Terapia conductual (autoayuda) y terapia cognitivo‐conductual proporcionadas por distintos medios para los trastornos por ansiedad en adultos
Resumen
Antecedentes
Los trastornos por ansiedad son los problemas de salud mental más frecuentes. Son crónicos y continuados. Están disponibles tratamientos eficaces, pero el acceso a los servicios es limitado. Las intervenciones conductuales y cognitivo‐conductuales (autoayuda) proporcionadas por distintos medios intentan administrar el tratamiento con menos participación de los profesionales en comparación con las terapias tradicionales.
Objetivos
Evaluar los efectos de las terapias conductuales y cognitivo‐conductuales proporcionadas por distintos medios para los trastornos por ansiedad en adultos.
Métodos de búsqueda
Se consideraron estudios publicados y no publicados sin restricción por idioma o fecha. Se hicieron búsquedas en el registro especializado del Grupo de Revisión Cochrane de Depresión, Ansiedad y Neurosis (Cochrane Depression, Anxiety and Neurosis, CCDANCTR), todos los años hasta el 1 de enero de 2013. El CCDANCTR incluye ensayos controlados aleatorios relevantes de las siguientes bases de datos bibliográficas: The Cochrane Library (todos los años),EMBASE (1974 hasta la fecha), MEDLINE (1950 hasta la fecha) yPsycINFO (1967 hasta la fecha). Se realizaron búsquedas complementarias en Ovid MEDLINE (1950 hasta el 23 de febrero de 2013) y PsycINFO (1987 hasta febrero, semana 2, 2013), junto con registros de ensayos internacionales (el trials portal of the World Health Organization (ICTRP) y ClinicalTrials.gov). Las listas de referencias de los metanálisis previos y los informes de los ensayos controlados aleatorios fueron revisados, y se estableció contacto con los autores para obtener datos no publicados.
Criterios de selección
Ensayos controlados aleatorios de terapia conductual y terapia cognitivo‐conductual proporcionadas por distintos medios en adultos con trastornos por ansiedad (diferente del trastorno por estrés postraumático) comparadas con ninguna intervención (incluidos controles de atención / relajación) o comparadas con la terapia presencial.
Obtención y análisis de los datos
Ambos autores de la revisión examinaron de forma independiente los títulos y resúmenes.Las características de los estudios y los resultados se extrajeron por duplicado. Los resultados se combinaron mediante modelos de efectos aleatorios y se realizaron pruebas para la heterogeneidad y para el sesgo de estudio pequeño. Se examinaron las diferencias de subgrupos por tipo de trastorno, tipo de intervención proporcionada, tipo de medios y métodos de reclutamiento utilizados.
Resultados principales
Se incluyeron 101 estudios con 8403 participantes; se incluyeron 92 estudios en la síntesis cuantitativa. Estos ensayos compararon varios tipos de intervenciones proporcionadas por distintos medios (con niveles variables de apoyo) con ningún tratamiento y con intervenciones presenciales. La inconsistencia y el riesgo de sesgo redujeron la confianza en los resultados generales. Para el resultado primario síntomas de ansiedad, pruebas de calidad moderada mostraron efectos medios en comparación con ninguna intervención (diferencia de medias estandarizada [DME] 0,67; intervalo de confianza [IC] del 95%: 0,55 a 0,80; 72 estudios, 4537 participantes) y pruebas de baja calidad de efectos pequeños que favorecieron a la terapia presencial (DME ‐0,23; IC del 95%: ‐0,36 a ‐0,09; 24 estudios, 1360 participantes). La intervención se asoció con una mayor respuesta que la que se observó con ningún tratamiento (cociente de riesgos [CR] 2,34; IC del 95%: 1,81 a 3,03; 21 estudios, 1547 participantes) y no fue significativamente inferior a la terapia presencial en estos estudios (CR 0,78; 95; IC del % 0,56 a 1,09; diez estudios, 575 participantes), pero la última comparación incluyó versiones de terapias que no eran tan exhaustivas como las proporcionadas en la práctica clínica habitual. Las pruebas indicaron efectos beneficiosos en las medidas de resultado secundarias (depresión, discapacidad relacionada con la salud mental, calidad de vida y abandono), pero estas pruebas fueron de calidad baja a moderada. Faltan pruebas con respecto a los efectos perjudiciales.
Conclusiones de los autores
La autoayuda puede ser útil en los pacientes que no son capaces o no están dispuestos a utilizar otros servicios para pacientes con trastornos por ansiedad; en los pacientes que pueden obtener acceso a esta intervención, la terapia cognitivo‐conductual presencial es probablemente superior desde el punto de vista clínico. Los análisis económicos estuvieron más allá del alcance de esta revisión.
Se observó heterogeneidad importante entre los ensayos. Intervenciones recientes para problemas específicos que incorporan el apoyo de los médicos pueden ser más eficaces que las intervenciones transdiagnósticas (es decir, intervenciones para trastornos múltiples) proporcionadas sin orientación, pero estos temas están sujetos a confusión en los ensayos disponibles.
Aunque se han realizado muchos ensayos pequeños, la generalizabilidad de los resultados es limitada. En su mayoría las intervenciones probadas no están disponibles para los consumidores. La autoayuda se ha recomendado como el primer paso en el tratamiento de algunos trastornos por ansiedad, pero no se ha establecido la efectividad a corto y a largo plazo de las intervenciones proporcionadas por distintos medios. Se necesitan ensayos pragmáticos grandes para evaluar y maximizar los efectos beneficiosos de las intervenciones de autoayuda.
PICOs
Resumen en términos sencillos
Autoayuda para los trastornos por ansiedad
Los trastornos de ansiedad son habituales. Dificultan la vida normal y tienden a no desaparecer sin tratamiento. Están disponibles tratamientos eficaces, que incluye la terapia cognitivo‐conductual. Se sabe que esta terapia funciona cuando se proporciona en persona, pero muchos pacientes pueden no tener acceso a la terapia presencial. Esta revisión examinó 101 ensayos clínicos de autoayuda y analizó estadísticamente 92 de ellos. En estos ensayos, 8403 pacientes recibieron autoayuda o se asignaron a una condición control.
En general, la autoayuda con algún apoyo de un profesional parece ser más eficaz que ningún tratamiento. Solamente la mitad de los pacientes que utilizaron la autoayuda estuvieron mejores al final del tratamiento, pero la autoayuda todavía se puede considerar eficaz porque los pacientes con ansiedad no tienden a mejorar sin tratamiento. La autoayuda puede ser menos eficaz que la terapia presencial normal. Algunos resultados fueron difíciles de interpretar porque los efectos de los tratamientos variaron y el riesgo de sobrestimación de los resultados fue grave debido a limitaciones en los métodos de estudio. Se concluye que la autoayuda es probablemente mejor que ningún tratamiento, pero muchos pacientes con un trastorno por ansiedad conseguirían mejores resultados del tratamiento proporcionado por un psicólogo capacitado. Además, la mayoría de los materiales de autoayuda utilizados en estos estudios están concebidos para estudios de investigación y no están disponibles para el público, de manera que los resultados presentados en esta revisión no se pueden aplicar a los productos comercialmente disponibles.
Authors' conclusions
Summary of findings
| Media‐delivered interventions compared with no intervention for anxiety disorders | ||||||
| Patient or population: adults with anxiety Settings: recruited from referrals and advertising Intervention: media‐delivered cognitive behavioural therapy or behavioural therapy Comparison: wait‐list, no intervention, treatment as usual | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | No. of participants | Quality of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| No intervention | Media‐delivered intervention | |||||
| Anxiety (any measure), self‐rated at post‐treatment | Across studies, the typical scores in intervention conditions were, on average, 0.67 standard deviations better (with a 95% CI ranging from 0.55 to 0.78 standard deviations) than scores in the control conditions | 4537 (72 studies) | ⊕⊕⊕⊝ | This is a medium effect | ||
| Response (specific), self‐rated at post‐treatment | 180 per 1000 | 421 per 1000 | RR 2.34 (1.81 to 3.03) | 1547 (21 studies) | ⊕⊕⊝⊝ | The probability of clinical improvement doubled, but most people did not improve |
| Depression, self‐rated at post‐treatment | Across studies, the typical scores in intervention conditions were, on average, 0.47 standard deviations better (with a 95% CI ranging from 0.36 to 0.58 standard deviations) than scores in the control conditions | 3682 (50 studies) | ⊕⊕⊕⊝ | This is a medium effect | ||
| Quality of life, self‐rated at post‐treatment | Across studies, the typical scores in intervention conditions were, on average, 0.36 standard deviations better (with a 95% CI ranging from 0.22 to 0.49 standard deviations) than scores in the control conditions | 1344 (18 studies) | ⊕⊕⊝⊝ | This is a medium effect | ||
| *The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence: | ||||||
| 1Downgraded for inconsistency. 2Downgraded for risk of bias. | ||||||
| Media‐delivered interventions compared with face‐to‐face intervention for anxiety disorders | ||||||
| Patient or population: adults with anxiety Settings: recruited from referrals and advertising Intervention: media‐delivered cognitive behavioural therapy or behavioural therapy Comparison: face‐to‐face interventions | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | No. of participants | Quality of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| Face‐to‐face intervention | Media‐delivered intervention | |||||
| Anxiety (any measure), self‐rated at post‐treatment | SMD ‐0.23 (‐0.36 to ‐0.09) Across studies, the typical scores in intervention conditions were, on average, 0.23 standard deviations worse (with a 95% CI ranging from 0.09 to 0.36 standard deviations) than scores in the control conditions | 1360 (24) | ⊕⊕⊝⊝ | This is a small effect in favour of face‐to‐face intervention | ||
| Response (specific), self‐rated at post‐treatment | 537 per 1000 | 419 per 1000 | RR 0.78 (0.56 to 1.09) | 575 (10) | ⊕⊝⊝⊝ | The difference was not statistically significant |
| Depression, self‐rated at post‐treatment | Across studies, the typical scores in intervention conditions were, on average, 0.01 standard deviations worse (with a 95% CI ranging from ‐0.21 to 0.22 standard deviations) than scores in the control conditions | 906 (13) | ⊕⊝⊝⊝ | The difference was not statistically significant | ||
| Quality of life, self‐rated at post‐treatment | Across studies, the typical scores in intervention conditions were, on average, 0.08 standard deviations better (with a 95% CI ranging from ‐0.23 to 0.38 standard deviations) than scores in the control conditions | 282 (4) | ⊕⊝⊝⊝ | The difference was not statistically significant | ||
| *The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence: | ||||||
| 1Downgraded for inconsistency. 2Downgraded for risk of bias. 3Downgraded for indirectness. | ||||||
Background
Description of the condition
Anxiety disorders include generalised anxiety disorder (GAD), obsessive compulsive disorder (OCD), panic disorder (with or without agoraphobia), social phobia (social anxiety disorder) and specific phobias (APA 1994). They are characterised by excessive fears and worries that are difficult to control. Avoidance of feared situations or stimuli (including thoughts) is a cardinal behavioural feature of most anxiety disorders.
Anxiety disorders are the most common class of mental disorders. They are seen across cultures (Demyttenaere 2004) and will be experienced by about 18% of adults every year and by about 32% of people during their lifetimes (Kessler 2005a; Kessler 2005b). Half of all anxiety disorders begin by the age of 11 years (Kessler 2005a; Kessler 2005b), and most are chronic and unremitting (Bruce 2005). Anxiety reduces quality of life (Cramer 2005; Rapaport 2005; Saarni 2007) and contributes to disability (Bloom 2011), causing severe occupational impairment and lost productivity (Greenberg 1999, Ormel 1994). Despite their severity and prevalence, few people with anxiety disorders ever obtain professional treatment, and fewer obtain specialty care (Wittchen 2002).
Description of the intervention
Cognitive behavioural therapy (CBT) and behavioural therapy (BT) are well‐established and effective treatments in the short and medium term (Mitte 2005a; Mitte 2005b; Butler 2006). They are the most researched therapies for anxiety, and they are recommended by the National Institute for Health and Care Excellence (NICE) as front‐line treatments for both adults and young people (Kendall 2011; NCCMH 2006; Pilling 2013). Despite its effectiveness, face‐to‐face therapy is expensive and difficult to access.
To increase access to care and to reduce burden on therapists, computer programmes, books and other types of media have been developed to deliver CBT and BT. Self‐help has been defined as "(a) a therapeutic intervention administered through text‐audiotape, videotape or computer text, or through group meetings or individual exercises such as ‘therapeutic writing’, and (b) designed to be conducted predominantly independently of professional contact" (Bower 2001). Guided self‐help usually involves minimal contact that is primarily "of supportive or facilitative nature, and is meant to support the patient in working through the standardised psychological treatment" (Cuijpers 2010); it is commonly prescribed in primary and secondary care (Clark 2009). An overwhelming majority of behavioural and cognitive therapists use self‐help materials, most often in addition to traditional therapy (Keeley 2002).
How the intervention might work
Behavioural therapy has proved effective in the treatment of people with anxiety. Exposure therapy puts participants in direct and prolonged contact with feared situations (Marks 1971). Exposure in vivo (Emmelkamp 1975) aims to evoke fearful responses and to help participants develop coping strategies that will generalise to everyday circumstances (Foa 1986). Some reviews suggest that progressive muscle relaxation (Jacobson 1938) and controlled breathing are efficacious (Conrad 2007). Little evidence of long‐term benefit has been found when relaxation is used alone; muscle relaxation is sometimes used as an attentional control in the evaluation of treatments for anxiety. Applied relaxation teaches users to use relaxation techniques as coping mechanisms during exposure therapy, and it may provide some other benefits (Öst 1987; Öst 1988).
Cognitive behavioural therapy often includes (1) behavioural experiments, which aim to test one’s fears rather than habituate to a stimulus (Bennett‐Levy 2004), and (2) cognitive restructuring, which teaches people to recognise negative thoughts as they occur and to challenge unrealistic thoughts (Wells 1997). Treatment for anxiety may include other specific techniques, including exposure with ritual prevention for OCD (Steketee 1982), attention training for social anxiety (Rapee 2009) and meta‐cognitive training for GAD (Wells 2006). CBT may also include many non‐specific components that are common to effective forms of psychotherapy (e.g. goal setting, psychoeducation).
Why it is important to do this review
Media‐delivered interventions are intended to be inexpensive and easily accessible. Many literature reviews (Bessell 2002; Bower 2001; Glasgow 1978; Kaltenthaler 2002; Newman 2003; Richards 2003b; Riordan 1989; Schrank 1981; Stevens 1982; Mataix‐Cols 2006; Przeworski 2006; Pull 2006; van Boeijen 2005b) and meta‐analyses (Andrews 2010; Bower 2001; Cuijpers 2009; Cuijpers 2010; den Boer 2004; Gould 1993b; Griffiths 2006; Griffiths 2010b; Haug 2012; Hirai 2006; Kaltenthaler 2008; Lewis 2012; Marrs 1995; Menchola 2007; Norton 2007; Reger 2009; Scogin 1990; Spek 2007) have suggested that media‐delivered CBT may be effective in treating psychological problems, including anxiety, alone or in conjunction with other interventions.
