Internet and Computer-Based Cognitive Behavioral Therapy for Anxiety and Depression in YouthA Meta-Analysis of Randomized Controlled Outcome Trials

Roughly eight in ten children and adolescents with anxiety or depression never receive any treatment. Sit with that number for a moment — not as a statistic, but as a waiting room that never fills because the kids never show up. The treatments exist. The evidence is there. The problem is the distance between the child and the care. So, the question Ebert and colleagues set out to answer is a simple and urgent one: what if the therapy could go to them instead? Anxiety and depression in young people are not minor or transient. Ebert and colleagues open by noting that these disorders frequently co-occur, carry substantial burdens, increase the risk of further mental health problems, and often persist into adulthood. That persistence is what makes the treatment gap especially costly. Effective care early in life can alter a trajectory that otherwise carries forward for decades. Yet the barriers to that care are real. The authors cite stigma, discomfort discussing mental health problems, a preference for self-help, and a simple shortage of available services. These aren't problems of clinical knowledge — cognitive behavioral therapy, or CBT, is well established as the treatment of choice for youth anxiety and depression. The gap is a problem of access and uptake, and it's systemic. That's where computer- and Internet-based CBT — which the authors call cCBT — enters the picture. The premise is straightforward: take the same evidence-based CBT principles — identifying unhelpful thought patterns, building behavioral strategies, practicing coping skills — and deliver them through a digital platform. Ebert and colleagues describe the practical appeal clearly. cCBT can be accessed at any time and place, provides a degree of anonymity that may reduce stigma, allows self-pacing, and eliminates travel time and cost. For younger, Internet-savvy generations, the interactivity of these platforms may itself be an asset. By 2013, a solid evidence base existed for cCBT in adults. One review of 19 randomized trials found a mean effect size around 0.56 for adult depression; a meta-analysis of 27 trials on adult anxiety found a pooled mean effect near 0.90. But for children and adolescents, the picture was much thinner. Earlier reviews had identified only three randomized trials in youth. Ebert and colleagues set out to build a more systematic picture. Their method was carefully specified. They searched PubMed, the Cochrane Controlled Trials Register, and PsycINFO up to December fourth, 2013, looking for randomized controlled trials — the gold standard for causal evidence — that compared a computer, Internet, or mobile-delivered CBT intervention against a control condition in young people up to age 25. Thirteen trials met the criteria, enrolling 796 participants in total: 420 in active treatment and 376 in control groups. Ten comparisons used a no-treatment waitlist control; three used a placebo comparator. Two independent assessors handled data extraction and risk-of-bias coding, reaching 100 percent agreement on inclusion decisions and near-perfect reliability on bias items, with a kappa of 0.94. The primary effect-size metric was Hedges' g — a standardized measure that expresses how much the treatment group improved relative to the control group, in units of the study's own variability. A g around 0.2 is typically considered small, 0.5 medium, and 0.8 large. The authors also converted each effect size into a number needed to treat, or NNT — how many young people must receive the intervention for one additional person to benefit beyond what the control would have produced. And to account for variability across the diverse set of trials, they used a random-effects pooling model, which assumes the studies are estimating a distribution of related effects rather than a single fixed truth. The headline result is this: across all 13 trials, the overall effect of cCBT on symptoms of anxiety or depression was Hedges' g equal to 0.72, with a 95 percent confidence interval from 0.55 to 0.90, and an NNT of 2.56. That's a moderate-to-large effect. For every two and a half young people treated, one additional person benefits beyond what the control condition would have produced. And critically, heterogeneity was low — I squared equal to 20.14 percent, with a confidence interval spanning zero to 58 percent. In plain terms, the studies pointed in the same direction with limited variation between them. That consistency strengthens the finding. The breakdown by condition is worth dwelling on. Seven trials targeted anxiety and returned a pooled g of 0.68 — consistent and precise, with I squared essentially at zero, meaning the anxiety results were strikingly uniform across studies. Four trials targeted depression and showed a slightly larger effect, g equal to 0.76, but with noticeably more between-study variability: I squared equal to 61.42 percent. That higher heterogeneity means the depression subgroup is less stable — some studies showed larger benefits, some smaller — and the single pooled number should be read with more caution. Two trials took a transdiagnostic approach, targeting both anxiety and depression simultaneously, and produced the largest effect of all: g equal to 0.94, with an NNT of just over two. Age emerged as a meaningful moderator. Adolescent-focused trials produced the strongest effects — g equal to 0.95, NNT equal to 2.01 — with zero heterogeneity. Child-only trials showed more modest effects: g equal to 0.56, NNT equal to 3.25. Mixed-age samples landed lower still, at g equal to 0.46, NNT equal to 3.91. The difference between adolescent and child subgroups was statistically significant, with a p-value of 0.007. Whether this reflects something about adolescent engagement with digital formats, cognitive readiness for CBT techniques, or simply the nature of the studies in each subgroup is an open question — but the pattern is consistent enough to be meaningful. The authors also examined whether publication bias could be inflating these numbers. Funnel plots and Egger's test suggested some asymmetry. Applying the Duval-Tweedie trim-and-fill procedure — which imputes hypothetical unpublished studies to correct for that asymmetry — added four studies and reduced the overall effect to g equal to 0.64, with an NNT of 2.86. That's a real reduction, and it matters. Even after adjustment for potential publication bias, the effect remains moderate and clinically meaningful. Sensitivity analyses confirmed stability: restricting the sample to participants 18 and under brought the pooled g to 0.61, and removing any single trial left the overall pattern intact. Now the limitations, and Ebert and colleagues are direct about them. Thirteen trials is a small base of evidence. Ten of the thirteen compared cCBT to a waitlist — meaning the question being answered is primarily whether digital therapy beats nothing, not whether it matches face-to-face care. That's a lower bar, and the results should be interpreted accordingly. Long-term evidence was scarce; only two trials assessed follow-up outcomes in both groups simultaneously, so whether effects persist over months or years simply cannot be determined from this data. Most depression trials excluded young people with severe symptoms, limiting generalizability to the kids who may need help most. All studies came from high-income countries, so the findings say nothing definitive about settings with different resources or different cultural contexts. Future research, the paper argues, needs active treatment comparators, longer and balanced follow-ups, and trials in lower-income settings. Return to where we started. Eight in ten kids with anxiety or depression receive no care. The meta-analysis by Ebert and colleagues shows that cCBT produces real, consistent effects — g equal to 0.72 overall, holding at g equal to 0.64 even after correcting for publication bias — in a population that, in the absence of digital options, would largely go untreated. The authors frame this not as a replacement for human care but as a scalable layer of access: available at any hour, approachable without a waiting room, and usable without disclosing to anyone that you're struggling. One way to think about it is as infrastructure — the kind that can reach a child at two in the morning, in a town with no therapist, in a family where seeking help still carries a stigma. The evidence isn't yet complete. But it's enough to make the question of what we do next feel genuinely consequential. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.