Treating Anxiety with CBT: The Evidence

Generally considered a short-term therapy, Cognitive Behavioural Therapy (CBT) often consists of about 8 to 12 sessions in which client and therapist work collaboratively to identify problem thoughts and behaviours (click here to learn more about CBT’s principles and practices). CBT is considered the gold standard in the psychotherapeutic treatment of anxiety disorders and several meta-analyses (i.e., studies reviewing the results of multiple studies) have been published in recent years regarding the efficacy and effectiveness of CBT for anxiety (Stewart & Chambless, 2009; Hoffman & Smits, 2008; Norton & Price, 2007).

In this article, we briefly explore the evidence for treating anxiety with CBT.

Controlled and uncontrolled effect sizes

When researchers perform a meta-analysis, treatment efficacy is quantified in terms of an effect size, indicating the magnitude of an observed effect in a standard unit of measurement. Different types of effect sizes can be used to evaluate the available evidence. A “controlled” effect size expresses the magnitude of a specific treatment effect as compared with alternative treatments or control conditions. It is usually calculated by subtracting the post-treatment mean of the control group from the post-treatment mean of the treatment group divided by the pooled standard deviation. This effect size is called “Cohen’s d” (Otte, 2011). An “uncontrolled” effect size, in contrast, expresses the magnitude of improvement within a group from pre-treatment to post-treatment. This is calculated by subtracting a group’s post-treatment mean from its pre-treatment mean divided by the pooled standard deviation. Uncontrolled effect sizes are less preferable than controlled ones, as they are vulnerable to threats of internal validity (Otte, 2011).

This brief look at the relevant literature will present meta-analytically-derived controlled and uncontrolled effect sizes reflecting the efficacy and effectiveness results for the chief anxiety disorders: GAD, Panic Disorder (PD), and Social Anxiety Disorder (SAD). We also include Post-traumatic Stress Disorder (PTSD) and Obsessive-Compulsive Disorder (OCD). When the DSM-5 came out in 2013 (click here learn more about the differences between the DSM-IV and the DSM-5), it added a category called Trauma and Stressor-Related Disorders and PTSD is now listed there. It also added a category of Obsessive-Compulsive and Related Disorders, and OCD is listed there (APA, 2013a, 2013b). At the time the meta-analyses were conducted, however, clinicians were working from the DSM-IV, which lists both PTSD and OCD as anxiety disorders; thus we include here the results about them.

Large effect sizes, but there are limitations

Meta-analytic reviews of studies treating CBT for anxiety disorders have typically produced large effect sizes for most of the studies reviewed, concluding that CBT is highly effective (Olatunji, Cisler, & Deacon, 2010; Butler, Chapman, Forman, & Beck, 2006). The meta-analyses are not without limitations, however. Norton and Price (2007) noted that many meta-analyses focused only on one diagnosis (say, GAD or PD), thereby disallowing diagnostic comparisons. The main problem flagged by Otte (2011) in his review of current meta-analyses is that the studies included in the analyses have varied greatly with respect to control procedures. These have included waitlist, alternative treatments, and placebo interventions evaluated with or without randomisation.

Some studies did not even include control groups, yet including a control condition (and the specific nature of it) may greatly impact the efficacy results of CBT in anxiety-disorder treatments. Moreover, many studies have not taken into account how results derived from research studies in well-controlled research designs (that is: the question of efficacy) will generalise to real-world settings in naturalistic surroundings such as the counsellor’s therapy rooms (the question of effectiveness). In this brief look at where the research stands on the use of CBT for anxiety disorders, we focus on several recent meta-analyses: those by Norton and Price (2007), Hoffman & Smits (2008), and Stewart and Chambless (2009).

Norton and Price: Strong treatment effects across all diagnoses

Including in their meta-analysis studies of randomised, clinical trials of CBT for adults with any anxiety orders except specific phobias, Norton and Price reviewed 108 studies which had at least one treatment component that could be defined as cognitive therapy, exposure therapy, or some combination of either or both of those with added relaxation training (which was not deemed to be CBT and thus not included if it was the only treatment component. Relaxation was included in some studies as the control treatment when it was given to different clients than those who received the cognitive and/or exposure treatments). Weighted effect sizes of the various combinations of CBT treatments were computed from pre- to post-treatment and from post-treatment to follow-up. The authors ultimately had effect sizes from 169 separate treatment conditions and 109 post-treatment to follow-up effect sizes.

