‐
EBSCOhost Research Databases
Search Screen ‐ Advanced Search
Database ‐ Criminal Justice Abstracts
S1 (sex* N3 (abus* or aggress* or assault* or coerc* or delinquen* or
devian* or exploit* or homicid* or masochis* or molest* or murder* or
offen* or predator* or recidiv* or reoffend* or re‐offend* or torture*
or violen*)) Search modes ‐ Boolean/Phrase Interface ‐ EBSCOhost
Research Databases
Search Screen ‐ Advanced Search
Database ‐ Criminal Justice Abstracts 15,394

Social Care Online (www.scie‐socialcareonline.org.uk; searched 26 June 2019)

AllFields:'(adolescent* or juvenile* or young or teen*)'
AND AllFields:'("sex offender" OR "sex offenders" OR "sex offence" OR "sex offense" OR "sex offences" OR "sex offenses")'
AND AllFields:'(random* OR control OR group*)'

ProQuest Dissertations & Theses: UK & Ireland (searched 26 June 2019)

(sex* offen*) AND (juvenile OR minor OR adolescen* or teen*)

Networked Digital Library of Theses and Dissertations (NDLTD; www.ndltd.org/resources/find‐etds; searched 26 June 2019).

(sex* offen*) AND (juvenile OR minor OR adolescen* or teen*)

WorldCat (www.worldcat.org; searched 26 June 2019)

kw:sex offender* AND random*

ClinicalTrials.gov (clinicaltrials.gov/ct2/home; searched 26 June 2019)

sex offence OR sex offenders | Interventional Studies | cognitive OR behavioural OR CBT

UK Clinical Research Network (UKCRN; www.nihr.ac.uk/about‐us/how‐we‐are‐managed/managing‐centres/crn/; searched 26 June 2019)

Speciality: All Research summary: sex offenders

World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) (apps.who.int/trialsearch/Default.aspx; searched 26 June 2019)

Advanced search Title contains: sex* offen* OR Condition contains: sex* offen* AND Intervention contains: cognit* OR behav* OR CBT

Basic search sex* offen* AND CBT OR sex* offen* AND cognitive OR sex* offen* AND behavioural

Appendix 2. Data extraction form

Study procedures, including recruitment, diagnosis, dosage, duration and setting

• Study design (for example, randomised or quasi‐randomised)

• Randomisation method (including list generation)

• Method of allocation concealment

• Blinding participants

• Blinding of investigators

• Blinding of outcome assessors

Participants

• Inclusion/exclusion criteria

• Number (total/per group)

• Age at time of treatment

• Gender

• Learning disability

• Official categorisation of sexual offending behaviour

Treatment

• Location of treatment

• Modality of treatment

• Type of intervention

Data

Primary outcomes

• Recidivism

  • Any sexual offence

  • Any nonsexual offence

  • Time before re‐offence

• Adverse events (e.g. Juvenile Risk Assessment Scale (Hiscox 2007)

  • Increase in sexual offending

  • Increased seriousness of sexual offending

  • Self‐harm

  • Suicide attempt

  • Suicide

Secondary outcomes

• Reactions to the offending behaviour

  • Offence accountability: accepting responsibility for actions

  • Denial/minimization: acknowledging they engaged in the behaviour or some aspect of the offence

• Coping skills

• Psychological well‐being

  • Self‐esteem

  • General mental state

• Sexual attitudes and behaviour

  • Cognitive distortions about sexual behaviour

  • Deviant sexual interests, preferences or arousal

• Thinking patterns

• Victim empathy

• Aggression

• Social functioning

• Emotional self‐regulation and impulse control

• Substance use

• Programme engagement

  • Completion of treatment programme

  • Dropouts

  • Treatment refusers

• Economic outcomes

  • Direct costs

  • Indirect costs

Follow‐up data

• Duration of follow‐up

• Loss to follow‐up

Analysis data

• Methods of analysis (intention‐to‐treat/per‐protocol analysis)

• Comparability of groups at baseline (yes/no)

Appendix 3. 'Risk of bias' domains

Sequence generation

We described the method used to generate the allocation sequence in detail to assess whether it should have produced comparable groups.

