The Effectiveness of Psychological Therapies for Anxiety Disorders in Adolescents: A Meta-Analysis

Anxiety disorders are common in adolescence (Costello et al., 2011; Polanczyk et al., 2015; Vizard et al., 2018), with around 8% of 11- to 16-year olds (Vizard et al., 2018) meeting diagnostic criteria for an anxiety disorder. Anxiety disorders during this developmental stage are particularly problematic due to their negative impact on psychosocial functioning including education, social interaction, school refusal and school dropout (Van Ameringen et al., 2003). They are likely to persist into adulthood and are associated with a risk of developing other mental health problems (Pine et al., 1998), such as depression, alcohol dependence and suicidal behaviour (Kendall & Peterman, 2015; Kendall et al., 2004; Woodward & Fergusson, 2001), as well as negative long-term impairments in social and occupational functioning (Woodward & Fergusson, 2001). Consequently, the high prevalence and substantial impact of anxiety disorders during adolescence highlight the vital need for effective treatments.

There is a continuing growth in psychological treatments for child and adolescent anxiety disorders, including both cognitive behaviour therapy (CBT) and non-CBT-based approaches (e.g. mindfulness and acceptance-based therapies Dunning et al., 2019; Vøllestad et al., 2012), delivered in a range of formats such as individual, group and computer based (cCBT), in clinic and school settings, with varying degrees of parent/carer involvement. CBT is the most extensively evaluated treatment for anxiety disorders among children and adolescents, with generally good outcomes, across different formats of delivery. When outcomes have been examined across broad age ranges (2–19 years of age), 49.4% of children and adolescents who have had CBT (not including cCBT) have been found to be in remission from their primary anxiety disorder at the end of treatment (James et al., 2020).

It is unclear, however, to what extent these findings can be generalised to adolescents with anxiety disorders, who have typically been underrepresented in treatment outcome studies. Reynolds et al. (2012) examined the six studies in their review that recruited only adolescents (aged 14–19 years) with either elevated anxiety symptoms or an anxiety disorder diagnosis and found the reduction in symptoms post-treatment to be in the very large range (d =  − 1.38), although with very wide confidence intervals (95% CI 2.65, − 0.11). Although specific outcomes for the remission of anxiety disorders in adolescents are not reported, in two meta-analyses (Bennett et al., 2013; James et al., 2020) that have examined age as a moderator of outcome, they found no significant differences between studies comparing remission of anxiety disorders for adolescents and younger age groups. While James et al. (2020) found larger treatment effects for CBT (vs. no waitlist/ no treatment) among adolescents aged 12 years or more compared to children 12 years or less, they highlighted the substantial heterogeneity in findings. The majority of the studies in their review used mixed child and adolescent samples, with less than 20% of included studies focussing specifically on adolescents.

Examining adolescents in their own right is important, as adolescence is a unique stage of development and factors associated with this developmental period may influence the effectiveness of treatment for anxiety disorders. Findings from both animal and human research suggest that during adolescence, fear expression and extinction are temporarily impaired (Ganella & Kim, 2014; Waters et al., 2017) making it more difficult to retain new, non-fearful, inhibitory information. In addition, adolescents may have severe symptoms, comorbid depression and difficulty attending school (Hudson et al., 2002; Waite & Creswell, 2014). Taken together, these factors may influence the overall effectiveness of treatment and questions about what works for whom. Notably, to date, no studies have gone beyond examining age as a moderator to investigate what moderates outcomes for adolescents.

