A systematic review and meta-analysis of transdiagnostic cognitive behavioural therapies for emotional disorders

Emotional disorders are highly prevalent and comorbid, sharing cognitive, neuropsychological and genetic processes. Transdiagnostic CBT (TD-CBT) targets shared mechanisms across diagnostic categories and can be delivered via unified protocols to address comorbidity efficiently. Prior reviews have examined TD treatments across varied settings and definitions (unified vs tailored; group, individual, internet) with mixed conclusions, particularly regarding comparisons with disorder-specific CBT (DS-CBT), generalizability across formats, and durability of effects. Given rapid growth in TD-CBT research, the authors aimed to provide an updated comprehensive systematic review and meta-analysis focused on unified TD-CBT for adult emotional disorders (anxiety and depression outcomes) across individual, group and internet-based settings, comparing TD-CBT against waitlist, treatment-as-usual (TAU), DS-CBT, and other active interventions at posttreatment and follow-ups. The key research question: In adults with emotional disorders, what is the effect of unified TD-CBT on anxiety and depression relative to these control conditions at posttreatment and at 3, 6, 12, and 24 months follow-up?

Previous meta-analyses have varied in scope and inclusion: some focused on face-to-face anxiety treatments, others compared TD to DS-CBT, and several investigated unified and tailored internet-based interventions or specific TD protocols like the Unified Protocol (UP). A recent meta-analysis aggregated transdiagnostic treatments for depression but did not focus exclusively on CBT nor include anxiety-focused transdiagnostic treatments. Earlier comprehensive reviews including all settings were limited by small numbers of RCTs or inclusion of only clinician-guided internet interventions, omitting self-guided formats. Overall, prior work suggested large pre- to posttreatment effects for unified TD treatments but left unresolved questions about equivalence to DS-CBT, comparability across delivery formats, and long-term outcomes. This review targets these gaps by restricting to unified TD-CBT, including all formats (individual, group, guided and unguided internet), and emphasizing both anxiety and depression outcomes in RCTs.

Design and registration: Systematic review and meta-analysis following PRISMA and Cochrane Handbook; protocol registered on PROSPERO (CRD42019141512). Ethics approval was not required. Search strategy: Databases searched up to 16 June 2023 included PubMed, MEDLINE, PsycINFO, Google Scholar, medRxiv (including bioRxiv) and OSF Preprints. Search combined terms for 'transdiagnostic', 'CBT', 'emotional disorder', and 'RCT', including MeSH terms and commonly used terms (e.g., 'unified'). Reference lists were screened. Timeframe for inclusion: January 2000–June 2023. Eligibility criteria: Adults seeking treatment with at least one clinician-established diagnosis of an emotional disorder (SAD, panic disorder, agoraphobia, GAD, OCD, unipolar depressive disorders, adjustment disorders). Interventions were unified TD-CBT protocols delivered individually, in groups, or internet-based (with/without guidance), including UP, false safety behavior elimination therapy, emotion regulation therapy, affect regulation training, transdiagnostic behavior therapy, and common-elements CBT approaches targeting shared mechanisms (e.g., cognitive restructuring). Exclusions: non-CBT approaches (e.g., psychodynamic, process-experiential), standalone mindfulness-based, metacognitive therapy, ACT; tailored/modularized treatments; protocols focused on phenomena not considered shared mechanisms (e.g., self-worth, loneliness). Comparators: waitlist control, TAU, DS-CBT, or other active psychological interventions (e.g., behavioral activation). Outcomes: continuous self-report of anxiety and/or depression pre- and posttreatment and at follow-up (if available). Design: RCTs. Study selection and data extraction: Two independent reviewers screened titles/abstracts and full texts, resolved disagreements by consensus. Data were requested from authors when missing (35 contacted; 69% responded). Extracted means and SDs at pre, post, and grouped follow-ups (3, 6, 12, 24 months). When multiple measures existed, the study’s primary outcome or the most common measure across studies (e.g., BAI, GAD-7 for anxiety; BDI-II, PHQ-9 for depression) was used. Studies were grouped by control type and setting. Risk of bias: Assessed with RoB 2.0 across five domains; two raters independently coded with high agreement; visualized via robvis. Statistical analysis: Conducted in R (metafor, meta, dmetar). Controlled effects (relative efficacy) computed as Hedges’ g with 95% CI at posttreatment (and follow-ups) using random-effects models due to anticipated heterogeneity. Heterogeneity assessed via Q and I2. Subgroup analyses by treatment format (individual, group, internet-based) performed when k≥3. Outlier analyses (studies whose 95% CI did not overlap the overall effect’s 95% CI) evaluated impact on heterogeneity. Uncontrolled effects (absolute efficacy) estimated from pre- to post and pre- to follow-ups using dSMc with heteroscedastic variance and retest correlations (from study data if available; otherwise default 0.5, with sensitivity at 0.3 and 0.7). Publication bias assessed with funnel plots, rank correlations (Kendall’s tau), Egger’s tests, and Trim and Fill. PRISMA: 1,069 database records identified (996 screened), 56 RCTs included in review (53 in meta-analysis).

