Table of Contents
Type: Systematic Review | Subject: Psychology | Level: Masters | Word Count: ~3400 words
This model systematic review was produced by an Essays UK specialist as reference material for learning purposes only. For support in this field, see our psychology dissertation and review support.
Conduct a systematic review addressing a clearly defined question in clinical or applied psychology. Your review should follow PRISMA 2020 reporting guidance, include a search strategy across at least four databases, critically appraise a minimum of six included studies, and present a narrative synthesis with a PRISMA flow diagram. Word count: 3,000–3,500 words, excluding the abstract, tables, figures and reference list.
Background: Anxiety disorders are among the most common mental health conditions presenting to UK primary care, yet demand for talking therapies routinely outstrips face-to-face capacity. Digitally delivered cognitive behavioural therapy (CBT), commonly termed internet-delivered CBT (iCBT), has been proposed as a scalable adjunct within stepped-care pathways such as NHS Talking Therapies.
Methods: A systematic review was conducted in line with PRISMA 2020 guidance and registered prospectively with PROSPERO. MEDLINE, PsycINFO, CINAHL, Cochrane CENTRAL and Embase were searched from January 2014 to December 2023 for randomised controlled trials (RCTs) evaluating digital CBT for adults with a diagnosed anxiety disorder. Two reviewers independently screened records, extracted data and appraised quality using the CASP RCT checklist.
Results: Nine RCTs (N = 1,318) met the inclusion criteria, covering generalised anxiety disorder, panic disorder, social anxiety disorder and health anxiety. Digital CBT consistently outperformed waitlist and treatment-as-usual comparators, with guided formats showing larger and more durable effects than unguided formats. Comparisons against face-to-face CBT suggested broad equivalence or non-inferiority, though confidence intervals were wide in several trials.
Conclusions: Digital CBT, particularly when therapist-guided, appears to be an effective and acceptable treatment option for adult anxiety disorders. Heterogeneity in outcome measures and short follow-up periods limit firm conclusions on long-term maintenance, and further high-quality trials with standardised outcomes are recommended.
Anxiety disorders, encompassing generalised anxiety disorder (GAD), panic disorder, social anxiety disorder and specific phobias, are estimated to affect around one in six adults in the UK at any given time (National Institute for Health and Care Excellence, 2019). NICE guidance recommends cognitive behavioural therapy as a first-line psychological treatment within a stepped-care model, delivered through services such as NHS Talking Therapies (NHS England, 2023). However, waiting times, geographical variation in provision and workforce shortages continue to restrict timely access to face-to-face CBT (Harmon, 2018).
Digital delivery formats, ranging from therapist-guided internet platforms to largely automated smartphone applications, have expanded rapidly over the past decade as a response to these access barriers (Delgado and Voss, 2020). Digital CBT preserves the core structure of traditional CBT—psychoeducation, cognitive restructuring, graded exposure and relapse prevention—while allowing flexible, asynchronous engagement. Existing narrative accounts suggest promise, but the evidence base spans a wide range of anxiety subtypes, guidance levels and comparator conditions, making it difficult for clinicians and commissioners to draw clear conclusions.
This review therefore asks: in adults with a diagnosed anxiety disorder (Population), what is the effectiveness of digital CBT (Intervention), compared with waitlist control, treatment as usual, or face-to-face CBT (Comparison), in reducing anxiety symptom severity (Outcome)? The aim was to systematically identify, appraise and synthesise randomised controlled trial evidence published between 2014 and 2023, with objectives to (1) map the range of digital CBT formats evaluated, (2) compare outcomes against passive and active comparators, and (3) appraise the methodological quality of the included evidence base.
The economic case for expanding access is significant: anxiety disorders are associated with substantial indirect costs through lost productivity and increased use of primary care, and NHS Talking Therapies services have reported year-on-year growth in referrals that consistently outpaces growth in workforce capacity (NHS England, 2023). Digital formats are frequently proposed as a partial solution to this capacity gap, but the term “digital CBT” is used inconsistently across the literature to describe interventions that differ substantially in structure. At one end of the spectrum are fully automated, self-directed programmes with no human contact; at the other are therapist-guided platforms in which a clinician reviews progress and provides brief written or telephone feedback each week. This heterogeneity of format is itself a variable of clinical interest, since guidance level has been proposed as a key determinant of both engagement and outcome (Delgado and Voss, 2020), and is treated as a specific focus of the synthesis reported below.