Factors that have been found to be associated with better treatment outcomes across broad age ranges include having a greater number of treatment sessions (i.e. more than 11 sessions), (Reynolds et al., 2012), treatments targeting a specific disorder rather than being transdiagnostic (Reynolds et al., 2012), clinical treatment-seeking populations rather than those recruited from the general community (Weisz et al., 2015), being from White ethnic backgrounds compared to those from other ethnic groups (Weisz et al., 2017) and comparisons with wait list controls, rather than active controls or treatment as usual (James et al., 2015, 2020; Reynolds et al., 2012). In contrast, poorer outcomes have been found for children and adolescents with social anxiety disorder than those with other anxiety disorders (Hudson et al., 2015). There have been mixed findings for delivery format; while Zhou et al., (2019) concluded that group formats (of CBT) are particularly effective compared to passive control groups as well as to most other psychotherapies, James et al. (2020) concluded that the evidence does not provide clear and consistent support for group CBT having an advantage over other delivery formats and highlighted that studies that differed in terms of treatment delivery format also differed on other key characteristics. Other factors, such as gender and parental involvement, have not significantly moderated treatment outcomes in studies to date (James et al., 2015; Manassis et al., 2014; Reynolds et al., 2012). Although children and adolescents from socio-economically disadvantaged backgrounds are significantly more likely to develop mental health problems than those from less disadvantaged backgrounds (Reiss, 2013; Reiss et al., 2019), whether this disadvantage specifically moderates treatment outcomes for anxiety disorders has not yet been examined. The extent to which these factors moderate outcomes specifically among adolescents has not been evaluated, however, there are clear developmental reasons that may lead to differences. For example, adolescent patterns of fear expression and extinction (Ganella & Kim, 2014; Waters et al., 2017) may lead to different effects based on the amount of treatment hours/sessions provided. Other factors such as the effectiveness of different modes of treatment delivery (e.g. individual, group or online) may also be unique in adolescence, due to their desire for autonomy (Zimmer-Gembeck & Collins, 2008), high levels of self-consciousness (Sebastian et al., 2008) and heightened sensitivity to others’ perceptions of themselves (Kilford et al., 2016). Understanding potential moderators of treatment within this unique developmental period is vital in understanding who does and does not benefit from psychological treatments for anxiety disorders to then develop more effective treatments in the future.

This meta-analysis aims to address the current gap in the literature by examining treatment outcomes and moderators of treatment outcome for adolescents with an anxiety disorder. It specifically focuses on RCTs of any psychological treatment (i.e. not just CBT-based approaches) using any delivery format (including cCBT), for anxiety disorders among adolescents. We defined the adolescent age range as 11–18 years (inclusive) based on 11 being the average age at which external indicators of puberty become apparent (American Psychological Association, 2002) and 18 being both the legal age of adulthood and the age at which child and adolescent mental health services end in many countries. In addition, typically 11–18 is the age range when young people are in secondary education, therefore adolescents in this age range have broadly similar educational and social demands and roles (Perry et al., 1993).

We aimed to answer the following research questions:

How effective are psychological therapies in (i) reducing anxiety disorder symptoms and (ii) achieving remission from the primary anxiety disorder, when compared with controls, at post-treatment and follow-up time points? (iii) Is the effectiveness of psychological therapies for treating anxiety disorders in adolescents moderated by the following treatment/demographic variables: CBT (including cCBT) vs non-CBT intervention, mode of treatment delivery (individual, group, mixed group and individual, cCBT), age, number of treatment hours, disorder-specific vs. generic anxiety treatment, active vs. passive control group, clinic vs. community sample, type of primary anxiety disorder, ethnicity (white or other ethnicity), gender (percentage female), parental involvement (involvement vs no involvement) and socio-economic status, at post-treatment and follow-up time points?

We also examined study quality as a moderator of treatment outcome. Very little is known about adverse events in RCTs of psychological treatments due to underreporting (Duggan et al., 2014). In evaluating the effectiveness of treatments, it is crucial to understand any potential harms as well as the benefits of therapy in terms of treatment outcomes. We therefore also examined the presence of adverse events reported within the identified studies. Finally, in addition to the aims specified in our protocol, we also examined to what extent interventions were developed or adapted to be developmentally sensitive to adolescents.

The review protocol was pre-specified and registered on the International Prospective Register of Systematic Reviews (PROSPERO; protocol number: CRD42018091744 https://​www.​crd.​york.​ac.​uk/​prospero/​). PRISMA guidelines were followed throughout (Moher et al., 2009) (Fig. 1).

