The authors have declared that no competing interests exist.
Conceived and designed the experiments: DDE HR. Performed the experiments: DDE PC HR. Analyzed the data: DDE HR PC. Contributed reagents/materials/analysis tools: DDE MB. Wrote the paper: DDE AZ HC PC YS MB HR.
Anxiety and depression in children and adolescents are undertreated. Computer- and Internet-based cognitive behavioral treatments (cCBT) may be an attractive treatment alternative to regular face-to-face treatment.This meta-analysis aims to evaluate whether cCBT is effective for treating symptoms of anxiety and depression in youth.
We conducted systematic searches in bibliographical databases (Pubmed, Cochrane controlled trial register, PsychInfo) up to December 4, 2013. Only randomized controlled trials in which a computer-, Internet- or mobile-based cognitive behavioral intervention targeting either depression, anxiety or both in children or adolescents up to the age of 25 were compared to a control condition were selected. We employed a random-effects pooling model in overall effect analyses and a mixed effect model for sub-group analyses. Searches resulted in identifying 13 randomized trials, including 796 children and adolescents that met inclusion criteria. Seven studies were directed at treating anxiety, four studies at depression, and two were of a transdiagnostic nature, targeting both anxiety and depression. The overall mean effect size (Hedges’ g) of cCBT on symptoms of anxiety or depression at post-test was g=0.72 (95% CI:0.55-0.90, numbers needed to be treated (NNT)=2.56). Heterogeneity was low (I²=20.14%, 95% CI: 0-58%). The superiority of cCBT over controls was evident for interventions targeting anxiety (g=0.68; 95% CI: 0.45-0.92; p < .001; NNT=2.70) and for interventions targeting depression (g=0.76; 95% CI: 0.41-0.12; p < .001; NNT=2.44) as well as for transdiagnostic interventions (g=0.94; 95% CI: 0.23-2.66; p < .001; NNT=2.60).
Results provide evidence for the efficacy of cCBT in the treatment of anxiety and depressive symptoms in youth. Hence, such interventions may be a promising treatment alternative when evidence based face-to-face treatment is not feasible. Future studies should examine long-term effects of treatments and should focus on obtaining patient-level data from existing studies, to perform an individual patient data meta-analysis.
Depression and anxiety are common in children and adolescents (in the following referred to collectively as
In the past decades, a number of psychological treatments for anxiety and depression in youth have been developed with demonstrated efficacy in a substantial number of clinical trials [
However, up to 80% of children and adolescents with mental health needs receive no treatment [
Using the computer or the Internet to provide CBT may overcome some of the limitations of traditional treatment services. Advantages of computer and Internet-based CBT (cCBT) include (1) availability (2) anonymity, (3) accessibility at any time and place, (4) flexibility in self-direction and self-pacing, (5) reduced travel time and costs for both participants and clinicians, and (6) the appeal of interactivity and visual attractiveness of Internet-based programs [
cCBT is effective in the treatment of depressive [
Less is known, however, about the effectiveness of cCBT for anxiety and depression in youth. In a recent systematic review [
Several strategies were used to identify relevant studies. Two independent assessors (Ebert & Zarski) searched three major bibliographical databases (PubMed, PsycInfo, Cochrane) by combining terms indicative of each of the disorders with terms indicative of psychological treatment and randomized controlled trials. The searches were performed December 04, 2013. We also checked the references of the identified studies for earlier publications. Details of the searches are provided in
We included (a) randomized controlled trials in which (b) a computer or Internet- or mobile-based (c) cognitive behavioural intervention (d) targeting either depression, anxiety or both (e) in children or adolescents up to the age of 25 (f) were compared to a non-active control condition (wait-list; placebo). We only included studies (g) that used an anxiety or depression measure as the primary outcome measure and (h) in which participants were selected because of elevated levels of depressive/anxiety symptoms based on either a standardized self-report measure or a formal diagnosis of major depressive or any anxiety disorder (including OCD), and (i) effect sizes could be calculated from the information provided in the paper. If effect sizes could not be calculated, authors were contacted to retrieve the necessary information. Studies were not included when they had only an active treatment comparison (e.g., face-to-face CBT) or only included participants older than 18 years. Comorbid general medical and psychiatric or substance/alcohol use disorders were not used as an exclusion criterion. No language restrictions were applied. Studies that evaluated interventions in which the Internet/computer was not the primary treatment modality (e.g., blended treatments with face-to-face psychotherapy and some online elements) were excluded.
