A Meta-Analysis on the Efficacy of Technology Mediated CBT for Anxious Children and Adolescents

  • Ioana R. Podina
  • Cristina Mogoase
  • Daniel David
  • Aurora Szentagotai
  • Anca Dobrean


Several meta-analyses indicate that cognitive-behavioral therapy (CBT) via electronic/technological devices or applications (i.e., eCBT) is an effective alternative to standard therapist-delivered CBT for anxious adults. However, we know little about the efficacy of eCBT interventions for anxious children and adolescents. The present meta-analysis set out to investigate the efficacy of eCBT in comparison to standard CBT or waitlist control for anxious children and adolescents. Eight randomized controlled studies (N = 404 participants) that targeted anxiety at post-intervention and follow-up were included in the analysis. The results indicated that eCBT was as effective as standard CBT (g = .295) and more effective than waitlist (g = 1.410) in reducing anxiety symptoms. Moderation analyses revealed that anxious children and adolescents benefited the most from eCBT in the minimal therapist involvement condition (g = 2.682) in contrast to the significant therapist involvement group (g = .326). Furthermore, older participants seemed to extract greater clinical benefits from eCBT in contrast to younger participants (slope = .514). Current eCBT interventions for anxious children and adolescents appear to be promising, but require further investigation.


Anxiety Children Adolescents Technology CBT Meta-analysis 


Anxiety disorders are among the most common emotional disorders in children and adolescents (Beesdo et al. 2009). Prevalence estimates for anxiety in this age group range from 9 to 32 % (Essau and Gabbidon 2013). Overall, anxiety causes significant and long lasting impairment in social, academic, occupational, and general day-by-day functioning (Kendall et al. 2004). Left untreated in youth, it frequently extends into adulthood and becomes accompanied by a number of other adverse mental health issues (Creswell et al. 2014), such as major depression and suicidal behavior (Woodward and Fergusson 2001).

In light of the high prevalence and chronic nature of anxiety disorders in children and adolescents, early intervention is recommended. In this respect, cognitive-behavioral therapy (CBT) is usually the first-line treatment (Walkup et al. 2008), with remission rates among anxious youth of 67 % or greater at post-treatment (Siqueland et al. 2005).

Despite scientific support and clinical efficacy, standard therapist-delivered CBT can pose serious financial and time-related challenges. Additionally, CBT may be unavailable due to lack of access to adequately trained professionals (for a review, see Newman et al. 2011). Technology can provide a solution to all the above issues, introducing a new wave of technologically mediated treatments.

Technological Enhancements of CBT: eCBT

In contrast to other therapeutic modalities, CBT is fitted for the new technological wave in psychotherapy, as it is highly structured and implemented in a sequential manner (Heimberg and Coles 1999; Proudfoot et al. 2004). Currently, three major platforms are being used to deliver CBT services. These are computer and Internet-delivered programs (including multimedia programs and the Internet/web conference environments), virtual reality programs (VR), and smartphone or handheld device applications (e.g., tablets).

General advantages of electronic/technologically mediated CBT or eCBT1 (Spurgeon and Wright 2010) include increased mastery and control for the user and the therapist, ease of access to therapy, portability (i.e., technology for handheld devices), confidentiality, reproducibility, and safety (Bouchard 2011; Stallard et al. 2010). Altogether, the above technological advantages are extremely valued by children and adolescents (Stallard et al. 2010) and come at a time when technology is easily accessible.

Furthermore, electronic/technologically mediated therapies provide a way around time-consuming traveling means and logistic challenges that are often brought on by face-to-face techniques. Studies have estimated savings of up to $540–$630 per client when compared with standard CBT (Newman et al. 1997, 1999). Reduced costs include VR tools, as in some cases, such as flight phobia, it is less expensive to follow a VR exposure than to pay for flight tickets during therapy sessions.