They found that pre- to post-treatment, there was no main effect of treatment component (that is, cognitive therapy was not, for instance, deemed better than exposure therapy), nor was there an interaction of treatment component by diagnosis (i.e., significantly greater effect size for the treatment condition of, say, cognitive therapy plus exposure with PD than for GAD). For pre- to post-treatment, effect sizes for CBT were significantly greater than the no-treatment effects sizes, and this effect size did not vary by diagnosis. Further, pre- to post-treatment effects sizes for CBT were significantly greater than the expectancy control effects sizes, and this main effect did not vary by diagnosis. There were no differences in how well studies of different diagnoses maintained their gains post-treatment to follow-up, nor were there significant differences post-treatment to follow-up among the various treatment components.

In summary, the main results of the meta-analysis supported the efficacy of CBT techniques, showing significantly larger treatment effect sizes than no-treatment or placebo across all of the anxiety disorders for cognitive therapy or exposure therapy alone, in combination, or combined with relaxation. Norton and Price concluded that any CBT component is likely to be effective, irrespective of diagnosis, and these effects are likely to be sustained after treatment cessation (Norton and Price, 2007).

Hofmann/Smits and Stewart/Chambless meta-analyses: Efficacy and effectiveness

Although the Norton/Price meta-analysis attempted to enforce rigorous research standards including, for example, only studies published in peer-review journals which used random assignments of clients to conditions and had at least one control or comparison condition, the meta-analysis nevertheless was limited by biased control condition effect sizes. This occurred in that, given the small number of conditions from which the effect size estimates were obtained, the values were appropriately used only as “best guess” estimates of no-treatment, expectancy control, and relaxation effects.

Hofmann and Smits: Psychological placebo controls

Conversely, the meta-analysis by Hofmann/Smits limited the included studies to randomised, placebo-controlled trials: the gold standard in clinical outcome research. In the Hofmann/Smits work, the criterion for inclusion was not only that there was a placebo control group. In addition, the placebo had to be a psychological one. That is, no study was included which administered a pill to the control group. Moreover, the psychological placebos allowed for inclusion had to involve interventions to control for nonspecific factors. This means that there had to be regular contact with a therapist, a reasonable rationale for the intervention, and discussions of the psychological problem the client was experiencing.

The study authors acknowledged that it is almost impossible to protect the “blind” in placebo-controlled psychotherapy trials (that is: to know who was getting the “treatment” and who was in the “control” group), but such a design is still the most rigorous and conservative test of the effects of an active treatment. Thus, the meta-analysis is able to assess the overall efficacy of CBT in anxiety disorders under well-controlled research conditions. The Hofmann/Smits meta-analysis found 27 studies meeting inclusion criteria: 7 for SAD, 6 for PTSD, 5 for PD, 3, for OCD and 2 for GAD. The researchers used a variation of Cohen’s d to take into account the small sample sizes (Hofmann & Smits, 2008).

Stewart and Chambless: Effectiveness in session

The discerning student of research might well comment that high-calibre laboratory studies are crucial to establish the efficacy of a given treatment, such as CBT, but what happens in the real world of clinical settings, where disorder severity may be greater, there may be greater co-morbid conditions, and/or other factors may be less amenable to control than in strict research settings? How does CBT work in these? Studies in naturalistic environments are impacted by the reality that, unlike randomised, controlled trials which are manualised and strictly adhered to, clinicians in practice settings may not use manuals at all.

Beyond that, practitioners are unlikely to have access to the intensive training, monitoring, and supervision available to therapists in research settings. Clinicians in research settings may be especially expert in the administration of certain treatments under study, and are motivated to stay consistent with protocols because of adherence measures. All of these factors may mean that treatments delivered in naturalistic settings are not as rigorous in terms of content or quality and studies of controlled research results are thus less able to be generalised to actual clinical practice. It is therefore important to examine how well findings from research studies (efficacy) translate into real-world settings (effectiveness).

Well aware of the difference in effect sizes between efficacy studies (in controlled research settings) and those of effectiveness (in naturalistic, “normal” clinical environments), Otte (2011) extensively reviewed both the Hofmann/Smits meta-analysis and that of Stewart and Chambless (2009), who examined 56 effectiveness studies of CBT. In these, CBT was defined broadly and included any treatment with cognitive, behavioural (e.g., exposure), or a combination of therapy components. The 56 were comprised of: 17 for PD, 11 each for SAD, OCD, and GAD, and 6 for PTSD. The following results contrast the effect sizes (gained through the meta-analyses) showing the efficacy and effectiveness results for each of the anxiety disorders (Otte, 2011).