Review authors' judgement: was the allocation concealment sequence adequately generated?

Allocation concealment

We described the method used to conceal the allocation sequence in sufficient detail to assess whether intervention schedules could have been foreseen in advance of, or during, recruitment.

Review authors' judgement: was allocation adequately concealed?

Blinding of participants and personnel

We described any measures used to blind participants and personnel from knowledge of which intervention a given participant might have received.

Review authors' judgement: was knowledge of the allocated intervention adequately prevented during the study?

Blinding of outcome assessment

We described any measures used to blind outcome assessors' from knowledge of which intervention a given participant might have received.

Review authors' judgement: was knowledge of the allocated intervention adequately prevented during the study?

Incomplete outcome data

We extracted and reported data on attrition and exclusions, as well the numbers involved (compared with total randomised), reasons for attrition/exclusion (where reported or obtained from investigators) and any re‐inclusions in the analyses performed by the review authors.

Review authors' judgement: were incomplete outcome data adequately addressed?

Selective outcome reporting

We attempted to assess the possibility of selective outcome reporting by investigators.

Review authors' judgement: were reports of the study free of suggestion of selective outcome reporting?

Other sources of bias

We described any important concerns we had about other possible sources of bias.

Review authors' judgement: was the study apparently free of other problems that could put it at risk of bias?

Data and analyses

Comparison 1. CBT versus no intervention or TAU.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Secondary outcome: psychological well being ‐ self‐esteem (CBT vs no treatment)	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
1.2 Secondary outcome: sexual attitudes and behaviour ‐ cognitive distortions about sexual behaviour (general) (CBT vs no treatment)	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
1.3 Secondary outcome: victim empathy ‐ attitudes towards women (CBT vs no treatment)	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
1.4 Secondary outcome: psychological well being ‐ depression (CBT vs TAU)	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
1.5 Secondary outcome: sexual attitudes and behaviour ‐ cognitive distortions about sexual behaviour pertaining to rape (CBT vs TAU)	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected

Comparison 2. CBT versus alternative interventions.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 Primary outcome: sexual aggression	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
2.2 Secondary outcome: psychological well‐being ‐ CBCL	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
2.2.1 Internalizing	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
2.2.2 Externalizing	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
2.2.3 Total	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
2.3 Secondary outcome: sexual attitudes and behaviour ‐ cognitive distortions about sexual behaviour pertaining to justifications (CBT vs sexual education programme)	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
2.4 Secondary outcome: sexual attitudes and behaviour ‐ cognitive distortions about sexual behaviours pertaining to apprehension confidence (CBT vs sexual education programme)	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
2.5 Secondary outcome: sexual attitudes and behaviour ‐ cognitive distortions about sexual behaviours pertaining to inappropriate sexual fantasies (CBT vs sexual education programme)	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
2.6 Secondary outcome: sexual attitudes and behaviour ‐ cognitive distortions about sexual behaviour pertaining to social‐sexual desirability (CBT vs sexual education programme)	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
2.7 Secondary outcome: psychological well‐being ‐ DSMD	1		Other data	No numeric data
2.7.1 Internalizing	1		Other data	No numeric data
2.7.2 Externalizing	1		Other data	No numeric data
2.7.3 Critical pathology	1		Other data	No numeric data
2.7.4 Total	1		Other data	No numeric data

2.7. Analysis.

Comparison 2: CBT versus alternative interventions, Outcome 7: Secondary outcome: psychological well‐being ‐ DSMD

Secondary outcome: psychological well‐being ‐ DSMD
Study	CBT	MDT	SST
Internalizing
Apsche 2005	Mean = 70.5 (range = 62 to 84)	Mean = 71.3 (range = 64 to 83)	Mean = 72.10 (range = 62 to 84)
Externalizing
Apsche 2005	Mean = 73.1 (range = 64 to 86)	Mean = 72.5 (range = 67 to 84)	Mean = 71.25 (range = 60 to 86)
Critical pathology
Apsche 2005	Mean = 68.7 (range = 58 to 88)	Mean = 70.5 (range = 60 to 86)	Mean = 72.33 (range = 68 to 86)
Total
Apsche 2005	70.77	71.50	71.79 (range = 62 to 84)

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Apsche 2005.