Clinical Efficacy of eCBT

Internet (iCBT) or computer-delivered CBT (cCBT) has been shown to be an effective alternative to traditional CBT for anxiety in adults (Newman et al. 2011) and has yielded patient satisfaction with treatment and attrition rates comparable to standard interventions (Cuijpers et al. 2009; Andrews et al. 2010). These results are similar in studies comparing CBT with virtual reality exposure (otherwise known as virtual reality exposure therapy; VRET) to evidence-based interventions in adult populations (Opris et al. 2012).

Considering the success of eCBT with anxious adults, the next question in line is “Are eCBT interventions efficacious for anxious children and adolescents, as well?” Two recent meta-analyses have attempted to answer this question. One focused exclusively on anxiety in children (Rooksby et al. 2015) and the other on anxiety and depression in children and young people up to 25 years old (Pennant et al. 2015). However, because these papers focus either exclusively on children and/or emerging adults, we believe that the childhood-adolescence age range has yet to be investigated. Of note, to our knowledge, previous work did not exclude protocols for transdiagnostic treatment of anxiety and depression and there was little to no mention of the moderators of the efficacy of eCBT.

In contrast to previous work, the current meta-analysis restricted the analyses (a) to children and adolescents, (b) to eCBT protocols that were specific to anxiety and that were not transdiagnostic treatments of anxiety and depression or (c) to studies where anxiety was a primary and not a secondary outcome. We believe that this precise focus will render more rigorous and more reliable conclusions. Furthermore, this restricted focus enabled moderation analyses of the efficacy of eCBT. Finally, we included all the major platforms on which CBT is currently being delivered (e.g., VR), not limiting the generalization of results to iCBT or cCBT.

Such a meta-analysis has the potential to (1) inform subsequent research and (2) to contribute towards the understanding of the clinical utility of eCBT, thus setting the stage for its inclusion into clinical practice.

Overview of the Present Study

The purpose of the present study is twofold. Firstly, the study aims to compare the relative efficacy of eCBT in contrast to a control group (i.e., waitlist or standard therapist-delivered CBT) with respect to anxiety at post-intervention and follow-up. Secondly, it aims to investigate potential moderators of the difference in efficacy between eCBT and control. Previous reviews on eCBT for children and adolescents (Reger and Gahm 2009; Elkins et al. 2011) suggested several potential moderators. These are the degree of therapist involvement, the source of anxiety assessment, and the participants’ age (Elkins et al. 2011).

Degree of Therapist Involvement

Therapist contact is an important variable to consider in understanding the cost-effectiveness of eCBT. Thus far, the degree of therapist involvement has been shown to impact the efficacy of eCBT in adults, namely interventions relying on therapist support yielded greater improvements than self-help interventions (Spek et al. 2007). Given that we identified various degrees of therapist involvement in eCBT, we are able to assess whether eCBT interventions are cost-effective in terms of therapist time and whether therapist involvement moderates the efficacy of eCBT for anxious children and adolescents.

Source of Anxiety Assessment

Standard procedures for working with child and adolescent clients require the clinician to interview parents and youth, and incorporate their reports into the clinical assessment. Information obtained from multiple sources (i.e., self-report, parent-report, and clinician-rated anxiety) is considered to be complementary, although not always convergent (Campbell and Rapee 1996). Given the potential inconsistency between sources, we considered the source of anxiety assessment as a potential moderator of the efficacy of eCBT for anxious children and adolescents.

Participants’ Mean Age

Although CBT is the treatment of choice for anxious youth, a number of authors have questioned whether the observed benefits apply equally to children and adolescents (Hudson 2005). Effect size estimates based on this age range may mask age-related heterogeneity, especially for individuals who respond more favorably to CBT. Recent results indicate that despite initial concerns, children who received standard CBT had reductions in anxiety comparable with adolescents (Bennett et al. 2013). However, the technological component of CBT could contribute to a differential impact of eCBT on anxious youth. Perhaps, the technological manner of delivering CBT will appeal more to adolescents than to younger children due to their facilitated access to various communication gadgets. Given that the literature is not conclusive in this respect, we are interested in investigating whether there is an age-related differential efficacy of eCBT in anxious children and adolescents.