Differences in inclusion criteria and methodological inconsistencies contribute to different conclusions across these reviews. An analysis of interventions for GAD found a large overall effect (g = 0.80, 95% CI 0.29 to 1.30) but significant heterogeneity (Mitte 2005a), similar to a recent review (Haug 2012), which reported a large effect across disorders (g = 0.78, 95% CI 0.67 to 0.90). However, another review (Bower 2001) reports a much smaller effect of self‐help manuals (d = 0.41; 95% CI 0.09 to 0.72), similar to an earlier review of bibliotherapy for many conditions (Marrs 1995). Several reviews reference the Cochrane Handbook for Systematic Reviews of Interventions, but one summed items on the risk of bias tool and failed to follow Cochrane methods for assessing each item (Andrews 2010). Despite the title, one recent review was not systematic (Wade 2010). A recent meta‐analysis (Lewis 2012) missed many studies that should have been included and suggested that media‐delivered interventions have large effects. The most recent analysis used single coding and included no assessment of bias (Haug 2012). Research in this area is advancing quickly, with many studies reported each year, and a rigorous and up‐to‐date review is needed to estimate the effects of media‐delivered interventions provided with or without professional support.
Objectives
To assess the effects of media‐delivered behavioural and cognitive behavioural therapies for anxiety disorders in adults.
Methods
Criteria for considering studies for this review
Types of studies
Randomised controlled trials and cluster‐randomised trials with a parallel‐group design, in which intervention and control groups were enrolled concurrently.
Types of participants
Adults (older than 18 years) with an anxiety disorder (as measured through diagnostic interview or questionnaire) other than post‐traumatic stress disorder (PTSD) or acute stress disorder were included.
We included studies of health anxiety because it shares the cardinal features of anxiety disorders, and it is included in the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM‐V), as “illness anxiety disorder”.
We excluded PTSD and acute stress disorder, which differ from other anxiety disorders. They may involve physical injury, difficulty separating past from present events and feelings of guilt. Reactions to severe stress are treated separately in the Tenth Revision of the International Classification of Diseases (ICD‐10) and in DSM‐V.
Studies of people experiencing worries or stress as the result of upcoming events (e.g. medical procedures) were excluded unless participants demonstrated symptoms of an underlying anxiety disorder (e.g. phobic avoidance of dental examinations). Other fears were excluded (e.g. "test anxiety", "fear of public speaking") unless participants met criteria for an anxiety disorder (e.g. specific phobia, social anxiety).
Studies including participants with other problems (e.g. depression) were eligible if participants with anxiety (including those with co‐morbid problems) could be separated from other participants (i.e. those whose problems do not include anxiety), or if most of the population had an anxiety disorder. In the absence of diagnostic interviews, studies were included when average scores on validated measures of anxiety were indicative of caseness.
No restrictions by setting were applied.
Types of interventions
Experimential intervention
Interventions must have delivered CBT or BT using printed materials, audio recordings, video recordings, or computers (including Internet), or via some combination of media. It must have been possible to use them as stand‐alone interventions (i.e. without a therapist), but we included studies with researcher or therapist contact and support. Computer programmes that could be used without the aid of a therapist were included whilst virtual reality systems, which use technology to facilitate therapist‐delivered interventions, were excluded. Interventions that encouraged users to enlist the help of a spouse or a friend were included.
We included studies of media‐delivered interventions alone and studies of media‐delivered interventions as adjuncts to treatment as usual or another intervention.
Comparator
We compared media‐delivered interventions with no intervention and with face‐to‐face behavioural or cognitive behavioural interventions. Only intervention arms in which the difference was a media‐delivered intervention were compared. Thus, '10 hours of therapy' versus '10 hours of therapy plus booklet‐delivered CBT' was eligible; '10 hours of therapy' versus '5 hours of therapy plus booklet‐delivered CBT' was not eligible. If data had been available, we also would have compared media‐delivered interventions with medication, psychoeducation and other forms of media‐delivered interventions.
If it had been possible to analyse the effects of media‐delivered interventions with and without adjunct medication, we would have treated these separately; most trials included participants with and without concurrent medication, so it was not possible to separate “standalone” and “adjunct” treatments. Instead, we report the number of participants in each trial who are taking medication.
Types of outcome measures
Clinician‐ and self‐rated measures were analysed separately because service users and clinicians may have different views about a person's improvement, and because these outcomes may be at risk of bias for different reasons.
Primary outcomes
1. Symptoms of anxiety:
a) Continuous symptom measures (e.g. Agoraphobic Cognitions Questionnaire (Chambless 1984), Beck Anxiety Inventory (Beck 1988), Liebowitz Social Anxiety Scale (Liebowitz 1987));
b) Response (e.g. Clinical Global Impression (Zaider 2003));
c) Recovery (e.g. no longer met criteria for diagnosis).
Secondary outcomes
2. Behavioural test (level of hierarchy achieved).
3. Behavioural test (subjective units of distress).
4. Depression.
5. Mental‐health related disability.
6. Quality of life.
7. Dropout.
Timing of outcome assessment
Outcome measures were grouped by length of follow‐up. We analysed outcomes at post‐treatment, at follow‐up less than six months after randomisation and at follow‐up longer than six months after randomisation. For outcomes measured at multiple time points in a single period, we included the longest time point after randomisation.
Search methods for identification of studies
Published and unpublished studies were considered without language restriction, although all searches were conducted in English and all contacts with authors were initiated in English. No date restrictions were applied.
Electronic searches
Searches were conducted using the following databases (Appendix 1).
-
The Cochrane Depression, Anxiety and Neurosis Review Group's Specialized Register (CCDANCTR‐Studies and CCDANCTR‐References) (all years to 1 January 2013). This search was conducted at the editorial base by the Group's Trials Search Co‐ordinator (TSC).
-
The Cochrane Central Register of Controlled Trials (CENTRAL) (all years to 2 April 2012) restricted to records submitted from the CCDANCTR‐References Register (SR‐DEPRESSN). Original search was conducted by the author team.
-
Ovid MEDLINE (all years to 23 February 2013).
-
Ovid PsycINFO (all years to February, Week 2, 2013).
International trial registries were also searched via the World Health Organization's trials portal (ICTRP) and ClinicalTrials.gov to identify unpublished or ongoing studies (all years to 1 Januray 2013).
Searching other resources
References from previous meta‐analyses and from articles retrieved during the search were examined for additional studies, and authors of included studies were contacted for unpublished data.
Data collection and analysis
Selection of studies
Two review authors independently screened titles and abstracts to determine which were eligible for inclusion. We were not blind to study authors, institutions, journal of publication or results. Disagreements regarding eligibility were resolved by seeking additional information and through discussion.
Data extraction and management
Data regarding methodology and outcomes were extracted independently in Excel by EM‐W and a second rater.
We included the following information from each study:
-
Year;
-
Location (country, urban/rural);
-
Method of recruitment;
-
Diagnosis (DSM category, method of assessment);
-
Inclusion criteria;
-
Risk of bias (see later).
Participants:
-
Socio‐demographics (e.g. age, sex);
-
Co‐morbidities.
For each intervention and comparison group of interest:
-
Type of intervention;
-
Type of media;
-
Duration;
-
Contact with participants.
For each outcome of interest:
-
Time points (collected and reported);
-
Loss to follow‐up.
Main comparisons
-
Media‐delivered intervention versus no treatment.
-
Media‐delivered intervention versus face‐to‐face intervention.
-
Media‐delivered intervention versus psychoeducation.
-
Media‐delivered intervention versus medication.
Assessment of risk of bias in included studies
EM‐W and a second rater (JJ, RC or AH) coded each included study using the Cochrane Collaboration's tool for assessing risk of bias (Higgins 2011). We judged whether each study was at low, high or unclear risk of bias in relation to sequence generation; allocation concealment; blinding of personnel; blinding of outcome assessors; incomplete outcome data; selective outcome reporting; and other sources of bias. Disagreements were resolved through discussion and by seeking further information.
Measures of treatment effect
Studies often report outcomes using multiple definitions and outcome measures. We gave preference to data that involved the least manipulation by authors or inference by review authors, that is, we extracted raw values at endpoint (e.g. means, standard deviations) rather than calculated effect sizes (e.g. Cohen’s d).
Unit of analysis issues
Cluster‐randomised trials
For each cluster‐randomised trial, we would have determined whether or not its data incorporated sufficient controls for clustering (e.g. robust standard errors, hierarchical linear models). If the data did not have proper controls, then we would have attempted to obtain an appropriate estimate of the data’s intra‐cluster correlation coefficient (ICC). If we could not find an estimate in the report of the trial, then we would have requested an estimate from the trial report authors. If the authors could not provide an estimate, then we would have obtained one from a similar study and conducted a sensitivity analysis to determine whether the results were robust when different values were imputed. We would have used the ICC estimate to control the trial’s data for clustering (Higgins 2011).
Cross‐over trials
For cross‐over trials, we extracted and analysed data from the first period only.
Studies with multiple treatment groups
For factorial studies, we included all comparisons. In other studies with multiple eligible intervention groups (e.g. interventions with or without a specific component), all eligible intervention groups were combined to estimate an average treatment effect compared with the control.
Dealing with missing data
Missing data, and methods for imputing such data, may affect the magnitude and direction of effects. For all analyses, we attempted to include all randomly assigned study participants. When analyses were reported for completers and controlled for the effects of missing cases (e.g. using mixed effects), we extracted the latter. If data were missing for some cases, or if reasons for dropout were not reported, we contacted study authors to request missing data and further information about dropouts. We conducted a sensitivity analysis that excluded studies at high risk of bias (see Sensitivity analysis).
Assessment of heterogeneity
Differences among included studies are discussed in terms of their participants, interventions, outcomes and methods. For each meta‐analysis, we also visually inspected forest plots to see whether the confidence intervals of individual studies had poor overlap (a rough indication of statistical heterogeneity); conducted a Chi2 test; and calculated the I2 statistic (Higgins 2011). We considered meta‐analyses to have heterogeneity when the P value for Chi2 was less than 0.10 and I2 was greater than 25%.
Assessment of reporting biases
For each meta‐analysis that included 10 or more studies, we drew a funnel plot and looked for asymmetry to assess the possibility of small study or reporting bias. We also compared effects using random‐effects and fixed‐effect models (see Sensitivity analysis), and we identified unpublished studies through correspondence and trial registration databases.
Data synthesis
In studies of transdiagnostic interventions (i.e. interventions for multiple anxiety disorders) that reported outcomes separately for participants with specific diagnoses (e.g. panic disorder, social anxiety), symptoms of anxiety were extracted for each diagnostic group where possible.
We used Review Manager (RevMan) Version 5.1 (Review Manager 2011) to conduct all meta‐analyses. Risk ratios (RRs) and 95% confidence intervals (CIs) were calculated for dichotomous outcomes and were combined with the use of Mantel‐Haenszel methods. Standardised mean differences (SMDs) and 95% CIs were calculated for continuous measures and were combined by using inverse variance methods. Random‐effects models were used because studies included different interventions and populations. When studies reported more than one measure of a particular outcome (e.g. symptoms of social anxiety measured using two scales), we averaged the results in Comprehensive Meta‐Analysis Version 2 software (Borenstein 2005) before entering them into RevMan (e.g. Imdad 2010); the specific measures included in each analysis for each study are listed in the Characteristics of included studies. For all analyses, the area to the right of the 'line of no effect' (SMD > 0, RR > 1) indicates a favourable outcome for media‐delivered interventions.
Subgroup analysis and investigation of heterogeneity
For the primary outcome, we conducted the following subgroup analyses to explore possible differences across the disorders and interventions included in this review. Because many factors co‐varied (e.g. several research groups contributed multiple studies in which the intervention and the population were very similar), these were considered hypothesis generating rather than hypothesis testing.
-
Types of controls.
-
Wait‐list, attention controls, treatment as usual, adjunct to a treatment versus other treatment alone.
-
Group psychotherapy, individual psychotherapy.
-
-
Types of disorders: GAD, health anxiety, mixed populations, OCD, panic disorder, social anxiety, specific phobias.
-
Types of interventions: CBT, exposure/desensitisation, relaxation.
-
Types of media: book, computer/Internet, other.
-
Types of recruitment: advertising only, referral only, other.
In future updates, we plan to investigate the level of therapist contact as a subgroup analysis.
Sensitivity analysis
We conducted the following sensitivity analyses to determine whether findings were robust to methodological decisions made throughout the review process.
-
We repeated the analyses using fixed‐effect models.
-
To investigate the influence of bias, we repeated the primary meta‐analysis by excluding studies at high risk of bias as the result of incomplete outcome data, which we considered the most important source of bias in the available data.
Results
Description of studies
Results of the search
The searches retrieved 1456 (CCDANCTR) and 1323 (Medline and PsycINFO) citations (Figure 1). After duplicates were removed electronically, two review authors screened 2801 abstracts.
Study flow diagram.
Included studies
One hundred and one studies reported in 159 papers were included. Eighteen studies included multiple eligible comparisons, which we treated separately, and 21 studies included multiple eligible intervention groups, which we combined for analysis.
Included studies randomly assigned 8403 participants (mean 83, median 60); the largest included 274 participants with anxiety, depression or mixed anxiety and depression (Proudfoot 2004a). Across studies, approximately 4277 participants received media‐delivered intervention, 2850 received no intervention, 916 received face‐to‐face intervention and 383 were assigned to a group not included in this review.
Ten included studies were not included in the quantitative synthesis because they reported no outcome data that could be analysed (Barrera 1977; Botella 2009; Holdsworth 1996; Koszycki 2011; Salaberria 1995; Tyrer 1988) or compared only two eligible interventions (Fraser 2001; Matthews 2011; Titov 2010b; Tortella‐Feliu 2011). Thus, characteristics are described for 101 studies, and 91 are included in a meta‐analysis.
Design
All studies randomly assigned individuals (Characteristics of included studies). One study randomly assigned general practices, then randomly assigned participants within practices (van Boeijen 2005a).
Setting
Studies were reported between 1973 and 2012; more were reported in 2011 than in any previous year. Most were conducted in four countries: UK (26), Australia (21), USA (17) and Sweden (20).
Participants
Studies included participants who were predominantly female (67% mean of means), white (94%) and in their 30s (37 years), who had more than a high school education (13 years). Most had long‐standing disorders (14 years), and many participants were taking an antidepressant or some other psychiatric medication (35%).
Interventions targeted panic (19 panic disorder, 3 panic disorder with agoraphobia, 3 panic attacks), social anxiety (19 unspecified, 2 specific subtype, 2 general subtype), specific phobias (11 spider, 2 snake, 1 flying, 1 heights), GAD (10), OCD (5), health anxiety (2), several diagnoses (11) or unspecified symptoms of anxiety (14). Some studies included participants with symptoms of anxiety or depression (9); these were included because a minority of participants had depression only, and further details are included in the Characteristics of included studies (Fletcher 2005; Grime 2004; Holdsworth 1996; Maunder 2009; Mead 2005; Proudfoot 2004a; Seekles 2011; Tyrer 1988; Zetterqvist 2003). Most studies assessed participants through a clinical interview (61 face‐to‐face interviews, 28 telephone interviews) or relied on a previous diagnosis (2), although some used questionnaires (10).
