does cochrane rate the evidence for cbt as at best low to moderate certainty?
“The Cochrane reviews, which are a systematic attempt to assess how strong the evidence for particular approaches to disease of all sorts are say that at best the evidence for CBT is of low and medium confidence, you know?”
a guest, historian of psychiatry (Andrew Scull) said on Lex Fridman Podcast, 2026-09-20, at 17:37. hear it in context → the speaker cited: Cochrane reviews.
first, the part that holds
The speaker has this right. In every Cochrane review of cognitive behavioral therapy we opened, the best evidence was rated moderate (Cochrane's word for what the speaker called "medium"). Most outcomes were rated low or very low, and none were rated high.
what we found when we went and looked
We read five Cochrane reviews of cognitive behavioral therapy, covering anxiety in children, anxiety in older adults, chronic pain, tinnitus and insomnia in people with cancer. No outcome in any of them was rated high certainty. The best rating was moderate: for example, 49.4% remission with therapy against 17.8% on a waitlist in childhood anxiety, and small benefits for pain and distress in chronic pain. Many other outcomes were low or very low. So the speaker's "at best low and medium" matches what we read. What the claim leaves out is that most of these reviews still found a benefit. A low or moderate rating means Cochrane isn't sure how big the benefit is. It does not mean the therapy was shown not to work.
assertion by assertion
| assertion | verdict | what the record says |
|---|---|---|
| Cochrane reviews rate the evidence for cognitive behavioral therapy as, at best, low to moderate certainty (no high-certainty ratings) | confirmed | No outcome in any of the five reviews was rated high. The top rating was moderate: remission against waitlist in childhood anxiety, pain and distress in chronic pain, tinnitus impact against audiological care, and time to fall asleep in cancer-related insomnia. Limitation: five reviews is not every Cochrane review of this therapy. [1][2][3][4][5] |
| the speaker's 'medium confidence' matches Cochrane's 'moderate certainty' category | confirmed | Cochrane grades evidence on four levels using a system called GRADE: high, moderate, low and very low. The speaker's 'medium' is the moderate level. Many outcomes in these reviews fall below it, at very low, mostly for harms, long-term follow-up and some comparisons against active treatments. [2][3][4][5] |
| popular framing: low/moderate certainty ratings mean Cochrane found CBT doesn't work | false | Most of the reviews found a benefit and then rated how sure they could be of it. Childhood anxiety: probably increases remission (49.4% vs 17.8%, number needed to treat 3, moderate). Chronic pain: small benefits for pain and distress against no treatment (moderate). Older adults with anxiety: a small-to-moderate reduction after treatment (low certainty), with little to no difference at three to six months. [1][2][5] |
| context: the wider review literature on CBT is methodologically weak | confirmed | A meta-review of 494 systematic reviews of this therapy judged 351 of them to be of low or critically low quality. That grades how well each review was conducted, which is a different scale from Cochrane's certainty ratings, and it covers reviews from many publishers, not only Cochrane. [6] |
both directions
for the speaker: Each Cochrane review we read stops at moderate certainty, and most outcomes sit lower. Reasons given include risk of bias, heterogeneity, imprecision and poorly reported harms. Chronic pain: moderate for pain and distress, low for disability, very low for adverse effects. Older adults with anxiety: low or very low on every outcome. Cancer-related insomnia: one moderate outcome; the rest low or very low. Tinnitus: the summary itself says no high-certainty evidence was found. A separate meta-review of 494 reviews rated 351 of them as low or critically low quality.
against: Ranking evidence as low or moderate certainty is not the same as finding the therapy has no effect. The reviews mostly found benefits, some of them clinically meaningful. In childhood anxiety, roughly one extra child reaches remission for every three treated compared with a waitlist (moderate quality). Moderate certainty is also the usual ceiling for psychotherapy trials, because patients and therapists can't be blinded to which treatment is being given. So "at best moderate" is common across psychological treatments and is not a mark against this therapy in particular. The quote doesn't say which conditions it means, and Cochrane's verdicts differ by condition and by what the therapy is compared against.