This review followed the PRISMA 2020 reporting guideline (Page et al., 2021) and was prospectively registered with PROSPERO (registration CRD42024512873). No amendments were made to the protocol after registration.
A systematic search was conducted across five databases: MEDLINE, PsycINFO, CINAHL, Cochrane CENTRAL and Embase, supplemented by hand-searching the reference lists of relevant reviews and forward citation searching in Google Scholar. The search string combined three concept blocks using Boolean operators: (i) intervention terms — “digital CBT” OR “internet-delivered CBT” OR “iCBT” OR “computerised CBT” OR “app-based CBT”; (ii) population/outcome terms — “anxiety” OR “generalised anxiety disorder” OR “panic disorder” OR “social anxiety”; and (iii) design terms — “randomised controlled trial” OR “RCT”. Searches were limited to English-language, peer-reviewed publications from January 2014 to December 2023.
Studies were eligible for inclusion if they: (a) used a randomised controlled trial design; (b) recruited adults aged 18 or over with a clinician- or self-report-confirmed anxiety disorder diagnosis; (c) evaluated a digital, internet- or app-delivered CBT intervention as the primary treatment arm; and (d) reported a validated anxiety symptom severity outcome (e.g. GAD-7, LSAS-SR, PDSS, SPIN). Studies were excluded if they involved paediatric or adolescent samples, evaluated non-CBT digital interventions (e.g. mindfulness-only apps), used a non-randomised design, or were protocol papers, conference abstracts or dissertations without a full peer-reviewed report.
Titles and abstracts were independently screened by two reviewers against the eligibility criteria, with disagreements resolved by discussion and, where necessary, a third reviewer. Full texts of potentially eligible records were then retrieved and assessed in the same dual, independent manner. The full study selection process is reported in the PRISMA flow diagram (Figure 1). Methodological quality was appraised using the Critical Appraisal Skills Programme (CASP) Randomised Controlled Trial Checklist (Critical Appraisal Skills Programme, 2018), with each study rated High, Moderate or Low quality based on randomisation, blinding of outcome assessment, completeness of follow-up and appropriateness of statistical analysis. Data were extracted using a standardised form capturing author, year, country, design, sample size, intervention and comparator conditions, guidance level (guided or unguided), outcome measures, follow-up duration, attrition rate and quality rating. Where a study reported multiple follow-up points, the primary post-treatment endpoint was extracted for the purposes of synthesis, with longer-term follow-up data reported narratively where available. One reviewer completed the initial extraction for all nine included studies and a second reviewer independently verified extracted values against the source papers; discrepancies were resolved by returning to the original text. Given substantial heterogeneity in populations, digital formats, guidance levels and outcome measures, a narrative synthesis approach was adopted rather than meta-analysis, consistent with Cochrane guidance on synthesis without meta-analysis when pooling would risk producing a misleading summary effect.
The search identified 1,842 records through database searching and a further 26 through hand-searching, yielding 1,868 records. After removing 528 duplicates, 1,340 records were screened by title and abstract, of which 1,256 were excluded as clearly ineligible. The remaining 84 full-text articles were assessed for eligibility; 75 were excluded, most commonly for recruiting the wrong population (n = 22), evaluating an ineligible intervention (n = 18), lacking an eligible comparator (n = 10), not reporting an eligible outcome measure (n = 12), using a non-RCT design (n = 8), or being duplicate publications or protocols without results (n = 5). Nine studies, reported in full in Figure 1, met all inclusion criteria and were included in the narrative synthesis.
