Meta-analysis on cognitive behavioral treatment and behavioral intervention technologies for anxious youth: more than a BIT effective

Anxiety disorders in youth are linked to impaired social and academic functioning; however, only 20% of individuals with these disorders receive treatment. The current stigma of seeing a psychologist or the accessibility of resources may represent barriers to treatment. This study aimed to review and compare the effects of Faceto-Face Cognitive Behavioral Therapy (F2FCBT) and Behavioral Intervention Technologies (BITs) for youth anxiety disorders. BITs are electronic interventions (e.g. mobile phones, internet, virtual reality) that aim to help the user improve mental and physical health. Little research comparing the effects of the two types of modalities has been conducted previously. Methods: Systematic review methods and meta-analysis techniques were used to analyze thirty-four randomized controlled trials that included F2FCBT and/or BITs therapy. Results: Both types of interventions were effective for the reduction of childhood anxiety, with the overall effect for F2FCBT indicating less post-test anxiety (F2FCBT g = -.84 and BITs g = -.40). A subgroup analysis revealed that there were significant effect sizes for F2FCBT and for the combination of the two modalities, but not for BITs alone. Conclusion: Although F2FCBT has shown a larger effect size, BITs with guided therapist help can be an effective modality for delivering treatment for youth with anxiety disorders. Correspondence to: Bunge EL, Department of Clinical Psychology, Faculty of Palo Alto University, 1791 Arastradero Road, Palo Alto, CA 94304, USA, Tel/ Fax: 800-818-6136, 650-433-3888; E-mail: ebunge@paloaltou.edu


Introduction
Approximately three percent of children and adolescents suffer from an anxiety disorder before adulthood [1]. Anxiety lifetime prevalence rates span from 5.9% to 25.1%, often leading to poor outcomes in social and academic functioning, increased risk of serious mental illness and substance use disorder [1][2][3]. Although anxiety disorders cost more than one-third of the United States mental health bill (42 billion dollars), only 20% of youth with an anxiety disorder receive treatment [4]. The current stigma of seeing a psychologist and the accessibility of resources represent barriers to treatment. Early, effective, and engaging delivery of treatment for youth anxiety disorders are necessary due to the associated level of impairment. Currently, the leading psychosocial method for delivering treatment for youth anxiety is face-to-face (F2F) psychotherapy. Cognitive Behavioral Therapy (CBT) is the F2F treatment with the strongest empirical support [5][6][7].
Behavioral Intervention Technologies (BITs) have been developed as an alternative method of delivering evidence based treatments for anxious youth. BITs are electronic interventions (e.g. mobile phones, internet, virtual reality) that aim to help the user improve mental and physical health [8]. BITs have been shown to increase the reach of mental health resources, thus representing a potential benefit [9]. Additionally, there is an increase of applications being developed surrounding health (e.g. there are over 160,000 new health applications) [10], but most of these applications have not been tested, therefore there is a need to know the impact and efficacy of such applications.
Electronic interventions such as Cool Teens, BRAVE-ONLINE, and Camp Cope-a-Lot are just a few of the examples of evidence-based BITs specifically designed to reduce anxiety symptoms in children and adolescents [11][12][13].
In the last 10 years, several meta-analyses and reviews have focused on the utilization of BITs [2,3,7,[14][15][16][17][18][19][20]. Most of the metaanalyses addressed the effectiveness of BITs for comorbid anxiety and depression. Ebert and colleagues [14] concluded that computerized CBT (cCBT) is an effective treatment for both disorders and an attractive alternative for younger generations. Similar results were found when using computerized therapies (e.g. self-help interventions) for anxiety and depression in youth [3]. Twenty-seven studies were analyzed, concluding that there were small positive effects for symptoms of anxiety and depression in young people but uncertainty around the effectiveness of cCBT in children. Likewise, Richardson,Stallard,and Velleman [16] found that cCBT yielded reductions in behaviors and cognitions related to anxiety and depression with moderate to high user satisfaction. Ye and colleagues [19] found that Internet-based interventions increased remission rates and decreased symptom severity for anxiety but did not affect depression symptoms.
Although there have been several meta-analyses reviewing BITs interventions for comorbid disorders, only one completed a systematic review and meta-analysis looking at Internet-delivered CBT for childhood anxiety [17]. The review included seven studies and used interventions that were either fully delivered through technology (computer assisted program plus an online therapist) or technological interventions that included an in person contact (computer assisted program plus an in person facilitator). Though the results suggested that both therapies were effective, the researchers acknowledged that the diverse range of samples, heterogeneity in study designs, and inclusion of only seven studies were significant limitations. Additionally, the meta-analysis did not compare the effectiveness of BITs directly to any form of F2F treatment [17].
The BITs meta-analyses described above have several limitations. The majority used participants with co-morbid disorders (e.g. anxiety and depression) thereby limiting the known treatment effect on anxiety. Although anxiety and depression are highly comorbid, anxiety onset tends to be earlier developmentally in children and adolescents than depression. Also, there is a population of youth with solely anxiety and not depression, thus the efficacy of anxiety BITs treatments must be researched [21]. Many of the meta-analyses above mentioned treatment for youth and included young adults over the age of 17, making the results less representative of the children and adolescent population. Children and adolescents have varying cognitive skills and developmental differences exist that could change the outcome of treatment. Thus, separating the children and adolescent population from the young adult population and researching the differences in outcome for treatment using BITs modalities is imperative [22]. Additionally, the reviewed meta-analyses compared identical modalities (e.g. F2F vs. F2F; BITs vs. BITs) without cross comparison to determine the most effective treatment systems for youth with anxiety. The meta-analysis done by Rooksby et al. [17] focused only on childhood anxiety, not adolescent anxiety, and did not compare BITs vs. F2F treatment. Lastly, previous meta-analyses did not compare inactive control groups to treatments, leading to data that does not identify the total effect of BITs treatment. This meta-analysis aims to address and expand on these limitations by comparing F2FCBT to BITs and BITs + F2FCBT interventions for anxious youth across age groups.