Participants were recruited through advertising (56), referral (24), mixed advertising and referral (15), college classes (Lewis 1978; Muller 2011; Shoenberger 2008), prison (Maunder 2009) or workplace (Grime 2004); the recruitment strategy was unclear in one study (Tolin 2011). Exclusion rates varied with method of recruitment and ranged from 0% to 99% (mean 48%). Studies reported that participants were excluded for suicidality (66), depression (54), drug or alcohol misuse (53) or physical health problems (33). Studies also explicitly excluded participants who had received psychotherapy previously (20) or were receiving it currently (64). Most studies (66) restricted the amount or type of medication that participants could be taking or asked participants to remain on a stable dose throughout the study.
Interventions
Of the included studies, 81 compared a media‐delivered intervention with no treatment (including wait‐list and attention controls). Additionally, 29 compared a media‐delivered intervention with a face‐to‐face intervention. Four compared a media‐delivered intervention only with another form of media‐delivered intervention; studies including more than one type of media‐delivered intervention and those comparing multiple forms of media‐delivered intervention only were idiosyncratic in design and objectives, thus a systematic synthesis was not possible. No studies that included a medication group reported outcomes that could be synthesised (Koszycki 2011; Tyrer 1988). Some studies provided information about anxiety or relaxation instructions to control for non‐specific intervention factors (Al‐Kubaisy 1992; Carlbring 2003; Furmark 2009b; Gilroy 2000; Greist 2002; Marks 2004; Schneider 2005); we treated these as attention controls rather than as active comparisons because the comparisons were described with the use of terms like “relaxation placebo” (Gilroy 2000) and “self‐relaxation as a placebo” (Marks 2004).
In most studies, interventions were developed by the authors (89). Exposure and behavioural experiments were the most common components (84). Most studies provided a version of cognitive behavioural therapy (70); others tested exposure alone (19), desensitisation alone (2), muscle relaxation (6), applied relaxation (3) or “self‐examination therapy”, which the author described as a type of CBT but which appears similar to problem solving (Bowman 1997). Interventions were designed to take up to 12 weeks to complete (mean 57 days to post‐treatment assessment), but some consisted of a single session (Heading 2001; Matthews 2011), and others did not specify how long the interventions should take to complete, with some lasting up to six months before assessment (Carlbring 2003; Wright 1997a).
Almost as many studies examined the use of books and booklets (39) as studied Internet‐delivered interventions (43). Most Internet‐delivered interventions did not exceed the capabilities of printed materials, that is, they provided static content on a Website or in a PDF. Some studies tested computer‐delivered interventions (11), mixed media (8) or a computerised telephone system (Greist 2002), and some of these were interactive. Interventions in 13 studies (including 10 of the computerised interventions) had to be used in a general practice office, a clinic, or a workplace.
Therapist contact
In most studies, participants had some contact with providers during treatment, and this is an important source of heterogeneity across studies (82). Contact occurred face‐to‐face (35), by e‐mail (22), by e‐mail and phone (11), by e‐mail and short message service (SMS) (1), by Internet forum only (1) or by telephone (12). When providers had contact with participants, this ranged from 10 minutes to about 6 hours. When studies with no contact and studies in which some contact was common to both groups were excluded (Kenardy 2003; Salaberria 1995), about eight (mean of means) contacts during treatment (six median of means) lasted about 103 (standard deviation (SD) 92) minutes (median 81). A few studies included automated reminders (e‐mail or SMS), which we recorded separately.
Excluded studies
We immediately excluded studies that were not randomised controlled trials, treatments for post‐traumatic stress, treatments for stress associated with medical procedures and other interventions that were not relevant to this review. In other studies, participants did not have anxiety disorders (64); these included studies of hoarding, perfectionism, public speaking, specific fears, stress, test anxiety and trichotillomania (currently listed as an impulse control disorder). Additionally, we excluded studies in which participants were younger than 18 years of age (21), interventions were not stand‐alone (27) or no eligible comparison was available (18). In others, the intervention was not BT or CBT, including a book about snakes (2) and interventions that were educational only (2), attentional retraining alone (2) or designed for group leaders (1; Characteristics of excluded studies). We also excluded a relapse prevention study (Wright 1997) that followed another study in this review (Febbraro 1997a). Ninety studies were identified through correspondence and the checking of reference lists and were excluded.
Two studies appeared likely to be eligible but were discontinued early because of poor recruitment; neither reported any results (Hetherton 2004; Jones 2002b).
Ongoing studies
We identified 11 studies that appear to be ongoing (Characteristics of ongoing studies).
Studies awaiting classification
We identified 15 studies that appear to be completed but not yet reported (Characteristics of studies awaiting classification).
Risk of bias in included studies
Because of the nature of the intervention, it was impossible to blind participants, so all studies were at high risk of bias. Other sources of bias are described later (see also Figure 2).
Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.
Allocation
Sequence generation
All included studies were described as randomised controlled trials: 57 specified the method of sequence generation and were at low risk of bias, but the method of randomisation was not described in 34. Ten studies had high risk of bias because randomisation appeared to be inadequate or undermined.
Allocation concealment
Allocation concealment was adequate in most studies reporting the method of randomisation, and 54 studies were at low risk of bias; risk of bias was unclear in 31 and high in 16.
Blinding
Providers
In nine studies, participants had no contact with researchers or clinicians, or providers were blind. Providers were not blind in 89 studies, which were at high risk of bias, and provider blinding was unclear in three studies.
Assessors
Most studies either did not report assessor‐rated outcomes or used blind assessors, thus 85 were at low risk of bias. Blinding of assessors was at high risk of bias in 13 studies and unclear in three. Of the studies that used assessor ratings (46), most (30) reported that assessors were blind.
Incomplete outcome data
For response and recovery (dichotomous) data, we assumed that dropouts in both groups did not improve.
For continuous outcomes, 60 studies reported no dropout or analysed outcomes in a manner that mitigated the effect of missing cases and measures. Methods of analysis were at high risk of bias in 38 studies. It was unclear how three studies handled missing data.
Selective reporting
Most of the studies in the review included multiple outcome measures. Only 20 studies appeared to be free of selective outcome reporting, and 40 clearly omitted measured outcomes and were at high risk of bias. Risk of selective outcome reporting was unclear in 41 studies that reported all outcomes mentioned in the text but did not reference a protocol.
Other potential sources of bias
In addition to the sources of bias already discussed, 11 studies were at risk of bias for other reasons, which are described in the Characteristics of included studies.
Effects of interventions
See: Summary of findings for the main comparison ; Summary of findings 2
Comparison 1: Media‐delivered interventions versus no intervention
Primary Outcomes
1.1 Symptoms of anxiety
1.1.a Continuous symptom measures
Any measure of anxiety (mean effect of all measures of anxiety in each study)
At post‐treatment, 72 studies, including 4537 participants, reported any self‐rated measure of anxiety (Figure 3). A medium effect was noted (SMD 0.67, 95% CI 0.55 to 0.78), although heterogeneity was substantial (Chi² = 227.97, degrees of freedom (df) = 75, P < 0.00001, I² = 67%). Only 13 studies, including 684 participants, reported controlled follow‐up within six months; a slightly smaller overall effect was reported (SMD 0.49, 95% CI 0.24 to 0.75) with substantial heterogeneity (Chi² = 32.96, df = 14, P = 0.003, I² = 58%). Long‐term follow‐up (> 6 months) was reported in five studies, including 287 participants; the overall effect was not significant (SMD 0.18, 95% CI ‐0.15 to 0.52), though heterogeneity approached significance (Chi² = 6.60, df = 4, P = 0.16, I² = 39%).
1 Self‐help compared with no treatment.
1.3 Anxiety (any measure), self‐rated at post‐treatment (by disorder).
Specific and general symptoms of anxiety
Effects described previously (any measure of anxiety) include the average of all measures of anxiety for each study, which are listed in the Characteristics of included studies. Most studies in the main analysis measured symptoms of a particular disorder (e.g. Leibowitz Social Anxiety Scale), but to ensure that the results were not influenced by the inclusion of broader measures (e.g. Beck Anxiety Inventory), we also analysed specific and general measures separately.
At post‐treatment, 55 studies including 3192 participants reported a self‐rated measure of a specific anxiety disorder. A medium to large effect was noted (SMD 0.68, 95% CI 0.57 to 0.80), though heterogeneity was moderate (Chi² = 128.49, df = 58, P < 0.00001, I² = 55%). Only 10 studies including 439 participants reported controlled follow‐up within six months, and a slightly smaller overall effect was found (SMD 0.61, 95% CI 0.27 to 0.94) with substantial heterogeneity (Chi² = 31.13, df = 11, P = 0.0004, I² = 65%). Long‐term follow‐up was reported in two studies including 194 participants; the overall effect was not significant (SMD 0.20, 95% CI ‐0.22 to 0.62), although heterogeneity approached significance (Chi² = 5.90, df = 4, P = 0.21, I² = 32%). The pattern of results in subgroup analyses was unchanged.
Additionally, 13 studies including 629 participants reported a clinician‐rated measure of a specific anxiety disorder. There was a large effect (SMD 1.04, 95% CI 0.50 to 1.58), though heterogeneity was considerable (Chi² = 124.76, df = 14, P < 0.00001, I² = 89%). Only five studies including 98 participants reported controlled follow‐up within six months, and a larger overall effect was described (SMD 1.12, 95% CI 0.70 to 1.53). No studies reported long‐term follow‐up. At post‐treatment, studies reporting a clinician‐rated measure were consistent with all self‐rated measures (SMD 0.70, 95% CI 0.46 to 0.95); larger effects on clinician‐rated measures may be explained by bias.
At post‐treatment, 48 studies including 3101 participants reported a self‐rated measure of general symptoms. There was a medium effect (SMD 0.58, 95% CI 0.43 to 0.73), though heterogeneity was considerable (Chi² = 176.38, df = 47, P < 0.00001, I² = 73%). Only nine studies including 569 participants reported controlled follow‐up within six months, and a small effect was described (SMD 0.22, 95% CI 0.06 to 0.39) with no important heterogeneity (Chi² = 7.46, df = 8, P = 0.49, I² = 0%). Long‐term follow‐up was reported in two studies including 194 participants; the overall effect was not significant (SMD ‐0.07, 95% CI ‐0.63 to 0.49), but the studies were inconsistent (Chi² = 3.19, df = 1, P = 0.07, I² = 69%).
1.1.b Response
At post‐treatment, 21 studies including 1547 participants reported response to treatment according to study criteria (Figure 4). Overall, intervention participants were twice as likely to respond (RR 2.34, 95% CI 1.81 to 3.03), although heterogeneity was moderate (Chi² = 40.30, df = 20, P = 0.005, I² = 50%). Only two studies including 128 participants reported controlled follow‐up within six months, and the effect was maintained (RR 2.12, 95% CI 1.18 to 3.79) with no heterogeneity (Chi² = 0.35, df = 1, P = 0.55, I² = 0%). No studies reported long‐term follow‐up. Several studies reported exceptional effect sizes, but only 18% of all participants in the comparison groups were classified as responders at the end of treatment; in absolute terms, media‐delivered interventions were associated with a 31% (95% CI 0.22 to 0.40) increase in response (number needed to treat for an additional beneficial outcome (NNTB = 3)).
1 Self‐help compared with no treatment.
1.26 Response at post‐treatment (by disorder).
1.1.c Recovery
Only nine studies with 605 participants reported the number of participants no longer meeting criteria for diagnosis at the end of treatment. Overall, media‐delivered interventions were associated with a fivefold increase in recovery (RR 5.36, 95% CI 1.62 to 17.68), but heterogeneity was considerable (Chi² = 74.23, df = 8, P < 0.00001, I² = 89%). Only one study reported more than 20% recovery in the comparison group, which could be explained by participants seeking treatment outside the study (Baillie 2002); in the other studies, only 6% of participants in the comparison groups recovered, and media‐delivered interventions were associated with a 45% (95% CI 0.26 to 0.65) increase in recovery (NNTB = 2).
Secondary outcomes
1.2 Behavioural test (level achieved)
Some studies of animal or situational phobias also reported behavioural measures. At post‐treatment, seven studies including 250 participants reported the level completed in a behavioural assessment; a large overall effect was described (SMD 1.11, 95% CI 0.36 to 1.87), but heterogeneity was considerable (Chi² = 37.20, df = 6, P < 0.00001, I² = 84%). When one study (Hassan 1992) that reported an implausible effect was excluded (SMD 8.17), a medium to large effect was still described (SMD 0.76, 95% CI 0.32 to 1.21). At follow‐up, one study including 30 participants reported a large difference (SMD 1.23, 95% CI 0.47 to 1.99). No studies reported long‐term follow‐up.
1.3 Behavioural test (symptoms)
At post‐treatment, five studies including 191 participants reported symptom ratings, and the overall effect was not significant (SMD 0.59, 95% CI ‐0.33 to 1.51), but heterogeneity was considerable (Chi² = 30.48, df = 4, P < 0.00001, I² = 87%). At follow‐up, one study including 30 participants reported no significant difference (SMD 0.60, 95% CI ‐0.11 to 1.31). No studies reported long‐term follow‐up.
1.4 Depression
At post‐treatment, 50 studies including 3682 participants reported a self‐rated measure of depression. A medium effect was described (SMD 0.47, 95% CI 0.36 to 0.58), although heterogeneity was substantial (Chi² = 120.40, df = 50, P < 0.00001, I² = 58%). Only seven studies including 526 participants reported controlled follow‐up within six months, and the combined effect was small (SMD 0.28, 95% CI 0.05 to 0.52), but heterogeneity was substantial (Chi² = 9.89, df = 6, P = 0.13, I² = 39%). When the largest study in the review, which targeted both anxiety and depression (Proudfoot 2004a), was excluded, the effect at follow‐up was not significant (SMD 0.20, 95% CI ‐0.04 to 0.43), and the results were not significantly heterogeneous (Chi² = 6.00, df = 5, P = 0.31, I² = 17%). Long‐term follow‐up was reported in two studies including 194 participants; the overall effect was not significant (SMD ‐0.25, 95% CI ‐0.58 to 0.08), and heterogeneity was not significant (Chi² = 1.11, df = 1, P = 0.29, I² = 10%).
1.5 Mental‐health related disability
At post‐treatment, 20 studies including 1629 participants reported a self‐rated measure of disability. A medium effect was found (SMD 0.48, 95% CI 0.33 to 0.63), although heterogeneity was substantial (Chi² = 37.21, df = 19, P = 0.007, I² = 49%). Only seven studies including 513 participants reported controlled follow‐up within six months, and no significant effect was found (SMD 0.21, 95% CI ‐0.15 to 0.57), but heterogeneity was substantial (Chi² = 22.09, df = 6, P = 0.001, I² = 73%). Long‐term follow‐up was reported in three studies including 147 participants; the overall effect was not significant (SMD ‐0.14, 95% CI ‐0.33 to 0.60), and heterogeneity was not significant (Chi² = 3.50, df = 2, P = 0.17, I² = 43%).
1.6 Quality of life
At post‐treatment, 18 studies including 1344 participants reported a self‐rated measure of quality of life. A medium effect was found (SMD 0.36, 95% CI 0.22 to 0.49), although heterogeneity was moderate (Chi² = 24.58, df = 17, P = 0.10, I² = 31%). Only three studies including 211 participants reported controlled follow‐up within six months, and the combined effect was not significant (SMD 0.08, 95% CI ‐0.19 to 0.35) with no heterogeneity (Chi² = 0.95, df = 2, P = 0.62, I² = 0%). One study reported no effect at long‐term follow‐up (SMD ‐0.06, 95% CI ‐0.48 to 0.36).