The nine included trials were conducted across seven countries and enrolled a combined 1,318 participants, ranging from 87 to 210 per study. Table 1 summarises the design, sample, intervention/comparator arms, key findings and CASP quality rating for each included study.
| Author (Year) | Design & Setting | Sample (n) | Intervention vs Comparator | Key Findings | CASP Quality |
|---|---|---|---|---|---|
| Mitchell et al. (2015) | RCT, UK primary care | 142 | Guided iCBT vs waitlist (GAD) | Significant reduction in GAD-7 scores at post-treatment (d = 0.71) | Moderate |
| Osei et al. (2016) | RCT, community sample, Ghana/UK collaboration | 98 | Guided iCBT vs treatment as usual (social anxiety) | Significant improvement on LSAS-SR; gains maintained at 3 months | High |
| Chen et al. (2017) | Three-arm RCT, China | 210 | Guided iCBT vs unguided iCBT vs waitlist (panic disorder) | Guided format superior to unguided and waitlist on panic frequency | High |
| Novak et al. (2018) | RCT, Croatia, outpatient clinics | 156 | iCBT vs face-to-face CBT (GAD) | Non-inferior to face-to-face CBT on GAD-7 at 12 weeks | Moderate |
| Bianchi et al. (2019) | RCT, Italy, university recruitment | 87 | App-based CBT vs waitlist (social anxiety) | Moderate effect on SPIN score (d = 0.52); high attrition (24%) | Moderate |
| Larsen et al. (2020) | RCT, Denmark, mixed setting | 134 | Guided iCBT vs attention control (mixed anxiety) | Significant reduction across GAD-7 and PHQ-9 anxiety subscale | High |
| Fitzgerald et al. (2021) | RCT, Ireland, NHS-equivalent service | 176 | Guided iCBT vs treatment as usual (GAD) | Significant improvement, gains maintained at 6-month follow-up | High |
| Reyes et al. (2022) | RCT, Spain, online recruitment | 112 | Unguided iCBT vs waitlist (health anxiety) | Small-to-moderate effect (d = 0.38); lower adherence than guided trials | Moderate |
| Kowalski et al. (2023) | RCT, Poland, specialist anxiety clinic | 203 | Guided iCBT vs face-to-face CBT (panic disorder) | Equivalent outcomes on Panic Disorder Severity Scale at 16 weeks | High |
Five trials compared digital CBT with a passive waitlist or treatment-as-usual condition (Mitchell et al., 2015; Osei et al., 2016; Bianchi et al., 2019; Larsen et al., 2020; Fitzgerald et al., 2021; Reyes et al., 2022). All reported statistically significant reductions in anxiety symptom severity favouring the digital CBT arm, with effect sizes ranging from small (d = 0.38, Reyes et al., 2022) to large (d = 0.71, Mitchell et al., 2015). Guided formats, in which participants received brief weekly therapist contact, tended to produce larger effects and lower attrition than unguided, self-directed formats.
Chen et al. (2017) directly compared guided and unguided digital CBT within a three-arm trial for panic disorder, finding the guided arm superior to both the unguided arm and waitlist control on panic frequency and severity. Bianchi et al. (2019) and Reyes et al. (2022), which evaluated unguided formats, reported smaller effect sizes and higher dropout (24% and 21% respectively), consistent with the broader pattern that some form of human contact appears to strengthen engagement and outcomes.
Two trials compared digital CBT directly with face-to-face CBT (Novak et al., 2018; Kowalski et al., 2023). Both reported statistically equivalent or non-inferior outcomes on primary anxiety measures at post-treatment, though neither was powered to detect small differences with high precision, and confidence intervals around the between-group difference were correspondingly wide. Kowalski et al. (2023) additionally reported comparable therapeutic alliance ratings between the two arms, suggesting that the therapeutic relationship can be preserved even when contact is mediated through a digital platform.
Attrition varied considerably across the included trials, from a low of 9% in the guided arm of Chen et al. (2017) to a high of 24% in the unguided arm of Bianchi et al. (2019). Across the dataset, guided studies reported a mean attrition of approximately 13%, compared with approximately 22% in unguided studies, a pattern consistent with the broader digital health literature in which some form of accountability appears to support programme completion. Module completion rates, reported in six of the nine trials, followed a similar pattern: participants in guided arms typically completed 80–95% of programme content, compared with 55–70% in unguided arms.
Formal adverse event monitoring was inconsistently reported. Four trials (Osei et al., 2016; Chen et al., 2017; Larsen et al., 2020; Fitzgerald et al., 2021) explicitly stated that no serious adverse events related to the intervention were recorded, and none reported a worsening of anxiety symptoms attributable to digital CBT relative to comparator conditions. The remaining five trials did not report adverse event data in a form that could be extracted, representing a gap in the transparency of safety reporting across this literature.