Study characteristics
Methods	Randomised controlled trial
Participants	Setting: inpatient treatment facility in the USA Sample size: total of 60 male adolescents (44 Black, 13 European Americans, 4 Hispanic): CBT n = 19: 14 black, 4 white, 1 Hispanic Mode deactivation therapy (MDT) n = 21: 15 black, 5 European Americans, 1 Hispanic Social skills training (SST) n = 20: 14 black, 4 white, 2 Hispanic Withdrawals: no attrition Mean age: CBT = 16.5 years MDT = 16.5 years SST = 16.1 years Inclusion criteria: documented incidences of physical and sexual aggression. All had been diagnosed with conduct or personality disorder, or both. MDT: The principal Axis I diagnoses for this group included conduct disorder (n = 15), oppositional defiant disorder (n = 2), post‐traumatic stress disorder (n = 7), and major depressive disorder, primary or secondary (n = 5). Axis II diagnoses for the group included mixed personality disorder (n = 6), borderline personality traits (n = 3), and narcissistic personality traits (n = 2). CBT: The principal Axis I diagnoses for this group included conduct disorder (n = 14), oppositional defiant disorder (n = 4), and post‐traumatic stress disorder (n = 7). Axis II diagnoses for the group included mixed personality disorder (4), borderline personality disorder (2), narcissistic personality disorder (1) and dependent personality disorder (n = 1). SST: The principal Axis I diagnoses for this group included conduct disorder (n = 17), oppositional defiant disorder (n = 3), post‐traumatic stress disorder (n = 5). Axis II diagnoses for the group included mixed personality disorder (n = 4), borderline personality traits (n = 1), narcissistic personality traits (n = 1), and avoidant personality traits. Exclusion criteria: none described
Interventions	Comparison: CBT with MDT or SST Experimental group (n = 19): CBT. The CBT methodology used for this group employed a published treatment curriculum and workbook system for adolescents who had committed sexual offences called “Thought Change” (quote; Apsche 1999; Apsche 2004). This structured treatment program is specifically designed for personality disordered and conduct‐ordered youth with psychosexual disturbances and high levels of aggression and violence. Components of this psycho‐educational treatment curriculum included daily recording of negative thoughts, cognitive distortions, cognitive restructuring, sexual offence patterns and beliefs, aggressive patterns and beliefs, mood management, dysfunctional beliefs, taking responsibility, mental health maintenance, substance abuse issues, and victim empathy. Control groups: MDT (n = 21): MDT is a methodology designed to treat conduct disordered youths who have co‐occurring personality disorders or traits. The methodology is completed in individual groups and family therapy. There is a clinician’s manual and a client work book to assure adherence to the MDT methodology in individual, group and family therapy. SST: (n = 20) Social Skills Training program included identification and reinforcement of appropriate behaviours, target skill identification, modelling, practicing skills, and role playing. The youth in this condition were encouraged to practice skills and were reinforced by shaping and fading procedures. All staff and therapists were trained and supervised in SST by a doctoral‐level psychologist. All skill training was performance based and evaluated each individual. SST was chosen as a control because it was the treatment as usual for part of the residents at the residential treatment centre and an accepted method of intervention at many facilities Duration and format of interventions not described. Average length of residential treatment across all conditions across all 3 conditions was 11 months.