1.7 Dropout
Of those receiving media‐delivered interventions and no‐intervention, 855 (20%) and 463 (16%) did not complete the study. A small difference favoured no treatment (RR 0.96, 95% CI 0.94 to 0.99), and heterogeneity was moderate (Chi² = 129.33, df = 77, P = 0.0002, I² = 40%).
Comparison 2: Media‐delivered interventions versus face‐to‐face interventions
Primary Outcomes
2.1 Symptoms of anxiety
2.1.a Continuous symptom measures
Any measure of anxiety (mean effect of all measures of anxiety in each study)
At post‐treatment, 24 studies, including 1360 participants, reported any self‐rated measure of anxiety (Figure 5). A small effect favoured face‐to‐face intervention (SMD ‐0.23, 95% CI ‐0.36 to ‐0.09), and moderate heterogeneity approached significance (Chi² = 31.46, df = 23, P = 0.11, I² = 27%). Only 11 studies including 677 participants reported controlled follow‐up within six months, and a slightly smaller effect favoured face‐to‐face intervention (SMD ‐0.14, 95% CI ‐0.30 to ‐0.01) with no heterogeneity (Chi² = 6.50, df = 10, P = 0.77, I² = 0%). Long‐term follow‐up (> 6 months) was reported in eight studies including 423 participants; the overall difference was not significant (SMD ‐0.16, 95% CI ‐0.50 to 0.18), although heterogeneity was substantial (Chi² = 18.09, df = 7, P = 0.01, I² = 61%) and was attributable to one study that reported a large effect favouring face‐to‐face intervention (Ost 1991); in that study, values for participants in the media‐delivered intervention group were consistent with those of other studies, and the face‐to‐face intervention appeared to be very effective.
2 Self‐help compared with face‐to‐face therapy.
2.3 Anxiety (any measure), self‐rated at post‐treatment (by disorder).
Specific and general symptoms of anxiety
At post‐treatment, 23 studies including 1255 participants reported a self‐rated measure of a specific anxiety disorder. A small effect favoured face‐to‐face intervention (SMD ‐0.23, 95% CI ‐0.37 to ‐0.10), and moderate heterogeneity approached significance (Chi² = 30.57, df = 22, P = 0.11, I² = 28%). Only 10 studies including 573 participants reported controlled follow‐up within six months, and the overall difference was not significant (SMD ‐0.15, 95% CI ‐0.31 to ‐0.02) with no heterogeneity (Chi² = 8.61, df = 9, P = 0.47, I² = 0%). Long‐term follow‐up (> 6 months) was reported in eight studies including 423 participants; the overall difference was not significant (SMD ‐0.20, 95% CI ‐0.54 to 0.14), although heterogeneity was substantial (Chi² = 18.08, df = 7, P = 0.01, I² = 61%) and was attributable to one study that reported a large effect favouring face‐to‐face intervention (Ost 1991). The pattern of results in subgroup analyses was unchanged.
Additionally, six studies including 418 participants reported a clinician‐rated measure of a specific anxiety disorder. The difference was not significant (SMD ‐0.13, 95% CI ‐0.33 to 0.06), and no heterogeneity was noted (Chi² = 3.97, df = 5, P = 0.55, I² = 0%). Only three studies including 185 participants reported controlled follow‐up within six months, and the effect was similar to the effect at post‐treatment (SMD ‐0.19, 95% CI ‐0.69 to 0.30). Only one study including 98 participants reported long‐term follow‐up, and the effect was similar to the effect at post‐treatment (SMD ‐0.13, 95% CI ‐0.53 to 0.28).
At post‐treatment, 11 studies including 627 participants reported a self‐rated measure of general symptoms. No significant difference between groups was noted (SMD ‐0.08, 95% CI ‐0.36 to 0.20), although heterogeneity was substantial (Chi² = 30.19, df = 10, P = 0.0008, I² = 67%). Only five studies including 308 participants reported controlled follow‐up within 6 months, and no significant difference (SMD 0.10, 95% CI ‐0.30 to 0.09) and no important heterogeneity were described (Chi² = 1.24, df = 4, P = 0.87, I² = 0%). Long‐term follow‐up was reported in four studies including 241 participants; no significant difference was noted (SMD 0.04, 95% CI ‐0.22 to 0.30), and no heterogeneity was reported (Chi² = 1.34, df = 3, P = 0.72, I² = 0%).
2.1.b Response
At post‐treatment, 10 studies including 575 participants reported response to treatment (Figure 6). Overall effects were not significant (RR 0.78, 95% CI 0.56 to 1.09), but heterogeneity was substantial (Chi² = 32.38, df = 9, P = 0.0002, I² = 72%). The absolute difference also was not significant (RD ‐0.15, 95% CI ‐0.35 to 0.04), but the absolute rate of response was only 47% overall. Only four studies including 224 participants reported controlled follow‐up within six months, and although no significant difference was found (RR 0.99, 95% CI 0.73 to 1.35), heterogeneity was substantial (Chi² = 7.97, df = 3, P = 0.05, I² = 62%). Three studies including 116 participants reported long‐term follow‐up and no significant difference (RR 0.58, 95% CI 0.20 to 1.64), but heterogeneity was considerable (Chi² = 8.68, df = 2, P = 0.01, I² = 77%).
Forest plot of comparison: 2 Media‐delivered intervention compared with face‐to‐face therapy, outcome. 2.21 Response (specific), self‐rated at post‐treatment (by disorder).
2.1.c Recovery
Only six studies including 587 participants reported the number of participants no longer meeting criteria for diagnosis at the end of treatment. No significant difference in recovery was found (RR 1.05, 95% CI 0.88 to 1.24), and no heterogeneity was noted (Chi² = 2.65, df = 5, P = 0.75, I² = 0%). Only three studies including 354 participants reported controlled follow‐up within six months; no significant difference was found (RR 1.04, 95% CI 0.70 to 1.56), but heterogeneity was substantial (Chi² = 4.02, df = 2, P = 0.13, I² = 50%). Only two studies including 147 participants reported long‐term follow‐up; no significant difference was found (RR 0.90, 95% CI 0.64 to 1.25), but the studies were inconsistent (Chi² = 3.60, df = 1, P = 0.06, I² = 72%).
Secondary outcomes
2.2 Behavioural test (level achieved)
Six studies included behavioural tests. At post‐treatment, five studies including 211 participants reported the level completed; a small to medium effect was found (SMD ‐0.44, 95% CI ‐0.76 to ‐0.13), and no important heterogeneity was noted (Chi² = 4.75, df = 4, P = 0.31, I² = 16%). Two studies including 60 participants reported the level competed at follow‐up; no significant difference was noted (SMD ‐0.02, 95% CI ‐0.74 to 0.71), but the studies were inconsistent (Chi² = 2.02, df = 1, P = 0.16, I² = 50%). Two studies including 122 participants reported the level completed at long‐term follow‐up; no significant difference was found (SMD ‐0.1, 95% CI ‐0.97 to 0.76), but the studies were not consistent (Chi² = 3.83, df = 1, P = 0.05, I² = 74%).
2.3 Behavioural test (symptoms)
At post‐treatment, five studies including 209 participants reported self‐rated symptoms, the overall difference was not significant (SMD ‐0.57, 95% CI ‐1.35 to 0.21) and considerable heterogeneity was noted (Chi² = 26.16, df = 4, P < 0.0001, I² = 85%). One study including 30 participants reported symptoms at follow‐up, and no significant difference was noted (SMD ‐0.09, 95% CI ‐0.79 to 0.60). Three studies including 162 participants reported symptoms at long‐term follow‐up, and no significant difference was found (SMD ‐0.46, 95% CI ‐1.19 to 0.27), but heterogeneity was considerable (Chi² = 8.93, df = 2, P = 0.01, I² = 78%).
2.4 Depression
At post‐treatment, 13 studies including 906 participants reported a self‐rated measure of depression. No overall difference was found (SMD ‐0.01, 95% CI ‐0.22 to 0.21), lthough heterogeneity was substantial (Chi² = 28.27, df = 12, P = 0.005, I² = 58%). Only five studies including 403 participants reported controlled follow‐up within six months, and the combined effect was not significant (SMD 0.09, 95% CI ‐0.19 to 0.36), but heterogeneity was moderate (Chi² = 7.11, df = 4, P = 0.13, I² = 44%). Long‐term follow‐up was reported in five studies including 339 participants; the overall effect was not significant (SMD 0.08, 95% CI ‐0.14 to 0.30), and no important heterogeneity was noted (Chi² = 4.08, df = 4, P = 0.40, I² = 2%).
2.5 Mental‐health related disability
At post‐treatment, 10 studies including 670 participants reported a self‐rated measure of disability. No significant difference was found (SMD ‐0.07, 95% CI ‐0.33 to 0.20), although heterogeneity was substantial (Chi² = 25.58, df = 9, P = 0.002, I² = 65%). Only five studies including 337 participants reported controlled follow‐up within six months, and the combined effect was not significant (SMD ‐0.10, 95% CI ‐0.39 to 0.19), but heterogeneity was moderate (Chi² = 6.80, df = 4, P = 0.15, I² = 41%). Long‐term follow‐up was reported in three studies including 242 participants; the overall effect was not significant (SMD ‐0.21, 95% CI ‐0.47 to 0.04), and no heterogeneity was noted (Chi² = 0.11, df = 2, P = 0.95, I² = 0%).
2.6 Quality of life
At post‐treatment, four studies including 282 participants reported a self‐rated measure of disability. No significant difference was found (SMD 0.08, 95% CI ‐0.23 to 0.38), although heterogeneity was moderate (Chi² = 4.82, df = 3, P = 0.19, I² = 38%). Only two studies including 143 participants reported controlled follow‐up within six months, and the combined effect favoured media‐delivered intervention (SMD 0.40, 95% CI ‐0.07 to 0.73) with no heterogeneity noted (Chi² = 0.64, df = 1, P = 0.43, I² = 0%). One study reported no difference at long‐term follow‐up (SMD ‐0.15, 95% CI ‐0.41 to 0.70).
2.7 Dropout
Of those receiving media‐delivered interventions and face‐to‐face interventions, 855 (20%) and 166 (18%) did not complete the studies. This difference was not significant (RR 1.08, 95% CI 0.90 to 1.30), and no heterogeneity was noted (Chi² = 25.77, df = 26, P = 0.48, I² = 0%).
Comparison 3: Media‐delivered interventions versus psychoeducation
No outcomes could be analysed.
Comparison 4: Media‐delivered interventions versus medication
No outcomes could be analysed.
Subgroup analyses
Comparison 1: Media‐delivered interventions versus no intervention
We examined the impact of methodological differences.
Types of controls
The specific control condition appeared related to relative effects (Chi² = 21.84, df = 3, P = 0.0001, I² = 86%). Specifically, interventions had similar effects compared with wait‐list and attention controls (Chi² = 3.99, df = 2, P = 0.14, I² = 50%), but the overall effect was small when participants were recruited from clinical settings and interventions were compared with treatment as usual (SMD 0.21, 95% CI 0.02 to 0.40). Without studies of treatment as usual, the overall effect was not importantly different from the primary result (SMD 0.72, 95% CI 0.61 to 0.84), and heterogeneity remained substantial (Chi² = 183.67, df = 68, P < 0.00001, I² = 63%).
Types of disorders
Effects were significantly different across types of disorders (Chi² = 17.04, df = 6, P = 0.009, I² = 65%). Studies of mixed disorders reported the smallest combined effect (SMD 0.36, 95% CI 0.19 to 0.52). After these were excluded, subgroups were no longer heterogeneous (Chi² = 3.96, df = 5, P = 0.56, I² = 0%); the overall effect remained medium to large (SMD 0.77, 95% CI 0.64 to 0.91), and heterogeneity remained substantial (Chi² = 157.63, df = 57, P < 0.00001, I² = 64%).
Types of interventions
Effects were not significantly different across types of interventions (Chi² = 1.88, df = 3, P = 0.60, I² = 0%), although the effect for relaxation was not significant. A medium effect for media‐delivered CBT was found (SMD 0.62, 95% CI 0.51 to 0.73), along with substantial heterogeneity (Chi² = 127.24, df = 55, P < 0.00001, I² = 57%).
Types of media
Effects were significantly different across types of media (Chi² = 7.70, df = 3, P = 0.05, I² = 61%) and were largest for Internet‐delivered interventions (SMD 0.79, 95% CI 0.62 to 0.96), although heterogeneity within this group was substantial (Chi² = 132.37, df = 35, P < 0.00001, I² = 74%).
Types of recruitment
Studies that advertised through mass media (including the Internet) reported larger effects than studies recruiting only through clinical referrals (Chi² = 14.98, df = 2, P = 0.0006, I² = 87%), although heterogeneity remained substantial among studies that used advertising (Chi² = 168.47, df = 54, P < 0.00001, I² = 68%).
Comparison 2: Media‐delivered interventions versus face‐to‐face interventions
Types of disorders
Effects were not significantly different across types of disorders (Chi² = 6.22, df = 4, P = 0.18, I² = 35.7%), although studies of specific phobias reported the largest difference between media‐delivered and face‐to‐face interventions (SMD ‐0.39, 95% CI ‐0.73 to ‐0.06), and studies of social anxiety reported the smallest (SMD 0.02, 95% CI ‐0.18 to 0.22).
Types of interventions
Disorders varied with types of therapy, and the difference between media‐delivered exposure and face‐to‐face exposure might be larger than the difference between different deliveries of CBT (Chi² = 2.36, df = 1, P = 0.12, I² = 57.6%), but no between‐group heterogeneity was noted after one study was removed from the exposure group (Ost 1991). Effects did not vary between individual and group CBT.
Types of media
Effects were not significantly different across types of media (Chi² = 5.07, df = 3, P = 0.17, I² = 40.9%).
Types of recruitment
Consistent with the first comparison, studies recruiting through advertisements found smaller differences than studies recruiting only through clinical referrals (Chi² = 2.94, df = 1, P = 0.09, I² = 66.0%), that is, recruiting participants through advertising was associated with more favourable outcomes for media‐delivered therapy.
Discussion
Summary of main results
This review suggests that media‐delivered interventions may be superior to no intervention for people with anxiety; there were positive effects on symptoms of anxiety and depression, response and recovery from illness, disability, and quality of life. Face‐to‐face interventions may be superior to media‐delivered interventions; there were differences favouring face‐to‐face treatment for symptoms of anxiety, but no significant differences in response and recovery from illness, disability and quality of life. Statistical heterogeneity suggests that effects probably vary across participants and interventions. Effects may be maintained after cessation of treatment, but few studies included follow‐up after six months.
Overall, interventions for specific disorders may be superior to interventions for any anxiety disorder. However, one recent transdiagnostic study reported effects that were consistent with the effects of interventions for specific disorders (Titov 2010a). The largest effects were reported in studies of Internet‐based CBT, which may be a superior medium for delivering interventions; however, many of these studies offered only documents for download and printing (PDFs) or access to non‐interactive Websites. More recent trials follow current models for treatment, so changes in intervention components and content may correlate with changes in disorders studied and technologies used, thus confounding these effects.
