The last 20 years have seen increased research into sexual recidivism by adolescents who have sexually offended (ASO). The three most recent meta-analyses of sexual recidivism by adolescents have reported mean rates of 5% (Caldwell, 2016; N = 106 studies), 7% (Caldwell, 2010; N = 63 studies) and 12% (McCann and Lussier, 2008; N = 18 studies); the mean follow-up period for all these was ~5 years. The vast majority of these studies were conducted using adjudicated samples and official data on recidivism. As a result, recidivism rates are possibly underestimated.

The identification of predictors of sexual recidivism is a critically important issue, since a single recidivist may be responsible for a great many victims. Unfortunately, there is still little consensus on reliable predictors of sexual recidivism, especially because of low base rates reported. McCann and Lussier's (2008) meta-analysis identified variables associated with an increased risk of sexual recidivism by ASO: a criminal record for either sexual or non-sexual offending, specific victim characteristics (male, stranger, child/adult but not peer), and evidence of sexual deviance and antisociality. However, these results must be interpreted with caution, given the low effect sizes, and the heterogeneity of the definitions of variables in the studies included. Nevertheless, these results are consistent in large part with those of Worling and Långström's (2006) systematic review of the literature on sexual recidivism. Those authors identified a total of six empirically supported risk factors for sexual recidivism—that is, factors that were statistically associated with sexual recidivism in at least two empirical studies and whose relevance has not been contested by the scientific community: (1) a criminal record for sexual aggression; (2) deviant sexual interests; (3) assault of a stranger; (4) assault of more than one victim; (5) social isolation; and (6) dropping out of specialized treatment. It should be noted that the last two factors were not studied in McCann and Lussier's (2008) meta-analysis.

Victim age is the most frequently encountered classification criterion in the literature, but has produced inconsistent results. The relationship between recidivism and victim age is unclear. On the one hand, ASO of peers/adults have been reported to exhibit significantly higher rates of sexual (Nisbet et al., 2004; Parks and Bard, 2006), violent non-sexual (Krause et al., 2020) non-sexual (Nisbet et al., 2004), and general (Vandiver, 2006; Fanniff and Kolko, 2012) recidivism. On the other, they have also been reported to exhibit significantly lower rates of sexual recidivism (Kemper and Kistner, 2007), as well as comparable rates of sexual (Krause et al., 2020), violent (Aebi et al., 2012), non-sexual (Kemper and Kistner, 2007), and general (Hagan et al., 2001; Parks and Bard, 2006; Aebi et al., 2012; Krause et al., 2020) recidivism. In addition, ASO against peers/adults have been reported to be more likely than ASO against children to drop out of treatment (Parks and Bard, 2006). From a theoretical perspective, it has been argued that ASOs against peers/adults are more generalist offenders than specialists, and that sexual offending is one manifestation of a general antisocial orientation (Seto and Lalumière, 2010; Leroux et al., 2016). This hypothesis could explain higher rates of recidivism and poor adherence in treatment (Olver et al., 2011).

Few studies have analyzed mixed ASO (i.e., ASO who had assaulted both children and peers/adults) as a distinct subgroup, despite the empirical evidence in favor of such an analytical strategy (see Kemper and Kistner, 2007; Joyal et al., 2016; Leroux et al., 2016). Mixed ASO are the subgroup of ASO with the greatest treatment needs (Leroux et al., 2016) and are the most likely to drop out of treatment (Kemper and Kistner, 2007).

Victim relationship has not been widely used for classification purposes in ASO, in contrast to adults who have sexually offended (Tidefors et al., 2010). In general, sexual recidivism rates have been reported to be lower in intrafamilial adult sexual aggressors than in extra familial ones (Hanson and Bussière, 1998); however, the reverse has been reported for incestuous young adults (18–24 years), (Hanson, 2002). Among ASO, assault of a stranger victim is generally associated with a higher risk of sexual recidivism (Worling and Långström, 2006; McCann and Lussier, 2008). A recent meta-analysis comparing extra familial and intrafamilial ASOs found higher scores on antisociality indicators among the first group, suggesting a more general pattern of criminality (generalist theory; Martijn et al., 2020). The only reported study to analyses the effect of treatment of ASO on recidivism as a function of victim relationship was that of Latzman et al. (2011; N = 166), who used the Estimate of Risk of Adolescent Sexual Offense Recidivism (ERASOR; Worling and Curwen, 2001) and the Youth Level of Service Case Management Inventory (YLS-CMI; Hoge et al., 2002) to measure the risk of recidivism in intrafamilial and extra familial ASO. Intrafamilial and extra familial ASO did not significantly differ in their global or subscale scores (including the score on the ERASOR treatment subscale) on these two instruments. However, the real rates of recidivism in these two groups were not measured, and mixed ASO were included in the intrafamilial subgroup, rather than considered a distinct group. Finally, the study included only ASO who had received treatment; there was no comparison group (untreated ASO).