Selection process of articles
The search was limited to publications in English that evaluated the efficacy and effectiveness of F2FCBT and/or BITs treatment for youth anxiety. To identify relevant articles, synonyms of the term "children and adolescents" crossed with "anxiety" were searched. This was followed by the search term "cognitive or behavioral or psychotherapy or treatment." The age range of participants within the studies was limited to birth through 17 years. Only peer-reviewed articles published between 1990 and 2015 were used. This search yielded an initial pool of 2824 articles. Duplicate articles were automatically removed by the search program, and titles and abstracts were screened for relevance  by the authors, resulting in a pool of 226 articles (See Figure 1). The articles were each individually read and several were removed, as they did not meet the inclusion/exclusion criteria. Authors of articles with incomplete data sets were contacted to obtain means and standard deviations necessary for effect size calculations. A final set of 34 articles was included in this meta-analysis. These articles were divided into two groups: 24 F2FCBT publications and 10 BITs publications.

Inclusion/exclusion criteria
The search process for this meta-analysis was modeled after the one used by Richardson et al. [16]. Individuals who were 17 and younger and diagnosed with an anxiety disorder were included in the analysis. Studies that included participants who had a learning disability or a neurological disorder were excluded. Studies consisting of participants with comorbid disorders such as depression or attention-deficit hyperactivity disorder were excluded to isolate the effect on anxiety alone. In order to assess the direct effects of F2FCBT and BITs interventions, studies including medication were excluded, as this was an expected confounding variable. Studies that included a waitlist control condition (WLC) or a non-evidence based control condition (NEB) were included, in order to reduce potential confounding variables relating to treatment type. NEBs are defined as conditions currently not supported by the literature (e.g. Education Support Therapy, Computer-assisted Education/Support/Attention, or Familybased Education/Support/Attention). Finally, only articles that assessed the outcome of anxiety following CBT and/or BITs were included; pilot studies, case studies, and prevention research were excluded.
An article was considered BITs research, when the target of the intervention received active treatment alone using BITs. It was considered F2FCBT if the target client received active treatment from another person, either therapist or parent. If an intervention was targeted at child behavior but acted via the parents of that child, it was included as a F2FCBT study. Finally, a study was considered a combined treatment when the target child or adolescent at various times throughout the intervention received treatment through BITs or from another person.