Overall completeness and applicability of evidence
In general, studies were conducted in high‐income, English‐speaking countries among white, female, middle‐aged participants; these results may not generalise to other settings or participants. Trials had high exclusion rates, for example, one study excluded 100 of 333 applications for high scores on a measure of depression, or for risk of suicide (Titov 2008a). Despite the large numbers of studies and participants in this review, the combined results of efficacy studies cannot demonstrate effectiveness in the general population, and few controlled comparisons suggest that benefits may be maintained after cessation of treatment.
For people seeking treatment, the magnitude of these effects should be interpreted with caution. For example, standardised effects appear similar to effective pharmacotherapies (Kapczinski 2003; Schneier 2001), but there was virtually no response among wait‐list and no‐treatment controls. It is also true that wait‐list participants who receive many face‐to‐face psychotherapies do not respond, but within‐group changes in these studies were not comparable with evidence‐based face‐to‐face interventions in which most participants respond or recover (e.g. Clark 2006).
Differences between media‐delivered interventions and face‐to‐face interventions were small, but some studies (particularly the older ones) provided face‐to‐face interventions designed to control for specific effects of media‐delivered interventions rather than face‐to‐face interventions as they would be provided in usual services. Established interventions may be more effective than these diminished controls. For example, one study demonstrated that a single session of therapist‐guided exposure was greatly superior to media‐delivered intervention for people with specific phobias (Ost 1991).
On the other hand, these interventions increase the number of people receiving help. Anxiety disorders are very common, most people with anxiety never receive treatment and these disorders are naturally chronic and unremitting. For people who would not receive another intervention, NNTBs of 2 and 3 may be valid estimates of treatment efficacy.
Among trials that included some contact with participants, the amount of therapist input varied greatly. Interventions in these studies may not be representative of the interventions that users can access outside of clinical trials or outside of particular services that are designed to offer guided self‐help.
Quality of the evidence
This review includes large numbers of studies and participants. Most trials reported the primary outcome, self‐rated symptoms of anxiety, and the evidence was of moderate quality compared with no treatment (summary of findings Table for the main comparison) but of low quality compared with face‐to‐face intervention (summary of findings Table 2).
Overall, methods of randomisation and allocation concealment appear unlikely to influence overall effects. However, participants and providers were not blind, and assessors were blind in only 64% of studies using clinician ratings. Compared with self‐ratings, clinician ratings suggest a greater effect compared with no treatment and a smaller difference compared with face‐to‐face therapy. Studies reporting self‐ratings and clinician ratings have reported effects on self‐ratings that were consistent with the average effect for all studies. Additionally, the post‐treatment effect for recovery (SMD 0.75, 95% CI 0.61 to 0.90) was slightly larger than the effect for symptoms of anxiety. Few studies referenced a protocol and reported all outcomes, thus the results of this review may be overestimated because of selective outcome reporting, which could explain the relatively larger effects on clinician‐rated outcomes and recovery. Compared with no intervention and compared with face‐to‐face intervention, 10 of 82 studies and 5 of 29 studies did not report a measure of anxiety that could be included in the main analysis. Incomplete outcome data may also lead to overestimation of continuous outcomes; however, for dichotomous outcomes, we assumed that dropouts did not respond, thus the response and recovery outcomes are unlikely to be overestimated due to missing data. Anxiety disorders tend to be chronic and naturally unremitting, so the assumption that dropouts did not improve may represent the true outcome for most missing cases. Subgroup analyses should be interpreted with caution, as some variables may correlate with risk of bias, for example, average effects of exposure alone may be overestimated compared with the effects of CBT.
Safety data were inadequately reported. Whilst a few studies mentioned adverse events in intervention or control groups, a systematic synthesis was not possible. It is unclear whether adverse events were not reported because they do not occur, or if they were not reported because they were not monitored; most trials excluded participants at greatest risk. Similarly, measures of participant satisfaction, the need for alternative interventions and costs were reported so selectively and incompletely that a formal synthesis would be inappropriate.
Statistical heterogeneity was substantial in most analyses (i.e. I2 > 50%). We attempted to explain this heterogeneity through prespecified analyses, but inconsistency remained within subgroups. Many of the individual studies were not significantly different from zero, but most reported positive effects compared with no intervention.
The quality of most outcomes was low according to GRADE criteria (GRADE 2004), although the quality of response and recovery data was moderate. Outcomes were downgraded for risk of selective outcome reporting. Continuous data were downgraded due to risk of bias, especially for incomplete outcome data.
Potential biases in the review process
This review included a comprehensive search for published and unpublished data. We obtained some unpublished data for 25 studies: Seven completely unpublished studies were included, we obtained an additional report of seven published papers (e.g. a thesis) and authors provided unpublished outcomes for nine studies reported only in published papers. To examine the effects of small study bias, the main analysis was repeated for each comparison using a fixed‐effect model. The magnitude and the precision of the effects were essentially unchanged compared with no treatment (SMD 0.64, 95% CI 0.58 to 0.70) and compared with face‐to‐face therapy (SMD, ‐0.21, 95% CI ‐0.32 to ‐0.10). Funnel plots appeared symmetrical; however, these tests for small study bias were unlikely to identify bias in a review with a large number of small studies, each of which received similar weight in the analysis. The Characteristics of studies awaiting classification include several trials that may be complete, and the Characteristics of excluded studies include two trials that were discontinued.
Several trials included participants with anxiety or depression. Two did not report symptoms of anxiety (Holdsworth 1996; Tyrer 1988), and one excluded participants with depression from the analysis because so few people with depression were recruited (Maunder 2009). Six others contributed to the primary analysis (Fletcher 2005; Grime 2004; Mead 2005; Proudfoot 2004a; Seekles 2011; Zetterqvist 2003). Excluding these studies did not importantly change the overall effect (SMD 0.70, 95% CI 0.58 to 0.82; Chi² = 210.97, df = 69, P < 0.00001, I² = 67%); furthermore, these studies were not included in an analysis restricted to symptoms of a specific anxiety disorder, which was also consistent with the primary analysis.
Agreements and disagreements with other studies or reviews
Many recent reviews have focused on Internet‐ and computer‐delivered interventions, and this review has a wider scope. After accounting for differences in inclusion criteria, this review still includes more trials than previous syntheses of self‐help for anxiety. As a result, few trials appear in both this review and most other meta‐analyses, but overlap is difficult to characterise for other reasons. For example, one review aimed to include studies with less than an hour of contact (Lewis 2012); by contacting the authors, we discovered that many included studies did not meet this criterion. In addition to studies of PTSD, one review (Haug 2012) included a study that was not randomised (the wait‐list taken from another trial in Tillfors 2008), one study that the author described as quasi‐randomised (Lucock 2008) and one study that combined self‐help with another intervention (i.e. Andersson 2006 compared self‐help with in vivo exposure vs waiting list). One meta‐analysis refers to studies that report outcome data but do not appear in the quantitative summary (Spek 2007). Another meta‐analysis omitted many studies that we included, and it included studies that met neither our inclusion criteria nor its stated criteria (Hirai 2006).
Effects on symptoms of anxiety were somewhat smaller than those reported in recent reviews (Andrews 2010; Cuijpers 2010; Griffiths 2010b; Haug 2012; Lewis 2012; Reger 2009), and we report effects for secondary outcomes that have not been widely analysed. Furthermore, this review includes more information about the characteristics of included studies than was provided in previous reviews. This review also includes a comprehensive assessment of risk of bias, which identifies some important limitations in the results that have not been fully considered elsewhere.
Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.
1 Self‐help compared with no treatment.
1.3 Anxiety (any measure), self‐rated at post‐treatment (by disorder).
1 Self‐help compared with no treatment.
1.26 Response at post‐treatment (by disorder).
2 Self‐help compared with face‐to‐face therapy.
2.3 Anxiety (any measure), self‐rated at post‐treatment (by disorder).
Forest plot of comparison: 2 Media‐delivered intervention compared with face‐to‐face therapy, outcome. 2.21 Response (specific), self‐rated at post‐treatment (by disorder).
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 1 Anxiety (any measure), self‐rated at post‐treatment.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 2 Anxiety (any measure), self‐rated at post‐treatment (type of control).
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 3 Anxiety (any measure), self‐rated at post‐treatment (by disorder).
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 4 Anxiety (any measure), self‐rated at post‐treatment (by media).
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 5 Anxiety (any measure), self‐rated at post‐treatment (by therapy).
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 6 Anxiety (any measure), self‐rated at post‐treatment (by setting).
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 7 Anxiety (any measure), self‐rated at post‐treatment (sensitivity analysis for bias).
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 8 Specific symptoms, self‐rated at post‐treatment.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 9 Specific symptoms, self‐rated at post‐treatment (type of control).
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 10 Specific symptoms, self‐rated at post‐treatment (by disorder).
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 11 Specific symptoms, self‐rated at post‐treatment (by media).
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 12 Specific symptoms, self‐rated at post‐treatment (by therapy).
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 13 Specific symptoms, self‐rated at post‐treatment (by setting).
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 14 Specific symptoms, self‐rated at post‐treatment (sensitivity analysis for bias).
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 15 Specific symptoms, self‐rated at post‐treatment (studies also reporting clinician rated).
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 16 Specific symptoms, clinician‐rated at post‐treatment.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 17 Specific symptoms, clinician‐rated at post‐treatment.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 18 General symptoms, self‐rated at post‐treatment.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 19 Behavioural Test (level), clinician‐rated at post‐treatment.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 20 Behavioural Test (symptoms), self‐rated at post‐treatment.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 21 Response (specific), self‐rated at post‐treatment.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 22 Response (specific), self‐rated at post‐treatment (type of control).
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 23 Response (specific), self‐rated at post‐treatment (by disorder).
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 24 Response (specific), self‐rated at post‐treatment (by media).
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 25 Response (specific), self‐rated at post‐treatment (by therapy).
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 26 Response (specific), self‐rated at post‐treatment (by setting).
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 27 Response (specific), clinician‐rated at post‐treatment.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 28 Response (general), self‐rated at post‐treatment.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 29 Recovery, clinician‐rated at post‐treatment.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 30 Depression, self‐rated at post‐treatment.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 31 Depression, clinician‐rated at post‐treatment.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 32 Depression improvement, self‐rated at post‐treatment.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 33 Disability, self‐rated at post‐treatment.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 34 Disability, clinician‐rated at post‐treatment.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 35 Quality of Life, self‐rated at post‐treatment.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 36 Dropout at post‐treatment.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 37 Anxiety (any measure), self‐rated at follow‐up.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 38 Specific symptoms, self‐rated at follow‐up.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 39 Specific symptoms, clinician‐rated at follow‐up.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 40 General symptoms, self‐rated at follow‐up.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 41 Behavioural Test (level), clinician‐rated at follow‐up.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 42 Behavioural Test (symptoms), self‐rated at follow‐up.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 43 Response (specific), self‐rated at follow‐up.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 44 Response (specific), clinician‐rated at follow‐up.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 45 Response (general), self‐rated at follow‐up.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 46 Recovery, clinician‐rated at follow‐up.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 47 Depression, self‐rated at follow‐up.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 48 Depression, clinician‐rated at follow‐up.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 49 Disability, self‐rated at follow‐up.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 50 Disability, clinician‐rated at follow‐up.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 51 Quality of Life, self‐rated at follow‐up.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 52 Anxiety (any measure), self‐rated at >6m.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 53 Specific symptoms, self‐rated at >6m.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 54 General symptoms, self‐rated at >6m.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 55 Response (specific), self‐rated at >6m.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 56 Depression, self‐rated at >6m.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 57 Disability, self‐rated at >6m.
Comparison 1 Media‐delivered interventions compared to no intervention, Outcome 58 Quality of Life, self‐rated at >6m.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 1 Anxiety (any measure), self‐rated at post‐treatment.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 2 Anxiety (any measure), self‐rated at post‐treatment (group versus individual).
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 3 Anxiety (any measure), self‐rated at post‐treatment (by disorder).
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 4 Anxiety (any measure), self‐rated at post‐treatment (by media).
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 5 Anxiety (any measure), self‐rated at post‐treatment (by therapy).
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 6 Anxiety (any measure), self‐rated at post‐treatment (by setting).
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 7 Anxiety (any measure), self‐rated at post‐treatment (sensitivity analysis for bias).
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 8 Specific symptoms, self‐rated at post‐treatment.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 9 Specific symptoms, self‐rated at post‐treatment (group versus individual).
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 10 Specific symptoms, self‐rated at post‐treatment (by disorder).
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 11 Specific symptoms, self‐rated at post‐treatment (by media).
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 12 Specific symptoms, self‐rated at post‐treatment (by therapy).
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 13 Specific symptoms, self‐rated at post‐treatment (by setting).
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 14 Specific symptoms, self‐rated at post‐treatment (sensitivity analysis for bias).
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 15 Specific symptoms, clinician‐rated at post‐treatment.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 16 General symptoms, self‐rated at post‐treatment.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 17 Behavioural Test (level), clinician‐rated at post‐treatment.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 18 Behavioural Test (symptoms), self‐rated at post‐treatment.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 19 Response (specific), self‐rated at post‐treatment.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 20 Response (specific), self‐rated at post‐treatment (group versus individual).
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 21 Response (specific), self‐rated at post‐treatment (by disorder).
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 22 Response (specific), self‐rated at post‐treatment (by media).
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 23 Response (specific), self‐rated at post‐treatment (by therapy).
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 24 Response (specific), self‐rated at post‐treatment (by setting).
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 25 Response (specific), clinician‐rated at post‐treatment.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 26 Response (general), self‐rated at post‐treatment.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 27 Recovery, clinician‐rated at post‐treatment.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 28 Depression, self‐rated at post‐treatment.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 29 Depression, clinician‐rated at post‐treatment.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 30 Depression improvement, self‐rated at post‐treatment.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 31 Disability, self‐rated at post‐treatment.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 32 Disability, clinician‐rated at post‐treatment.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 33 Quality of Life, self‐rated at post‐treatment.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 34 Dropout at post‐treatment.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 35 Anxiety (any measure), self‐rated at follow‐up.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 36 Specific symptoms, self‐rated at follow‐up.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 37 Specific symptoms, clinician‐rated at follow‐up.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 38 General symptoms, self‐rated at follow‐up.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 39 Behavioural Test (level), clinician‐rated at follow‐up.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 40 Behavioural Test (symptoms), self‐rated at follow‐up.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 41 Response (specific), self‐rated at follow‐up.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 42 Response (specific), clinician‐rated at follow‐up.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 43 Recovery, clinician‐rated at follow‐up.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 44 Depression, self‐rated at follow‐up.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 45 Depression, clinician‐rated at follow‐up.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 46 Depression improvement, self‐rated at follow‐up.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 47 Disability, self‐rated at follow‐up.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 48 Disability, clinician‐rated at follow‐up.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 49 Quality of Life, self‐rated at follow‐up.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 50 Anxiety (any measure), self‐rated at >6m.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 51 Specific symptoms, self‐rated at >6m.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 52 Specific symptoms, clinician‐rated at >6m.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 53 General symptoms, self‐rated at >6m.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 54 Behavioural Test (level), clinician‐rated at >6m.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 55 Behavioural Test (symptoms), self‐rated at >6m.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 56 Response (specific), self‐rated at >6m.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 57 Response (specific), clinician‐rated at >6m.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 58 Response (general), self‐rated at >6m.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 59 Recovery, clinician‐rated at >6m.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 60 Depression, self‐rated at >6m.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 61 Depression improvement, self‐rated at >6m.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 62 Disability, self‐rated at >6m.