- Study yield and characteristics: 56 RCTs (N=6,916) included in review; 53 RCTs (N=6,705) in meta-analysis. Most studies post-2019; internet-based from 2010 onward. Settings: group (n=21), individual (n=18), internet-based (n=17). Common diagnoses: GAD (79%), SAD (70%), MDD (55%). Mean sessions ~11 (range 4–20). Comparators: waitlist (n=25), TAU (n=18), other active (n=15), DS-CBT (n=8). - Overall controlled posttreatment effects versus controls: Depression g=0.74 (95% CI 0.57–0.92; P<0.001; I2≈88%); Anxiety g=0.77 (95% CI 0.56–0.97; P<0.001; I2≈92%). - By comparator at posttreatment (all formats): • TD-CBT vs waitlist: Depression g=1.23 (0.80–1.66; P<0.001); Anxiety g=1.24 (0.82–1.67; P<0.001). • TD-CBT vs TAU: Depression g=0.90 (0.66–1.14; P<0.001); Anxiety g=0.98 (0.63–1.33; P<0.001). • TD-CBT vs DS-CBT: Depression g=0.09 (−0.07–0.25; P=0.269); Anxiety g=0.09 (−0.01–0.20; P=0.091) — no significant differences; Bayesian analyses corroborated (Depression g=0.09 [−0.12–0.27]; Anxiety g=0.09 [−0.04–0.24]). • TD-CBT vs other active: Depression g=0.27 (0.13–0.42; P<0.001; small); Anxiety g=0.14 (−0.04–0.31; P=0.128; ns). - By format at posttreatment: Individual, group, and internet-based TD-CBT each outperformed waitlist and TAU; effects against DS-CBT generally non-significant, indicating comparability. Formats showed broadly comparable efficacy. - Follow-up controlled effects (vs controls): • 3 months: Depression g=0.55 (0.30–0.80; P<0.001); Anxiety g=0.48 (0.18–0.79; P=0.002). TD-CBT superior to waitlist, TAU; comparable to DS-CBT. • 6 months: Depression g=0.20 (0.10–0.30; P<0.001) for individual format; pooled comparisons show superiority to TAU (Dep g=0.26; Anxiety g=0.32) and some other actives; DS-CBT comparable. • 12 months: Depression g=0.24 (0.13–0.35; P<0.001); Anxiety g=0.22 (0.12–0.32; P<0.001). TD-CBT superior to TAU and some other actives; comparable to DS-CBT. • 24 months: No significant differences vs controls overall (Depression g=0.20 [−0.05–0.46]; P=0.111; Anxiety g=0.14 [−0.02–0.31]; P=0.092). - Heterogeneity: High across analyses (e.g., I2≈88–92% at post), reduced somewhat by format subgrouping and outlier removal; results robust in sensitivity analyses. - Publication bias: Funnel plot asymmetry indicated (Kendall’s tau=0.36–0.38; P<0.001; Egger’s Z=6.89–7.45; P<0.001). Trim-and-fill shown in supplements. - Attrition at posttreatment: Individual M=12% (SD=13%), group M=21% (SD=14%), internet-based M=20% (SD=10%). - Risk of bias: Most studies had some concerns; blinding of therapists and assessors not feasible; missing data handling often incomplete; limited preregistered analysis plans.

Findings demonstrate that unified TD-CBT is effective for adult emotional disorders across delivery formats, producing large advantages over waitlist and TAU for both depression and anxiety and small advantages over other active interventions for depression. Critically, TD-CBT yields outcomes comparable to DS-CBT on both depression and anxiety, addressing a key debate and supporting TD-CBT as a viable, flexible alternative to diagnosis-specific care. Comparable efficacy across individual, group, and internet-based settings highlights scalability. Group formats can conserve therapist resources; internet-based delivery (including unguided) can expand access, reduce barriers (e.g., geography, mobility), and enable blended-care models. Benefits are durable through 3, 6, and 12 months; however, superiority vs controls was not detected at 24 months, suggesting attenuation or insufficient power at the longest follow-up. Despite high heterogeneity and publication bias indications, sensitivity analyses support the robustness of main conclusions. Overall, results align with and extend prior reviews by including unified TD-CBT across formats in RCTs, reinforcing TD-CBT’s role in addressing comorbidity efficiently with broad dissemination potential.

This comprehensive review and meta-analysis of 56 RCTs provides strong evidence that unified TD-CBT effectively reduces anxiety and depression symptoms across individual, group, and internet-based settings. TD-CBT is superior to waitlist and TAU, comparable to DS-CBT, and yields small advantages over other active treatments for depression. Benefits are maintained up to 12 months, with mixed evidence at 24 months. The findings support transdiagnostic approaches as scalable, efficient options to address comorbidity and treatment gaps. Future research should: (1) conduct individual participant data meta-analyses to identify moderators/mediators and optimize patient-treatment matching; (2) examine effectiveness across specific diagnoses and broader outcomes (functioning, quality of life); (3) assess long-term durability beyond 24 months across independent groups; (4) evaluate implementation strategies, training efficiencies, and cost-effectiveness; (5) adopt open science practices to reduce bias; and (6) explore personalization within unified TD-CBT and blended-care models.

- Diagnostic scope: Included multiple emotional disorders, but evidence is concentrated on GAD, SAD, and MDD; cannot determine efficacy for each specific diagnosis or for less-studied emotional disorders (e.g., somatic symptom disorders, PTSD, borderline personality disorder). - Generalizability: Under-representation of studies from South America and Africa. - Risk of bias: Widespread practical constraints on blinding (therapists/assessors), limited preregistered analysis plans, and incomplete handling/reporting of missing data. - Heterogeneity: High between-study heterogeneity remained, even after subgrouping by format; outliers influenced variability. - Publication bias: Funnel plot asymmetry and significant Egger’s tests indicate potential bias. - Long-term outcomes: 24-month findings are limited and largely derived from a few internet-based trials from the same research group; superiority vs controls not detected. - Outcomes: Focused on self-reported anxiety/depression; clinical relevance, functioning, quality of life, symptom deterioration, comorbidity changes, and dropout were not systematically assessed. - Population: Adults only; findings do not generalize to children/adolescents.