Process of study selection
The initial search of eight databases in April 2018 retrieved 12,339 references, and the updated search in January 2019 retrieved 916 references. In total, 7738 titles and abstracts were screened after deduplication, from which 2948 reviews were selected for full-text analysis. On full-text analysis, 494 systematic reviews32–523 were selected for final inclusion. Data extraction for the mapping was done for all these reviews, the synthesis of which is presented in this chapter. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram describing each of these stages is presented in .
The PRISMA flow diagram describing review selection for mapping. Adapted from Fordham et al. © The Author(s) 2021. Published by Cambridge University Press. This is an Open Access article, distributed under the terms of the Creative Commons Attribution (more...)
Excluded studies
We excluded 2454 reviews at the full-text screening stage. Nearly half of these exclusions (1108/2454, 45%) were because the review did not provide a synthesis of CBT trials. Ten per cent (237/2454) of reviews were excluded because they were not available in English. References for the excluded reviews along with their reasons for exclusion are presented in Appendix 6.
Description of the included systematic reviews
We included 494 systematic reviews, which reported 2052 RCTs involving 221,128 participants.32–523 The included reviews were synthesised in three main formats: summary tables, bubble charts and gap maps.
Summary tables
The summary tables provide comprehensive details of all the 494 included reviews, split into the ICD-11 codes (see Appendix 7, Tables 5–33).
Bubble charts
The 494 systematic reviews identified by the search examine the effectiveness of CBT on HRQoL, depression, anxiety or a condition-specific outcome in conditions represented in 14 out of 20 primary (physical) and 13 out of 20 secondary (mental) ICD-11 codes. This equates to 68% of all ICD-11 categories (27/40). ‘Mood disorders [6A60-80]’ were the most researched condition (92 reviews, 272 RCTs, 42,676 participants). The primary and secondary ICD-11 categories that are represented in the included reviews are listed in the unshaded rows presented in Box 2; those that are not represented (i.e. evidence gaps) are listed in the shaded rows in Box 2. The volume of reviews, trials and participants are represented in two bubble maps in : (1) conditions that include < 1000 participants and (2) conditions that include > 1000 participants.
Primary and secondary ICD-11 codes represented and not represented in the CBT evidence map
Bubble map representing the volume of systematic reviews, RCTs and participants included in the qualitative synthesis (n = 494 reviews). (a) Conditions with < 1000 participants in total; and (b) conditions with > 1000 (more...)
Gap maps
We produced a gap map that details the context and population characteristics of all the reviews conducted within each ICD-11 category (see Appendix 8, Tables 34–37). We have summarised the information in the following section.
Context characteristics of the included reviews
In , we present the number of reviews that included trials conducted in different contexts. One review could include some trials conducted in one context and also include trials conducted in another context. Therefore, that one review could represent two or more context characteristics. Consequently, the percentages presented across the rows will not always add up to 100% (n = 494 reviews). The shaded cells represent how many reviews did not report on this characteristic.
Context characteristics of the included reviews
Context characteristics well reported
Nearly all the included reviews (486/494, 98%) reported whether or not they examined high- or low-intensity CBT. The majority were conducted on high-intensity CBT, but low-intensity CBT trials were included in 28% (139/494) of reviews across 14 out of 40 (35%) ICD-11 categories. Nearly all reviews (487/494, 99%) reported when follow-up data were collected. One-third of reviews (130/494, 26%) included a long-term (> 12 months) follow-up time point.
Context characteristics poorly reported
Only half of the reviews (247/494, 50%) reported on the severity of participants’ symptoms. Of these, the majority described participants as having a clinical diagnosis (216/494, 44%), with no further description on the severity of the symptoms (i.e. chronic or severe). Only 3% (16/494) of reviews examined participants with subclinical symptoms.
Over half of the included reviews (283/494, 57%) did not report from which care setting they had recruited their samples. Of the reviews that did report this, the majority recruited their samples from outpatient settings (114/494, 23%).
Context characteristics rarely examined
All the included reviews reported when the intervention was delivered (494/494, 100%); only 7% (36/494) examined the use of CBT in a preventative context.
Population characteristics of the included reviews
In , we present the number of reviews that included trials with samples representing different characteristics. One review could include some trials conducted with one type of population (e.g. children and adolescents) and other trials conducted with another population (e.g. adults), or one trial that included children, adolescents and adults. Therefore, that one review could represent two (or more) sample characteristics. Consequently, the percentages presented across the rows will not always add up to 100% (n = 494 reviews). The shaded cells represent how many reviews did not report on this characteristic.
Population characteristics of the included reviews
Population characteristics well reported
Most reviews reported the age of their samples (475/494, 96%); of these, only 6% (30/494 reviews, 81 RCTs and 6629 participants) were conducted with an older adult population.
Population characteristics poorly reported
Most reviews (458/494, 93%) did not report the ethnicity of the samples of their included trials. Of the 36 reviews that did report the ethnicity of their samples, we found an equal number of reviews that reported more white than non-white participants (10/494, 2%) and that reported more non-white than white participants (10/494, 2%).
Nearly half of the reviews (218/494, 44%) did not report on the sex of their trial samples or the country where their included trials were conducted. When reported, a higher number of reviews had a greater representation of female participants (167/494, 34%) in their trial samples, and most reviews included trials conducted in Europe, North America and Australasia (231/494, 47%).
The AMSTAR-2 review quality rating
Every review (n = 494) was assessed twice (by KE and TS) using the AMSTAR-2 checklist. The agreement between reviewers (KE and TS) in assessing the quality of reviews using the AMSTAR-2 checklist was good (327/494, 66%) (κw = 0.63, 95% CI 0.62 to 0.65). presents the proportion of reviews conducted over the preceding 20 years, classified into the four AMSTAR-2 quality categories.
Publication year and AMSTAR-2 quality rating of the included reviews.
Over the previous 20 years, the quality of systematic reviews has improved; however, in the latest time epoch (2015–19), we still identified that 36% of the included reviews were of critically low quality and only 29% of reviews were classified as being of moderate or high quality.
represents the item summaries from the AMSTAR-2 checklist. Of the ‘critical’ items on the checklist, 68% (336/494) of the reviews failed to register the protocol before commencement of the review (item 2), 76% (373/494) failed to provide the list of excluded studies along with the reasons for exclusion of each (item 7) and 50% (248/494) of reviews did not report an adequate search strategy.
The AMSTAR-2 items summary (n = 494)
Patient perspective and safety data
Of the 494 reviews, 118 (24%) reviews reported data on dropout rates, adherence and satisfaction analyses. Twenty reviews32,53,56,68,78,103,133,153,165,198,219,234,244,251,266,367,376,402,464,469 searched for safety data, of which nine reviews included reports of adverse events occurring in the CBT groups. We have summarised all the patient perspective and safety data under the relevant conditions in Box 3.
Patient perspective and safety data
Overall, there does not seem to be a great difference in dropout rates between CBT and active or non-active comparator groups. However, it appears that more participants drop out of low-intensity internet-based CBT than out of face-to-face CBT, and patients reported greater satisfaction with the therapeutic relationship in face-to-face CBT. Older adults appeared to drop out more than younger adults, but they also reported preferring psychological therapies, such as CBT, over pharmacological therapies. There may be certain groups who do not find CBT acceptable; for example, one review of mainly male veterans reported very low completion rates.58 Participants with common mental and physical conditions seemed generally satisfied with CBT, but schizophrenic patients seemed more likely to find CBT an unacceptable treatment option. In relation to adverse events, there is a lack of reporting on safety data from CBT reviews. However, the evidence we found does not suggest that participating in CBT could cause harm to participants.