Comparison 2 Media‐delivered interventions compared to face‐to‐face interventions, Outcome 63 Quality of Life, self‐rated at >6m.
| Media‐delivered interventions compared with no intervention for anxiety disorders | ||||||
| Patient or population: adults with anxiety Settings: recruited from referrals and advertising Intervention: media‐delivered cognitive behavioural therapy or behavioural therapy Comparison: wait‐list, no intervention, treatment as usual | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | No. of participants | Quality of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| No intervention | Media‐delivered intervention | |||||
| Anxiety (any measure), self‐rated at post‐treatment | Across studies, the typical scores in intervention conditions were, on average, 0.67 standard deviations better (with a 95% CI ranging from 0.55 to 0.78 standard deviations) than scores in the control conditions | 4537 (72 studies) | ⊕⊕⊕⊝ | This is a medium effect | ||
| Response (specific), self‐rated at post‐treatment | 180 per 1000 | 421 per 1000 | RR 2.34 (1.81 to 3.03) | 1547 (21 studies) | ⊕⊕⊝⊝ | The probability of clinical improvement doubled, but most people did not improve |
| Depression, self‐rated at post‐treatment | Across studies, the typical scores in intervention conditions were, on average, 0.47 standard deviations better (with a 95% CI ranging from 0.36 to 0.58 standard deviations) than scores in the control conditions | 3682 (50 studies) | ⊕⊕⊕⊝ | This is a medium effect | ||
| Quality of life, self‐rated at post‐treatment | Across studies, the typical scores in intervention conditions were, on average, 0.36 standard deviations better (with a 95% CI ranging from 0.22 to 0.49 standard deviations) than scores in the control conditions | 1344 (18 studies) | ⊕⊕⊝⊝ | This is a medium effect | ||
| *The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence: | ||||||
| 1Downgraded for inconsistency. 2Downgraded for risk of bias. | ||||||
| Media‐delivered interventions compared with face‐to‐face intervention for anxiety disorders | ||||||
| Patient or population: adults with anxiety Settings: recruited from referrals and advertising Intervention: media‐delivered cognitive behavioural therapy or behavioural therapy Comparison: face‐to‐face interventions | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | No. of participants | Quality of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| Face‐to‐face intervention | Media‐delivered intervention | |||||
| Anxiety (any measure), self‐rated at post‐treatment | SMD ‐0.23 (‐0.36 to ‐0.09) Across studies, the typical scores in intervention conditions were, on average, 0.23 standard deviations worse (with a 95% CI ranging from 0.09 to 0.36 standard deviations) than scores in the control conditions | 1360 (24) | ⊕⊕⊝⊝ | This is a small effect in favour of face‐to‐face intervention | ||
| Response (specific), self‐rated at post‐treatment | 537 per 1000 | 419 per 1000 | RR 0.78 (0.56 to 1.09) | 575 (10) | ⊕⊝⊝⊝ | The difference was not statistically significant |
| Depression, self‐rated at post‐treatment | Across studies, the typical scores in intervention conditions were, on average, 0.01 standard deviations worse (with a 95% CI ranging from ‐0.21 to 0.22 standard deviations) than scores in the control conditions | 906 (13) | ⊕⊝⊝⊝ | The difference was not statistically significant | ||
| Quality of life, self‐rated at post‐treatment | Across studies, the typical scores in intervention conditions were, on average, 0.08 standard deviations better (with a 95% CI ranging from ‐0.23 to 0.38 standard deviations) than scores in the control conditions | 282 (4) | ⊕⊝⊝⊝ | The difference was not statistically significant | ||
| *The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence: | ||||||
| 1Downgraded for inconsistency. 2Downgraded for risk of bias. 3Downgraded for indirectness. | ||||||
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Anxiety (any measure), self‐rated at post‐treatment Show forest plot | 76 | 4537 | Std. Mean Difference (Random, 95% CI) | 0.67 [0.55, 0.78] |
| 2 Anxiety (any measure), self‐rated at post‐treatment (type of control) Show forest plot | 76 | 4537 | Std. Mean Difference (Random, 95% CI) | 0.67 [0.55, 0.78] |
| 2.1 Adjunct | 1 | 93 | Std. Mean Difference (Random, 95% CI) | 0.33 [‐0.08, 0.75] |
| 2.2 Attention controls | 22 | 1028 | Std. Mean Difference (Random, 95% CI) | 0.65 [0.48, 0.82] |
| 2.3 Treatment as Usual | 7 | 745 | Std. Mean Difference (Random, 95% CI) | 0.21 [0.02, 0.40] |
| 2.4 Wait‐List | 46 | 2671 | Std. Mean Difference (Random, 95% CI) | 0.77 [0.61, 0.92] |
| 3 Anxiety (any measure), self‐rated at post‐treatment (by disorder) Show forest plot | 76 | 4537 | Std. Mean Difference (Random, 95% CI) | 0.67 [0.55, 0.78] |
| 3.1 GAD | 10 | 649 | Std. Mean Difference (Random, 95% CI) | 0.95 [0.44, 1.45] |
| 3.2 Health | 2 | 120 | Std. Mean Difference (Random, 95% CI) | 1.01 [0.24, 1.78] |
| 3.3 Mixed | 18 | 1376 | Std. Mean Difference (Random, 95% CI) | 0.36 [0.19, 0.52] |
| 3.4 OCD | 2 | 220 | Std. Mean Difference (Random, 95% CI) | 0.68 [0.41, 0.95] |
| 3.5 Panic | 21 | 797 | Std. Mean Difference (Random, 95% CI) | 0.62 [0.45, 0.79] |
| 3.6 Social | 15 | 1152 | Std. Mean Difference (Random, 95% CI) | 0.73 [0.59, 0.87] |
| 3.7 Specific | 8 | 223 | Std. Mean Difference (Random, 95% CI) | 0.99 [0.50, 1.48] |
| 4 Anxiety (any measure), self‐rated at post‐treatment (by media) Show forest plot | 76 | 4537 | Std. Mean Difference (Random, 95% CI) | 0.67 [0.55, 0.78] |
| 4.1 Book | 27 | 1046 | Std. Mean Difference (Random, 95% CI) | 0.48 [0.30, 0.65] |
| 4.2 Computer | 8 | 458 | Std. Mean Difference (Random, 95% CI) | 0.69 [0.31, 1.07] |
| 4.3 Internet | 36 | 2522 | Std. Mean Difference (Random, 95% CI) | 0.79 [0.62, 0.96] |
| 4.4 Other | 5 | 511 | Std. Mean Difference (Random, 95% CI) | 0.51 [0.32, 0.71] |
| 5 Anxiety (any measure), self‐rated at post‐treatment (by therapy) Show forest plot | 76 | 4537 | Std. Mean Difference (Random, 95% CI) | 0.67 [0.55, 0.78] |
| 5.1 CBT | 56 | 3550 | Std. Mean Difference (Random, 95% CI) | 0.62 [0.51, 0.73] |
| 5.2 Exposure/ Desensitization | 13 | 448 | Std. Mean Difference (Random, 95% CI) | 0.79 [0.49, 1.09] |
| 5.3 Relaxation | 6 | 504 | Std. Mean Difference (Random, 95% CI) | 0.84 [‐0.06, 1.73] |
| 5.4 Self‐examination | 1 | 35 | Std. Mean Difference (Random, 95% CI) | 0.92 [0.22, 1.62] |
| 6 Anxiety (any measure), self‐rated at post‐treatment (by setting) Show forest plot | 75 | 4506 | Std. Mean Difference (Random, 95% CI) | 0.67 [0.55, 0.78] |
| 6.1 Advertising | 55 | 3438 | Std. Mean Difference (Random, 95% CI) | 0.77 [0.63, 0.90] |
| 6.2 Referral Only | 16 | 884 | Std. Mean Difference (Random, 95% CI) | 0.32 [0.14, 0.51] |
| 6.3 Other | 4 | 184 | Std. Mean Difference (Random, 95% CI) | 0.46 [0.10, 0.82] |
| 7 Anxiety (any measure), self‐rated at post‐treatment (sensitivity analysis for bias) Show forest plot | 76 | 4537 | Std. Mean Difference (Random, 95% CI) | 0.67 [0.55, 0.78] |
| 7.1 High or Unclear Risk for Incomplete Outcome Data | 30 | 1481 | Std. Mean Difference (Random, 95% CI) | 0.68 [0.46, 0.90] |
| 7.2 Low Risk for Incomplete Outcome Data | 46 | 3056 | Std. Mean Difference (Random, 95% CI) | 0.66 [0.53, 0.79] |
| 8 Specific symptoms, self‐rated at post‐treatment Show forest plot | 59 | 3192 | Std. Mean Difference (Random, 95% CI) | 0.68 [0.57, 0.80] |
| 9 Specific symptoms, self‐rated at post‐treatment (type of control) Show forest plot | 59 | 3192 | Std. Mean Difference (Random, 95% CI) | 0.68 [0.57, 0.80] |
| 9.1 Adjunct | 1 | 93 | Std. Mean Difference (Random, 95% CI) | 0.31 [‐0.10, 0.73] |
| 9.2 Attention controls | 19 | 873 | Std. Mean Difference (Random, 95% CI) | 0.68 [0.47, 0.89] |
| 9.3 Treatment as Usual | 2 | 118 | Std. Mean Difference (Random, 95% CI) | 0.02 [‐0.35, 0.40] |
| 9.4 Wait‐List | 37 | 2108 | Std. Mean Difference (Random, 95% CI) | 0.74 [0.60, 0.88] |
| 10 Specific symptoms, self‐rated at post‐treatment (by disorder) Show forest plot | 58 | 3201 | Std. Mean Difference (Random, 95% CI) | 0.69 [0.57, 0.80] |
| 10.1 GAD | 8 | 514 | Std. Mean Difference (Random, 95% CI) | 0.68 [0.29, 1.07] |
| 10.2 Health | 2 | 120 | Std. Mean Difference (Random, 95% CI) | 0.85 [‐0.50, 2.21] |
| 10.3 Mixed | 4 | 231 | Std. Mean Difference (Random, 95% CI) | 0.21 [‐0.15, 0.57] |
| 10.4 OCD | 2 | 220 | Std. Mean Difference (Random, 95% CI) | 0.68 [0.41, 0.95] |
| 10.5 Panic | 20 | 740 | Std. Mean Difference (Random, 95% CI) | 0.61 [0.43, 0.79] |
| 10.6 Social | 15 | 1153 | Std. Mean Difference (Random, 95% CI) | 0.76 [0.63, 0.89] |
| 10.7 Specific | 8 | 223 | Std. Mean Difference (Random, 95% CI) | 0.99 [0.50, 1.48] |
| 11 Specific symptoms, self‐rated at post‐treatment (by media) Show forest plot | 59 | 3192 | Std. Mean Difference (Random, 95% CI) | 0.68 [0.57, 0.80] |
| 11.1 Book | 19 | 691 | Std. Mean Difference (Random, 95% CI) | 0.53 [0.32, 0.74] |
| 11.2 Computer | 7 | 261 | Std. Mean Difference (Random, 95% CI) | 0.79 [0.29, 1.28] |
| 11.3 Internet | 29 | 1829 | Std. Mean Difference (Random, 95% CI) | 0.77 [0.61, 0.92] |
| 11.4 Other | 4 | 411 | Std. Mean Difference (Random, 95% CI) | 0.60 [0.39, 0.82] |
| 12 Specific symptoms, self‐rated at post‐treatment (by therapy) Show forest plot | 59 | 3192 | Std. Mean Difference (Random, 95% CI) | 0.68 [0.57, 0.80] |
| 12.1 CBT | 44 | 2662 | Std. Mean Difference (Random, 95% CI) | 0.67 [0.54, 0.80] |
| 12.2 Exposure/ Desensitization | 13 | 440 | Std. Mean Difference (Random, 95% CI) | 0.77 [0.46, 1.08] |
| 12.3 Muscle relaxation | 2 | 90 | Std. Mean Difference (Random, 95% CI) | 0.66 [0.22, 1.09] |
| 13 Specific symptoms, self‐rated at post‐treatment (by setting) Show forest plot | 59 | 3192 | Std. Mean Difference (Random, 95% CI) | 0.68 [0.57, 0.80] |
| 13.1 Advertising | 48 | 2749 | Std. Mean Difference (Random, 95% CI) | 0.74 [0.62, 0.86] |
| 13.2 Referral Only | 9 | 350 | Std. Mean Difference (Random, 95% CI) | 0.38 [0.09, 0.67] |
| 13.3 Other | 2 | 93 | Std. Mean Difference (Random, 95% CI) | 0.53 [‐0.25, 1.31] |
| 14 Specific symptoms, self‐rated at post‐treatment (sensitivity analysis for bias) Show forest plot | 59 | 3192 | Std. Mean Difference (Random, 95% CI) | 0.68 [0.57, 0.80] |
| 14.1 High or Unclear Risk for Incomplete Outcome Data | 23 | 949 | Std. Mean Difference (Random, 95% CI) | 0.60 [0.44, 0.76] |
| 14.2 Low Risk for Incomplete Outcome Data | 36 | 2243 | Std. Mean Difference (Random, 95% CI) | 0.73 [0.57, 0.88] |
| 15 Specific symptoms, self‐rated at post‐treatment (studies also reporting clinician rated) Show forest plot | 59 | 3192 | Std. Mean Difference (Random, 95% CI) | 0.68 [0.57, 0.80] |
| 15.1 Not reporting clinician rated | 47 | 2674 | Std. Mean Difference (Random, 95% CI) | 0.68 [0.55, 0.81] |
| 15.2 Also reporting clinician rated | 12 | 518 | Std. Mean Difference (Random, 95% CI) | 0.70 [0.46, 0.95] |