Database searches yielded a total of 5283 records. After removing duplicates and screening, a final total of 17 studies were assessed as eligible for inclusion. One study (O'Brien et al., 2007) could not be included in the meta-analysis because insufficient data (including the number of participants and means and standard deviations for intervention and control conditions post-treatment) were reported in the paper and we were unable to obtain the necessary data from the authors. Therefore, 16 studies were included in the analysis of continuous symptom severity outcomes (Baer & Garland, 2005; Ebrahiminejad et al., 2016; Ginsburg & Drake, 2002; Hayward et al., 2000; Herbert et al., 2009; Ingul, 2014; Masia-Warner et al., 2007; Masia-Warner et al., 2005; Masia Warner et al., 2016; Olivares et al., 2002; Pincus et al., 2010; Spence et al., 2011; Stjerneklar et al., 2019; Swain, 2015; Waite et al., 2019; Wuthrich et al., 2012), with a total sample of 766 adolescents. Nine studies were included in the meta-analysis of dichotomous remission data (Hayward et al., 2000; Herbert et al., 2009; Masia Warner et al., 2016; Masia-Warner et al., 2005, 2007; Spence et al., 2011; Stjerneklar et al., 2019; Waite et al., 2019; Wuthrich et al., 2012) with a total sample of 563 adolescents. Eight studies could not be included as they did not report remission data for the primary anxiety disorder (Baer & Garland, 2005; Ebrahiminejad et al., 2016; Ginsburg & Drake, 2002; Ingul, 2014; O'Brien et al., 2007; Olivares et al., 2002; Pincus et al., 2010; Swain, 2015).

Further information about the characteristics of included studies is provided in Table 1. Over half the studies (k = 9) had samples consisting of adolescents with a primary diagnosis of SAD. Three quarters (k = 12) recruited participants from the community (e.g. schools or advertisements), the remaining four were clinic samples. Fifteen studies looked at CBT (including one study of mindfulness-based cognitive therapy), with only one non-CBT study of acceptance and commitment therapy (ACT). Nine delivered CBT in a group, four delivered individual sessions, two combined group and individual sessions, while four studies examined cCBT. The number of treatment hours delivered ranged from 4 to 29 h. Half the studies (k = 8) used and reported ITT analysis for diagnostic outcomes (Ingul, 2014; Masia-Warner et al., 2007; Pincus et al., 2010; Spence et al., 2011; Stjerneklar et al., 2019; Swain, 2015; Waite et al., 2019; Wuthrich et al., 2012).

Examining adolescents in their own right is important, as adolescence is a unique stage of development and factors associated with this developmental period may influence the effectiveness of treatment for anxiety disorders. We identified sixteen RCTs that examined the effectiveness of psychological treatments for anxiety disorders specifically in the adolescent age range. For adolescents who had completed a psychological treatment, compared to controls, there was a moderate and significant effect on symptom severity post-treatment. Just over half the studies examined remission from the primary anxiety disorder and both those randomised to, and those who completed, a psychological treatment were significantly more likely than controls to be in remission from their primary anxiety disorder post-treatment, with large effects. Despite this apparent positive finding, in the treatment group, only 36% of adolescents no longer met diagnostic criteria for their primary anxiety disorder at the end of treatment. We were unable to identify any treatment/demographic moderators that were significantly associated with outcomes.

There are a number of reasons to be cautious, however, when drawing conclusions from the available studies. Of concern, over half the studies were rated as ‘poor’ quality. Although study quality was not a significant moderator of symptom severity outcomes, our subgroup analyses revealed poor study quality was associated with larger treatment effects than those of fair quality, indicating that biases may have led to overestimated treatment effects. Forest plots showed high heterogeneity between studies and we were unable to statistically identify the source of this. Furthermore, there are limits to the extent that findings can be generalised given that three quarters of the studies recruited participants from the community (e.g. through schools) and half focussed specifically on social anxiety disorder. Conclusions cannot be drawn about the relative efficacy of different treatment types, as CBT (delivered in a variety of formats) was the treatment approach in all but one study.