Risk of bias assessment and data extraction was conducted independently by two assessors (Ebert & Zarski). We assessed the validity of included studies using four criteria of the
Variables related to the study design included the target condition (anxiety vs. depression vs. transdiagnostic) and the comparison condition (no treatment vs. placebo). The comparison group was coded as placebo control group if cCBT was compared to an active condition with no intended therapeutic properties that would be expected to produce change in the target diagnosis [
We also coded the outcome informant (youth, parent, teacher), that is, whether the information about symptoms were derived from the youth report, parent report, or teacher report. We further coded the type of outcome assessment (self-report vs. observer based) and whether the studies included a follow-up (yes/no; duration).
In line with previous meta-analyses on psychotherapy in children and adolescents [
Interventions were classified as with or without parental involvement (yes/no) and with or without professional guidance (therapeutic or simply administrative) (yes/no). We also coded the technology on which the intervention was delivered (computer: cCBT; Internet: iCBT; mobile: mCBT), the setting in which the intervention was conducted (at home/not at home, e.g., clinic, school, university), as well as the number of treatment sessions.
For each comparison between cCBT for anxiety or depression or both and a control group, we calculated the effect size indicating the difference between the two groups in change from baseline to post-test (Hedges’ g). To calculate pooled mean effect sizes, we used the Comprehensive Meta-Analysis (CMA) software (Biostat, Inc.). In the main outcome analysis, we examined the overall mean effects of cCBT on symptoms of depression or anxiety. To generate a mean effect size, we only used instruments that related to the principal measure of the disorder. If interventions were transdiagnostic, targeting both anxiety and depression, we used the pooled effect size of these outcomes. If more than one measure per outcome was used, the mean of the effect sizes was calculated, so that each study provided only one effect size. Given that youth depression and anxiety symptoms highly correlate [
We expected considerable heterogeneity among the studies, hence we employed a random-effects pooling model (DerSimonian-Laird) in all analyses. Given the difficulty of interpreting the standardized mean difference (Hedges’ g) from a clinical perspective, we transformed these values into the number-needed-to-treated (NNT) using the formula provided by Kraemer and Kupfer [
As a test of homogeneity of effect sizes, we calculated the
Publication bias was tested by visually inspecting the funnel plot on primary outcome measures. We also conducted Egger’s test to quantify the bias captured by the funnel plot and to test whether it was significant [
We further conducted a series of subgroup-analyses, according to the mixed-effect model [
Thirteen studies met inclusion criteria for the meta-analysis of cCBT for depression and anxiety in youth.