Results
The studies included in this meta-analysis that are considered F2FCBT, used CBT in varying modalities including manuals such as "Coping Koala" [19] and "Coping Cat" [20]. Treatments were given in several formats including individual, group, and family. Treatment length ranged from 3 to 18 weekly sessions that were 60 minutes to 2 hours long. Included BITs studies used CBT in varying modalities such as computer programs and Internet video games. Treatments were given individually and ranged from single 3 hours sessions to 12 weekly hour sessions (refer to Table 1 for more information on the components of each treatment).

Methodological quality
Methodological quality was calculated using a 19-item authordeveloped measure that was adapted from relevant items from the Quality of Cohort Studies (Q-Coh) [21] and a previous meta-analysis with a researcher-created quality rating system [22]. Adaptation of these previous established measures helped ensure consistency in the items with relevancy to the studies included in the analysis. The final quality rating form included categories of study design (2 points), comparability of groups (2 points), treatment quality (3 points), measurement (8 points), attrition (3 points), and analyses (3 points).
Two raters (MJ & BD) used a blind rating system in order to calculate an overall methodological quality score for each study. A two-way mixed Intraclass Correlation Coefficient (ICC) was calculated using a twoway mixed, absolute design [23], which indicated consistent inter-rater reliability (ICC= .919). The overall mean of study quality was 15.61 with the lowest study receiving a score of 10 and the highest a score of 18. Further review revealed themes in methodological soundness, including agreement on 33 studies that did not include alternate forms on retest, 19 that did not include intention-to-treat, and 13 with treatments that were not manualized, as the highest categories of lost points. Studies had many different patterns of scores across domains, indicating the need for a standardization of F2F and BITs randomized, controlled designs in future research. However, the mean score of 15.61 indicated that most studies included in the final analysis had sound theoretical designs.

Data synthesis
The final analysis was conducted using REVMAN [24] and an inverse-variance weighted effect size model [25]. Effect sizes were calculated using post-test scores for the most relevant measure of childhood anxiety. A hierarchy of outcome measurements was created based on the most robust and reliable measures related to the construct in order to calculate a single effect size for each study. Pooled effect sizes were calculated for 11 studies that included more than one intervention within a subgroup category (e.g. family and individual F2FCBT). The overall analysis included 34 studies and a total of 1889 participants.

Heterogeneity and risk of bias
A random effects model was used due to significant heterogeneity in the overall model (T 2 =0.20, X 2 (33) =104.84, p < .05, I 2 = 69%). Standard mean differences (Hedges' adjusted g) were calculated as unbiased estimates of effect size in order to account for small sample sizes [26]. Subgroup analyses were conducted in order to identify specific areas of heterogeneity based on treatment type, control comparison, and age. Inconsistency was measured using the I 2 statistic, which ranged from 56-86%, indicating substantial to considerable heterogeneity across all groups [27]. It is also possible that the heterogeneity was present due to the variety of measures used in the assessment of childhood anxiety (i.e. wideband vs. specific, such as ADIS-CSR vs. FSSCR), however studies were not separated by type of measure. Subgroup differences are described below and in Table 2.
A funnel plot was generated in order to determine evidence of bias for F2FCBT and BITs. An overall analysis of risk of bias was tested using Egger's regression test for asymmetry [28] through the Meta-Essentials program for independent groups [29]. The test indicated no bias for small study sizes based on standard error estimates (t = 1.61, p = .118). Interpretation of the funnel plot indicated an overall acceptable distribution. A visual representation of the funnel plot is included in Figure 2.

Primary comparison
Face-to-face CBT was compared to BITs using two distinct models. The first model included 24 F2FCBT studies of single or pooled interventions that did not include any technological intervention. The second group (BITs Total) included 10 studies that had any degree of technological intervention [5,13,[30][31][32][33][34][35][36][37]. Both types of intervention were effective for the reduction of childhood anxiety. Subgroup differences in this category were significant (χ 2 (1) =5.78, p = 0.02), with the overall effect for F2FCBT indicating less post-test anxiety (g = Four weeks of child-only psychoeducation, four weeks of parent-only psychoeducation, four weeks of parent-and-child sessions focused on exposures and relapse prevention, and four weeks of parent-only sessions focused on conducting exposures at home.

RCMAS
[53] 6-10 40 FGCBT WLC 53 12 Treatment was based on the Coping Cat CBT program. Ten weekly sessions and two booster sessions at 1 and 3 month follow-up.

