Table 2.

Key findings of the included studies.

Reference	Effectiveness	Contributing factors	Cost-effectiveness	Inclusion of data on low- and middle- income countries	Quality of included studies	Quality of review (AMSTARa)
Barnes et al (2018) [28]	Although early findings suggest that therapeutic games have the potential to lead to clinically measurable reductions in symptoms in adolescents with anxiety, evidence on the effectiveness is extremely limited. On the basis of 2 RCTsb included in this review, no difference in anxiety outcomes is found between the intervention and control groups (alternative nontherapeutic videogame).	Not discussed	Not discussed	Limited (China and Hong Kong)	Mean rating of 75% using mixed methods appraisal tool. Only 2 RCTs included in the review.	Critically low
Bry et al (2018) [29]	Evidence-based treatment content within consumer smartphone apps marketed for child and adolescent anxiety is scant, and only a few comprehensive anxiety self-management apps are identified. Half of the sampled apps for anxiety include any evidence-based treatment component, and 23% included two or more evidence-based components.	Not discussed	Low cost but effectiveness unknown	No	N/Ac	N/A
Grist et al (2017) [30]	Authors conclude that there is currently no evidence to support the effectiveness of apps for adolescents with mental health problems. In 2 RCTs on mobile app for depression, anxiety, and stress, no significant effect is found between intervention (app with self-monitoring) and control (no self-monitoring) groups. Acceptability is generally rated average to high, with adherence ranging from 65% to 83%.	Specific factors: privacy, safety, discretion, and data security; credibility of design and visual appearance; engaging and interactive content; concise, interesting, and trustworthy information; reminders to use; and personalization allowed	Not discussed	No	Issues with quality, including small sample size. Only 2 small RCTs included in the review, both without adequate control group.	Critically low
Hollis et al (2017) [24]	cCBTd provides clinical benefits for depression and anxiety when compared with inactive control (waitlist). The benefits for attention deficit hyperactivity disorder and autism are inconsistent, for psychosis are unknown, and eating disorders are no better than waitlist control in regard to symptomology.	Self-guided cCBT has poor uptake and adherence. Human involvement is positively associated with adherence. Adolescents and young people prefer face-to-face over web-based interventions. Specific factors: privacy, safety, discretion, and anonymity; providing concise, interesting, and trustworthy information; and ability to complete interventions on own terms and pace.	Authors note a considerable lack of evidence	Limited (China)	Most studies (18/21) rated as moderate quality, 2 rated as low quality, and 1 rated as high quality using AMSTAR. Methodological issues and high level of heterogeneity in the included studies.	Critically low
Davies et al (2014) [31]	Web-based and computer-delivered interventions are found effective in improving students’ depression (pooled SMDe −0.43; 95% CI −0.63 to −0.22; P<.001), anxiety (pooled SMD −0.56; 95% CI −0.77 to −0.35; P<.001), and stress (pooled SMD −0.73; 95% CI −1.27 to −0.19; P=.008) outcomes when compared with inactive controls (no treatment, waitlist). When compared with active controls (alternative materials), no benefits are found for depression, anxiety, and stress.	Not discussed	Not discussed	No	A moderate risk of bias. Quality issues with reporting of methodology, data, and outcome measures. Only 3 studies with active control, with reported skewed data. Heterogeneity of interventions.	Moderate
Farrer et al (2013) [32]	Approximately half (24/51) of the technology-based mental health interventions targeting tertiary students with anxiety or depression are associated with at least one significant positive outcome, and approximately one-third (15/51) fail to find a significant effect. Effect size for interventions targeting symptoms of depression and anxiety range from −0.07 to 3.04 (median 0.54; depression=0.48; anxiety=0.77). Effect size for interventions targeting symptoms of anxiety range from 0.07 to 2.66 (median 0.84). cCBT was the most deployed therapy in 25 of 51 of the interventions.	Not discussed	Included studies do not report cost-effectiveness	Limited (China)	Mean rating 4.42 out of 9 using Cochrane Effective Practice and Organisation of Care Group. Methodological issues with reporting on randomization, intended outcomes, and heterogeneity of interventions. Insufficient data in more than half of the studies (14/27) to calculate effect sizes.	Low