Current research data suggests that cognitive-behavioral and multi-systemic approaches are more effective in reducing recidivism for ASOs (Walker et al., 2004; Fanniff and Becker, 2006; Kim et al., 2016; Ter Beek et al., 2018). The cognitive-behavioral therapeutic approach (CBT) is more widely used, and is considered to be a first-line treatment, whether in groups or individually (Bereiter and Mullen, 2012; Kim et al., 2016; ATSA, 2017). CBT aims to teach adolescents to identify and modify their thoughts and feelings which support or precipitate abusive sexual behavior (Kim et al., 2016). To this end, CBT develops emotional and behavioral self-regulation skills, in addition to social skills (Rich, 2011). Intervention involves work on accountability for abusive behaviors, the development of empathy, cognitive restructuring, exploring one's own sexual victimization, if applicable, and reducing deviant sexual interests (Center for Sex Offender Management, 2006). Generally, CBT includes relapse prevention strategies, which aim to shed light on the adolescent cycle of abuse, identification of high-risk situations, and the development of internal and external control strategies that reduce risk recidivism (Rich, 2011). This allows the learning and consolidation of new and more adapted behaviors (Sion and Blondeau, 2012). Family involvement and support in the therapeutic process is crucial and can help reduce the risk of recidivism and promote better general functioning (Worling and Curwen, 2000; ATSA, 2017). According to the Risk-Need-Receptivity (RNR) model, the intensity of services should be tailored to the level of risk of recidivism and criminogenic needs (Bonta and Andrews, 2007).

Caldwell's (2007) meta-analysis revealed that sexual recidivism by ASO has significantly decreased over the last 15 years, falling from a mean rate of 10.3% in the studies published between 1980 and 1995 to 2.75% in the studies published between 2000 and 2015. In Caldwell's view, at least some of this significant decrease is due to the effectiveness of specialized treatment programs and increased accessibility to these programs. Indeed, the great majority of meta-analyses on this topic have found that ASO who have received specialized treatment of any kind exhibit lower sexual recidivism rates than do those who have not received any treatment (Walker et al., 2004; Reitzel and Carbonell, 2006; Hanson et al., 2009; Schmucker and Lösel, 2015; Ter Beek et al., 2018). For example, in Reitzel and Carbonell's (2006) meta-analysis of nine studies (N = 2,986), the mean sexual recidivism rate after a mean follow-up period of 59 months was 7.37% in ASO who had received specialized treatment (n = 1,655), and 18.93% in ASO who had received no treatment (n = 1,331). While the effect size varied widely between the studies, the results nevertheless suggest that specialized treatment is effective in reducing sexual recidivism. Walker et al.'s (2004) meta-analysis of 10 studies (N = 644 ASO) arrived at a similar conclusion. Kim et al.'s (2016) conducted a systematic review that included five meta-analyses published since 2002 (including the two previously described). They found that specialized treatment of any type lowered general recidivism rates (i.e., reoffending of any type) by a mean of 24% compared to no treatment, and was more effective in ASO than in adult sexual offenders. However, sexual recidivism was not specifically analyzed.

The authors of the most recent meta-analysis on this question also concluded that specialized treatment was effective in reducing the rate of general, but not sexual, recidivism in ASO (Kettrey and Lipsey, 2018). Moreover, the authors call in to question the reliability of the conclusions of previous meta-analyses, the majority of which included studies with non-experimental designs. They also note that their own meta-analysis' statistical power—and by extension, ability to detect more modest effects on sexual recidivism—was limited by the small number of studies they analyzed that had used, an experimental design (N = 7), and the modest effect sizes.

