During childhood and adolescence, externalising, or disruptive behavioural disorders become a significant risk factor for later juvenile delinquency, adult crime, aggressiveness, violence (1), and internalising disorders (2, 3). Both childhood externalising behaviours and juvenile delinquency are currently considered a serious public health concern (4). Consequently, youth violence prevention and intervention are considered one of our society's most pressing issues today (5–7). Identifying essential therapeutic change mechanisms and mediators of outcome in psychological interventions for externalising behaviours problems is of paramount importance as it would help in the treatment and prevention of externalising problems in adolescence and adulthood (8). Previous reviews on psychotherapy with adolescents (10–19 years old) for the treatment of externalising symptomatology have been published to identify potential mediators (6). However, there is still a lack of knowledge of the involved mechanisms of change.

The externalising symptomatology is a category of psychiatric social functioning disturbances (1) that have been categorised as externalising disorders (ED) (9). They can also be present in other specific disorders such as conduct disorder (CD), oppositional defiant disorder (ODD), attention deficit hyperactivity disorder (ADHD) (10), and substance abuse (11). Essentially, almost all ED are characterised by under-controlled (12), impulsive (13), or/and aggressive (14), or rule-breaking behaviours, which have been labelled as externalising behaviours (13) or externalising symptomatology (15).

Traditionally, Achenbach and colleagues introduced the construct of externalising vs. internalising disorders (ID). ED are oriented toward behavioural maladjustment, which affects social functioning. That is, externalising behaviours involve acting negatively on one's external environment. Conversely, ID are oriented toward explaining maladjustment in the child and adolescent's personal functioning (16, 17). However, ED and ID have a substantial overlap, as they share common aetiologies (18). For example, long-term externalising behaviours can lead to internalising symptoms later in life (19, 20) and vice versa (21) and are highly correlated (22). Despite the comorbidity of ED and ID, distinctive pathways have been described for externalising disorders from ODD via CD to anti-social-personality disorder in adulthood (22).

Some authors have pointed out a lack of consensus on including ADHD and drug abuse in the same category as oppositional defiant disorder and conduct disorder (23). ODD and CD are related more to disruptive behaviours, while ADHD is related to inattention, and alcohol abuse to dopaminergic problems of addiction (17). As stated in the DSM-5 (24), ODD co-occurs with ADHD in approximately half of the cases in the general population in children with a combined presentation (inattention and hyperactivity-impulsivity) and roughly a quarter of children and adolescents with a predominantly inattentive presentation. CD occurs in about one-quarter of children and adolescents with a combined display [e.g., ADHD, see (5)], depending on the age and setting (25). The impulsivity trait that typically manifests in children with ADHD confers considerable risk for the manifestation and development of externalising behaviours (26). Therefore, ADHD is highly comorbid with these disorders and predicts externalising behaviours but is not a disorder caused by the same distal (e.g., parenting style) or proximal (e.g., personality traits) causes (5, 27). Besides, from a neurological point of view, ADHD is considered a neurodevelopmental disorder linked with frontal lobe maturity (28) and substance abuse underlies neurobiochemical mechanisms (29). For the purpose of this review, we have decided to focus only on externalising behaviours and we therefore excluded those articles with a primary diagnosis of ADHD when externalising symptoms were not part of the primary outcomes.

Psychological treatments need to be tailored to the specific needs of individuals with ED, as most psychological interventions have been developed for ID and do not sufficiently address ED (6). Moreover, treatment response and motivation to change may be altered in individuals with ED in comparison to ID (22). Thus, there is a call for a deeper understanding of the mechanisms of change in ED treatment that could inform the development of more effective psychological therapies.

Meta-analyses have concluded that psychological treatments have a more substantial positive effect on externalising symptomatology than punishment-oriented or pedagogical interventions (30). Several psychological therapies have been designed for adolescents' externalising behaviours. The most effective programs include the family and/or peer systems such as multi-systemic (family) therapy, parent training, and multidimensional foster care (24, 26). Third-wave treatments have recently been modified for ED, such as dialectic behavioural treatment (31) and mentalisation-based treatment (32). However, different therapies are based on diverse change theories for addressing externalising symptoms and behaviours among adolescents. Hence, we are interested in reviewing therapies that may have different significant mediators related to externalising outcomes.