Valimaki et al (2017) [33]	Web-based mental health interventions yield statistically significant effect on depressive (P=.02; median 1.68; 95% CI 3.11 to 0.25) and anxiety symptoms (P<.001; median 1.47; 95% CI 2.36 to 0.59) when compared with control group (type not specified), but not on stress (P=.14; median 1.06; 95% CI 2.44 to 0.33). After 6 months of intervention, significant improvement is found on depressive symptoms (P=.01; median 1.78; 95% CI 3.20 to 0.37), on anxiety symptoms (P<.001; median 1.47; 95% CI 2.36 to 0.59), and on moods and feelings (P=.04; median 5.55; 95% CI 10.88 to 0.22). Dropout of those in intervention groups was higher than those in control groups.	Interventions with human elements, such as face-to-face guidance or telephone follow-ups, are associated with adherence and effect.	Included studies do not assess costs. Authors note a considerable lack of evidence	Limited (China)	Some risk of bias using Review Manager. Issues include biases related to attrition rates, selective reporting, and small sample sizes. Mixed control groups.	High
Harrer et al (2019) [34]	Internet interventions for university students’ mental health have a small effect on anxiety (Hedges g=0.27; 95% CI 0.13 to 0.40), depression (Hedges g=0.18; 95% CI 0.08 to 0.27), and stress (Hedges g=0.20; 95% CI 0.02 to 0.38) when compared with nonactive controls. Moderate effects were found on eating disorder symptoms (Hedges g=0.52; 95% CI 0.22 to 0.83) and role functioning (Hedges g=0.41; 95% CI 0.26 to 0.56). Effects on well‐being are nonsignificant (Hedges g=0.15; 95% CI −0.20 to 0.50).	Guidance does not significantly affect intervention efficacy (P≥.05).	Not discussed	Limited (Romania)	Half of the studies with high risk of bias. Moderate to substantial level of heterogeneity and selective reporting.	Low
Garrido et al (2019) [35]	Digital interventions work better than no intervention (Cohen d=0.33; 95% CI 0.11 to 0.55) but not better than active alternatives (alternative web-based materials; Cohen d=0.14; 95% CI −0.04 to 0.31) in improving depression in young people, when results of different studies are pooled together. Most interventions were based on CBTf. Authors conclude that interventions may be clinically significant only if supervised. Engagement and adherence rates are low.	Interventions with supervision have a higher pooled effect size than those without supervision (studies with no intervention controls: Cohen d=0.52; 95% CI 0.23 to 0.80 and studies with active controls: Cohen d=0.49; 95% CI −0.11 to 1.01). Specific factors: credibility of design and visual appearance; concise, interesting, and trustworthy resources; engaging and interactive tools and content; esthetically attractive; relatable situations, characters, or avatars; and reflect local and cultural differences and needs. Technical glitches as a barrier to complete interventions.	Not discussed	Limited (China, Hong Kong, and Chile)	On the basis of Joanna Brigg Institute appraisal tool and CONSORT (Consolidated Standards of Reporting Trials), 32 of 41 studies with high or unclear overall bias and 9 of 41 with low overall bias.	Low
Pretorius et al (2019) [36]	N/A	Young people value web-based services because of anonymity, accessibility, self-reliance, and ease of use. Theoretical frameworks, including self-determination theory and help-seeking model, should be deployed in research. Specific factors: anonymity, privacy, safety, and discretion; site moderation by professionals; credibility of design and visual appearance; concise, interesting, and trustworthy information; esthetically attractive; flexibility, self-reliance, and control; and 24-h availability.	Not discussed	No	Moderate to strong using Critical Appraisal Skills Program. Heterogeneity of interventions. Only 1 RCT included in the review.	Critically low
Ridout et al (2018) [37]	Social networking sites targeting mental health have significant improvement in mental health knowledge and a number of depressive symptoms in young people, but no improvement in anxiety or psychosis symptoms. The results are not compared with a control group.	Young people value involvement of professionals and peers in social networking sites.	Authors conclude that web-based interventions are cost-effective but provide no evidence	Limited (China and Hong Kong)	No quality assessment performed. On the basis of descriptive studies, no RCTs included in the review.	Critically low
Podina et al (2016) [38]	cCBT is as effective as standard CBT (Hedges g=0.295) and more effective than waitlist (Hedges g=1.410) in reducing anxiety symptoms in anxious children and adolescents.	Not discussed	Not discussed	No	No quality assessment performed. No publication bias found. Only 8 RCTs included in the review.	Critically low