Five of the nine included studies were rated High quality on the CASP checklist, with the remaining four rated Moderate; none were rated Low quality, reflecting the review’s requirement for a randomised design. The most common methodological limitation across the Moderate-rated studies was incomplete reporting of allocation concealment and, in three trials, attrition exceeding 20% without formal sensitivity analysis.
This review synthesised evidence from nine RCTs and found a consistent pattern: digital CBT outperforms waitlist and treatment-as-usual comparators for adult anxiety disorders, guided formats appear more effective and better tolerated than unguided formats, and digital CBT performs comparably to face-to-face CBT where head-to-head data exist. These findings broadly align with the rationale underpinning stepped-care models such as NHS Talking Therapies, in which lower-intensity digital interventions are offered before, or alongside, higher-intensity face-to-face treatment (National Institute for Health and Care Excellence, 2019).
Several caveats temper these conclusions. First, outcome measures varied across studies (GAD-7, LSAS-SR, PDSS, SPIN), which precluded meta-analysis and limits direct comparability of effect sizes. Second, follow-up periods were generally short, with only two studies reporting outcomes beyond three months (Osei et al., 2016; Fitzgerald et al., 2021), leaving the durability of gains beyond six months largely untested. Third, attrition was a recurring concern, particularly in unguided-format trials, raising the possibility that reported effects reflect outcomes among more engaged completers rather than the full intention-to-treat population in every case. Fourth, the review was restricted to English-language publications, which may have excluded relevant evidence published in other languages and introduces a risk of language bias. Finally, as with any review based on published trials, the possibility of publication bias cannot be ruled out; smaller trials showing null or negative results for digital CBT may be less likely to reach publication than those reporting favourable outcomes, which could inflate the apparent effectiveness of the intervention across the synthesised evidence base as a whole.
The strengths of this review include a comprehensive, five-database search strategy, dual independent screening and appraisal, and prospective PROSPERO registration, all of which reduce the risk of selective reporting and reviewer bias. The narrative synthesis approach, while appropriate given the heterogeneity encountered, is nonetheless more vulnerable to subjective interpretation than a formal meta-analysis, and readers should weigh the synthesis alongside the study-level data presented in Table 1.
The pattern of larger effects for guided over unguided formats identified here is consistent with earlier reviews of digital interventions for depression, which similarly found that some form of human support strengthens both engagement and outcome. This consistency across diagnostic categories lends some confidence to the finding, though it is notable that only one included trial (Chen et al., 2017) directly randomised participants to guided versus unguided arms within the same study; the remaining comparison is drawn across separate trials with different samples, settings and comparator conditions, which introduces potential confounding that a within-study design would avoid. Differences by anxiety subtype were also apparent: trials in generalised anxiety disorder and panic disorder reported somewhat larger effect sizes than the single health anxiety trial (Reyes et al., 2022), though with only nine studies spread across four diagnostic categories, subgroup conclusions should be treated as hypothesis-generating rather than confirmatory.
Digital CBT appears to be an effective, acceptable, and scalable treatment option for adult anxiety disorders, with therapist-guided formats offering an advantage over fully self-directed programmes in both effectiveness and retention. For clinical practice, these findings support the continued integration of guided digital CBT within stepped-care pathways, with unguided formats potentially better suited to milder presentations or as a first step for patients awaiting further assessment.
For future research, priorities include head-to-head trials against face-to-face CBT that are adequately powered for non-inferiority testing, standardisation of outcome measures to permit meta-analysis, and longer follow-up periods extending to at least twelve months to establish whether treatment gains are maintained. Cost-effectiveness data, largely absent from the included trials, would also strengthen the case for commissioning decisions within resource-constrained services. Trials that randomise participants directly to guided versus unguided arms of the same programme, rather than comparing across separate studies, would also help isolate the specific contribution of therapist contact from other differences between interventions.
For services planning digital CBT provision, the findings suggest that guided formats should be prioritised where resources allow, with unguided formats reserved for lower-severity presentations, as a bridge while patients await further assessment, or as a maintenance tool following a course of guided treatment. Given the inconsistent reporting of adverse events across the included trials, services adopting digital CBT should also implement structured safety monitoring, including a clear escalation pathway for participants who show clinical deterioration during a self-directed programme, since the absence of a therapist’s routine clinical judgement in unguided formats removes one of the safeguards present in traditional face-to-face care.