The 13 studies on cCBT for symptoms of anxiety and depression in youth included 796 participants, 420 in the treatment groups and 376 in the control groups. Selected characteristics of the included studies are displayed in
Study | Cond | Age Range | Inclusion | Diagn | Recr | Setting | Mean Age (SD), % Boys | Conditions | N | Nmod | Par Inv | Guid-ance | Outcomes and Method | FU | Bias |
Country |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Fleming, | Dep | 13–16 | CDRS-R >29, excl: severe depr | N | School | School | 14.9 (0.79) 56 | cCBT (SPARX) | 20 | 7 | N | Y | D: CDRS-R (Y, O), RADS-2 (Y, SR); A: SCAS-C (Y, SR) | N | ++++ | NZ |
2012 [ |
WL | 12 | ||||||||||||||
Keller, | Anx | 6–12 | GAD > 8, SepA > 4 / SP > 7 on SCARED-P, excl:depr T-score > 65 on CDI-S | N | Com | Home | 8.41 (1.59), 51.35 | iCBT | 22 | 12 | Y | Y | A: SCARED-P (P, SR), DISC IV-P (P, O), RCMAS (Y, SR); D: CDI-S (Y, SR) | N | ++++ | USA |
2009 [ |
||||||||||||||||
WL | 15 | |||||||||||||||
Khanna, 2010 [ |
Anx | 7–13 | SepA, SoP, GAD, SP / PD on ADIS-P CSR > 3 | Y | Com | Uni | 10.1 (1.6), 67.35 | cCBT (Camp Cope-A-Lot) | 17 | 12, Par:2 | Y | Y | A: ADIS-P CSR (P, O), MASC (Y, SR) | N | +?++ | USA |
Placebo | 16 | D: CDI (Y, SR) | ||||||||||||||
Makarushka, 2012 [ |
Dep | 11–15 | CES-D > 15, excl: no MD, D / M on K-SADS-E | Y | Com | Home | 12.7 (1.24) 44 | iCBT (Blues Blaster) | 76 | 6 | N | N | D: CES-D (Y, SR) | 6M | +?++ | USA |
Placebo | 85 | |||||||||||||||
March, | Anx | 7–12 | Anx Dis (other than OCD, PD / PTSD) on ADIS-C/-P CSR > 3 | Y | Com | Home | 9.45 (1.37), 51.95 | iCBT, (BRAVE for Children—ONLINE) | 40 | 10 + 2 Boost Par:6 | Y | Y | A: ADIS-C (Y, O), /-P CSR (P, O),SCAS-C (Y, SR), /-P (P, SR), D: CES-D (Y, SR) | N | ++++ | AU |
2009 [ |
||||||||||||||||
WL | 33 | |||||||||||||||
Sethi, | Trans | 15–25 | low/ moderate depr/anx on DASS-21, excl: severe Depr/Anx | N | School | Uni | 19.47 (1.57) 34.21 | iCBT | 9 | 5 | N | Y | D+A: DASS-21 (Y, SR), K, K10 (Y, SR) | N | ??+? | AUS |
2010 [ |
||||||||||||||||
No treatment | 10 | |||||||||||||||
Spence, | Anx | 12–18 | GAD, SoP, SepA, SP on ADIS-C/-P CSR > 3, excl: prim diag of PD, OCD/ PTSD, moderate—severe mood disturbance on ADIS-C > 5 | Y | Com | Home | 13.98 (1.63), 40.87 | iCBT, BRAVE for Children—ONLINE | 44 | 10 + 2 Boost Par: 5 + 2 Boost | Y | Y | A: ADIS-C (Y, O) /-P CSR (P, O), SCAS-C (Y, SR), /-P (P, SR) | N | ++++ | AU |
2011 [ |
||||||||||||||||
WL | 27 | |||||||||||||||
Stallard, | Trans | 11–16 | GAD, SP, SoP / PD, excl: PTSD; mild/moderate depr, excl: severe depr | N | Clin | Home | Median 12, 66.7 | cCBT (Think, Feel, Do) | 10 | 6 | N | Y | A: SCAS (Y, SR), D: AWS (Y, SR) | N | +++- | UK |
2011 [ |
||||||||||||||||
WL | 10 | |||||||||||||||
Stasiak, | Dep | 13–18 | CDRS-R > 29, RADS-2 > 75 | N | School | School | 15.2 (1.6) 59 | cCBT (The Journey) | 17 | 7 | N | N | D: CDRS-R (Y, O), RADS-2 (Y, SR) | 1M | ++++ | NZ |
2012 [ |
Placebo | 17 | ||||||||||||||
Storch, 2011 [ |
Anx | 7–16 | OCD on ADIS-IV-C/-P CSR > 3, CY-BOCS > 15 | Y | Clin Com | Home | 11.10 (2.59), 61 | cCBT WL | 16 15 | 14 | Y | Y | A: ADIS-IV-C (Y, O) /-P CSR (P, O), CY-BOCS (Y, O), COIS-C (Y, SR) /-P (P, SR), MASC (Y, SR), | N | +?-+ | USA |
D: CDI (Y, SR) | ||||||||||||||||
Tillfors, | Anx | 15–21 | SoP on SPSQ-C, public speaking fears, Depr < 30 on MADRS-S | Y | Com | Home | 16.5 (1.6), 10.53 | iCBT | 10 | 9 | N | Y | A: LSAS-SR (Y, SR); SPS (Y, SR), SIAS (Y, SR), SPSQ (Y, SR), BAI (Y, SR); D: MADRS-S (Y, SR) | N | ??+- | SWE |
