Effectiveness of Unguided Internet-Based Cognitive Behavioral Therapy for Subthreshold Social Anxiety Disorder in Adolescents and Young Adults: Multicenter Randomized Controlled Trial

Social anxiety disorder (SAD) is characterized by excessive anxiety and fear of negative evaluation in social situations, with onset typically in adolescence and substantial functional impacts on relationships, academics, and career. Subthreshold social anxiety symptoms are common among youths and confer a markedly increased risk of developing full-blown anxiety disorders. Given barriers to face-to-face CBT for adolescents—such as fear of therapist interaction, limited access, and resource constraints—unguided internet-based CBT (ICBT) may provide an accessible early intervention. This study investigates whether fully self-help, unguided ICBT can reduce social anxiety symptoms and is acceptable to adolescents and young adults with subthreshold SAD, potentially preventing progression to clinically diagnosed SAD.

CBT is an established, effective treatment for SAD in adults and youth by targeting maladaptive cognitive-behavioral cycles (Clark and Wells model). Access barriers limit face-to-face delivery, especially in Japan. Internet-based CBT can overcome logistical obstacles; guided ICBT has shown efficacy for diagnosed SAD, and unguided programs may be particularly acceptable to socially anxious individuals due to reduced need for therapist interaction. Prior unguided ICBT studies, including mobile delivery, report moderate to large effects (e.g., Cohen d around 0.8 for LSAS) and promising adherence, though high dropout is typical for self-help. Evidence on unguided ICBT for subthreshold SAD is lacking; treatment response may differ from diagnosed SAD due to lower impairment. Early intervention for subthreshold anxiety may prevent onset of full disorders, supporting investigation of unguided ICBT in high-risk adolescent and young adult populations.

Design: Multicenter randomized controlled trial (RCT) conducted in Japan from November 2022 to October 2023, reported per CONSORT for nonpharmacologic trials. Trial registration: UMIN000050064. Ethics: Approved by Kagoshima University Hospital Clinical Research Management Center (ID 220196). Written informed consent obtained from all participants; for high school students, guardian consent also required. Setting and participants: Students from 6 universities and 1 high school. Inclusion: age 15-25 years, LSAS total score ≥30, owning a smartphone. Exclusion: diagnosed psychiatric disorders (e.g., depression), CBT within past 2 years, IQ <85, imminent suicide risk, serious medical conditions (e.g., cancer). Sample and randomization: 89 consented; 77 eligible and randomized 1:1 to intervention (ICBT) or control (no treatment) using minimization with randomization generator, adjusting for LSAS (<50 or ≥50), sex, and facility. Blinding: None. Intervention: Fully self-help unguided ICBT delivered via smartphone on an e-learning platform (learningBOX). Program based on Clark and Wells model, comprising 10 text-based modules: (1) psychoeducation and case formulation; (2) safety behaviors and self-focused attention experiments; (3) video feedback to correct negative self-image; (4) attention training in social situations; (5) behavioral experiments testing negative predictions; (6) opinion survey follow-up to update assumptions; (7) handling anticipatory worry and postevent rumination; (8) image description to update self-images; (9) schema work targeting dysfunctional beliefs; (10) relapse prevention and coping strategies. One module per week recommended; automated email reminders on Monday and Saturday. Control: no treatment; asked to refrain from accessing CBT information. Compensation: University participants received coupons (~JP ¥5000) for data submissions (baseline and start for intervention; baseline and 10 weeks for control). High school participants received no remuneration. Outcomes: Primary—self-rated Liebowitz Social Anxiety Scale (LSAS) total score. Response: ≥28% LSAS reduction; remission: LSAS <35; worsening: ≥28% LSAS increase. Secondary—Social Phobia Inventory (SPIN), Patient Health Questionnaire-9 (PHQ-9), Generalized Anxiety Disorder-7 (GAD-7), EQ-5D-5L (QALY). Recovery (IAPT-style): SPIN <19 and PHQ-9 <10 simultaneously. Sample size: Estimated 78 using G*Power (α=.05, power=80%, effect size=0.8), accounting for ~50% dropout; minimum required n=52. Statistical analysis: Baseline characteristics compared via independent two-tailed t tests. Missing data imputed using mice in R 4.3.2. ANCOVA assessed group differences in change scores for primary and secondary outcomes with baseline PHQ-9 as covariate. Fisher exact tests computed ORs (95% CIs) for response, remission, recovery, and worsening. Risk metrics for worsening: risk ratio (RR), relative risk reduction (RRR), absolute risk reduction (ARR), and number needed to treat (NNT). Significance threshold α=.05.