Ebert et al (2015) [39]	cCBT for youth is associated with significant moderate to large effects on symptoms of anxiety (Hedges g=0.68; 95% CI 0.45 to 0.92; P<.001) and depression (Hedges g=0.68; 95% CI 0.45 to 0.92; P<.001) in comparison with nonactive controls. Effect size on symptoms of anxiety or depression for cCBT was similar to face-to-face CBT (Hedges g=0.72 vs Hedges g=0.66) and higher than face-to-face CBT targeting depression (Hedges g=0.35).	No association between parental involvement and better outcomes (without parental involvement: Hedges g=0.83; 95% CI 0.53 to 1.13; P<.001; NNTg=2.26 and with parental involvement: Hedges g=0.64; 95% CI 0.40 to 0.88; P<.001; NNT=2.86)	Not discussed	No	Low risk of bias overall. Low heterogeneity	Low
Pennant et al (2015) [40]	cCBT has positive effects for symptoms of anxiety (SMD 0.77; 95% CI 1.45 to 0.09; n=6; number of participants=220) and depression (SMD 0.62; 95% CI 1.13 to 0.11; n=7; number of participants=279) for young people with risk of diagnosed anxiety and depression disorders. cCBT has lower effect size on anxiety (SMD 0.15; 95% CI 0.26 to 0.03; number of participants=1273) and depression (SMD 0.15; 95% CI 0.26 to 0.03; number of participants=1280) in the general population. Evidence for interventions other than cCBT is sparse and inconclusive.	Not discussed	Not discussed	Limited (China)	On the basis of Grading of Recommendations, Assessment, Development and Evaluation evidence quality review, most studies rated from very low (1/17) to low (11/17) to moderate (5/17). Heterogeneity associated with number of outcomes.	Critically low
Ye et al (2014) [41]	When compared with inactive controls, cCBT is effective in reducing anxiety symptoms (SMD −0.52; 95% CI −0.90 to −0.14) but not depression (SMD −0.16; 95% CI −0.44 to 0.12). No significant difference is found when compared with standard face-to-face CBT, suggesting it is as effective.	Not discussed	Included studies do not report on cost-effectiveness	No	On the basis of Quality Assessment Tool for Quantitative Studies, studies rated high (3/7) and moderate (4/7) quality. Only 7 RCTs included in the review.	Critically low
Grist et al (2019) [42]	A small effect (n=8; Hedges g=0.41; 95% CI 0.08 to 0.73; P<.01) is found in technology-delivered mental health interventions related to attention bias modification when compared with waitlist controls. Although cCBT interventions yield a medium effect size, attention bias modification programs yield a small effect size, and cognitive bias modification programs yield no effect size.	Therapist support (Cochran Q=27.28; P<.001) as well as parental involvement (Cochran Q=24.43; P<.001) have a significant effect on effectiveness of and adherence to an intervention. Therapist involvement yields a higher effect size (n=9; Hedges g=0.87; 95% CI 0.68 to 1.06; P<.001) than predominantly or purely self-administered interventions.	Authors note a considerable lack of evidence	Limited (China)	Most studies rated as low quality and unclear risk using Cochrane Risk of Bias Tool. Most studies (29/34) conducted by program developer. Methodological limitations, small sample size, and nonblinding participants.	Low
Vigerland et al (2016) [43]	cCBT yields moderate effects when compared with waitlist controls (Hedges g=0.62; 95% CI 0.41 to 0.84).	Not discussed	Authors note a considerable lack of evidence	No	Quality varied largely across the studies; Moncrieff mean 30.2 of 46. Heterogeneity of measures included.	Low
Clarke et al (2015) [44]	There is some evidence that skills-based interventions presented in a module-based format can have a significant impact on promoting adolescent mental health and that cCBT has significant positive effects on adolescents’ anxiety and depression symptoms; however, research is limited. Improvements of symptoms are maintained at 6 and 12 months.	Face-to-face and web-based support are associated with improved program completion and outcomes.	Not discussed	Limited (China)	On the basis of Quality Assessment Tool for Quantitative Studies, quality varied significantly from weak (12/20) to moderate or strong (7/20). Issues include a small number of studies, poor sampling, and heterogeneity across interventions.	Low

^aAMSTAR: A Measurement Tool to Assess Systematic Reviews.

^bRCT: randomized controlled trial.

^cN/A: not applicable. This is a systematic review of apps and not studies, and therefore, quality assessment is not applicable.

^dcCBT: computerized cognitive behavioral therapy.

^eSMD: standardized mean difference.

^fCBT: cognitive behavioral therapy.

^gNNT: number needed to treat.