While the vast majority of studies of the effectiveness of treatment programs have compared ASO who received treatment to those who had not (either no treatment at all or unspecialized treatment), Edwards et al.'s (2005) compared ASO who had completed specialized treatment programs to those who had dropped out. In their sample (N = 49), general, violent, and sexual recidivism rates after a mean follow-up period of 3 years were higher in ASO who had dropped out of treatment than in those who had completed treatment: in fact, none of the ASO who completed their treatment programs committed a sexual offense in the follow-up period. These results are interesting, as the significant differences observed between ASO who completed treatment and those who dropped out suggest that it is important to study these two subgroups as distinct entities, something few studies have done to date. Unfortunately, it is difficult to generalize from these results, given the study's small sample size, the relatively short follow-up period, and the low recidivism rates.

In their review of studies published between 1986 and 2005 on recidivism by ASO, Fortune and Lambie (2006) concluded that the majority of studies were characterized by methodological weaknesses, and that these weaknesses were primarily in three domains: (1) characteristics of the populations studied, and the absence of control groups; (2) definition of recidivism; (3) duration of follow-up period. Their criticisms remain pertinent today, as very few studies have addressed all these issues (Kettrey and Lipsey, 2018).

First, Fortune and Lambie (2006) observed that most studies do not use appropriate control groups (ASO who had not received specialized treatment or who had not completed treatment). In addition, they acknowledge that obvious ethical considerations render randomization of subjects to groups (specialized treatment vs. no treatment or unspecialized treatment) for the purposes of comparison difficult, if not impossible. They therefore suggested that a more useful approach to evaluating the effectiveness of treatment in reducing recidivism by ASO would be to compare groups who had received treatment to those that had not received treatment or had not completed treatment (p. 1,090). They also note the importance of taking into consideration the criteria that govern the (non-random) allocation of participants to treatment groups. More specifically, they recommend taking into account not only the usual variables (e.g., age, gender, socio-economic status) but also variables that may influence selection of candidates for treatment or candidates' ability to complete treatment (e.g., comorbid psychiatric conditions, history of sexual victimization). Finally, they emphasize the importance of ensuring samples are large enough to ensure representability and the construction of groups with comparable personal, familial, and offending characteristics.

Second, given the low base rates of sexual recidivism and the higher rates of non-sexual and general recidivism, Fortune and Lambie recommend measuring the effectiveness of treatment in terms of all three recidivism rates. In addition, they suggest using several sources of information on recidivism (official and non-official); should this be impossible, they suggest using an official source that best reflects real offending behavior (e.g., charges rather convictions, to avoid the effects of plea bargaining). Third, they recommend using follow-up periods that are long enough to allow evaluation of the effects of treatment on short-, medium-, and long-term recidivism.

The current study was designed to avoid the three major weaknesses outlined above. First, it uses a large (N = 351) sample of ASO, all of whom had undergone a multidisciplinary (e.g., psychology, psychiatry, criminology) evaluation that was adapted to sexual offenders and intended to direct them to specialized treatment. Second, the study comprises three groups: (1) ASO who had completed specialized treatment (completed treatment); (2) ASO who received no treatment (no treatment); and (3) ASO who had commenced treatment but dropped out before the treatment was completed (dropped out). The sexual, violent (including sexual), and general recidivism rates of the completed treatment group (experimental group) was compared to those of the other two (control) groups. Third, the mean follow-up period was almost 8 years (range = 2–13 years), long enough to allow evaluation of recidivism. Fourth, data on recidivism was collected from multiple official sources, namely Youth Court (recidivism in adolescence), and municipal, provincial, and federal courts (recidivism in adulthood). In keeping with the recommendation of Fortune and Lambie (2006) and Caldwell (2007), recidivism was defined in terms of charges laid, to avoid the effect of plea bargaining. Finally, the current study not only takes into consideration the principal methodological recommendations of Fortune and Lambie (2006), but also analyses recidivism and participation in treatment in terms of two types of typologies, victim age (children, peers or adults, mixed) and victim relationship (intrafamilial, extra familial, intra/extra familial). The validity of those subtypes has been supported by some studies (Kemper and Kistner, 2007; Latzman et al., 2011; Joyal et al., 2016; Leroux et al., 2016; Krause et al., 2020; Martijn et al., 2020). This is the first published study aiming to measure the effect of this unique treatment offered in Quebec (Canada) on reducing the recidivism rates of ASOs.

The primary objective of this study was to evaluate the effectiveness of a specialized treatment program designed to reduce recidivism by ASO. A secondary objective was to determine whether completion of treatment and recidivism are associated with victim age or victim relationship.