2011 [ |
||||||||||||||||
WL | 9 | |||||||||||||||
Wuthrich, 2012 [ |
Anx | 14–17 | Any Anx Dis on ADIS-IV-C/-P CSR | Y | Com | Home | 15.17 (1.11) 37.21 | cCBT, Cool TeensWL | 24 19 | 8 | Y | Y | A: ADIS-C (Y, O) /-P CSR (P, O), SCAS-C (Y, O) /-P (P, O) | N | ++++ | AU |
Van der Zanden, | Dep | 16–25 | CES-D >10 < 45 | N | Com | Home | 20.9 (2.2), 15.6 | iCBT, Master your Mood | 121 | 6 | N | Y | D: CES-D (Y, SR), A: HADS-A (Y, SR) | N | ++++ | NL |
2012 [ |
WL | 123 |
a) Bias: Risk of Bias: In this column a positive or negative sign is given for four risk of bias criteria, respectively: allocation sequence; concealment of allocation to conditions; blinding of assessors; and intention-to-treat analyses. A? is given if there were not enough information to judge.
ADIS-C: Anxiety Disorders Interview Schedule for Children; ADIS-IV-C: for DSM-IV: Child Version; ADIS-IV-P: Parent Version; ADIS-P: Parent Version; Anx: Anxiety; Attention: Attention Control; AU: Australia; AWS: Adolescent Well Being Scale; BAI: Beck Anxiety Inventory; Boost: Booster Sessions; CBT: Cognitive Behaviour Therapy; cCBT: Computerized CBT; CDI-S: Children´s Depression Inventory, Short Form; CDI: Children's Depression Inventory; CDRS-R: Children’s Depression Rating Scale, Revised; CES-D: Center for Epidemiologic Studies Depression Scale; Clin: Clinical Sample; COIS-C: Child Obsessive Compulsive Impact Scale, Child Version; COIS-P: Parent Version; Com: Community Sample; Cond: Condition; CSR: Clinical Severity Rating; CY-BOCS: Children's Yale-Brown Obsessive-Compulsive Scale; D: Dysthymia; DASS-21: Depression Anxiety Stress Scale Dep: Depression; Diagn: Diagnostic Confirmation, Diagnostic Interview + all participant fulfil disorder (no subclinical); Dis: Disorders; DISC IV-P: Diagnostic Interview Schedule for Children; Excl: Excluded; FU: Follow-Up; GAD: Generalized Anxiety Disorder; Guidance: Therapeutic or Administrative Support by a Professional; HADS-A: Hospital Anxiety and Depression Scale (Anxiety subscale); iCBT: Internetbased CBT; K-SADS-E: Schedule for Affective Disorders and Schizophrenia for School-Age Children, Epidemiological Version; K10: Kessler Psychological Distress Scale; LSAS-SR: Liebowitz Social Anxiety Scale, Self-Report; M: Month: Trans: Transdiagnostic; M. Mania; MADRS-S: Montgomery-Åsberg Depression Rating Scale-Self-Rated; MASC: Multidimensional Anxiety Scale for Children; MD: Major Depression; N: NoN: Number of participants in each condition;NL: Netherlands;Nmod: Number of modules in the intervention;NZ: New Zealand; O: Observerrating (Clinician) OCD = Obsessive Compulsive Disorder; P: Parent Report; Par Inv: Parental Involvement;Par: Parent Sessions; PD: Panic Disorder; Prim Diag: Primary Diagnosis; PTSD: Posttraumatic Stress Disorder; RADS-2: Reynolds ´Adolescent Depression Scale, 2nd Edition; RCMAS: Revised Children ´s Manifest Anxiety Scale; Recr: Recruitment; SCARED-P: Screen for Child Anxiety Related Emotional Disorders, Parent Version; SCAS-C: Spence Children´s Anxiety Scale, Child Version; SCAS-P: Parent Version; SD: Standard Deviation, SepA: Separation Anxiety Disorder; SIAS: Social Interaction Anxiety Scale; SoP: Social Phobia; SP: Specific Phobia; SPS: Social Phobia Scale; SPSQ-C: Social Phobia Screening Questionnaire for Children; SR: Self Report; SWE: Sweden; UK: United Kingdom; Uni: University Lab; USA: United States of America; WL: Waiting List Control; Y: Yes; Y: Youth Report.