what this cannot say
We read five condition-specific Cochrane reviews, not the whole library. Cochrane Library full texts returned access errors, so we worked from Cochrane's own plain-language summaries on cochrane.org. We also could not check reviews that may hold high-certainty ratings, such as the schizophrenia and psychosis reviews, or older reviews published before Cochrane adopted its current grading system. Some reviews we found only through search results and did not read, including fear of falling and first-episode psychosis. We don't know which reviews the speaker had in mind. Certainty ratings measure how confident reviewers are in an estimate, not how large the effect is.
sources, each one fetched and read
- Cochrane review CD013162, Cognitive behavioural therapy for anxiety disorders in children and adolescents, Cochrane Database of Systematic Reviews, 2020 (plain-language summary) What we read there: CBT probably increases post-treatment remission of primary anxiety diagnoses (49.4% vs 17.8%, moderate quality, NNT 3); low-quality evidence vs treatment as usual and attention control. Fetched 2026-10-21 (HTTP 200). https://www.cochrane.org/evidence/CD013162_cognitive-behavioural-therapy-anxiety-disorders-children-and-adolescents
- Cochrane review CD007407, Psychological therapies for the management of chronic pain (excluding headache) in adults, Cochrane Database of Systematic Reviews, 2020 (plain-language summary) What we read there: CBT vs no treatment: small benefits for pain and distress (moderate quality), disability low quality; adverse effects very low quality; no high-quality ratings. Fetched 2026-10-21 (HTTP 200). https://www.cochrane.org/CD007407/SYMPT_what-are-benefits-and-risks-psychological-therapies-adults-persistent-and-distressing-pain-neither
- Cochrane review CD012614, Cognitive behavioural therapy for tinnitus, Cochrane Database of Systematic Reviews, 2020 (plain-language summary) What we read there: CBT probably reduces tinnitus impact on quality of life vs audiological care (moderate certainty); low certainty vs no intervention; anxiety very low to low. Fetched 2026-10-21 (HTTP 200). https://www.cochrane.org/evidence/CD012614_cognitive-behavioural-therapy-adults-tinnitus
- Cochrane review CD015176, Cognitive behavioural therapy for insomnia in people with cancer, Cochrane Database of Systematic Reviews, 2025 (plain-language summary) What we read there: Evidence is very low-certainty to moderate-certainty; the only moderate-certainty finding is reduced sleep onset latency vs inactive control. Fetched 2026-10-21 (HTTP 200). https://www.cochrane.org/evidence/CD015176_does-cognitive-behavioural-therapy-insomnia-cbt-i-help-people-cancer
- Cochrane review CD007674, Cognitive behavioural therapy and third-wave approaches for anxiety and related disorders in older people, Cochrane Database of Systematic Reviews, 2024 (plain-language summary) What we read there: CBT may result in a small-to-moderate reduction of anxiety post-treatment (SMD -0.51, low certainty); little to no difference at 3-6 months (low); clinical improvement very low certainty. Fetched 2026-10-21 (HTTP 200). https://www.cochrane.org/evidence/CD007674_cognitive-behavioural-therapy-and-third-wave-approaches-anxiety-and-related-disorders-older-people
- The evidence for cognitive behavioural therapy in any condition, population or context: a meta-review of systematic reviews and panoramic meta-analysis, Psychological Medicine, 2021, PMC7856415 What we read there: 494 reviews, 221,128 participants; 'The quality of the reviews we identified was poor (351/494 were low or critically-low).' Fetched 2026-10-21 (HTTP 200). https://pmc.ncbi.nlm.nih.gov/articles/PMC7856415/
speakers are described by role, not named. one sentence is quoted under fair use, with attribution and a link to the original at the timestamp; we do not re-host audio or video. nothing on this page is a reason to start, stop or change a medication. if you have the file behind anything marked unverifiable, the research board is where people can answer. adam