Nevertheless, the reason why only a third of adolescents are free of their primary anxiety disorder at the end of treatment warrants urgent and extensive evaluation. This may reflect severe anxiety symptoms/disorders, high levels of social anxiety disorder/symptoms, comorbid depression, and potentially chronic and entrenched problems (Essau et al., 2000; Kendall & Peterman, 2015; Pine et al., 1998; Waite, 2014; Woodward & Fergusson, 2001), that do not respond sufficiently to current treatments. There is some evidence that adolescents have difficulty retaining new non-fearful information during this developmental stage (Waters et al., 2017). This could potentially account for adolescents’ poor responses to predominantly exposure-based anxiety treatments in the ways that they are currently delivered. Clearly, treatment optimisation must be underpinned by a clear developmental understanding of the mechanisms that maintain anxiety disorders in adolescents. However, the role of the wider context that adolescents are living in, including acute social demands and academic pressures (Blakemore, 2008, 2018), are also likely to be important. It was notable that less than half the studies reported what aspects of the treatment were designed to be developmentally sensitive to adolescents. Where adaptations were made, this consisted of making language ‘age-appropriate’, giving adolescent-specific examples, or conducting exposure/social skills training within activities typical for adolescents. It would be helpful for future studies to explicitly report how interventions have been developed or adapted to consider specific developmental needs.

Notably, none of the treatment and demographic variables previously shown to moderate the effectiveness of treatment when examined among children and adolescents, i.e. group delivery format (Zhou et al., 2019); greater number of treatment hours, disorder-specific treatment, type of control (Reynolds et al., 2012); and ethnicity (Weisz et al., 2017) moderated treatment effects specifically in adolescents. Given that adolescents often have severe symptoms, it was of interest that the number of treatment hours did not significantly moderate outcomes. However, studies that differed in treatment length differed on other key characteristics, making it difficult to draw meaningful conclusions. For example, all five studies where the treatment was ≥18 h were with adolescents with social anxiety disorder, which is typically associated with poorer outcomes (e.g. Hudson et al., 2015). Although disorder type was not a significant moderator, we were unable to examine associations with specific anxiety disorders, as all but one study focussed on social anxiety disorder or mixed anxiety disorders including social anxiety disorder. Mode of delivery was also not found to significantly moderate outcomes. Notably, two studies in this meta-analysis compared group and individual CBT directly, and found no significant differences in outcomes between delivery formats (Herbert et al., 2009; Ingul, 2014). However, both studies involved the treatment of adolescents with social anxiety disorder from the community (e.g. through schools) identified through screening and so it is possible that the young people in these studies were less severe than those referred to clinical services and potentially more responsive to working in a group format. Clearly, there is a great deal more work to be done to understand what works for whom, to then develop more effective treatments.

Unfortunately, we were unable to draw conclusions about potential adverse effects of treatment as only two studies reported adverse events. Clinical trials of psychological interventions have been identified as insufficiently reporting harm arising from treatment, as unlike with drug trials, this is not mandatory (Duggan et al., 2014). In a review of National Institute for Health Research (NIHR) funded trials, none of the psychological intervention studies reported adverse events in their final reports, and where adverse events were mentioned (e.g. within trial protocols), reporting guidelines for drug trials were used, which may not be suitable for psychological treatments (Duggan et al., 2014). To date, the focus of research examining psychological interventions has been on the benefits of therapy, but in future must also include the potential harm it might cause (e.g. worsening of symptoms, self-harm, suicide).