Literature Search

Potentially relevant studies were identified following a systematic search of the PsychInfo, Web of Science, EBSCO, ProQuest, Scopus, and Medline databases through September 2015. The following keywords were entered: “computer”, “phone”, “tablet”, “internet”, “online”, “web”, “VR”, “exposure”, “CBT”, “adolescent”, “child” paired with “anxiety”. We also searched for relevant references within the most recent articles and reviews on eCBT (Wuthrich et al. 2012; Storch et al. 2011; Rickwood and Bradford 2012; Kendall et al. 2012; Reyes-Portillo et al. 2014; Pennant et al. 2015; Rooksby et al. 2015). Lastly, we sought out unpublished papers and data (i.e., dissertations, theses) by searching the above-mentioned databases and contacting several researchers who study eCBT.

I.R.P. screened the titles and abstracts in order to assess their relevance for the meta-analysis. Selected full-text articles were screened in congruence with the inclusion/exclusion criteria. Both I.R.P. and C.M. individually read each full-text article to assess its suitability for inclusion in the current meta-analysis. Disagreements between the two reviewers were resolved by consensus and consultation with A.D.

Selection of Studies

Upon a thorough search, we identified 142 records (see Fig. 1). After removing duplicates, we examined the abstracts of the remaining studies to determine their relevance to the purposes of the current study. We excluded four publications that described treatment protocols and/or qualitative reviews, and retained 45 potentially relevant articles. Remaining full-text papers were analyzed in detail, and the following inclusion criteria were applied: (a) selected studies included an age range that began with school-aged children (5–7 years old is regularly the lower age limit for CBT trials) and extended up to 18 years old; (b) the study aimed primarily to reduce diagnosed/clinical anxiety in children and adolescents by means of an eCBT intervention; (c) the intervention involved a full CBT protocol and/or CBT components, like exposure-based interventions; (d) selected studies were randomized controlled investigations; (e) a control group existed (i.e., standard face-to-face CBT or waitlist); (f) eCBT included multimedia platforms (computerized or Internet-delivered), web conferences, VR environments, handheld devices (phone or tablet)2; (g) post-intervention and/or follow-up measures of anxiety were provided; (h) sufficient data to compute between-group effect sizes; and (i) the studies were written in English.
Fig. 1

Flow diagram of the study selection process

We excluded studies that focused on reducing depression, where anxiety was a secondary outcome. We also excluded studies that were not tailored to address primarily anxiety symptoms and whose intervention protocol was transdiagnostic (i.e., for depression and anxiety) (see Fig. 1). In order to avoid carryover effects, we discarded crossover design studies (see Fig. 1). Only eight of the 45 studies met the inclusion criteria and were included in the current meta-analysis.


Coding Procedure

We retained the following variables: study identification data (i.e., author and year of publication), type of control group (i.e., waiting list or standard face-to-face CBT), type of eCBT platform, sample and intervention characteristics (i.e., sample size, clinical status, type of anxiety disorder, follow-up length in months), moderators (i.e., the degree of the therapist involvement, source of anxiety assessment, mean age), and anxiety outcomes (see Table 1).
Table 1

Characteristics of studies included in meta-analysis


Mean age (age range)

Contrast pair

Sample size

Type of eCBT platform

Clinical status

Therapist involvement

Source of anxiety assessment

Follow-up length (month)


1. Maldonado et al. (2009)

11.9 (10–15)




Analogue (School Phobia)




2. Khanna and Kendall (2010)

10.1 (7–13)




Diagnosed (primary diagnosis of anxiety disorder)


Self-reported, clinician-rated



3. March et al. (2009)

9.45 (7–12)




Diagnosed (primary diagnosis of anxiety disorder)


Parent-reported, clinician-rated



4. Spence et al. (2006)

9.93 (7–14)




Diagnosed (primary diagnosis of anxiety disorder)


Self-reported, parent-reported, clinician-rated



5. Spence et al. (2011)

13.98 (12–18)




Diagnosed (primary diagnosis of anxiety disorder)


Self-reported, parent-reported, clinician-rated



6. St-Jacques et al. (2010)

10.6 (8–15)