The study had four hypotheses based on previous empirical research and theoretical considerations:

The Institute Philippe-Pinel de Montréal (Quebec, Canada) offers a specialized outpatient evaluation and treatment program for adolescents (12–18 years old) who have committed sexual offenses. Referrals can be made by Youth Court case workers and by professionals in the health and social services network, including Quebec Youth Centers. Adolescents are referred from across the province, but the largest group comes from the Greater Montreal Area. Adolescents referred to the clinic undergo an initial evaluation, intended to establish their sexual issues, and treatment needs, by a multidisciplinary team of professionals (psychiatrist, psychologist, criminologist). If enrolment in a specialized treatment program is indicated and the adolescent exhibits at least minimal motivation to enter treatment or acknowledgment (acknowledgment of some facts or of the existence of sexual problems), he or she is admitted to a treatment program.

In all cases, the approach is cognitive-behavioral therapy targeting relapse prevention. The treatment program targets three main objectives: (1) fostering a better understanding of the offending process, i.e., the sequence of factors that predispose to, and trigger, sexual aggression; (2) developing control and avoidance strategies that prevent recidivism; and (3) favoring a return to healthy psychosexual development, particularly by addressing factors that are not specific to sexual aggression (e.g., self-control, emotional regulation, relational and social skills, prosocial cognitions, self-esteem, empathy). This group therapy is delivered weekly for ~26 weeks (90 minutes per week) by two therapists of the multidisciplinary team. The therapy groups are closed, and are composed of six to nine adolescents who have no developmental impairment; to the extent possible, the groups are stratified by age (12–15 years and 15–18 years). The program was implemented in the 1990s and has been periodically revised, based on ATSA guidelines and evidence from the scientific literature. Three of the program's founding therapists, including a psychiatrist, still conduct therapy and act as trainers and supervisors for the program. Themes are addressed in the form of theoretical or practical exercises (e.g., family history, significant life events, pre-abuse, abuse and post-abuse period, consequences of sexual assault, letter to the victim, healthy sexual development and intimacy). An unpublished treatment manual is available to therapists.

Progress in therapy is assessed at mid-therapy and at the end, in case management meetings with adolescents, parents and case workers. Adolescents who have not reached the treatment goals, who do not satisfy the basic criteria for group therapy (e.g., developmental impairment, commission of a non-contact sexual offense), or for whom group therapy is contraindicated are referred for individual cognitive-behavioral therapy, the duration of which depends on their needs. Individual therapy is delivered by a psychologist of the team, with the same objectives as the group therapy. In both the group and individual treatment programs, there are regular meetings with parents and case workers in contact with the adolescent, in order to foster teamwork and ensure that the adolescent can develop in an environment that is safe for him and others. The vast majority of adolescents also benefited from psychosocial services; some also benefited from other clinical services, such as rehabilitation and family interventions.

The initial sample was composed of 351 male ASO who were evaluated at the Institute Philippe-Pinel outpatient clinic between 1992 and 2002. The sample has been previously described (Carpentier and Proulx, 2011). Following exclusion of 24 initial participants due to the destruction or unavailability of files, the final sample comprised 327 participants. The mean age of the participants at the time of evaluation was 15.8 years (SD = 1.9). Almost all (95.1%) of the subjects had been born in Canada; 1.5% (n = 5) had been born in Africa, 2.3% (n = 8) had been born in Central America, and information on three participants was missing. At the time of their initial evaluation, 24.1% of the participants were facing a criminal charge under the Young Offenders Act, 49.6% were in care or were being followed under the Youth Protection Act, and 19.9% were followed under both acts¹. Only 5.5% (n = 8) of the participants were not facing charges or warrants. During initial evaluation, it was estimated that 208 adolescents had committed sexual assaults against children only (victim younger than 12 years and aggressor at least 3 years older) and that 88 had sexually assaulted only peers or adults (same age group as aggressor, or an adult, i.e., older than 17 years). The rest of the sample (n = 31) was composed of adolescents who had assaulted victims in both age categories (mixed aggressors). The victim relationship was intrafamilial only (biological sibling) in 30.0% of cases (n = 98), extra familial (e.g., sibling in a blended family, cousin, classmate, stranger) in 57.5% (n = 188), both intrafamilial and extra familial in 11.9% (n = 39, and unknown in 0.6% (n = 2). For the index sexual offense (the most recent offense at the time of evaluation), the mean age of the victim was 9 years (range = a few months to 41 years, SD = 5.42); age was unknown in five cases. The majority of the victims (65.4%) were female. The mean number of known victims per participant was 2.29 (range = 1–16, SD = 1.94), yielding a total of 748 victims for the entire sample.

