does cbt beat antidepressants for mild depression, with fewer side effects?
“for the milder forms of depression, I think CBT works better than the drugs without some of the side effects that the the drugs have.”
a guest, a historian of psychiatry said on Lex Fridman Podcast, 2026-09-21, at 09:39. hear it in context →
first, the part that holds
for milder depression the record does back two things: antidepressants barely beat a sugar pill, and cbt works about as well as the drugs while avoiding drug side effects like nausea and sexual problems. the speaker was right that there is no case for drugs being the better first choice for milder depression.
what we found when we went and looked
we found no head-to-head evidence that cbt beats antidepressants for mild depression in the short term. a 2023 meta-analysis of 409 trials found no significant difference at the end of treatment and an advantage for cbt only at 6-12 month follow-up (g=0.34), and that was across all severities, not mild cases alone. a pooled analysis of the raw patient data from 16 trials (1,700 outpatients) found that baseline severity made no difference to how cbt compared with drugs, and overall the drugs had a small edge on one clinician-rated scale. what does hold for mild depression is that the drugs add little: in a patient-level analysis of six placebo-controlled trials the drug-placebo effect was below d=0.20 for mild-to-moderate symptoms. the side-effect half holds up for drug side effects (on the sertraline label, 12% of patients stopped because of adverse reactions versus 4% on placebo), but cbt's own harms are rarely recorded, so the comparison is lopsided.
assertion by assertion
| assertion | verdict | what the record says |
|---|---|---|
| for mild depression, cbt is more effective than antidepressant drugs | overstated | in the comprehensive meta-analysis of 409 trials (52,702 patients), cbt 'did not differ significantly from' pharmacotherapies at the short term and was larger only at 6-12 month follow-up (g=0.34; 95% ci 0.09-0.58). the individual patient data meta-analysis of 16 trials (1,700 outpatients) found a small main effect of drugs over cbt on the clinician-rated ham-d scale (beta -0.88, p=.03), no difference in response or remission, and found that baseline severity did not moderate the comparison (beta 0.00, p=.96). none of this isolates mild depression, so 'better' fits the long-term data only, at modest size, and not specifically in mild cases. [1][2] |
| cbt avoids some of the side effects antidepressants have | confirmed | the fda-approved sertraline label reports nausea 26% vs 12% on placebo, diarrhea 20% vs 10%, ejaculation failure 8% vs 1%, and 12% of 3,066 patients stopping because of an adverse reaction vs 4% of 2,293 on placebo, plus a boxed warning on suicidal thoughts in young people. cbt does not cause these drug effects. caveat: a review of uk publicly funded trials found that no psychological-treatment trial reported any adverse event in its final report, so cbt's own harms go largely unmeasured. [4][5] |
| antidepressants offer little benefit over placebo in milder depression (popular framing behind the claim) | confirmed | a patient-level meta-analysis of six placebo-controlled trials (718 patients, paroxetine and imipramine) estimated drug-placebo effect sizes below d=0.20 at hamilton scores of 18 or less, rising to d=0.47 at 23 or more; the authors concluded benefit 'may be minimal or nonexistent, on average' in mild or moderate symptoms. this compares drugs with placebo, not drugs with cbt, and covers only two drugs. [3] |
both directions
for the speaker: at 6-12 months cbt beat medication (g=0.34) in the largest meta-analysis of cbt for depression, and adding cbt to drugs beat drugs alone, while adding drugs to cbt did not beat cbt alone. in milder depression the drugs barely separate from placebo (d below 0.20), so cbt has little to beat there. drug side effects are real and recorded on the label, with three times the placebo rate of stopping because of adverse reactions.
against: at the end of acute treatment, the main trial syntheses show cbt and antidepressants as roughly equal, not cbt as better. the patient-data meta-analysis found a small edge for drugs on the ham-d and found that severity did not change the cbt-versus-drug comparison, so there is no specific mild-depression advantage for cbt in that data. 'without side effects' also rests partly on missing data: psychological-treatment trials seldom look for harms, so the side-effect gap may be smaller than it appears.
what this cannot say
no source we read compares cbt with antidepressants only in people with mild depression, so the claim's mild-depression scope can only be pieced together from mixed-severity trials and drug-versus-placebo data. the placebo-controlled analysis covers two older drugs and six trials. cbt trials can't be blinded, and waitlist-heavy designs can make cbt look better. we read abstracts and label sections, not full texts. we couldn't load the uk guideline (nice ng222) page, which advises against routine first-line antidepressants for less severe depression, so it isn't cited here. we didn't check the side-effect profiles of drugs other than sertraline.
sources, each one fetched and read
- Cognitive behavior therapy vs. control conditions, other psychotherapies, pharmacotherapies and combined treatment for depression: a comprehensive meta-analysis including 409 trials with 52,702 patients. World Psychiatry, 2023. PMID 36640411 What we read there: "The effects of CBT did not differ significantly from those of pharmacotherapies at the short term, but were significantly larger at 6-12 month follow-up (g=0.34; 95% CI: 0.09-0.58)"; combined treatment beat pharmacotherapy alone but "was not more effective than CBT alone at either time point." Fetched 2026-10-17 (HTTP 203). https://pubmed.ncbi.nlm.nih.gov/36640411/
- Baseline depression severity as moderator of depression outcomes between cognitive behavioral therapy vs pharmacotherapy: an individual patient data meta-analysis. JAMA Psychiatry, 2015;72(11):1102-1109. PMID 26397232 What we read there: 16 trials, 1,700 outpatients: "There was a main effect of ADM over CBT on the HAM-D (β = -0.88; P = .03)... but no significant differences in response (OR, 1.24; P = .12) or remission (OR, 1.18; P = .22)"; "Baseline depression severity did not moderate differences between CBT and ADM." Fetched 2026-10-17 (HTTP 203). https://pubmed.ncbi.nlm.nih.gov/26397232/
- Antidepressant drug effects and depression severity: a patient-level meta-analysis. JAMA, 2010. PMC3712503 What we read there: six trials, 718 patients; drug-placebo effect sizes < .20 for hrsd ≤18, d=.47 for ≥23; benefit "may be minimal or nonexistent, on average, in patients with mild or moderate symptoms." Fetched 2026-10-17 (HTTP 200). https://pmc.ncbi.nlm.nih.gov/articles/PMC3712503
- DailyMed (US National Library of Medicine), sertraline tablets, film coated, prescribing information, section 6.1 and boxed warning What we read there: nausea 26% vs 12% placebo; diarrhea 20% vs 10%; ejaculation failure 8% vs 1%; "368 (12%) of the 3,066 patients who received sertraline discontinued treatment due to an adverse reaction, compared with 93 (4%) of the 2,293 placebo-treated patients." Fetched 2026-10-17 (HTTP 200). https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=2dac437f-b5ef-42dd-aa0b-1f4b9b45ef43
- The recording of adverse events from psychological treatments in clinical trials: evidence from a review of NIHR-funded trials. Trials, 2014. PMC4152561 What we read there: "None of the psychological intervention trials mentioned the occurrence of an adverse event in their final report"; "the reporting of adverse events in psychological treatments is weak." Fetched 2026-10-17 (HTTP 200). https://pmc.ncbi.nlm.nih.gov/articles/PMC4152561/
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