Outcomes	Physical and sexual aggression, assessed with Daily Behaviour Report Cards and Behaviour Incidence Reports forms completed by staff, which were analysed as the average number of incidences per week relating to physical and sexual aggression during the first 60 days following admission (baseline) and the 60 day period prior to discharge (post‐treatment) Psychological well‐being assessed with: Child Behaviour Checklist (CBCL; Achenbach 1991); a multiaxial assessment regarding the behaviours and competencies of 11‐ to 18‐year olds around internalising (i.e. withdrawn behaviours, somatic complaints, anxiety and depression), externalising (i.e. delinquent behaviour and aggressive behaviour), and total problems (i.e. the conglomerate of total problems and symptoms, both internal and external); higher scores on this measure represent increased severity of problems Devereux Scales of Mental Disorders (DSMD; Naglieri 1994), which illustrate the level of functioning in comparison to a normal group, via behavioural ratings. The 4 scales assess: 1) internalizing (which measures negative internal mood, cognition, and attitude); 2) externalizing (which measures prevalence of negative overt behaviour or symptoms); (3) critical pathology (which represents the severe and disturbed behaviour in children and adolescents); and total (which represent the conglomerate of all scores, including general Axis I pathology, delusions, psychotic symptoms, and hallucinations). Higher scores on this measure represent increased severity of problems. Scores are converted to T scores, which have a mean of deviation of 10; a score of 60 or higher indicates an area of clinical concern. Results All groups showed improvements in physical and sexual aggression. Baseline average rate of aggression was 1.56 (SD = + 0.501, SE = +0.65), which significantly reduced to a post‐treatment mean of 0.41 (SD = 0.495, SE = 0.65) (independent t test: T = 18, df = 59, P < 0.01). MDT showed statistically significant improvements in post‐treatment physical aggression (80.7%) compared to both CBT (72.6%) and SST (68.8%). There was no statistically significant difference between CBT and SST. The reduction in rates of sexual aggression from CBT (72%) and SST (70.6%) were not statistically significant from baseline to post‐treatment. MDT did show a statistical reduction in sexual aggression (84.5%). Residents who participated in MDT had lower scores on CBCL and DSMD measures than residents who participated in CBT. Mean DSMD scores for the internalising factor, externalising factor, critical pathology, and total score for the MDT group were at or near one SD below the CBT group. MDT was more than one SD more significant in reducing internal, external, and total scores on the CBCL. Timing of outcome assessment: outcomes were measured before and after the intervention
Notes	Study start date: not stated Study end date: not stated Funding sources and conflicts of interest: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Comment: described as randomly assigned to one of the three treatment conditions but on the basis of available openings in the caseload of the participating clinicians at the time of admission. This is a vague description.
Allocation concealment (selection bias)	Unclear risk	Comment: no specific information about allocation concealment provided
Blinding of participants and personnel (performance bias) All outcomes	High risk	Comment: no blinding of participants or personnel described
Blinding of outcome assessment (detection bias): subjective outcomes All outcomes	Unclear risk	Comment: no blinding of outcome assessment described
Incomplete outcome data (attrition bias) All outcomes	Low risk	Comment: attrition not reported
Selective reporting (reporting bias)	Low risk	Comment: information presented on all measures
Other bias	Low risk	Comment: groups described as similar at baseline for demographics and prior offence history