Participants: 77 randomized (ICBT n=38; control n=39). Analysis included ICBT n=31 and control n=38 after attrition. Intervention adherence: dropout 9% (3/34), implementation 91% (31/34), completion 65% (22/34). Primary outcome (LSAS): Greater reduction in LSAS from baseline to postintervention in ICBT vs control by mean difference 11.62 (95% CI 1.67-21.56; F=3.91; P=.02; Hedge g≈0.64). Pre- to postintervention change in ICBT exceeded control by 9.39 (95% CI 1.31-17.48; F=4.65; P=.01; Hedge g≈0.66). Secondary outcomes: SPIN—greater reduction baseline to post in ICBT vs control by 8.05 (95% CI 3.02-13.08; F=5.39; P=.007; Hedge g≈0.80); pre- to post difference 8.26 (95% CI 3.75-12.78; F=7.23; P=.01; Hedge g≈0.93). PHQ-9—differences not significant (baseline to post MD 1.15; F=1.19; P=.31; pre- to post MD 1.75; F=2.96; P=.06). GAD-7—baseline to post not significant (MD 1.24; F=1.61; P=.21), but pre- to post significant (MD 1.81; 95% CI 0.31-3.33; F=4.15; P=.02; Hedge g≈0.32). EQ-5D-5L—no significant differences. Clinically significant change: Response—61% (19/31) ICBT vs 24% (9/38) control; OR 4.97 (95% CI 1.61-16.53); P=.003. Recovery (SPIN<19 and PHQ-9<10)—68% (21/31) vs 34% (13/38); OR 3.95 (95% CI 1.32-12.56); P=.008. Remission (LSAS<35)—39% (12/31) vs 24% (9/38); OR 2.01 (95% CI 0.64-6.60); P=.20 (ns). Worsening—6% (2/31) vs 24% (9/38); OR 0.23 (95% CI 0.02-1.23); P=.10 (ns). Risk metrics for worsening: RR 0.26, RRR 73%, ARR 17%, NNT ≈6. Safety: No severe adverse events reported.

Unguided, fully self-help ICBT significantly improved social anxiety symptoms among adolescents and young adults with subthreshold SAD compared to no treatment, with moderate effect sizes comparable to meta-analytic effects for face-to-face CBT in SAD. The program, based on the Clark and Wells model, was well accepted, with high implementation and reasonable completion rates, supporting feasibility for youths who may avoid therapist interaction. Significant improvements were observed in LSAS and SPIN, with some evidence of benefit for general anxiety (GAD-7) on pre- to post measures, while depressive symptoms (PHQ-9) and QOL did not show significant between-group differences in the ANCOVA baseline-to-post analysis. Clinically meaningful response and recovery rates favored the intervention, and exploratory risk analyses suggested reduced likelihood of symptom worsening, indicating potential preventive benefits against progression to diagnosed SAD. These findings address the research question by demonstrating that unguided ICBT can effectively reduce subthreshold social anxiety in a population with known barriers to conventional care, thereby offering a scalable, accessible early intervention. Results align with prior ICBT efficacy for diagnosed SAD and suggest that removing therapist guidance does not preclude benefit when engagement is adequate. Given the high prevalence of adolescent anxiety and low help-seeking, unguided ICBT may fill critical gaps in youth mental health services.

This multicenter RCT shows that unguided, complete self-help ICBT delivered via smartphones reduces social anxiety symptoms in adolescents and young adults with subthreshold SAD and is acceptable and safe. The intervention achieved higher response and recovery rates than no treatment and may reduce risk of symptom deterioration, supporting its use as an early, accessible strategy in school and university settings. Future research should incorporate longer-term follow-up, larger high-risk samples, blinded assessments, and active control conditions (e.g., sham apps) to evaluate durability of effects and prevention of progression to clinically diagnosed SAD, including cohort designs to assess onset risk over several years.

Unblinded self-reported primary outcome (LSAS) may introduce expectancy bias; no-treatment control limits control for nonspecific effects and blinding; short-term study duration precludes conclusions about medium- to long-term efficacy and prevention of SAD onset; analysis set had attrition and uneven baseline depressive symptoms, though adjusted in ANCOVA; generalizability may be limited to Japanese student populations with smartphone access.