Generalised anxiety disorder (GAD)

This condition characterised by excessive, uncontrollable worry is maintained in a client by selective attention to stimuli which seem threatening and the use of tension and overly cautious behaviours to forestall catastrophic images and the associated bodily arousal. Practitioners treating GAD use cognitive therapy to address the worry and excessive attention to threat, relaxation to address tension, visualisations to expose clients to catastrophic images, and exposure to stressful situations. Meanwhile they prevent overly cautious responses. The effect sizes were: Efficacy: Controlled effect size: 0.51 (95 percent confidence interval (CI) 0.05-0.97), a medium effect. Uncontrolled effect size: 1.80, a large effect. Effectiveness: 0.92 percent (95 percent CI 0.77-1.07).

Panic Disorder (PD)

Panic attacks include sudden symptoms such as palpitations, chest pains, sweating, trembling, dizziness, nausea, shortness of breath, feelings of choking, chills or hot flushes, paraesthesias, depersonalisation or derealisation, and fear of dying or losing control. When the attacks have at least once been unexpected and followed by at least one month of fearful concern about the consequences of an attack, a diagnosis of PD can be made.

Agoraphobia (fear of being in situations from which one cannot easily escape) often accompany or follow the panic attacks. CBT for this anxiety disorder generally involves psychoeducation about the nature and physiology of the panic response, cognitive techniques to modify misinterpretations that panic symptoms and consequences will be catastrophic, and gradually increased exposure to avoided situations and to panic-related bodily sensations (called “interoceptive” exposure).

Efficacy: Controlled effect size: 0.35 percent (95 percent CI 0.04-0.65), a small to medium effect. Uncontrolled effect size: 1.53 (from the Norton & Price meta-analysis, 2007; the Hofmann & Smits, 2008, meta-analysis did not have a study using CBT with PD with an uncontrolled effect size). Effectiveness: Effect size: 1.01 (95 percent CI 0.77-1.25) for panic attacks and Effect size: 0.83 (95 percent CI0.60-1.06) for avoidance.

Social anxiety disorder (SAD)

Plagued by a strong fear of performance, excessive fear of scrutiny, and fear of acting in an embarrassing way, sufferers of SAD are oversensitive to the (assumed) opinion of others; they typically have low self-esteem. Enduring an actual feared situation or even anticipating it can trigger physical symptoms such as blushing, sweating, or trembling (which in themselves can trigger more anxiety as people fear the consequences of these).

Cognitive restructuring and in vivo exposure to feared social situations are generally the techniques employed, as clients identify and challenge their beliefs about their social competence and the probability that their social skills will be evaluated negatively. In vivo exposure affords opportunities to come face to face with feared, avoided social situations and to practice social skills.

Efficacy: Controlled effect size: 0.62 (95 percent CI 0.39-0.86), a medium effect. Uncontrolled effect size: 1.27 (from the Norton & Price meta-analysis, 2007). Effectiveness: Uncontrolled effect size: 1.04 (95 percent CI 0.79-1.29).

Post-traumatic stress disorder (PTSD)

As noted above, the DSM-5 includes PTSD as part of a new chapter on trauma- and stressor-related disorders. We include it here because it has for so long been regarded as an anxiety disorder that there have been numerous anxiety studies examining outcomes of various treatments on PTSD, and most meta-analyses examining anxiety disorders include a section on it. Because the use of CBT with anxiety disorders is so well studied, there were numerous studies included in all three of the CBT-focused meta-analyses reviewed here which included data on PTSD. The criteria in the DSM-5 vary somewhat from the DSM-IV. They are that the trigger to PTSD has been exposure to actual or threatened death, serious injury, or sexual violation and that the exposure must result from one or more of the following scenarios, in which the individual:

  • directly experiences the traumatic event;
  • witnesses the traumatic event in person;
  • learns that the traumatic event occurred to a close family member or close friend (with the actual or threatened death being either violent or accidental); or
  • experiences first-hand repeated or extreme exposure to aversive details of the traumatic event (not through media, pictures, television or movies unless work-related).