16
Children and parents met in separate groups that used natural group processes to discuss anxiety, selfevaluation, self-rewards, and facilitating exposure training at home. Participants conducted vicarious exposure by guiding computer avatars towards virtual spider pictures, plastic spiders, dead spiders, and live spiders. Participants used the computer for three sessions.

17
Used one-session treatment manualized by Ost & Ollendick. Treatment consisted of a forty-five minute session that consisted of rapport-building and a brief functional analysis of the child's phobia and one three hour sessions that included graduated exposures targeted towards the subject's phobia. A second analysis separated the groups into F2FCBT (n = 24), BITs only, (n = 5), and BITs and F2FCBT combined (n = 5). The differences between these groups reached marginal significance (χ 2 (2)=5.21, p = 0.07). However, interpretation of these results should be cautious due to few studies in the BITs only and combined BITs and F2FCBT, with moderate heterogeneity still persisted with subgroups using this model ( Table 2).

Secondary comparisons
Subgroup analyses were conducted for the control comparison groups as well as age differences following previous meta-analyses [14]. The control comparison groups were established to understand sources of heterogeneity based on NEB control or WLC comparisons across treatment type (F2FCBT or BITs total). The final model was also marginally significant (χ 2 (3) = 6.58, p = 0.09), however the comparison did provide information about sources of heterogeneity. Inconsistency was measured using I 2 , which can be visualized as the amount of overlap in the confidence intervals of study effects, and understood as the degree of variance in point estimates that is attributed to true heterogeneity [27]. The BITs with WLC group had the least amount of heterogeneity attributed to random error rather than differences between study groups (I 2 =48%) while the F2FCBT group with a NEB control had the most amount of error-related heterogeneity (I 2 =81%).

Discussion
Due to the high prevalence rates of anxiety and the effectiveness of CBT in youth, it is imperative that researchers develop novel interventions that could make way for the future of psychology. Technology is becoming more associated with psychotherapy and mental health with over 160,000 new applications for treatment existing, but little research avilable to test the efficacy and impact of such applications [10]. This meta-analysis analyzed different CBT modalities to further understand if new developments such as BITs are comparable to F2F psychotherapy for youth anxiety. Previous metaanalyses compared various BITs and F2F treatments separately, but none were focused on direct comparisons of these modalities. Also, previous meta-analyses included a small amount of BITs studies, whereas this meta-analysis included ten studies. Because CBT represents the most empirically supported approach for youth anxiety, this meta-analysis compared the efficacy of F2FCBT and BITs treatments for youth with anxiety disorders. Although anxiety and depression are often comorbid disorders, anxiety tends to be more prevalent in children and precedes the onset of depression [21]. Thus, assessing the impact of interventions only on anxiety may yield a better understanding of effectiveness of treatment for youth with anxiety disorders and potentially prevent the later onset of depression. This meta-analysis focused solely on the children and adolescent population as previous analyses have included young adults, thus making the results not completely representative of youth and not taking into consideration the differing cognitive skills between the children and adolescent population and young adults [22]. The results showed that both types of modalities were effective for the reduction of childhood anxiety, with the overall effect for F2FCBT indicating less post-test anxiety than BITs (F2FCBT g = -.84 and BITs g = -.40). Effect sizes of BITs for anxious youth were slightly lower than those found in a previous meta-analysis on computerized CBT for depression and anxiety in youth (g = 0.68) (14) and adults (g = 0.88) (38). In order to develop a better understanding of the unique contribution of BITs, we compared the studies using solely BITs with those combining BITs with some F2F component. A subgroup analysis revealed that there were significant effect sizes for F2FCBT (g = -0.84) and for the combination of the two modalities (g = -.41) but not for BITs alone (g = -.39). This is the first subgroup analysis done to identify the efficacy of the combined modality.
Based on these findings, although F2FCBT yields a higher effect size, a combined BITs and F2FCBT modality can be a useful treatment to decrease symptoms of anxiety in youth populations. Combined treatment has some potential benefits that may be taken into consideration such as greater accessibility, increased engagement of a tech savvy population, and decreased stigma of therapy with a mental health professional. For example, in cases where accessibility, engagement and stigma act as barriers to treatment, an online program such as BRAVE-ONLINE [33] can be effective in treating anxiety disorders. However, further research is necessary to determine these advantages.
Subgroup comparisons based on treatment and control type (F2FCBT, BITs, WLC, and NEB control) were conducted in order to gain insight into the sources of heterogeneity. The present study indicates a marginal effect for subgroup differences, with WLC groups representing a more homogenous effect; and studies that included NEB control groups had high levels of heterogeneity, which can be expected by the various methods employed as control comparisons across these studies. Studies on F2FCBT and BITs that used WLC had a larger effect size, than those that used NEB controls. Additionally, the F2FCBT studies that used NEB controls found significant effects of treatment, while BITs with NEB controls did not. This finding is congruent with findings that F2FCBT has a higher effect size than BITs. Another subgroup analysis showed that F2FCBT and BITs were effective across age groups with no significant difference between groups, which is expected given known effectiveness of CBT with this population [39]. This shows that BITs can be engaging for not only young children in elementary school but also adolescents in high school.
The inclusion of modern technologies familiar to today's youth may increase interest and engagement with mental health resources. Technologies that were used in BITs and combined treatments included in the present analysis may not be representative of modern technology. None of the studies in this meta-analysis included smartphones or tablets, which are currently a dominant form of Internet access for this population. It is possible that children and adolescents present differences in the utilization of technologies, thus affecting the impact of BITs used for the treatment of anxiety disorders. Thus assessing the impact and including up-to-date technologies in psychotherapy with youth may make BITs more user friendly and will increase the amount of interventions a mental health professional could deliver. Currently there is a wide selection of anxiety-directed smart phone and tablet applications that have yet to be empirically tested.