Karakosta 2015.

Study characteristics
Methods	Experimental study with random assignment
Participants	Setting: residential group treatment program in the northwestern part of the USA Sample size: 21 adjudicated, juvenile males who had committed sex offences, from 25 who were pre‐tested, completed the programme. Used purposive sampling used ‐ only participants willing to enter the programme and who agreed to participate in a thorough assessment procedure participated in the study. Only those residents from whom consent was obtained from their parents or legal guardians were allowed to participate in the study. Withdrawals: 4 (3 were discharged and 1 moved to higher level care during the 12‐week programme) Mean age: not reported (range = 12‐17 years) Inclusion criteria: adjudicated, juvenile males who committed sexual offences and were older than 12 years of age comprised the sample. Additionally, only volunteers for whom parents/guardians had given consent were allowed to take part in the study. 1 offender with a history of violent sexual offending and offenders with learning difficulties were also included in the sample. Exclusion criteria: none described
Interventions	Comparison: Integrated Sex Offender Treatment Programme (ISOTP) versus treatment as usual (TAU) Experimental group (n = not reported): ISOTP. This programme is a multi‐faceted treatment process that takes the offender through an initial phase of screening and assessment through cognitive behavioural health treatment interventions, leading to admission for successful discharge from the programme. This programme relies upon a standardized treatment curriculum facilitated in individual and structured sex‐offender‐specific group activities. The ISOTP is a 4‐phased approach, designed for 6‐12 months of care. Screening and assessment protocols ensure that participants receive a comprehensive behavioural health and sex‐offender‐specific needs assessment that is both timely and culturally sensitive. Behavioural health and sex‐offender‐specific treatment interventions utilized within the ISOTP program include: a) individual counselling and case management; b) family interventions (if feasible and appropriate); c) crisis intervention services; and d) structured, sex‐offender‐specific treatment, based upon curriculum lessons. Control group (n = not reported): TAU, which features 3 specific program curricula: 1) Pathways: A guided workbook for youth beginning treatment (Kahn 2011); 2) Elements of the Footprints: Steps to a Healthy Life (Hansen 2006; and 3) Elements of the Roadmaps to Recovery (Kahn 2007). Treatment is administered according to following regimen: a) 5 × 90‐minute sex‐offender‐specific group sessions per week; b) a life skills group meeting daily; c) individual therapy once per week; and d) family therapy once per month. Residents are assigned to specific groups based on standard pre‐assessment testing employing Juvenile Sex Offender Protocol‐III (JSOAP‐III) and assessment for a) suicide risk and b) risk of self‐harm. TAU does not rely upon a standardized treatment curriculum or a treatment‐phased approach. TAU regimens Pathways: A guided workbook for youth beginning treatment: the first of the TAU regimens, focuses on strength‐based methods to help individuals develop healthy and productive lifestyles reflecting the Good Lives Model discussed earlier in this work. Footprints: Steps to a Healthy Life is the second TAU, developed for disabled adults and adolescents with a variety of sexual behaviour problems, including learning disabilities, attention deficits, and cognitive disabilities. In its second edition, Footprints: Steps. To a Healthy Life focuses on a) a positive and goal‐oriented approach consistent with the Good Lives Model of treatment, b) self tests at the end of each chapter that reinforce the concepts discussed in the chapters, c) flash cards designed to foster retention of concepts and information that can also be personalized by the client, and d) a Clinician’s Guide provided separately to help clinicians implement the programme. Roadmaps to Recovery: also in the form of a workbook, targets boys and girls from the ages of 6 to 12 years and is designed to help them overcome sexual behaviour problems by emphasizing the goal of becoming a 'survivor'. The creators of Roadmaps to Recovery have designed the curriculum to help abuse‐reactive children learn to make changes that alter their perceptions from being victims to becoming survivors by developing healthier behaviours. The programme incorporates aspects of trauma‐focused CBT designed to help children learn by focusing on several components, including, among other goals, learning to express feelings, regulate emotions, manage sexual urges, manage stress, and distinguish appropriate from inappropriate sexual activity. Together, elements of these programs comprise TAU.
Outcomes	Depression, measured using the Children’s Depression Inventory 2 (CDI‐2); a self‐report scale assessing emotional problems and functional problems, higher scores indicate more severe depression, and higher levels of emotional or functional problems Anxiety, measured using the Revised Children’s Manifest Anxiety Scale (RCMAS); a total anxiety score is calculated on the basis of 28 items divided into 3 subscales, measuring: a) physiological anxiety (such as sleep difficulties, fatigue, and nausea); b) worry and over‐sensitivity (such as obsessive concerns about having one’s feelings hurt or feeling emotionally isolated); and c) social concerns related to interpersonal relations and levels of concentration. Higher scores on this measure represent higher levels of anxiety Negative attitudes, measured using the Hostility Towards Women Scale; total possible scores range from 0 to 30, higher scores indicate more problematic cognitive distortions Cognitive distortions of sexual behaviour measured using the Bumby Cognitive Cardsort Rape Scale; a self‐report instrument wherein respondents rate their agreement to 36 statements reflecting attitudes toward women Results The experimental group experienced statistically significant decreases in cognitive distortions related specifically to rape distortions compared to the control group Greater reductions within the experimental group in cognitive distortions related specifically to rape and molestation Timing of outcome assessment: outcomes measured before and after the intervention
Notes	Study start date: not stated Study end date: not stated Funding sources and conflicts of interest: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Comment: no description provided of how randomisation was conducted
Allocation concealment (selection bias)	Unclear risk	Comment: no specific information about allocation concealment provided
Blinding of participants and personnel (performance bias) All outcomes	High risk	Comment: no blinding took place, for either participants or personnel
Blinding of outcome assessment (detection bias): subjective outcomes All outcomes	Unclear risk	Comment: no information given; only subjective outcomes measured, not objective ones (e.g. recidivism, self‐harm)
Incomplete outcome data (attrition bias) All outcomes	High risk	Comment: 4 lost to attrition due to changes in circumstances (3 discharged and 1 moved to a higher‐level care)
Selective reporting (reporting bias)	High risk	Comment: according to the authors, they only reported significant differences. Data on non‐significant sub‐scales were presented partially. We attempted to contact the authors to obtain this information but did not receive a response.
Other bias	Unclear risk	Comment: no obvious signs of other bias

