do 40% or more of people with depression not respond to antidepressants?

“We're talking 40 or north of 40% who aren't responding.”

a guest, a historian of psychiatry (Andrew Scull) said on Lex Fridman Podcast, 2026-09-21, at 11:48. hear it in context → the speaker cited: a recent Lancet study.

first, the part that holds

the number holds up. in the largest real-world antidepressant study the US government funded, about half of patients did not respond to the first drug they were given, so "40 or north of 40%" is a fair, even conservative, way to put it.

what we found when we went and looked

we went and looked. NIMH's summary of STAR*D, its pragmatic trial in outpatients with depression, says that at level 1 (citalopram) "about one-third of the participants reached remission and about 10-15 percent more responded." that leaves roughly half with no response to the first drug, which fits the speaker's "40 or north of 40%." what we could not find is the source the speaker named. the 2018 Lancet network meta-analysis of 522 trials and 116,477 participants reports odds ratios against placebo (1.37 to 2.13), not a non-response percentage, and we found no recent Lancet paper that gives this figure.

assertion by assertion

both directions

for the speaker: STAR*D is the closest primary record to the claim, and it backs it at the stated size or a little above. after the first drug, about one-third were in remission and another 10-15% had responded, so about half had not responded. the 2018 Lancet meta-analysis says non-response is to be expected: its authors call the effect sizes "mostly modest" and write that "non-response to antidepressants will occur." the FDA's participant-level analysis of placebo-controlled trials found that only 24.5% of people on a drug had a large response.

against: "not responding" is not a fixed share. it depends on the definition and on how many treatments are tried. in STAR*D, almost 70% of participants who stayed in the study became symptom-free across four treatment levels. the share left without remission after repeated attempts is closer to 30% than 40%. the FDA participant-level analysis grouped outcomes into large, non-specific and minimal response. only 12.2% of people on a drug had a minimal response (21.5% on placebo), and its authors write that patients "are likely to improve substantially from acute treatment of their depression with drug or placebo." so if non-response means getting essentially nothing, the figure is far below 40%. if it means falling short of a 50% drop in symptoms on the first drug, it is about half. a single number hides that difference. finally, the source the speaker named does not appear to contain the number.

what this cannot say

we could not open the primary STAR*D paper (the publisher returned 403), so the STAR*D numbers come from NIMH's own summary of its trial, not the journal article. STAR*D had no placebo arm, so its non-responders mix people for whom the drug did not work with people who would not have improved anyway. the reverse also holds: responders include people who would have improved without the drug. the 2018 Lancet meta-analysis gives relative effects, not absolute response rates, in the sections we read. we did not read its supplementary tables. we cannot rule out a recent Lancet paper we did not find. trial populations, often screened for severity and other conditions, may differ from patients seen in everyday care.

sources, each one fetched and read

1. National Institute of Mental Health, Sequenced Treatment Alternatives to Relieve Depression (STAR*D) study: questions and answers, all medication levels What we read there: level 1 (citalopram): "about one-third of the participants reached remission and about 10-15 percent more responded"; "about half of participants in the STAR*D study became symptom-free after two treatment levels"; "over the course of all four treatment levels, almost 70 percent of those who did not withdraw from the study became symptom-free." Fetched 2026-10-06 (HTTP 200). https://www.nimh.nih.gov/funding/clinical-research/practical/stard/allmedicationlevels
2. Comparative efficacy and acceptability of 21 antidepressant drugs for the acute treatment of adults with major depressive disorder: a systematic review and network meta-analysis. The Lancet, 2018;391:1357-66. doi:10.1016/S0140-6736(17)32802-7 What we read there: "522 trials comprising 116 477 participants"; "all antidepressants were more effective than placebo, with ORs ranging between 2·13 ... for amitriptyline and 1·37 ... for reboxetine"; response defined as a reduction of 50% or more on a standardised observer-rated depression scale; "given the modest effect sizes, non-response to antidepressants will occur." Fetched 2026-10-06 (HTTP 200). https://pmc.ncbi.nlm.nih.gov/articles/PMC5889788/
3. Response to acute monotherapy for major depressive disorder in randomized, placebo controlled trials submitted to the US Food and Drug Administration: individual participant data analysis. BMJ, 2022;378:e067606 What we read there: "participants who were treated with a drug were more likely to have a large response (24.5% v 9.6%) and less likely to have a minimal response (12.2% v 21.5%)"; "patients with depression are likely to improve substantially from acute treatment of their depression with drug or placebo." Fetched 2026-10-06 (HTTP 200). https://pmc.ncbi.nlm.nih.gov/articles/PMC9344377/

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