Diagnosed (Spider Phobia)


Self-reported, clinician-rated


7. Storch et al. (2011)

11.1 (7–16)




Diagnosed (OCD)


Self-reported, parent-reported, clinician-rated



8. Wuthrich et al. (2012)

15.17 (14–17)




Diagnosed (primary diagnosis of anxiety disorder)


Self-reported, parent-reported, clinician-rated



CBT cognitive behavioral therapy, cCBT computer-delivered CBT, CGAS Children’s Global Assessment Scale (Shaffer et al. 1983), CGI Clinical Global Impressions Scales, COIS-C The Child Obsessive Compulsive Impact Scale Child Version (Piacentini et al. 2003), COIS-P The Child Obsessive Compulsive Impact Scale Parent Version (Piacentini et al. 2003), CSR clinician severity rating, CY-BOCS Children’s Yale-Brown Obsessive Compulsive Scale (Scahill et al. 1997), eCBT electronic/technologically mediated CBT, FSSC-R Fear Survey Schedule for Children-Revised (Ollendick 1983), iCBT internet-delivered CBT, MASC The Multidimensional Anxiety Scale for Children (March et al. 1997), RCMAS Revised Children’s Manifest Anxiety Scale (Reynolds and Richmond 1985), SCAS-C The Spence Children’s Anxiety Scale Child version (Spence 1998; Spence et al.1999), SCAS-P The Spence Children’s Anxiety Scale Parent version (Spence 1998; Spence et al. 1999), SPSQ-C Social Phobia Screening Questionnaire for Children (Gren-Landell et al. 2009); SPQ Spider Phobia Questionnaire (Klorman et al. 1974), W waitlist, VRET virtual reality exposure therapy)

The investigated moderators were coded as follows:
  1. (1)

    Degree of therapist involvement Newman et al. (2011) served as a guideline for the modality in which we split the degree of therapist involvement. To our knowledge, this modality is one of the most commonly employed classifications of the degree of therapist involvement. As such, this variable was divided into minimal (i.e., self-help with minimal therapist support via email or telephone) and significant therapist involvement (i.e., eCBT interventions supplemented by face-to-face meetings, online patient-therapist interactions or predominantly therapist-administered sessions).

  2. (2)

    Source of anxiety assessment We coded this moderator based upon three sources of anxiety assessment, namely (1) self-reported anxiety, (2) parent-rated anxiety, and (3) clinician-rated anxiety.

  3. (3)

    Participants’ mean age was retrieved from the article and treated as a continuous moderator. The resulting mean age interval varied from 9.45 to 15.17 years old.


Statistical Analysis

To estimate the effect size, we computed Hedges’s g, a coefficient that controls for discrepancies in sample size between studies (Hedges and Olkin 2014). Just like Cohen’s d coefficient, a 0.2–0.5 value indicates a small effect size, whereas a 0.5–0.8 value indicates a medium effect size and a 0.8 value or higher indicates a large effect size (Cohen 1988). The effect sizes were coded so that a positive value of Hedges’s g indicated a superior efficacy of the eCBT intervention compared to the control group.

We computed two effect sizes, one at post-intervention and one at follow-up. When a study reported multiple anxiety measures, we computed an average effect size of those outcomes for the corresponding time point (i.e., post-intervention and/or follow-up). When a study reported more than one level of a categorical variable (e.g., both parent and child-rated anxiety levels), dependencies were accounted for by randomly selecting one within-study level per study (Hunter and Schmidt 2004). This technique enabled an independent analysis at the moderator level. The data utilized for computing the average effect sizes for anxiety symptoms included the following: means and standard deviations; between-group t values and sample sizes; between-group p values and degrees of freedom.

For all sets of computed effect sizes, we used a random effects’ model, which assumes that studies come from populations in which the effect sizes differ. In order to examine the extent to which the effect sizes vary between studies, we tested for heterogeneity of effect sizes using the Q statistic and the I2 statistic (Borenstein et al. 2011). The Q statistic is an index of heterogeneity that relates true heterogeneity to random error. A significant Q indicates heterogeneity in effect sizes beyond random error. I2 statistic indicates the percentage of the observed heterogeneity. Unlike Q, I2 is not sensitive to the number of studies entered in the analysis (Borenstein et al. 2011).