The following are the most commonly used psychological therapies for externalising symptoms and behaviours:

All these therapies have highlighted the role of parenting styles as a mediating factor between treatments and outcomes compared to control groups to reduce externalising symptoms in adolescents (27, 35, 36). In summary, the different therapies address parenting styles to treat ED in adolescents. On the one hand, parental warmth, behavioural control, autonomy granting, and democratic parenting styles predict lower ED rates. On the other hand, strict control, authoritarian, permissive, and neglectful parenting styles are associated with more severe externalising problems, with stronger associations observed for stringent control and psychological control (49). However, there is a need to clarify how the theories of change and other mediators in therapies are involved in reducing ED in adolescents. Further research is needed to identify whether more pertinent mediators between treatments and outcomes with a comparator group exist. To our knowledge, there are no systematic reviews on general mediators (beyond parenting styles), both treatment-specific (e.g., psychotherapeutic intervention) and non-treatment specific (e.g., adolescents' capacities for affect regulation), which are involved in psychological treatments to reduce ED symptoms among adolescents. Even though the most common therapies originate from different therapeutic approaches, they overlap in addressing family factors, but they tended to neglect other potential mediators.

To our knowledge, this is the first comprehensive systematic review that evaluated mediation studies in various forms of treatment of ED in adolescents. A total of 14 mediation studies with 3,314 participants were included, which investigated 37 different mediators in nine psychological interventions based on four theoretical foundations (systemic, CBT, educational, and humanistic).

A positive picture was gained regarding the risk of bias and study quality assessment. Although only eight studies met all the criteria for RCT, four additional faced only one problem—the absence of assessors' blindness. This result may suggest that the standards for performing RCTs are well-established and satisfyingly applied in the psychotherapy field. However, standards for mediation studies with respect to Kazdin's criteria (100) were low, which might be related to the fact that most studies were not designed as mechanism of change studies but efficacy studies in which mediation analyses were secondary research questions. Two main problems have been identified in the mediation designs: only one-third of studies included temporal precedence of the mediator, while none of studies performed experimental manipulation of mediator. The low quality of mediation testing questions the role of the proposed mediators in the respective mechanisms of change since causality could not be established and this means that changes in the putative mediators are mainly associated with the outcomes change. However, the temporal chain remains unclear (i.e., what changes first, outcome, or mediator).

With regards to outcome measures, the lack of a consensual criterion for assessing externalising behaviours was demonstrated. Namely, several studies (58, 60, 65, 66, 68) used various items from different self-report instruments to statistically develop an adjusted and reliable CFA to represent externalising behaviours. Others used objective, and hypothetically more robust measures of externalising behaviours such as interviews for testing antisocial behaviours (59, 62), criminal activities and incarcerations (60), and the number of criminal referrals and days locked up (67). The rest applied the externalising self-reports that are well-known and widely used in the literature. Thus, there was possible variance due to the evaluation method in outcomes (self-report vs. objective measures) that may have influenced the absence of significant results and would have required different statistical techniques, such as the multimethod–multitrait statistical approaches (53). Besides, some of the results might be compromised by the absence of clear clinical cutoff points in the instruments used for the assessment of externalising behaviours (50), while this was not the case for measuring depression with the Beck scale [see (13)].

The majority of psychological interventions were based on systemic principles (38). Hence, it appeared that systemic (family) psychological approaches were widely used in psychotherapy for treating externalising behaviours among adolescents (38) and are also focusing on establishing change mechanisms. Nonetheless, our systematic review points out that these interventions focused more on family-based treatment than on a theoretical systemic therapy orientation. Three randomised interventions were blended with CBT approaches (59, 62, 76), and eight used systemic theoretical mechanisms only (56, 58, 60, 61, 63–65, 69). They were presented in different formats: family (56, 58–61, 63–65, 67–69), group (59, 64–66, 69), and individual (59, 64–66, 69), and all systemic approaches were delivered face-to-face. Interestingly, no psychodynamic nor third-wave therapies were identified among these studies.