Limitations
There are several limitations that must be considered in the current meta-analysis. This analysis only included publications focused on CBT for youth anxiety disorders and did not take into consideration other types of treatments (psychosocial or psychopharmacological) for this specific population. Because our focus was to compare F2FCBT and BITs, we have excluded articles that were comparing active and evidence based control conditions. Within the 34 articles included in this analysis, significantly more treatments were F2FCBT (n= 24) rather than BITs (n= 10). This could have resulted in heavily weighted inaccurate BITs studies. Across studies, sample size varied greatly, which may have resulted in biased comparisons and increased heterogeneity. Although this analysis attempted to compare studies using common assessments tools, the wide breadth of assessments used by the various studies reviewed was a critical source of heterogeneity. With regard to treatment design, though methodological quality was assessed, all studies that passed the inclusion criteria were included in the current analysis due to a limited number of overall treatment studies. Future analyses would benefit from looking at methodological quality across the field of BITS and F2F research. Finally, this metaanalysis implemented several exclusion criteria in order to maintain consistency across articles. Variables such as psychopharmacology, learning disabilities, non-English studies, follow-up analyses, and active controls were excluded in order to increase homogeneity within the analysis. Intent-to-treat samples were not included and only data showing the outcome of participants who completed treatment within a study were used.

Clinical implications
The current meta-analysis suggests that youth with anxiety disorders may benefit from F2FCBT and the combination of F2FCBT+BITs treatments. While F2FCBT has higher efficacy rates, the combination of F2FCBT+BITs show significant improvement in anxiety and can provide increased accessibility to clients. However, BITs alone do not currently have a significant effect size, future research including newer technologies (e.g., smartphones or tablets) may yield more positive outcomes and may help to increase therapy's reach. Various barriers exist to seeking treatment for children and adolescents with diagnosed anxiety. For example, those living in geographically remote locations may have difficulty attending treatments, which are generally focused within cities. In addition, F2F therapy can be expensive, inhibiting access for those with low-income levels. Additionally, the stigma of mental health treatment represents a barrier to the treatment-seeking process, especially among children and adolescents. In this respect, BITs have the ability to reach youth who may be struggling with these barriers and can simultaneously treat more individuals at a lower cost than F2F interventions.

Conclusion
This meta-analysis compared the available research of F2FCBT and BITs treatments for children and adolescents with diagnosed anxiety. While F2FCBT has a larger effect size, combined interventions (F2FCBT + BITs) show marginal significance and have the potential to lower barriers to effective anxiety treatments for youth. However, studies including BITs alone did not yield a significant effect size and were relatively small in number compared to F2FCBT studies. Therefore current evidence does not support the utilization of BITs as a stand-alone treatment and future research including newer BITs is imperative to draw further conclusions.

Conflict of interest
The author(s) confirm that this article content has no conflict of interest.