To address publication bias, we computed a fail-safe N for significant main effect sizes. Fail-safe N approximates the number of unpublished studies with an effect size of zero that would bring the computed effect size to non-significance (Rosenthal 1991). In addition, we produced a funnel plot and visually examined it for publication bias. The assumption of the funnel plot is that small effect sizes based on small sample sizes are prone to error. If there is a publication bias, the funnel plot will be asymmetrical, with studies dispersed unevenly above or below the mean.

Furthermore, we used Duval and Tweedie’s trim-and-fill procedure (Duval and Tweedie 2000) that estimates the number of missing studies that would correct for publication bias, computing an effect size free of publication bias. All the analyses were run using Comprehensive Meta-Analysis (Version 2.2.046; Borenstein et al. 2011).


Between-Group Analysis

In order to investigate the efficacy of eCBT compared to waitlist, we ran the analyses on data derived from 6 studies, totaling 340 participants. At post-intervention, in the absence of outlying studies, the pooled effect size indicated a significant difference in favor of eCBT, (g = 1.410, p = .008, 95 % CI [.375; 2.444]). These values indicated that participants in the eCBT group experienced less anxiety than 92 % of those in the waitlist group (McGough and Faraone 2009). The analyses revealed heterogeneity in results as reflected by Q (5) = 67.822 and p < .001, I2 = 92.628. The results from these analyses are detailed in the forest plot (Fig. 2).
Fig. 2

Forest plot of average effect sizes indicating the difference between eCBT and waitlist control group regarding youth anxiety

Next, we computed the post-intervention (k = 4 studies, N = 241 participants) and follow-up (k = 4, N = 179) average effect sizes for the eCBT-CBT contrast. In the absence of outliers, results indicated non-significant differences between eCBT and CBT, both at post-intervention, g = .295, p = .480, 95 % CI [−.525; 1.115], and follow-up, g = −.150, p = .327, 95 % CI [−.450; .150]. There was evidence of heterogeneity at post-intervention, Q (3) = 23.047, p < .001, I2 = 86.983, but not at follow-up Q (3) = 4.498, p = .212, I2 = 33.299.

Moderators of Anxiety Symptoms

We specifically tested for moderators of the efficacy of eCBT compared to waitlist, as the status of the literature is too premature to perform moderation analyses on the eCBT-CBT contrast (i.e., only four studies). Furthermore, moderation analyses were ran for post-intervention average effect sizes of anxiety. Follow-up data for the eCBT-W pair were unavailable (see Table 1).

Categorical Moderators

The degree of therapist involvement (minimal vs. significant) significantly moderated the average effect size for anxiety at post-intervention. Therefore, participants in the eCBT group with minimal therapist involvement experienced significantly less anxiety than their counterparts in the waitlist group. In the significant involvement group, the difference between eCBT and waitlist was only marginally significant (see Table 2). Irrespective of the degree of therapist involvement, the results were in favor of the eCBT intervention (see Table 2). In other words, in the minimal therapist involvement group, eCBT participants had lower levels of anxiety than 99 % of the individuals in the waitlist group. In the significant therapist involvement group, eCBT participants had lower levels of anxiety than 62 % of the people in the waitlist group.
Table 2

Moderation analysis with categorical variables




Condition pair

No. of contrasts



Q w



Q b




Minimal/significant therapist involvement







[.326; 4.038]








[−.035; .687]


Self-reported/parent rated/clinician rateda







[−.357; 1.597]








[.245; 2.000]






[.884; 4.263]

aEven when controlling for dependency issues, with one level per study, the source of anxiety remained a non-significant moderator of anxiety, Q (2) = 2.597, p = .273

eCBT electronic/technologically mediated CBT, Post post-intervention, W waitlist

The source of anxiety assessment (i.e., self-reported, parent-rated, and clinician-rated) was not a significant moderator of the post-intervention average effect size for anxiety (see Table 2).