| 16 Specific symptoms, clinician‐rated at post‐treatment Show forest plot | 15 | 629 | Std. Mean Difference (Random, 95% CI) | 1.04 [0.50, 1.58] |
| 17 Specific symptoms, clinician‐rated at post‐treatment Show forest plot | 15 | 629 | Std. Mean Difference (Random, 95% CI) | 1.04 [0.50, 1.58] |
| 17.1 Not reporting self‐rated | 3 | 112 | Std. Mean Difference (Random, 95% CI) | 1.39 [0.97, 1.80] |
| 17.2 Also reporting self‐rated | 12 | 517 | Std. Mean Difference (Random, 95% CI) | 0.97 [0.32, 1.61] |
| 18 General symptoms, self‐rated at post‐treatment Show forest plot | 48 | 3101 | Std. Mean Difference (Random, 95% CI) | 0.58 [0.43, 0.73] |
| 19 Behavioural Test (level), clinician‐rated at post‐treatment Show forest plot | 7 | 250 | Std. Mean Difference (Random, 95% CI) | 1.11 [0.36, 1.87] |
| 20 Behavioural Test (symptoms), self‐rated at post‐treatment Show forest plot | 5 | 196 | Std. Mean Difference (Random, 95% CI) | 0.59 [‐0.33, 1.51] |
| 21 Response (specific), self‐rated at post‐treatment Show forest plot | 21 | 1547 | Risk Ratio (M‐H, Random, 95% CI) | 2.34 [1.81, 3.03] |
| 22 Response (specific), self‐rated at post‐treatment (type of control) Show forest plot | 21 | 1547 | Risk Ratio (M‐H, Random, 95% CI) | 2.34 [1.81, 3.03] |
| 22.1 Adjunct | 1 | 95 | Risk Ratio (M‐H, Random, 95% CI) | 1.43 [0.91, 2.25] |
| 22.2 Attention controls | 7 | 526 | Risk Ratio (M‐H, Random, 95% CI) | 1.85 [1.39, 2.45] |
| 22.3 Treatment as Usual | 1 | 120 | Risk Ratio (M‐H, Random, 95% CI) | 1.33 [0.49, 3.61] |
| 22.4 Wait‐List | 12 | 806 | Risk Ratio (M‐H, Random, 95% CI) | 3.18 [2.16, 4.68] |
| 23 Response (specific), self‐rated at post‐treatment (by disorder) Show forest plot | 21 | 1751 | Risk Ratio (M‐H, Random, 95% CI) | 2.42 [1.88, 3.12] |
| 23.1 GAD | 4 | 342 | Risk Ratio (M‐H, Random, 95% CI) | 4.60 [2.75, 7.68] |
| 23.2 Mixed | 2 | 161 | Risk Ratio (M‐H, Random, 95% CI) | 2.00 [1.23, 3.24] |
| 23.3 OCD | 1 | 149 | Risk Ratio (M‐H, Random, 95% CI) | 2.36 [1.16, 4.82] |
| 23.4 Panic | 9 | 384 | Risk Ratio (M‐H, Random, 95% CI) | 1.73 [1.29, 2.30] |
| 23.5 Social | 4 | 695 | Risk Ratio (M‐H, Random, 95% CI) | 2.85 [1.59, 5.11] |
| 23.6 Specific | 1 | 20 | Risk Ratio (M‐H, Random, 95% CI) | 17.50 [1.16, 263.03] |
| 24 Response (specific), self‐rated at post‐treatment (by media) Show forest plot | 21 | 1751 | Risk Ratio (M‐H, Random, 95% CI) | 2.42 [1.88, 3.12] |
| 24.1 Book | 7 | 250 | Risk Ratio (M‐H, Random, 95% CI) | 1.85 [1.36, 2.52] |
| 24.2 Computer | 2 | 115 | Risk Ratio (M‐H, Random, 95% CI) | 3.75 [0.25, 56.38] |
| 24.3 Internet | 8 | 882 | Risk Ratio (M‐H, Random, 95% CI) | 3.76 [2.84, 4.97] |
| 24.4 Other | 4 | 504 | Risk Ratio (M‐H, Random, 95% CI) | 1.79 [1.08, 2.99] |
| 25 Response (specific), self‐rated at post‐treatment (by therapy) Show forest plot | 21 | 1547 | Risk Ratio (M‐H, Random, 95% CI) | 2.34 [1.81, 3.03] |
| 25.1 Applied relaxation | 0 | 0 | Risk Ratio (M‐H, Random, 95% CI) | 0.0 [0.0, 0.0] |
| 25.2 CBT | 19 | 1378 | Risk Ratio (M‐H, Random, 95% CI) | 2.30 [1.76, 3.01] |
| 25.3 Exposure/ Desensitization | 2 | 169 | Risk Ratio (M‐H, Random, 95% CI) | 4.36 [0.63, 29.98] |
| 25.4 Muscle relaxation | 0 | 0 | Risk Ratio (M‐H, Random, 95% CI) | 0.0 [0.0, 0.0] |
| 25.5 Self‐examination | 0 | 0 | Risk Ratio (M‐H, Random, 95% CI) | 0.0 [0.0, 0.0] |
| 26 Response (specific), self‐rated at post‐treatment (by setting) Show forest plot | 21 | 1547 | Risk Ratio (M‐H, Random, 95% CI) | 2.34 [1.81, 3.03] |
| 26.1 Advertising | 19 | 1386 | Risk Ratio (M‐H, Random, 95% CI) | 2.43 [1.82, 3.25] |
| 26.2 Referral Only | 2 | 161 | Risk Ratio (M‐H, Random, 95% CI) | 2.00 [1.23, 3.24] |
| 26.3 Other | 0 | 0 | Risk Ratio (M‐H, Random, 95% CI) | 0.0 [0.0, 0.0] |
| 27 Response (specific), clinician‐rated at post‐treatment Show forest plot | 11 | 922 | Risk Ratio (M‐H, Random, 95% CI) | 4.90 [3.12, 7.68] |
| 28 Response (general), self‐rated at post‐treatment Show forest plot | 8 | 622 | Risk Ratio (M‐H, Random, 95% CI) | 2.11 [1.13, 3.93] |
| 29 Recovery, clinician‐rated at post‐treatment Show forest plot | 9 | 605 | Risk Difference (M‐H, Random, 95% CI) | 0.40 [0.20, 0.60] |
| 30 Depression, self‐rated at post‐treatment Show forest plot | 51 | 3682 | Std. Mean Difference (Random, 95% CI) | 0.47 [0.36, 0.58] |
| 31 Depression, clinician‐rated at post‐treatment Show forest plot | 1 | 50 | Std. Mean Difference (Random, 95% CI) | 0.06 [‐0.49, 0.60] |
| 32 Depression improvement, self‐rated at post‐treatment Show forest plot | 5 | 224 | Risk Ratio (M‐H, Random, 95% CI) | 2.64 [1.36, 5.14] |
| 33 Disability, self‐rated at post‐treatment Show forest plot | 20 | 1629 | Std. Mean Difference (Random, 95% CI) | 0.48 [0.33, 0.63] |
| 34 Disability, clinician‐rated at post‐treatment Show forest plot | 5 | 293 | Std. Mean Difference (Random, 95% CI) | 0.56 [‐0.01, 1.14] |
| 35 Quality of Life, self‐rated at post‐treatment Show forest plot | 18 | 1344 | Std. Mean Difference (Random, 95% CI) | 0.36 [0.22, 0.49] |
| 36 Dropout at post‐treatment Show forest plot | 78 | 6059 | Risk Ratio (M‐H, Random, 95% CI) | 0.96 [0.94, 0.99] |
| 37 Anxiety (any measure), self‐rated at follow‐up Show forest plot | 15 | 684 | Std. Mean Difference (Random, 95% CI) | 0.49 [0.24, 0.75] |
| 38 Specific symptoms, self‐rated at follow‐up Show forest plot | 12 | 439 | Std. Mean Difference (Random, 95% CI) | 0.61 [0.27, 0.94] |
| 39 Specific symptoms, clinician‐rated at follow‐up Show forest plot | 5 | 98 | Std. Mean Difference (Random, 95% CI) | 1.12 [0.70, 1.53] |
| 40 General symptoms, self‐rated at follow‐up Show forest plot | 9 | 569 | Std. Mean Difference (Random, 95% CI) | 0.22 [0.06, 0.39] |
| 41 Behavioural Test (level), clinician‐rated at follow‐up Show forest plot | 1 | 30 | Std. Mean Difference (Random, 95% CI) | 1.23 [0.47, 1.99] |
| 42 Behavioural Test (symptoms), self‐rated at follow‐up Show forest plot | 1 | 30 | Std. Mean Difference (Random, 95% CI) | 0.60 [‐0.11, 1.31] |
| 43 Response (specific), self‐rated at follow‐up Show forest plot | 2 | 128 | Risk Ratio (M‐H, Random, 95% CI) | 2.12 [1.18, 3.79] |
| 44 Response (specific), clinician‐rated at follow‐up Show forest plot | 1 | 23 | Risk Ratio (M‐H, Random, 95% CI) | 3.93 [0.21, 73.71] |
| 45 Response (general), self‐rated at follow‐up Show forest plot | 1 | 95 | Risk Ratio (M‐H, Random, 95% CI) | 0.97 [0.51, 1.84] |
| 46 Recovery, clinician‐rated at follow‐up Show forest plot | 1 | 117 | Risk Ratio (M‐H, Random, 95% CI) | 0.91 [0.64, 1.28] |
| 47 Depression, self‐rated at follow‐up Show forest plot | 7 | 526 | Std. Mean Difference (Random, 95% CI) | 0.28 [0.05, 0.52] |
| 48 Depression, clinician‐rated at follow‐up Show forest plot | 1 | 50 | Std. Mean Difference (Random, 95% CI) | 0.32 [‐0.23, 0.87] |
| 49 Disability, self‐rated at follow‐up Show forest plot | 7 | 513 | Std. Mean Difference (Random, 95% CI) | 0.21 [‐0.15, 0.57] |
| 50 Disability, clinician‐rated at follow‐up Show forest plot | 3 | 108 | Std. Mean Difference (Random, 95% CI) | 0.53 [0.14, 0.93] |
| 51 Quality of Life, self‐rated at follow‐up Show forest plot | 3 | 211 | Std. Mean Difference (Random, 95% CI) | 0.08 [‐0.19, 0.35] |
| 52 Anxiety (any measure), self‐rated at >6m Show forest plot | 5 | 287 | Std. Mean Difference (Random, 95% CI) | 0.18 [‐0.15, 0.52] |
| 53 Specific symptoms, self‐rated at >6m Show forest plot | 5 | 287 | Std. Mean Difference (Random, 95% CI) | 0.20 [‐0.11, 0.51] |
| 54 General symptoms, self‐rated at >6m Show forest plot | 2 | 194 | Std. Mean Difference (Random, 95% CI) | ‐0.07 [‐0.63, 0.49] |
| 55 Response (specific), self‐rated at >6m Show forest plot | 2 | 148 | Risk Ratio (Random, 95% CI) | 1.51 [0.67, 3.36] |
| 56 Depression, self‐rated at >6m Show forest plot | 2 | 194 | Std. Mean Difference (Random, 95% CI) | ‐0.25 [‐0.58, 0.08] |
| 57 Disability, self‐rated at >6m Show forest plot | 3 | 147 | Std. Mean Difference (Random, 95% CI) | 0.14 [‐0.33, 0.60] |
| 58 Quality of Life, self‐rated at >6m Show forest plot | 1 | 115 | Std. Mean Difference (Random, 95% CI) | ‐0.06 [‐0.48, 0.36] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Anxiety (any measure), self‐rated at post‐treatment Show forest plot | 24 | 1360 | Std. Mean Difference (Random, 95% CI) | ‐0.23 [‐0.36, ‐0.09] |
| 2 Anxiety (any measure), self‐rated at post‐treatment (group versus individual) Show forest plot | 24 | 1360 | Std. Mean Difference (Random, 95% CI) | ‐0.23 [‐0.36, ‐0.09] |
| 2.1 Group CBT | 5 | 477 | Std. Mean Difference (Random, 95% CI) | ‐0.18 [‐0.44, 0.09] |
| 2.2 Individual CBT | 9 | 549 | Std. Mean Difference (Random, 95% CI) | ‐0.17 [‐0.34, 0.00] |
| 2.3 Individual Exposure | 10 | 334 | Std. Mean Difference (Random, 95% CI) | ‐0.39 [‐0.65, ‐0.13] |
| 3 Anxiety (any measure), self‐rated at post‐treatment (by disorder) Show forest plot | 24 | 1360 | Std. Mean Difference (Random, 95% CI) | ‐0.23 [‐0.36, ‐0.09] |
| 3.1 Mixed | 2 | 165 | Std. Mean Difference (Random, 95% CI) | ‐0.25 [‐0.56, 0.05] |
| 3.2 OCD | 2 | 147 | Std. Mean Difference (Random, 95% CI) | ‐0.26 [‐0.59, 0.06] |
| 3.3 Panic | 8 | 430 | Std. Mean Difference (Random, 95% CI) | ‐0.27 [‐0.51, ‐0.03] |
| 3.4 Social | 4 | 373 | Std. Mean Difference (Random, 95% CI) | 0.02 [‐0.18, 0.22] |
| 3.5 Specific | 8 | 245 | Std. Mean Difference (Random, 95% CI) | ‐0.39 [‐0.73, ‐0.06] |
| 4 Anxiety (any measure), self‐rated at post‐treatment (by media) Show forest plot | 24 | 1360 | Std. Mean Difference (Random, 95% CI) | ‐0.23 [‐0.36, ‐0.09] |
| 4.1 Book | 10 | 535 | Std. Mean Difference (Random, 95% CI) | ‐0.34 [‐0.60, ‐0.08] |
| 4.2 Computer | 4 | 136 | Std. Mean Difference (Random, 95% CI) | ‐0.38 [‐0.71, ‐0.04] |
| 4.3 Internet | 7 | 519 | Std. Mean Difference (Random, 95% CI) | ‐0.05 [‐0.22, 0.12] |
| 4.4 Other | 3 | 170 | Std. Mean Difference (Random, 95% CI) | ‐0.23 [‐0.53, 0.07] |
| 5 Anxiety (any measure), self‐rated at post‐treatment (by therapy) Show forest plot | 24 | 1360 | Std. Mean Difference (Random, 95% CI) | ‐0.23 [‐0.36, ‐0.09] |
| 5.1 CBT | 13 | 916 | Std. Mean Difference (Random, 95% CI) | ‐0.16 [‐0.31, ‐0.00] |
| 5.2 Exposure/ Desensitization | 10 | 419 | Std. Mean Difference (Random, 95% CI) | ‐0.37 [‐0.62, ‐0.13] |
| 5.3 Muscle relaxation | 1 | 25 | Std. Mean Difference (Random, 95% CI) | ‐0.14 [‐0.92, 0.65] |
| 6 Anxiety (any measure), self‐rated at post‐treatment (by setting) Show forest plot | 24 | 1360 | Std. Mean Difference (Random, 95% CI) | ‐0.23 [‐0.36, ‐0.09] |
| 6.1 Advertising | 17 | 916 | Std. Mean Difference (Random, 95% CI) | ‐0.15 [‐0.32, 0.02] |
| 6.2 Referral Only | 7 | 444 | Std. Mean Difference (Random, 95% CI) | ‐0.37 [‐0.56, ‐0.18] |
| 7 Anxiety (any measure), self‐rated at post‐treatment (sensitivity analysis for bias) Show forest plot | 24 | Std. Mean Difference (Random, 95% CI) | ‐0.23 [‐0.36, ‐0.09] | |
| 7.1 High or Unclear Risk for Incomplete Outcome Data | 11 | Std. Mean Difference (Random, 95% CI) | ‐0.28 [‐0.44, ‐0.13] | |
| 7.2 Low Risk for Incomplete Outcome Data | 13 | Std. Mean Difference (Random, 95% CI) | ‐0.18 [‐0.40, 0.04] | |