The strengths of this review include its specific focus on studies of the adolescent age range, examination of developmental adaptations used in treatments for adolescents, and examination of both diagnostic and symptom severity outcomes and potential moderators of symptom severity. For diagnostic outcomes, we were able to analyse ITT and completer data separately, allowing us to conclude that treatment effects were consistent across ITT and completer analyses for diagnostic outcomes. Nevertheless, our definition of the adolescent age range is a limitation that needs consideration. While we defined the adolescent age range as 11–18 years for the reasons outlined earlier, adolescence is an arbitrary definition and can be defined in multiple ways depending on the theoretical framework adopted (e.g. biological or psychosocial) (Curtis, 2015), anywhere between 9 and 26 years (American Psychological Association, 2002), with this upper end of the age span reflecting the neural development that continues beyond the age of 18 (Paus et al., 2008; Pfefferbaum et al., 1994). Older adolescents may have more in common with young adults than younger adolescents in terms of neurological development (Waters et al., 2017), and in the future, it will be important to consider the effectiveness of treatment for older adolescents and young adults, and at what stage adult-focussed treatment approaches become appropriate.

Results also need to be considered in light of several limitations of the included studies. The overall quality of studies in this review was poor. There were high levels of heterogeneity across study characteristics, outcome measures and reported outcomes (e.g. diagnostic remission status) and follow-up time points (where included). As pre-specified in our protocol, we only included studies that reported specifically on outcomes for adolescents aged 11–18 years, in order to examine the effectiveness of treatment and potential moderators of outcome during this unique stage of development. This also allowed us to examine to what extent interventions were developed or adapted to be developmentally sensitive to adolescents. Nevertheless, as a result of this approach, we are unable to draw direct conclusions about how the findings differ to those of children or adults, and therefore to what extent they are specific to adolescents. By selecting studies that only included adolescents, a large number of studies involving children and adolescents across broad age range were not included. Had we obtained data from these studies for participants within the 11–18-year age range, this is likely to have substantially increased the number of studies in the review and potentially made for a sample more representative of the wider literature, e.g. from a clinically referred population. Given the issues we have raised in this paper, where possible, we would encourage study authors to report outcomes separately for adolescent participants and provide open access to research data. Although we examined publication bias and found it unlikely to have had an impact on results, inclusion of only published works is a limitation, and we suggest future reviews include non-published works to address this. We analysed remission from primary anxiety diagnosis because this is the most commonly reported primary outcome measure in studies, however, it is likely the number of adolescents in remission from all anxiety diagnoses would be lower than the results of this meta-analysis show (Wuthrich et al., 2012) and Creswell et al. (2021) recommend that all anxiety disorders are assessed post-treatment and at follow-up in research trials. Very few studies were with clinically referred populations or active control groups. Furthermore, while 12 studies recruited participants from real world settings (e.g. schools), it remains unclear how generalisable the results of the studies are to adolescents with more severe clinical presentations of anxiety, who are seeking treatment in day or inpatient clinic settings. The majority of studies used passive or waitlist controls, potentially leading to an inflation of treatment effects, and there were insufficient studies to be able to conduct moderator analyses for diagnostic outcomes. We recommend that future studies fully report demographic factors (including socio-economic status and ethnicity) and participants’ clinical characteristics using consistent measures between studies to report baseline and treatment outcomes. In particular, we encourage the consistent use of assessment tools, outcome measures and reporting standards as set out by a recent international consensus statement on reporting treatment trials of child and adolescent anxiety disorders (Creswell et al., 2021). It is imperative that RCTs meet high methodological standards, and we recommend the use of active control groups, reporting of adverse events and reporting outcomes at follow-up to allow more rigorous examination of the effects of psychological interventions and potential moderating factors.

Although there is evidence of efficacy of psychological therapies (predominantly CBT) targeting anxiety disorders in adolescents compared to (largely waitlist) controls, they have only a moderate effect on symptom severity and only just over a third of adolescents are in diagnostic remission after receiving treatment. Within the studies, we were unable to identify any moderators that influenced treatment outcome. There is a clear need to develop more effective treatments, that reflect adolescents’ specific developmental needs and that are evaluated through high-quality RCTs incorporating active controls and follow-up assessments to address the high prevalence, and substantial negative impact of adolescent anxiety disorders.