Continuous Moderator

The mean age variable was tested via meta-regression analysis procedures. The results indicated that the participants’ mean age (slope = .514, p < .001) was a significant moderator of anxiety. Older participants experienced greater clinical benefits as opposed to younger participants.

Publication Bias

Publication bias analyses were carried out for anxiety at post-intervention and, when possible, at follow-up. Publication bias is reported for both contrasts (i.e., eCBT-W and eCBT-CBT), as follows.

We produced funnel plots and used trim-and-fill procedure to estimate the magnitude of the publication bias (Duval and Tweedie 2000). No evidence of publication bias was found for the eCBT-W pair. Regarding the eCBT-CBT contrast, trim-and-fill method estimated one study with an effect size higher than the mean, but that would not change the results significantly, g = .484, 95 % CI [−.220; 1.189], Q = 29.253. In line with the trim-and-fill results, the funnel plot showed some asymmetry, suggesting the presence of one missing study with an effect size above the mean and the possibility of obtaining slightly under-inflated estimates of the true differences (see Fig. 3).
Fig. 3

Funnel plot of publication bias for the eCBT-CBT contrast

Next, we computed the fail-safe N analysis of significant main effects, namely for the eCBT-W contrast. The number of studies that would reduce the effect size to non-significance was 87. This number supports the robustness of the computed effect size and it is in congruence with the recommendations set forth by Rosenthal (1991). As such, a computed fail-safe N should be larger than 5K + 10 (where K is the number of studies included in the meta-analysis) in order to indicate a robust effect size. This is also the case of the six studies on the eCBT-W contrast, where fail-safe N would be expected to be higher than 40.


The current meta-analysis aimed to investigate the efficacy of eCBT versus control groups (i.e., standard face-to-face CBT or waitlist) in the treatment of child and adolescent anxiety. We conducted a meta-analysis of eight randomized controlled studies that included post-intervention and follow-up data. Furthermore, we investigated possible moderators of differences in response to eCBT.

Main Effects

Results showed a significant post-intervention difference between eCBT and waitlist, with participants in the eCBT group experiencing less anxiety than participants in the waitlist condition. Therefore, anxious children and adolescents benefit more from receiving eCBT interventions than from receiving no treatment at all. These results are similar to other meta-analyses on adult anxiety (Reger and Gahm 2009) and two recent meta-analyses on anxious children and emerging adults (Pennant et al. 2015; Rooksby et al. 2015).

It is important to note that for ethical reasons articles contrasting eCBT to waitlist were mainly limited to post-intervention assessments of anxiety (see Table 1). As such, we were unable to draw conclusions about the maintenance of anxiety changes across time. However, the results on adult anxiety are promising, in that participants in the eCBT condition continue to experience at follow-up less anxiety than individuals in the waitlist (Reger and Gahm 2009; Christensen et al. 2009).

Notably, there was no significant difference in efficacy between eCBT and standard face-to-face CBT in terms of anxiety, either at post-intervention or at follow-up. Therefore, eCBT seems to be comparable to standard CBT in terms of efficacy. Similar results were reported by previous meta-analyses on the efficacy of eCBT for adult anxiety (Cuijpers et al. 2009; Opris et al. 2012; Andrews et al. 2010) and by related meta-analyses on cCBT for anxious youth up to 25 years old (Pennant et al. 2015; Rooksby et al. 2015).

There was significant heterogeneity among studies indicating potentially important moderators. We specifically tested for moderators of the efficacy of eCBT compared to waitlist, as the status of the literature is too premature to perform moderation analyses on the eCBT-CBT contrast.