| 8 Specific symptoms, self‐rated at post‐treatment Show forest plot | 23 | 1255 | Std. Mean Difference (Random, 95% CI) | ‐0.23 [‐0.37, ‐0.10] |
| 9 Specific symptoms, self‐rated at post‐treatment (group versus individual) Show forest plot | 23 | Std. Mean Difference (Random, 95% CI) | ‐0.23 [‐0.37, ‐0.10] | |
| 9.1 Group CBT | 4 | Std. Mean Difference (Random, 95% CI) | ‐0.17 [‐0.52, 0.19] | |
| 9.2 Individual CBT | 9 | Std. Mean Difference (Random, 95% CI) | ‐0.18 [‐0.35, ‐0.02] | |
| 9.3 Individual Exposure | 10 | Std. Mean Difference (Random, 95% CI) | ‐0.40 [‐0.65, ‐0.14] | |
| 10 Specific symptoms, self‐rated at post‐treatment (by disorder) Show forest plot | 23 | 1255 | Std. Mean Difference (Random, 95% CI) | ‐0.23 [‐0.37, ‐0.10] |
| 10.1 Mixed | 2 | 165 | Std. Mean Difference (Random, 95% CI) | ‐0.24 [‐0.54, 0.07] |
| 10.2 OCD | 2 | 145 | Std. Mean Difference (Random, 95% CI) | ‐0.29 [‐0.62, 0.03] |
| 10.3 Panic | 7 | 327 | Std. Mean Difference (Random, 95% CI) | ‐0.30 [‐0.56, ‐0.04] |
| 10.4 Social | 4 | 373 | Std. Mean Difference (Random, 95% CI) | 0.03 [‐0.18, 0.23] |
| 10.5 Specific | 8 | 245 | Std. Mean Difference (Random, 95% CI) | ‐0.40 [‐0.74, ‐0.06] |
| 11 Specific symptoms, self‐rated at post‐treatment (by media) Show forest plot | 22 | 1275 | Std. Mean Difference (Random, 95% CI) | ‐0.26 [‐0.39, ‐0.12] |
| 11.1 Book | 10 | 553 | Std. Mean Difference (Random, 95% CI) | ‐0.42 [‐0.65, ‐0.20] |
| 11.2 Computer | 4 | 136 | Std. Mean Difference (Random, 95% CI) | ‐0.38 [‐0.71, ‐0.04] |
| 11.3 Internet | 6 | 416 | Std. Mean Difference (Random, 95% CI) | 0.01 [‐0.18, 0.20] |
| 11.4 Other | 3 | 170 | Std. Mean Difference (Random, 95% CI) | ‐0.23 [‐0.54, 0.07] |
| 12 Specific symptoms, self‐rated at post‐treatment (by therapy) Show forest plot | 23 | 1255 | Std. Mean Difference (Random, 95% CI) | ‐0.23 [‐0.37, ‐0.10] |
| 12.1 CBT | 12 | 811 | Std. Mean Difference (Random, 95% CI) | ‐0.16 [‐0.33, 0.01] |
| 12.2 Exposure/ Desensitization | 10 | 419 | Std. Mean Difference (Random, 95% CI) | ‐0.38 [‐0.62, ‐0.14] |
| 12.3 Muscle relaxation | 1 | 25 | Std. Mean Difference (Random, 95% CI) | ‐0.14 [‐0.92, 0.65] |
| 13 Specific symptoms, self‐rated at post‐treatment (by setting) Show forest plot | 23 | 1255 | Std. Mean Difference (Random, 95% CI) | ‐0.23 [‐0.37, ‐0.10] |
| 13.1 Advertising | 17 | 917 | Std. Mean Difference (Random, 95% CI) | ‐0.16 [‐0.32, 0.00] |
| 13.2 Referral Only | 6 | 338 | Std. Mean Difference (Random, 95% CI) | ‐0.42 [‐0.64, ‐0.21] |
| 14 Specific symptoms, self‐rated at post‐treatment (sensitivity analysis for bias) Show forest plot | 23 | 1255 | Std. Mean Difference (Random, 95% CI) | ‐0.23 [‐0.37, ‐0.10] |
| 14.1 High or Unclear Risk for Incomplete Outcome Data | 10 | 523 | Std. Mean Difference (Random, 95% CI) | ‐0.29 [‐0.47, ‐0.12] |
| 14.2 Low Risk for Incomplete Outcome Data | 13 | 732 | Std. Mean Difference (Random, 95% CI) | ‐0.20 [‐0.41, 0.02] |
| 15 Specific symptoms, clinician‐rated at post‐treatment Show forest plot | 6 | 418 | Std. Mean Difference (Random, 95% CI) | ‐0.13 [‐0.33, 0.06] |
| 16 General symptoms, self‐rated at post‐treatment Show forest plot | 11 | 627 | Std. Mean Difference (Random, 95% CI) | ‐0.08 [‐0.36, 0.20] |
| 17 Behavioural Test (level), clinician‐rated at post‐treatment Show forest plot | 5 | 211 | Std. Mean Difference (Random, 95% CI) | ‐0.44 [‐0.76, ‐0.13] |
| 18 Behavioural Test (symptoms), self‐rated at post‐treatment Show forest plot | 5 | 209 | Std. Mean Difference (Random, 95% CI) | ‐0.57 [‐1.35, 0.21] |
| 19 Response (specific), self‐rated at post‐treatment Show forest plot | 10 | 575 | Risk Ratio (M‐H, Random, 95% CI) | 0.78 [0.56, 1.09] |
| 20 Response (specific), self‐rated at post‐treatment (group versus individual) Show forest plot | 10 | 575 | Risk Ratio (M‐H, Random, 95% CI) | 0.78 [0.56, 1.09] |
| 20.1 Group CBT | 1 | 126 | Risk Ratio (M‐H, Random, 95% CI) | 1.61 [1.07, 2.44] |
| 20.2 Individual CBT | 7 | 382 | Risk Ratio (M‐H, Random, 95% CI) | 0.73 [0.53, 1.01] |
| 20.3 Individual Exposure | 2 | 67 | Risk Ratio (M‐H, Random, 95% CI) | 0.39 [0.02, 7.29] |
| 21 Response (specific), self‐rated at post‐treatment (by disorder) Show forest plot | 10 | 575 | Risk Ratio (M‐H, Random, 95% CI) | 0.78 [0.56, 1.09] |
| 21.1 OCD | 3 | 217 | Risk Ratio (M‐H, Random, 95% CI) | 0.59 [0.40, 0.87] |
| 21.2 Panic | 4 | 165 | Risk Ratio (M‐H, Random, 95% CI) | 0.82 [0.55, 1.23] |
| 21.3 Social | 1 | 126 | Risk Ratio (M‐H, Random, 95% CI) | 1.61 [1.07, 2.44] |
| 21.4 Specific | 2 | 67 | Risk Ratio (M‐H, Random, 95% CI) | 0.39 [0.02, 7.29] |
| 22 Response (specific), self‐rated at post‐treatment (by media) Show forest plot | 10 | 575 | Risk Ratio (M‐H, Random, 95% CI) | 0.78 [0.56, 1.09] |
| 22.1 Book | 6 | 224 | Risk Ratio (M‐H, Random, 95% CI) | 0.58 [0.30, 1.10] |
| 22.2 Computer | 1 | 33 | Risk Ratio (M‐H, Random, 95% CI) | 1.06 [0.74, 1.52] |
| 22.3 Internet | 2 | 175 | Risk Ratio (M‐H, Random, 95% CI) | 1.17 [0.61, 2.25] |
| 22.4 Other | 1 | 143 | Risk Ratio (M‐H, Random, 95% CI) | 0.61 [0.39, 0.95] |
| 23 Response (specific), self‐rated at post‐treatment (by therapy) Show forest plot | 10 | 575 | Risk Ratio (M‐H, Random, 95% CI) | 0.78 [0.56, 1.09] |
| 23.1 CBT | 7 | 365 | Risk Ratio (M‐H, Random, 95% CI) | 0.87 [0.60, 1.26] |
| 23.2 Exposure/ Desensitization | 3 | 210 | Risk Ratio (M‐H, Random, 95% CI) | 0.54 [0.22, 1.34] |
| 24 Response (specific), self‐rated at post‐treatment (by setting) Show forest plot | 10 | 575 | Risk Ratio (M‐H, Random, 95% CI) | 0.78 [0.56, 1.09] |
| 24.1 Advertising | 7 | 430 | Risk Ratio (M‐H, Random, 95% CI) | 0.90 [0.64, 1.27] |
| 24.2 Referral Only | 3 | 145 | Risk Ratio (M‐H, Random, 95% CI) | 0.48 [0.30, 0.76] |
| 25 Response (specific), clinician‐rated at post‐treatment Show forest plot | 9 | 663 | Risk Ratio (M‐H, Random, 95% CI) | 0.49 [0.31, 0.77] |
| 26 Response (general), self‐rated at post‐treatment Show forest plot | 1 | Risk Ratio (M‐H, Random, 95% CI) | Totals not selected | |
| 27 Recovery, clinician‐rated at post‐treatment Show forest plot | 6 | 587 | Risk Ratio (M‐H, Random, 95% CI) | 1.05 [0.88, 1.24] |
| 28 Depression, self‐rated at post‐treatment Show forest plot | 13 | 906 | Std. Mean Difference (Random, 95% CI) | ‐0.01 [‐0.22, 0.21] |
| 29 Depression, clinician‐rated at post‐treatment Show forest plot | 1 | Std. Mean Difference (Random, 95% CI) | Totals not selected | |
| 30 Depression improvement, self‐rated at post‐treatment Show forest plot | 2 | 73 | Risk Ratio (M‐H, Random, 95% CI) | 1.21 [0.74, 1.97] |
| 31 Disability, self‐rated at post‐treatment Show forest plot | 10 | 670 | Std. Mean Difference (Random, 95% CI) | ‐0.07 [‐0.33, 0.20] |
| 32 Disability, clinician‐rated at post‐treatment Show forest plot | 3 | 200 | Std. Mean Difference (Random, 95% CI) | 0.24 [‐0.04, 0.52] |
| 33 Quality of Life, self‐rated at post‐treatment Show forest plot | 4 | 282 | Std. Mean Difference (Random, 95% CI) | 0.08 [‐0.23, 0.38] |
| 34 Dropout at post‐treatment Show forest plot | 28 | 1852 | Risk Ratio (M‐H, Random, 95% CI) | 0.99 [0.95, 1.03] |
| 35 Anxiety (any measure), self‐rated at follow‐up Show forest plot | 11 | 677 | Std. Mean Difference (Random, 95% CI) | ‐0.14 [‐0.30, 0.01] |
| 36 Specific symptoms, self‐rated at follow‐up Show forest plot | 10 | 573 | Std. Mean Difference (Random, 95% CI) | ‐0.15 [‐0.31, 0.02] |
| 37 Specific symptoms, clinician‐rated at follow‐up Show forest plot | 3 | 185 | Std. Mean Difference (Random, 95% CI) | ‐0.19 [‐0.69, 0.30] |
| 38 General symptoms, self‐rated at follow‐up Show forest plot | 5 | 308 | Std. Mean Difference (Random, 95% CI) | ‐0.10 [‐0.30, 0.09] |
| 39 Behavioural Test (level), clinician‐rated at follow‐up Show forest plot | 2 | 60 | Std. Mean Difference (Random, 95% CI) | ‐0.02 [‐0.74, 0.71] |
| 40 Behavioural Test (symptoms), self‐rated at follow‐up Show forest plot | 1 | Std. Mean Difference (Random, 95% CI) | Totals not selected | |
| 41 Response (specific), self‐rated at follow‐up Show forest plot | 4 | 224 | Risk Ratio (M‐H, Random, 95% CI) | 0.99 [0.73, 1.35] |
| 42 Response (specific), clinician‐rated at follow‐up Show forest plot | 3 | 280 | Risk Ratio (M‐H, Random, 95% CI) | 1.10 [0.83, 1.45] |
| 43 Recovery, clinician‐rated at follow‐up Show forest plot | 3 | 354 | Risk Ratio (M‐H, Random, 95% CI) | 1.04 [0.70, 1.56] |
| 44 Depression, self‐rated at follow‐up Show forest plot | 5 | 403 | Std. Mean Difference (Random, 95% CI) | 0.09 [‐0.19, 0.36] |
| 45 Depression, clinician‐rated at follow‐up Show forest plot | 1 | Std. Mean Difference (Random, 95% CI) | Totals not selected | |
| 46 Depression improvement, self‐rated at follow‐up Show forest plot | 1 | Risk Ratio (M‐H, Random, 95% CI) | Totals not selected | |
| 47 Disability, self‐rated at follow‐up Show forest plot | 5 | 337 | Std. Mean Difference (Random, 95% CI) | ‐0.10 [‐0.39, 0.19] |
| 48 Disability, clinician‐rated at follow‐up Show forest plot | 2 | 145 | Std. Mean Difference (Random, 95% CI) | 0.18 [‐0.30, 0.65] |
| 49 Quality of Life, self‐rated at follow‐up Show forest plot | 2 | 143 | Std. Mean Difference (Random, 95% CI) | 0.40 [0.07, 0.73] |
| 50 Anxiety (any measure), self‐rated at >6m Show forest plot | 8 | 423 | Std. Mean Difference (Random, 95% CI) | ‐0.16 [‐0.50, 0.18] |
| 51 Specific symptoms, self‐rated at >6m Show forest plot | 8 | 423 | Std. Mean Difference (Random, 95% CI) | ‐0.20 [‐0.54, 0.14] |
| 52 Specific symptoms, clinician‐rated at >6m Show forest plot | 1 | Std. Mean Difference (Random, 95% CI) | Totals not selected | |
| 53 General symptoms, self‐rated at >6m Show forest plot | 4 | 241 | Std. Mean Difference (Random, 95% CI) | 0.04 [‐0.22, 0.30] |
| 54 Behavioural Test (level), clinician‐rated at >6m Show forest plot | 2 | 122 | Std. Mean Difference (Random, 95% CI) | ‐0.11 [‐0.97, 0.76] |
| 55 Behavioural Test (symptoms), self‐rated at >6m Show forest plot | 3 | 162 | Std. Mean Difference (Random, 95% CI) | ‐0.46 [‐1.19, 0.27] |
| 56 Response (specific), self‐rated at >6m Show forest plot | 3 | 116 | Risk Ratio (M‐H, Random, 95% CI) | 0.58 [0.20, 1.64] |
| 57 Response (specific), clinician‐rated at >6m Show forest plot | 2 | 82 | Risk Ratio (M‐H, Random, 95% CI) | 0.54 [0.16, 1.80] |
| 58 Response (general), self‐rated at >6m Show forest plot | 1 | Risk Ratio (M‐H, Random, 95% CI) | Totals not selected | |
| 59 Recovery, clinician‐rated at >6m Show forest plot | 2 | 147 | Risk Ratio (M‐H, Random, 95% CI) | 0.90 [0.64, 1.25] |
| 60 Depression, self‐rated at >6m Show forest plot | 5 | 339 | Std. Mean Difference (Random, 95% CI) | 0.08 [‐0.14, 0.30] |
| 61 Depression improvement, self‐rated at >6m Show forest plot | 1 | Risk Ratio (M‐H, Random, 95% CI) | Totals not selected | |
| 62 Disability, self‐rated at >6m Show forest plot | 3 | 242 | Std. Mean Difference (Random, 95% CI) | ‐0.21 [‐0.47, 0.04] |
| 63 Quality of Life, self‐rated at >6m Show forest plot | 1 | Std. Mean Difference (Random, 95% CI) | Totals not selected | |