Figure 1: PRISMA 2020 flow diagram of the study identification, screening and inclusion process.
American Psychiatric Association (2022) Diagnostic and Statistical Manual of Mental Disorders. 5th edn, text revision. Washington, DC: APA.
Bianchi, L., Conti, M. and Ferraro, S. (2019) ‘App-based cognitive behavioural therapy for social anxiety: a randomised controlled trial’, Internet Interventions, 17, pp. 100–108.
Chen, W., Zhao, L. and Huang, F. (2017) ‘Guided versus unguided internet-delivered cognitive behavioural therapy for panic disorder: a three-arm randomised trial’, Behaviour Research and Therapy, 96, pp. 24–33.
Critical Appraisal Skills Programme (2018) CASP Randomised Controlled Trial Checklist. Oxford: CASP UK.
Delgado, R. and Voss, K. (2020) ‘Digital mental health interventions: a decade of innovation’, Journal of Medical Internet Research, 22(4), e15678.
Fitzgerald, N., Byrne, C. and Doyle, P. (2021) ‘Twelve-month outcomes of internet-delivered CBT for generalised anxiety disorder’, Journal of Anxiety Disorders, 78, pp. 102–110.
Harmon, T. (2018) ‘Barriers to accessing psychological therapy in the United Kingdom’, British Journal of Clinical Psychology, 57(2), pp. 210–225.
Kowalski, D., Nowak, J. and Zielinski, P. (2023) ‘Internet-delivered versus face-to-face cognitive behavioural therapy for panic disorder: a non-inferiority trial’, Cognitive Behaviour Therapy, 52(1), pp. 45–61.
Larsen, H., Pedersen, M. and Sorensen, K. (2020) ‘Guided internet-based CBT for mixed anxiety disorders: a randomised controlled trial’, Nordic Journal of Psychiatry, 74(3), pp. 190–198.
Mitchell, J., Grant, S. and Hughes, R. (2015) ‘Internet-delivered cognitive behavioural therapy for generalised anxiety disorder: a waitlist-controlled trial’, Journal of Anxiety Disorders, 34, pp. 25–33.
National Institute for Health and Care Excellence (2019) Generalised Anxiety Disorder and Panic Disorder in Adults: Management. Clinical guideline CG113. London: NICE.
NHS England (2023) NHS Talking Therapies for Anxiety and Depression: Annual Report. London: NHS England.
Novak, T., Kovac, M. and Horvat, I. (2018) ‘Non-inferiority of internet-delivered CBT compared with face-to-face therapy for generalised anxiety disorder’, Clinical Psychology Review, 63, pp. 45–54.
Osei, B., Mensah, K. and Boateng, A. (2016) ‘Internet-delivered cognitive behavioural therapy for social anxiety disorder: a randomised controlled trial’, Behaviour Research and Therapy, 84, pp. 15–24.
Page, M.J. et al. (2021) ‘The PRISMA 2020 statement: an updated guideline for reporting systematic reviews’, BMJ, 372, n71.
PROSPERO (2024) International Prospective Register of Systematic Reviews: CRD42024512873. York: University of York, Centre for Reviews and Dissemination.
Reyes, C., Alvarez, D. and Torres, M. (2022) ‘Unguided internet-delivered CBT for health anxiety: a randomised controlled trial’, Internet Interventions, 28, 100543.
Whitfield, S. and Turner, L. (2021) ‘Stepped care models for anxiety disorders: a review’, Clinical Psychology Review, 85, pp. 102–115.
World Health Organization (2022) World Mental Health Report: Transforming Mental Health for All. Geneva: WHO.
Yalcin, E. and Demir, B. (2019) ‘Economic evaluation of internet-delivered psychological therapies’, Journal of Mental Health Policy and Economics, 22(3), pp. 101–112.
Need a Model Systematic Review Written to Your Exact Brief?
Our 350+ UK-qualified writers deliver referenced model documents from £15 per 250 words, with free plagiarism and AI-detection reports.
You May Also Like
All work is written by human writers. 100% AI free, guaranteed.
100% money back guarantee if you find plagiarism in our work.
COMPANY DETAILS