Categorical Moderators

The degree of therapist involvement (i.e., minimal vs. significant) was a significant moderator of anxiety. Technologically delivered CBT significantly outperformed the waitlist group in the minimal therapist involvement condition. Results are in line with previous studies, showing that minimal therapist support appears to be enough to keep a patient engaged in the current treatment (Ghosh et al. 1988; Menchola et al. 2007). In support of this finding, recent evidence indicates that minimal therapist contact is also sufficient to establish an adequate working alliance (Anderson et al. 2012; Cuijpers et al. 2010), suggesting that face-to-face interactions are not indispensable to make a working alliance possible. Furthermore, given the affinity of children and adolescents for technology, it may be that they are more receptive to the information delivered mainly in a technological manner, rather than supported by face-to-face interactions with the therapist (Stallard et al. 2010).

The source of anxiety assessment was not a significant moderator of the outcome, meaning that there were no significant discrepancies between self-report, parent-ratings, and clinician-ratings in terms of anxiety. However, we did notice a growing difference in effect sizes from self-report to clinician-rated anxiety, clinician ratings being more favorable to eCBT than parent-rated or self-reported assessments of anxiety. In part, these discrepancies in assessments are to be expected, as they had been encountered before in the literature (Campbell and Rapee 1996). In addition, younger children might find it more difficult to identify their anxiety (Reynolds et al. 2012), which might explain why self-reports scored lower than the clinician or parent-rated assessments. Nevertheless, irrespective of the effect size, all the sources of anxiety rating indicated that participants in the eCBT group tended to experience less anxiety than those in the waitlist group.

Continuous Moderator

Age significantly moderated the effect of eCBT on anxiety, as older participants extracted greater clinical benefits from eCBT in contrast to younger participants. This finding is consistent with previous meta-analyses on face-to-face psychotherapy for anxious youth (Newman et al. 1997), namely therapeutic changes in children were smaller in magnitude in comparison to adolescents (e.g., Newman et al. 1997). This may be because older children and teenagers are more motivated by psychological therapy in general, or because they have the cognitive and interpersonal skills to engage in CBT specifically. Alternatively, older children and teenagers may be more skilled in assessing their symptoms of anxiety than their younger peers.

Limitations and Future Directions

The present meta-analysis has several limitations related to the current state-of-affairs in the literature. First, there were a limited number of studies comparing eCBT to standard CBT, which restricted the conclusions that can be drawn from this contrast. However, the similarities between our findings and previous meta-analyses on adult and young anxious populations provide more reliability to our findings.

Second, despite the significant heterogeneity for the eCBT-CBT contrast, no moderation analyses could be done. One reason was the limited number of studies and another, the similarities between studies; most being grouped around one type of moderator (see Table 1). As such, further eCBT-CBT studies do not only need to better test the efficacy of eCBT interventions, but also to investigate for whom and under which conditions is eCBT comparable to standard CBT.

Third, we had no physiological and/or behavioral information regarding the amount of anxiety experienced by anxious children and adolescents. This is an important drawback of the published research, as multimodal anxiety assessments provide a more reliable estimate of anxiety (Schwerdtfeger 2004).

Fourth, almost none of the eCBT-W studies included a follow-up for the waitlist group. As such, it remains highly relevant to address the issue of follow-up length in anxiety-related interventions, but without crossing ethical boundaries, as up to 30 % of individuals experience symptom return (for review, see Craske and Mystkowski 2006). Furthermore, longer follow-up durations are necessary. From the current selection of studies, it is noticeable that follow-up extends up to 12 months. However, longer follow-up durations (e.g., longer than 2 years) could provide a better assessment of relapse, and relapse triggers. This is a pertinent question, as studies on standard CBT show that, in some cases, symptoms return after a 2 year interval (e.g., for a review see, Arch and Craske 2009).

Fifth, the amount of parental support required by eCBT remains an issue that could not be explored by the current meta-analysis, as all but one of the eCBT-W studies included parental support (i.e., Maldonado et al. 2009). Based on the frequent use of parental support in eCBT interventions, one may think that parental support is a requirement for this type of interventions to be successful. However, given that eCBT is an intervention that intends to demand fewer time resources from the beneficiary, as well as the therapist and parents, it might be useful for future studies to use less to no parental support. There are some efforts in this direction, especially in the case of depression (Stasiak et al. 2012; Van der Zanden et al. 2012). Currently, we can only pinpoint to other meta-analyses on standard CBT, which show no additional benefit of treatments with parental involvement (Reynolds et al. 2012).