A non-treatment comparison was used in 10 studies, the other three studies applied a placebo control. The majority relied on a multi-method outcome assessment approach (i.e., observer-based outcome interview and self-report measure, n = 8). Only five studies used only self-report measures. Outcome informants were both youths and parents in seven studies and youth-only in six studies. Studies using youth-only as outcome informant were mainly directed at adolescents (n = 4), two studies on mixed-age range samples. Many studies included a follow-up assessment. However, only two studies assessed a follow-up in both the intervention and control group [
In most studies, the intervention was delivered via the Internet (n = 8), followed by computer (n = 5). The number of treatment sessions ranged from 6 to 14. Six studies were classified as brief treatment (6–7 sessions) and seven as long treatment (≥ 8 sessions). Parents were involved in the intervention within six studies and the majority of interventions included guidance by a professional (n = 11). Four Studies were conducted in the USA and Australia, respectively, two studies were conducted in New Zealand, two studies in Sweden and one study was conducted in the UK and the Netherlands, respectively. Thus, all studies were conducted in high-income countries. Cohen's kappa for agreement between independent raters was very good (κ = .84).
Overall risk of bias was low. Eleven studies reported an adequate sequence generation. Twelve studies reported blinding of outcome assessors or used only self-report outcomes, whereas five did not report blinding. Only eight studies reported the information needed to clarify whether allocation to conditions was performed by an independent (third) party. Ten studies reported adequate handling of missing data, using intention-to-treat principles. Seven studies met all four quality criteria, four studies met three of four criteria, and two studies met only one criterion. Inter-rater reliability between independent raters on the risk of bias was very good (κ = 0. 94).
The overall mean effect size of cCBT on symptoms of anxiety or depression when compared to a control group at post-test was g = .72 (95% CI: 0.55 to 0.90; p <. 001;
Nco | g | 95% CI | Z | I2 | P |
NNT | ||
---|---|---|---|---|---|---|---|---|
Overall effect (primary outcome) | 13 | 0.72 | 0.55~0.90 | 8.10 |
20.14 (0~58) | 2.56 | ||
Highest effect size removed | 12 | 0.70 | 0.53~0.87 | 8.11 |
15.11 (0~55) | 2.63 | ||
Lowest effect size removed | 12 | 0.80 | 0.64~0.96 | 9.65 |
0 (0~58) | 2.34 | ||
Only studies with all participants ≤18 | 10 | 0.61 | 0.43~0.80 | 6.52 |
4.55 (0~64) | 2.99 | ||
Overall on depressive symptoms | 11 | 0.56 | 0.31~0.82 | 4.34 |
53.88 |
3.25 | ||
Overall on anxiety symptoms | 10 | 0.65 | 0.40~0.90 | 5.13 |
52.09 |
2.82 | ||
Target condition | Anxiety | 7 | 0.68 | 0.45~0.92 | 5.68 |
0 (0~71) | .77 | 2.70 |
Depression | 4 | 0.76 | 0.41~1.12 | 4.21 |
61.42 |
2.44 | ||
Transdiagnostic | 2 | 0.94 | 0.23~1.66 | 2.60 |
0 |
2.02 | ||
Age group | Children | 3 | 0.56 | 0.21~0.91 | 3.13 |
0 (0~90) | .007 | 3.25 |
Adolescents | 6 | 0.95 | 0.76~1.17 | 9.22 |
0 (0~75) | 2.01 | ||
Mixed | 4 | 0.46 | 0.22~0.70 | 3.75 |
0 (0~85) | 3.91 | ||
Confirmation of Disorder | Confirmed diagnosis | 6 | 0.71 | 0.44~0.99 | 5.07 |
13.67 (0~78) | .75 | 2.63 |
Anxiety/depressive symptoms | 7 | 0.74 | 0.49~1.00 | 5.70 |
34.99 (0~73) | 2.50 | ||
Risk of bias score | Low (4) | 7 | 0.77 | 0.59~0.95 | 8.34 |
0 (0~71) | .97 | 2.42 |
Some risk (< 4) | 6 | 0.72 | 0.37~1.07 | 4.06 |
33.55 (0~73) | 2.56 | ||
Parental Involvement | No | 7 | 0.83 | 0.53~1.13 | 5.47 |
45.40 |
.33 | 2.26 |
Yes | 6 | 0.64 | 0.40~0.88 | 5.14 |
0 (0~77) | 2.86 |
Note: N comp, Number of comparison;
a This p values indicate weather differences between subgroups are significant
c calculation of 95% CI not possible because dfs are 1.