Mathѐ 2007.

Study characteristics
Methods	Randomised controlled trial
Participants	Setting: correctional services – Westville medium B prison with maximum security prisoners in South Africa Sample size: 18 sentenced juvenile offenders. Used purposive sampling ‐ only participants willing to enter the programme and who agreed to participate in a thorough assessment procedure participated in the study. Withdrawals: 0 Mean age: not reported; described as male juveniles, but no other age information given Inclusion criteria: 18 sentenced juvenile offenders who committed sexual offences against females older than 12 years of age Exclusion criteria: no exclusion criteria were stated in the paper beyond the characteristics of the sample
Interventions	Comparison: group‐based, cognitive behavioural intervention versus no treatment Experimental group (n = 9): CBT. The experimental group attended group sessions characterised by the facilitation of a treatment programme for people who had committed a sexual offence using cognitive‐behavioural interventions (self‐instructional training and social‐cognitive skills training). The programme was characterised by 6 modules with 23 sessions. The group met on a weekly basis for approximately 2 hours per session. The group was also the context for data collection and provided scope for clarification and during the sessions. These strategies were group discussions, assignments, homework, audio‐visual material and each research participant’s presentation. The style adopted de‐emphasised the researcher’s role as teacher. Instead, the role was to facilitate interactions among group members, set up discussions and sharing of experiences that would enhance members’ commitment to change and deal with discouragement. Control group (n = 9): no treatment
Outcomes	Self‐concept, measured by the Self‐Concept Scale. This scale was designed by the researchers. A score between 81 and 125 indicates that the person has quite a number of problem areas contributing to negative self‐concept. A score between 51 and 80 indicates that the person has a fair self‐concept and attention should be given to the problem areas. A score between 31 and 50 indicates that the person has a good self‐concept with few problem areas. A score between 20 and 30 indicates that the person has a very good self‐concept. Sexual attitudes and behaviour, measured by the Abel and Becker Cognition Scale; a 29‐item scale that measures cognitive distortions (marked on a 5‐point Likert scale). Scores range from 29 to 145 with higher scores showing fewer cognitive distortions Victim empathy, measured by the Spencer and Helmreich Attitude Towards Women Scale; a 15‐item scale that measures attitudes towards a number of aspects of women's roles including vocational, educational, and interpersonal relationships. Scores range from 0 to 45 and a high score indicates a more egalitarian attitude towards women Results CBT provided evidence of greater improvements to self‐concept compared to no treatment Participants receiving CBT more positive attitudes towards women and increased victim empathy compared to the control group All participants in the CBT group showed more rational, positive and improved cognitions in their post‐test scores Timing of outcome assessment: outcomes were measured before and after the intervention
Notes	Study start date: not stated Study end date: not stated Funding sources and conflicts of interest: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Comment: no description provided on how randomisation was done
Allocation concealment (selection bias)	Unclear risk	Comment: no description provided. We attempted to contact the authors of published reports to clarify allocation concealment issues, but none responded
Blinding of participants and personnel (performance bias) All outcomes	High risk	Comment: no blinding took place, for either participants or personnel
Blinding of outcome assessment (detection bias): subjective outcomes All outcomes	Unclear risk	Comment: no information given; only subjective outcomes measured, not objective ones (e.g. recidivism, self‐harm)
Incomplete outcome data (attrition bias) All outcomes	Low risk	Comment: no attrition
Selective reporting (reporting bias)	Unclear risk	Comment: no obvious signs of selective reporting, protocol not available for review
Other bias	High risk	Comment: researcher was one of the therapists delivering the intervention