Sixth, despite our extensive search, no articles on portable devices could be found on anxious youth, which limits the generalizability of our results. Probably, a reason behind the lack of such publications is that these platforms are still in a developmental stage for children and adolescents (Berry and Lai 2014).

In addition to addressing current limitations, future studies could focus on the following:
  1. (a)

    Testing transdiagnostic protocols in contrast to diagnostic specific interventions. There is a trend, namely in eCBT literature, to develop unified treatment protocols that can be used across anxiety disorders. However, these procedures have not yet been fully tested against diagnostic-specific protocols, and may lack the degree of specificity needed to treat different classes of anxiety disorders. Moreover, this type of approach may result in an inability to determine which of the pre-treatment diagnoses responded better to treatment and, which were more resistant to it.

  2. (b)

    Assessing the impact of the programs themselves. None of the inspected studies have delivered any data on the quality of their software, level of engagement, experience of immersion, and/or the capacity of the software to enable learning. This makes it harder to detect whether failures to show efficacy in some studies are caused by technological issues or the intervention per se.

  3. (c)

    Examining the impact of comorbid disorders on the efficacy of eCBT. A study by Berman et al. (2000) revealed that children and adolescents responded less well to treatment when, in addition to anxiety, they were diagnosed with depression. In the current meta-analysis, only 2 out of 8 studies specifically measured depression in addition to anxiety. However, the reported results did not allow for an analysis of comorbidities or their impact on outcomes of efficacy. Future studies should report more details on comorbidities, as one of the objectives is to find who might benefit the most from such interventions.


Another interesting point is that all the eCBT interventions included some form of therapist support, unlike much of the adult literature. This is curious in itself, as one could assume that therapist involvement is a requirement for eCBT interventions. However, the inclusion of some form of therapist support is in contradiction with the efforts to reduce the allocated time resources of both parents and therapists. Future studies need to clarify this aspect and endeavor to examine whether therapist support is needed or not for efficient eCBT interventions.


The current findings suggest that eCBT for anxious youth is a promising line of research and therapy. However, since eCBT’s therapeutic benefits are mainly limited to waitlist comparisons, we believe that it is too premature to speak of a first-line treatment for anxious youth or even a robust therapeutic alternative. The current state of affairs in the literature undoubtedly demands more controlled clinical trials before eCBT platforms are to be launched onto the market and disseminated among professionals.


eCBT stands for CBT healthcare practices supported by electronic/technological devices. Electronic/technologically mediated CBT (i.e., eCBT) is part of the broader concept of eMental Health (eMH).


Phone or tablet applications relying on CBT could not be added to the meta-analysis. Despite our extensive search, no articles on portable devices were found for anxious youth.



This work was supported by a Grant of the Romanian National Authority for Scientific Research, CNCS—UEFISCDI, Project number PN-II-PT-PCCA-2011-3.1-1500, contract number 81/2012, coordinated by Dr. Anca Dobrean. Furthermore, Ioana Podina’s contribution to this work was also possible with the financial support of the Sectorial Operational Program for Human Resources Development 2007–2013, co-financed by the European Social Fund, under the project number POSDRU/159/1.5/S/132400 with the title “Young successful researchers—professional development in an international and interdisciplinary environment”. The authors would also like to thank Mirela Mohan proof reading this paper.

Compliance with Ethical Standard

Conflict of interest

The authors declare that they have no conflict of interest.

Copyright information

© Springer Science+Business Media New York 2015

Authors and Affiliations

  • Ioana R. Podina
    • 1
  • Cristina Mogoase
    • 1
  • Daniel David
    • 1
    • 2
  • Aurora Szentagotai
    • 1
  • Anca Dobrean
    • 1
  1. 1.Department of Clinical Psychology and PsychotherapyBabeş-Bolyai UniversityCluj-NapocaRomania
  2. 2.Department of Oncological SciencesIcahn School of Medicine at Mount SinaiNew YorkUSA

Personalised recommendations