The disturbance, regardless of its trigger, causes clinically significant distress or impairment in the individual’s social interactions, capacity to work, or other important areas of functioning. It is not the physiological result of another medical condition, medication, drugs or alcohol (APA, 2013a). Typically, CBT for PTSD requires the components of: (1) psychoeducation about the nature of fear, anxiety, and PTSD; (2) controlled, prolonged exposure to stimuli related to the traumatic event; and (3) cognitive restructuring, processing, or challenging of maladaptive beliefs.

Efficacy: Controlled effect size: 0.62 (95 percent CI 0.28-0.96), a medium effect. Uncontrolled effect size: 1.86 (from a different meta-analysis, i.e., Norton & Price, 2007). Effectiveness: Uncontrolled effect size of 2.59 (95 percent CI 2.06-3.13).

Obsessive-compulsive disorder (OCD)

As we mentioned earlier, OCD, like PTSD, has a different home in the DSM-5. It is listed now in the new section on Obsessive-Compulsive and Related Disorders (APA, 2013a). When a person experiences the presence of recurrent obsessions (that is, persistent thoughts, impulses, or images) or compulsions (repetitive behavior or thought patterns engaged in an attempt to prevent anxiety) that take more than an hour a day or cause significant distress or impairment, that person is said to have OCD. The person recognises that the patterns are excessive. Thus treating OCD with CBT involves the components of exposure, response prevention, and cognitive interventions. In the meta-analyses under review, OCD had a large effect size:

Efficacy: Controlled effect size: 1.37 (95 percent CI 0.64-2.20), a large effect. Uncontrolled effect size: 1.50 (calculated in a separate meta-analysis). Effectiveness: Uncontrolled effect size: 1.32 (95 percent CI 1.19-1.45).

Summary of the evidence

Having scrutinised the evidence in three major meta-analyses concerning use of CBT treatments with the various anxiety disorders in randomised controlled and uncontrolled trials and in real-life settings, we see that both the efficacy and effectiveness of CBT for anxiety in adults appears to be solidly grounded.

The controlled effect sizes ranged from small in PD to large in OCD, meaning that CBT compared favourably to placebo conditions in all anxiety disorders. In the effectiveness studies of CBT with anxiety disorders in clinical practice (i.e., naturalistic) settings, uncontrolled effect sizes ranged from 0.92 for GAD to 2.59 in PTSD; these indicated that CBT also works “in the real world” in the treatment of anxiety. Thus the reviewed meta-analyses confirm that CBT is far and away the most widely studied psychotherapeutic treatment for anxiety disorders, and it thus re-earns its already considerable reputation as the “gold standard” treatment (Otte, 2011).

© 2014 Mental Health Academy

This article was adapted from the upcoming Mental Health Academy CPD course “Using CBT with Generalised Anxiety Disorder”. Click here to learn more about MHA.


  • American Psychiatric Association (APA). (2013a). Highlights of changes from DSM-IV-TR to DSM-5. American Psychiatric Publishing. Retrieved on 14 July, 2014, from: hyperlink.
  • American Psychiatric Association (APA). (2013b). Post-traumatic stress disorder. American Psychiatric Publishing. Retrieved on 3 July, 2014, from: hyperlink.
  • Butler, A.C., Chapman, J.E., Forman, E.M., & Beck, A.T. (2006). The empirical status of cognitive behavioral therapy: A review of meta-analyses. Clinical psychology review. 2006; 26: 17-31.
  • Hofmann, S.G. & Smits, J.A. Cognitive-behavioral therapy for adult anxiety disorders: a meta-analysis of randomized placebo-controlled trials. Journal of Clinical Psychiatry. 2008; 69: 621-632.
  • Norton, P.J. & Price, E.C. (2007). A meta-analytic review of adult cognitive-behavioral treatment outcome across the anxiety disorders. Journal of Nervous and Mental Disease. 2007; 195: 521-531.
  • Olatunji, B.O., Cisler, J.M., & Deacon, B.J. (2010). Efficacy of cognitive behavioral therapy for anxiety disorders: A review of meta-analytic findings. Psychiatry Clinics of North America. 2010; 33: 557-577.
  • Otte, C. (2011). Cognitive behavioral therapy in anxiety disorders: current state of the evidence. Dialogues in Clinical Neuroscience. 2011; 13(4): 413-421. Retrieved on 7 July, 2014, from: hyperlink.
  • Stewart, R.E. & Chambless, D.L. (2009). Cognitive-behavioral therapy for adult anxiety disorders: a meta-analysis of effectiveness studies. Journal of Consulting and Clinical Psychology. 2009; 77: 595-606.