Piliero 1994.

Study characteristics
Methods	Randomised controlled trial
Participants	Setting: community‐based setting in New Jersey, USA Sample size: 16 adolescents who had committed sex offences, from 20 selected, completed final treatment Withdrawals: 4 lost to attrition (all in control group and lost because of changes in their probation status) Mean age: not provided (range = 13‐18 years) Inclusion criteria: all charged with and convicted of sexual misconduct, or voluntarily admitted to a history of criminal sexual behaviour. All showed mental health stability (i.e. no psychosis) and all of average intelligence. Exclusion criteria: excluded if they had learning disability
Interventions	Comparison: community‐based, group, CBT versus a sex education programme Experimental group (n = 10): CBT. The CBT treatment protocol for experimental participants involved 4 sessions of victim empathy training, 6 sessions of covert sensitization and 2 sessions of masturbatory satiation. Duration was 12 weeks. There were 2 therapists, one male, one female, and they were both aware of which group (experimental or control) they were working with. A project assistant, who is the probation officer to the offenders, administered the evaluation scales to control for experimenter bias. Control group (n = 6): sex education, with no elements believed to be correlated to the outcomes of the study
Outcomes	Measures included (Piliero 1994): Changes in cognition, assessed by 3 measures (the Multiphasic Sex Inventory ‐ juvenile form, the Multiphasic Sex Inventory (Supplement) and a structured interview), all of which were administered and scored by a programme assistant (the probation officer for the offenders) and not the treatment providers, to control for experimenter bias. The Multiphasic Sex Inventory monitors sexual offending, fantasies, attitudes, cognitive distortions and knowledge about sex. Psychosexual characteristics are classified into 18 scales that are scored individually. The Multiphasic Sex Inventory (supplement) was designed by the researcher to assess sexual behaviour, cognitions and fantasies. The interview covered sexual fantasies, beliefs about their behaviour, attitudes towards victims and the general effect of the programmes. Results Participants who received CBT scored greater improvements than control participants in cognitive distortions, victim empathy and self‐esteem No difference was found between the groups on the Sexual Fantasy Scale Timing of outcome assessment: outcomes were measured before and after the intervention
Notes	Study start date: not stated Study end date: not stated Funding sources and conflicts of interest: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Comment: no description given of how randomisation was undertaken
Allocation concealment (selection bias)	Unclear risk	Comment: no description given. We attempted to contact the authors of the published reports to obtain clarifications about methodological issues, but none responded.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Comment: no blinding took place, for either participants or personnel
Blinding of outcome assessment (detection bias): subjective outcomes All outcomes	Unclear risk	Comment: no information given; only subjective outcomes measured, not objective ones (e.g. recidivism, self‐harm)
Incomplete outcome data (attrition bias) All outcomes	High risk	Comment: 4 lost to attrition in control group because of change in probation status; none lost to attrition in the treatment group
Selective reporting (reporting bias)	High risk	Comment: reporting of results is not entirely clear as data for a number of scales appear to be missing or incomplete
Other bias	High risk	Comment: researcher was one of the therapists delivering the intervention; testing instruments were administered by another practitioner who knew the participants

