Skip to content
MyFreud

By , Editorial Team Published · Updated

CBT measured across 375 trials, and it is not one number

A unified meta-analysis put CBT for eleven disorders on the same scale. Effects ranged from large for PTSD and phobias to modest for bipolar and psychosis.

3 min read

Pop-art illustration of two people talking across a small table in a therapy session.

Key takeaways

  • CBT is not one treatment with one effect. The size of the benefit varies several-fold depending on the disorder.
  • Largest effects were for PTSD and specific phobia. Smallest were for bipolar and psychotic disorders, where CBT is an addition to medication rather than a replacement.
  • Every comparison was against an inactive control such as a waiting list, which reliably makes effects look larger than comparisons against another active treatment.
  • The value of the study is the uniform method. Applying identical rules to all eleven disorders makes the numbers comparable in a way that separate reviews never are.
  • A smaller effect does not mean do not bother. For psychosis, CBT is added on top of medication, and any gain on top of an existing treatment is a different kind of result.

“CBT works” is one of the most repeated claims in mental health, and it is approximately true and not very informative. A series of meta-analyses published in JAMA psychiatry set out to make it precise by asking the same question, the same way, about eleven different conditions. [pubmed-trauma-jul13-2026-source]

What the researchers did

Separate meta-analyses of CBT already existed for most disorders, but each used its own inclusion criteria, its own approach to assessing study quality, and its own statistics. That made cross-disorder comparison impossible: a bigger number in the OCD review than the depression review might reflect the treatment or might reflect the methodology.

This team applied one uniform method to all of them: identical search procedures, identical risk-of-bias assessment using the Cochrane tool, and identical meta-analytic technique.

The result covers 375 trials, 423 comparisons and 32,968 patients, across four anxiety disorders, two eating disorders, major depression, OCD, PTSD, and psychotic and bipolar disorders. Average patient age was 43, and about two thirds were women.

Every comparison was CBT against an inactive control, meaning a waiting list or similar.

What they found

The effects sort into three tiers.

How large the CBT effect was, by disorder Reported figures
0 0 1 1 1 Effect size band 1.2 PTSD 1.2 Specific phobia
0 0 1 1 1 Effect size band 0.75 Panic disorder 0.75 Social anxiety 0.75 Generalised anxiety 0.75 Depression 0.75 OCD 0.75 Bulimia and binge eating
0 0 1 1 1 Effect size band 0.4 Bipolar disorder 0.4 Psychotic disorders

Bands as reported by Cuijpers and colleagues (2025), JAMA psychiatry. Bars show the band each disorder fell into rather than its exact point estimate, and all comparisons were against inactive controls.

Largest, above 1.0: PTSD and specific phobia.

Moderate, between 0.5 and 1.0: panic disorder, social anxiety, generalised anxiety, depression, OCD, bulimia nervosa and binge eating disorder.

Smallest, below 0.5: bipolar and psychotic disorders.

Reading the ranking without over-reading it

Three things shape those numbers besides how well CBT works.

What the control group does. These are comparisons against no treatment. PTSD and specific phobias tend not to improve much on their own over a trial’s duration, so the untreated group stays put and the gap at the end is wide. Depression, by contrast, often improves somewhat without treatment, which narrows the measured difference even if the therapy is working equally well.

What CBT is being added to. For bipolar and psychotic disorders, CBT is delivered alongside medication, not instead of it. The effect being measured is what talking therapy adds on top of a treatment that is already doing most of the work. That is a harder bar, and a small number there means something different from a small number where CBT is the whole treatment.

Inactive controls inflate everything. Every figure here is CBT versus nothing. Against another active treatment, differences shrink considerably. These are real numbers answering a real question; they are not estimates of how much better CBT is than the alternatives.

Why it is a useful paper anyway

Because it replaces a slogan with a shape.

“CBT works” flattens a several-fold difference in expected benefit into a single claim. Knowing that the evidence is strongest for PTSD and phobias, solid for the common anxiety and mood disorders, and more modest as an add-on for bipolar and psychosis is genuinely better information for anyone deciding what to ask for or what to expect.

It also demonstrates something about evidence generally: the comparison group and the method determine the number at least as much as the treatment does. Two reviews reporting different effects for the same therapy may simply have counted differently.

If you are considering CBT

The practical reading is that CBT has good evidence across a wide range of conditions, that the size of the benefit depends on which condition, and that “evidence-based” describes a range rather than a promise to any one person.

Our trauma and PTSD guide covers the trauma-focused therapies at the top of this list, and our anxiety guide covers the disorders in the middle band.

The source

These findings are drawn from “Cognitive Behavior Therapy for Mental Disorders in Adults: A Unified Series of Meta-Analyses” (Cuijpers P, Harrer M, Miguel C, et al., 2025), published in JAMA psychiatry. Read the full study on PubMed.

Frequently asked questions

What did this study add that previous reviews did not?

Comparability. There were already meta-analyses of CBT for depression, for OCD, for PTSD and so on, but each used its own inclusion rules, its own bias assessment and its own statistical choices, so their numbers could not honestly be placed side by side. This team applied one method to all eleven disorders at once, which is what makes the ranking meaningful rather than an artefact of methodology.

Why does comparing against a waiting list inflate the effect?

Because people on a waiting list receive nothing, know they are receiving nothing, and have no expectation of improvement. The gap therefore includes everything that comes with getting treatment at all, not just the specific techniques. Comparisons against another active treatment produce consistently smaller differences. These numbers are best read as CBT versus nothing, which is a real and useful question, but not the same as CBT versus an alternative.

Why is the effect for PTSD so large?

Partly because trauma-focused CBT is genuinely effective, and partly because untreated PTSD tends not to improve much on its own over the timescale of a trial. When the comparison group stays roughly where it started, the difference at the end looks bigger. The same logic contributes to specific phobia, where an untreated fear of needles or heights does not fade by itself.

Does a smaller effect mean CBT is not worth it for bipolar or psychosis?

No, and the number needs reading in context. For those conditions CBT is delivered alongside medication rather than instead of it, so the effect measured is what talking therapy adds on top of a treatment that is already working. An improvement on top of existing care is a different and harder thing to achieve than an improvement over nothing at all.

How should I use this if I am choosing treatment?

As a rough guide to expectations rather than a decision rule. It tells you that CBT has strong evidence across a wide range of conditions and that its size varies. What it does not tell you is how CBT compares with medication, with another therapy, or with a combination for your particular situation, because those comparisons were not part of this analysis.

References

  1. 1.Cuijpers P, Harrer M, Miguel C, et al. ( 2025). Cognitive Behavior Therapy for Mental Disorders in Adults: A Unified Series of Meta-Analyses. JAMA psychiatry. pubmed.ncbi.nlm.nih.gov . doi:10.1001/jamapsychiatry.2025.0482

Related articles

Trauma and PTSD: Symptoms, Causes, and How It's Treated

A 2026 guide to psychological trauma and PTSD: how it presents, how it differs from complex PTSD, and which therapies currently have the best evidence.

Does Talking About Trauma Help? What Research Shows

Talking about a trauma right after it happens can make things worse. What the evidence says about debriefing, which therapies work, and when to seek help.

Hypervigilance: A Setting, Not a Personality

Hypervigilance is threat detection stuck switched on. It is a symptom rather than a diagnosis, it was correct once, and the treatable part is the threshold.

Coercive Control: Abuse That Leaves No Marks

A pattern of control, monitoring and isolation is abuse, and a growing number of countries now treat it as a crime with no violence required at all.