The review's main findings were related to the mediators that have been selected and proven to be significant. Family variables were found to be the largest group of investigated mediators (22 out of 37). The following groups are composed of a range of different relational (8 out of 37), behavioural (4 out of 37), and emotional (3 out of 37) mediators. No cognitive mediators were investigated. The lack of testing multiple mediators from different categories (e.g., family-related with emotional) was noted (see Table 1). Furthermore, none of the significant or non-significant mediators were treatment-specific and results in all these mediator categories showed both significant and non-significant results. We expect that some results are possibly coincidences (false-positive and false-negative) due to insufficient powering and not following Kazdin's recommendations.

With regards to the type of mediators, findings indicated that treatment approaches were mostly interested in evaluating mediators related to changes in the family system. Hence, family mediators were the largest group. Thus, they were closely related to the systemic therapeutic orientation and its findings (38). The commonalities observed with mediators showed how interventions improved family-based relationships (family cohesion, positive youth-adult relationships, lax caregiver discipline, mother-adolescent conflict, positive parenting, and parent-adolescent communication) and/or how-to-implement positive parenting (parental monitoring, parental discipline, parenting sense of competence, positive discipline, and effective parenting). These mediators were related to the presumed systemic mechanisms of change, which ranged from ineffective or negative parental styles to positive or more effective ones (58, 59, 62, 63, 65). Besides, studies also based on CBT approaches (59, 62, 67) included mechanisms of parental monitoring aimed at two primary purposes: to control/decrease risky peer relations (59–61, 63, 67) or/and to improve parental discipline (58, 59, 62, 64, 65). Both aim to help adolescents consider the consequences of their choices (especially with delinquent peers) before making decisions (1, 36, 57, 61, 88, 89, 95). These were followed by allowing adolescents to bear the consequences of poor decisions without bailing them out (5, 41, 60, 101, 102). Three studies (59, 61, 68) included family cohesion measures, i.e., strength of the emotional bonds between family members' mutual support under distressed environments and situations (103). Three studies demonstrated that increasing family cohesion might decrease ED (60, 63, 69). Therefore, it can be summarised that psychological ED treatments for adolescents often focused on these two mechanisms of change, i.e., increasing family relationships and improving effectiveness of parenting. Indeed, the majority of family-related mediators turned out to be significant in both domains of parenting and parent–adolescent relationship (16 out of 22). This points toward a putative change mechanism in the treatment of ED independent of the treatment approach.

Furthermore, our systematic review identified studies with relational mediators as the second largest group of mediators; however, only three out of eight proved to have significant effects. Significant relational mediators were related to the negative influence from peers adolescents with ED keep: deviant peer association, delinquent peer affiliation, and bad friends. Maintaining such relationships increases the likelihood of disruptive, delinquent or aggressive behaviour in adolescents with externalising symptoms (63, 67, 79). However, two studies investigating the influence of delinquent peers did not find a mediation effect on outcome. Furthermore, two studies reported on non-significant influences of social competences and also the therapeutic alliance.

Concerning behavioural mediators, two educational studies, which were oriented to decrease antisocial behaviours, were also reviewed (66, 68). One of them was a preventive study without a significant mediator (self-regulation) (68), and another had non-significant effects on lowering the level of participants' antisocial behaviour through the development of interpersonal skills and accomplishment motivation (66). Another rather unexpected significant mediator was reducing insomnia symptoms (CBT, individual and online settings), leading to lower internalising and externalising symptomatology (56).

Our review demonstrated a striking lack of studies on emotional mediators in research on treating externalising symptoms (100) as only three studies included mediators relating to emotional changes. Changes in empathy and emotion regulation turned out to be not significant in two studies, although increasing adolescents' emotional abilities is assumed to be a plausible mechanism that has been observed to improve the social (101, 104) and personality functioning (102, 103) of adolescents. The lack of emotional mediators as well as the absence of cognitive mediators may be a reflection of reductionist change theories in externalising disorders or the lack of effective treatments aside from family therapy that target at the system level more than on the adolescent him/herself.