In order to verify the homogeneity of the three groups, the groups were compared on several variables of interest (Table 1). No significant difference was observed for three of the four continuous variables: age of participants [H(2) = 2.38, p = 0.305], duration of follow-up (H = 5.01, p = 0.082) and age at first sexual assault [H(2) = 4.45, p = 0.108]. However, the mean number of known victims at the time of initial evaluation was slightly higher in the dropped out group than in the other two groups [2.8 vs. 2.1 and 2.05, H(2) = 8.90, p < 0.05]. The groups were comparable for almost all dichotomous variables. The only significant differences observed were for placement [χ²(2) = 13.99, p = 0.001, ES = 0.208] and male victim [χ²(2) = 15.70, p < 0.001, ES = 0.220], and the effect sizes for these variables were moderate. ASO in the dropped out group were significantly more likely than ASO in the other two groups to have been in out-of-home care (placement) at the time of initial evaluation and to have assaulted at least one male. While significant differences were observed for victim age, the effect size (0.14) was below the threshold of clinical significance. No significant difference was observed between groups for victim relationship (intrafamilial, extra familial, intra/extra familial).

The recidivism rates of each of the three ASO groups are presented in Table 2. The mean follow-up period was 7.8 years. The completed treatment group exhibited a sexual recidivism rate similar to the no treatment group, and lower violent and general recidivism rates than the other two groups. However, no significant intergroup differences were found.

As suggested by Worling and Curwen (2000), Kaplan-Meier survival analyses (Log rank) were performed, in order to take into account the period of time elapsed between initial evaluation and recidivism. The three groups (completed, dropped out, no treatment) were compared on sexual, violent, and general recidivism. No significant difference was found between the groups for sexual [χ² (2, N = 323) = 3.80, p = 0.150] and general recidivism [χ² (2, N = 323) = 1.59, p = 0.451], but a significant difference was found for violent recidivism [χ² (2, N = 323) = 6.09, p = 0.048]. The rate of violent recidivism in the treatment group was almost half that in the dropped out group and the no treatment group (16.1 vs. 29.8% and 31.1%). The two comparison groups (dropped out and no treatment) were merged to form a single group (n = 261), which was then compared to the completed treatment group (n = 62). Again, violent recidivism differed significantly in the two groups, with the completed treatment group exhibiting a violent recidivism rate almost half that of the comparison group [16.1 vs. 30.7%, χ² (1, N = 323) = 5.98, p = 0.014]. The rates of sexual and general recidivism were lower in the completed treatment group (8.1 vs. 10.3%, and 38.7 vs. 44.8%), but not significantly so (Figure 1).

To determine whether completion of the specialized treatment reduces the probability of reoffending, Cox regression analyses were performed, with control for placement, number of known victims at the time of initial evaluation, and male victim, all of which exhibited a moderate effect size and an acceptable level of significance (p < 0.05). The variables were entered into two blocks—the control variables in block 1 and the “treatment completed” variable in block 2—using Wald's stepwise descending method (Table 3). The results indicate that completion of the specialized treatment program did not significantly decrease the probability of sexual or general reoffending, but did reduce the probability of violent (including sexual) reoffending by almost half. The final model is significant [χ² (2, N = 323) = 8.883, p = 0.012] and includes the “history of placement” control variable.

Supplemental analyses were performed (Hypothesis 3 and 4) to determine whether victim age and relationship were associated with: (1) the completion of treatment (logistic regressions); and (2) sexual, violent, or general recidivism rates (Cox regressions). Neither of the two logistic regression models was significant: neither victim age (p = 0.163) nor victim relationship (p = 0.751) had any effect on the completion of treatment. However, while victim age did not have a significant effect on sexual (p = 0.767) or general (p = 0.159) recidivism, it did influence the probability of violent recidivism [χ² (2, N = 327) = 6.247, p = 0.044]: ASO who had sexually assaulted peers/adults were almost twice as likely to violently (including sexually) reoffend [Exp (B) = 1.754, p = 0.014]. Victim relationship had no significant effect on any form of recidivism (sexual: p = 0.287; violent: p = 0.421; general: p = 0.724).