CBT: cognitive‐behavioural therapy
DPT: dynamic play therapy
SD: standard deviation
SE: standard error

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Bonner 1999	The majority of the participants were younger than this review's target group (i.e. under 10 years of age). It was not possible to obtain separate data for the cohort within the eligible age range for this review
Dunham 2009	Not a randomised controlled trial. Not CBT
Feilzer 2004	Not a randomised controlled trial
Graham 1998	Not a randomised controlled trial
Gretton 2005	Not a randomised controlled trial
Hird 1996	Not a randomised controlled trial
Hout 2002	Not a randomised controlled trial
Jones 1998	Not a randomised controlled trial
Lab 1993	Not a randomised controlled trial
Langdon 2007	Not a randomised controlled trial. Adult participants
Linday 1999	Not a randomised controlled trial
Marshall 2008	Not CBT. Adult participants
Pérez 2012	Not a randomised controlled trial
Thoder 2011	Not a randomised controlled trial
Viens 2012	Not a randomised controlled trial
Waite 2005	Not a randomised controlled trial
Weinrott 1997	Not CBT as defined in this review
Worling 2000	Not a randomised controlled trial

CBT: cognitive‐behavioural therapy.

Differences between protocol and review

The authors are different in the protocol and the review: Mike Ferriter and Avery Bowser contributed to the protocol (Sneddon 2012) but not the final review; NL and DG contributed to the final review but not the protocol. Only HS and GM contributed to both the protocol and the final review.

We were not able to use in this review all of the methods set out in the published protocol (Sneddon 2012). For further detail, please see Table 4.

When we began searching in 2014, we no longer had access to National Criminal Justice Reference Service Abstracts Database and instead we searched Criminal Justice Abstracts. When we updated our searches in 2019, we could not access either database.

We updated our references to chapters of the Cochrane Handbook for Systematic Reviews of Interventions, to refer to the most recent chapter.

We now refer to certainty of the evidence, instead of quality of the evidence, in line with current practice.

Contributions of authors

Helga Sneddon (HS) drafted the protocol (Sneddon 2012), with input from GM. HS led the search for grey literature, including any ongoing and unpublished studies, and contacted experts in the field. DG contacted authors of published studies for further clarification. GM supervised HS's Cochrane Fellowship and assisted with the screening of the first search. HS, DGG and NL independently screened records for eligible studies and GM arbitrated where there was disagreement. Working independently, HS and DGG extracted the data, resolving disagreements amongst themselves, and assessed the risk of bias and the certainty of the evidence. GM resolved any disagreements. HS, DGG, NL and GM wrote the review. HS is the guarantor for the review.

Sources of support

Internal sources

• None, Other

External sources

• Health and Social Care Research and Development Division, Public Health Agency, UK

  HS was supported by a Cochrane Fellowship.

Declarations of interest

Helga Sneddon's institution received a Cochrane Fellowship from the Health and Social Care Research and Development Division of the Public Health Agency in Northern Ireland to cover payment for her salary while undertaking the review.
Dina Gojkovic Grimshaw ‐ none known.
Nuala Livingstone is an Associate Editor with the Cochrane Editorial & Methods Department, and an Editor with DPLP.
Geraldine Macdonald is the Co‐ordinating Editor of DPLP.

Edited (no change to conclusions)