O: p <. 1;
* p <. 05;
*** p <. 001.
Excluding studies that also assessed participants above 18 years of age [
Including also anxiety outcomes of interventions targeting depression, the effect size on anxiety was g = 0.65 (95% CI 0.40 to 0.90) with an NNT of 2.82. Including also depression outcomes of interventions targeting anxiety resulted in a pooled effect size of g = .56 (95% CI 0.31 to 0.82), NNT = 3.25.
We conducted a series of subgroup analyses (
The inspection of the funnel plot and Egger’s test indicated some possible publication bias. However, after adjustment for missing studies using the Duval-Tweedie trim and fill procedure (four imputed studies) Hedge’s g for the main outcome analyses was 0.64 (95% CI 0.46 to 0.82), corresponding to an NNT of 2.86.
We found cCBT for youth was associated with significant and moderate to large effects on symptoms of anxiety and depression, with an NNT of 2.56. Overall, risk of bias was low and the seven studies that met all quality criteria showed a significant effect comparable with the overall effect size. Heterogeneity was low in most analyses, suggesting that most studies pointed in the same direction, with no major outliers. We found some indications for publication bias, but adjusting effect sizes resulted in no major changes.
Effects sizes were slightly lower than those found for cCBT for anxiety and depressive disorders in adults (g = 0.88) [
We also found significant, moderate but somewhat smaller effects compared to the main-outcome analysis on symptoms of depression when including also outcomes of interventions targeting anxiety. Likewise, we also found significant and moderate effects on symptoms of anxiety when also including outcomes of interventions targeting depression. These findings are consistent with growing evidence indicating that (a) youth depression and anxiety are closely associated [
Participant age was the only significant moderator of treatment effects. The finding was consistent also when using different classifications of children (< 13 vs. ≤ 13 years of age). In line with the results of the most recent meta-analyses on psychotherapy for anxiety disorders in youth [
We found no association between the inclusion of parents and better treatment outcomes. Although it is often argued that parental involvement is important in the treatment of anxiety and depressive disorders in youth [
When interpreting results of this study, several limitations should be considered.
First, 11 of 13 studies reviewed reported no follow-up assessment with treatment versus control comparison. Hence, we could not examine long-term effects of treatments. This shortcoming points to the need for future studies with follow-up assessments. Second, the interventions evaluated in the studies were very heterogeneous regarding format of treatment, ranging from therapist focused group-based chat interventions based on face-to-face treatment manuals [
Findings from the present study provide evidence for the efficacy of cCBT in the treatment of anxiety and depressive symptoms in youth. Hence, such intervention might be a promising treatment alternative when evidence based face-to-face treatment is not available or simply not wished. Applying cCBT at large scale may thus serve as one (out of several) strategies to bridge the enormous gap between the need and provision of evidence-based treatments for anxiety and depression in children and adolescents. Given that access to evidence-based psychotherapeutic treatment is not only a problem in high, but even more in low- and middle-income countries [
(DOC)
(DOC)