There are some limitations to be considered when it comes to the methodology and findings of this systematic review. Firstly, we used general research terms and did not focus on externalising disorders in the systematic search—although it is expected that this kind of strategy would reveal any adolescent population in the psychotherapy process, more specific search terms would find at least, to some extent, different results. Second, the inclusion of only English literature could lead to some studies being excluded and some cultural biases. Thirdly, in order to obtain the most valid and reliable data, we excluded all the non-RCT studies that may limit or narrow the interpretability of available data. Hence, we found mainly US-based studies—and primarily systemic. Moreover, this review discussed the mediators that have been investigated so far, which turned out to be selective and biassed by the therapeutic school that was investigated. Finally, we aimed to give a narrative synthesis of the mediators. Due to their nature and variety, we were not able to conduct a meta-analysis and thus the theories of change specifically tested within the studies and related to specific therapies were not carried out.

Another issue to consider is the overlapping nature of mediators. In some ways, the types of mediators share characteristics. For example, family cohesion has both familial, relational, behavioural, and emotional connotations (104–106).

Due to a lack of methodological quality in mediation designs, as stated above, results should be interpreted tentatively.

After reviewing the existing studies that followed an RCT design, therapy-related biases and restrictions in the mediator selection process were found. Restrictions may also be related to a narrow clinical change theory of ED. As a result, emotional and cognitive mediators were strikingly neglected. Furthermore, some inconsistences regarding the mediators' significance were identified. Namely, several family and relational mediators were significant in some studies and not in others, which prevented us from developing an explanatory pattern with more strength. Therefore, with these caveats in mind, our conclusions are tentative pending further research on psychotherapeutic and mediator efficacy. In principle, it appeared that both adequate behavioural control of adolescents' peer behaviour and a better positive balance in their relationships with their parents seemed to buffer the effects of externalising behaviours in adolescents with ED.

We tried to derive three groups of recommendations for future research based on the open questions that emerged after the synthesis of findings. Firstly, future studies should extend the existing knowledge by investigating other plausible mediators (i.e., including other emotional mediators, cognitive mediators). Also, the statistical comparisons of explanatory power of different mediators and investigation of their (complex) relationship is needed. Furthermore, all psychotherapy schools still face the challenge to explain how they work since no data on treatment-specific mediators has been found. Moreover, some therapeutic schools, like psychodynamic or third-wave therapies, so far have not been engaged in examining potential mediators of their ED treatment. CBT treatments have not been studied with regard to cognitive mediators either. Furthermore, within systemic and family therapy approaches, it would be interesting to explore further mechanisms (for instance adolescent's emotional or cognitive capacities) that are built during the therapy process and that would be effective in preventing ED after the adolescent develops more autonomy and starts to live an independent and separate life from their family. Moreover, it would be important to examine which mediators are significant in adolescents with severe ED (e.g., those in correctional institutions) and whose families are not available or have no contact or influence on the adolescents' behaviour.

The second group of suggestions relates to research practises and is based on studies' methodological limitations that were recognised in this review. Namely, several design issues could be improved in order to have stricter and more rigorous testing of mediation according to established criteria. Also, a greater consensus on how to assess externalising behaviours that represents the common symptoms of the different mental disorders that belong to the ED would enable more reliable and valid conclusions. Looking into gender as a moderator of mediation effects could be another challenging question addressed in the future.

Finally, additional efforts could be made in order to understand the relationship between internalising and externalising symptoms. Similarly, the use of personality functioning as a construct consisting of self and interpersonal domains instead of externalising vs. internalising could be used to avoid stigmatisation during adolescence.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

JM, ST, MRH, CPM, AS, EH, SR, and SP drafted the first and second version of the manuscript. All co-authors provided a substantial contribution to the conception and design of the work by developing the research questions, the search string, and carrying out the stage 1 screening (SC-C, ST, EH, AS, SP, JV, AA, RB, DD, YI, JM, FM, CPM, MRH, JR, SS, TSP, RU, CS, and PM-P). The current manuscript was corrected and finally approved by the other authors (JV, AA, RB, DD, YI, SC-C, PM-P, FM, CPM, MRH, JR, SS, TSP, RU, and CS). RU coordinates the overall COST initiative. All authors agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

This publication is based upon work from COST Action: CA16102, European Network on Individualised Psychotherapy Treatment of Young People with Mental disorders, supported by COST (European Cooperation in Science and Technology). https://www.cost.eu/actions/CA16102/. The publication fee was paid by Heidelberg University (Germany) and Universidad de Càdiz (Spain).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.