does fda approval need only two winning trials, no matter how many failed?
“to get FDA approval of efficacy and safety, you needed two trials. You might have conducted 15 trials and 13 of them had been failures, but if you had two that worked, that was enough.”
a guest, a historian of psychiatry (Andrew Scull) said on Lex Fridman Podcast, 2026-09-20, at 18:19. hear it in context →
first, the part that holds
The speaker is right about the old rule. For decades the FDA's stated position was that the 1962 law required "at least two adequate and well-controlled studies, each convincing on its own." The speaker is also right that drugs have been approved even though many of their trials failed. FDA reviews of antidepressant approvals show about half of the submitted placebo-controlled trials came out positive.
what we found when we went and looked
The FDA's 1998 guidance says it was the agency's position that Congress intended "at least two adequate and well-controlled studies." That much of the quote matches the record. The rest goes too far in three ways. First, the two-trial rule was about efficacy only, not safety. The FDA's June 2026 draft guidance says trials that prove a drug works "may not include a sufficient number of participants" to show it is safe, and that a separate safety database is needed. Second, the FDA says "all relevant trials should be considered," and a well-run trial that shows no effect "could call into question" the positive ones. Failed trials are weighed, not ignored, even though in practice drugs do get approved alongside them: 15 of 30 trials for four newer antidepressants were positive, and about half of 74 trials for twelve older ones. Third, two is not the floor. Since 1997 the law has allowed one trial plus confirmatory evidence, and 36.8% of new-drug indications approved from 2005 to 2012 rested on a single pivotal trial.
assertion by assertion
| assertion | verdict | what the record says |
|---|---|---|
| the FDA's standard for proving a drug works was two adequate and well-controlled trials | confirmed | The FDA's 1998 guidance: "it has been FDA's position that Congress generally intended to require at least two adequate and well-controlled studies, each convincing on its own, to establish effectiveness." The basis is the plural "investigations" in section 505(d). The same guidance notes that the FDA sometimes accepted a single study even before 1997. [1] |
| two successful trials establish both efficacy and safety | misattributed | The two-trial rule applied to the efficacy standard only. The June 2026 draft guidance says substantial evidence of effectiveness is "necessary but not sufficient" for approval and that "there must be sufficient safety data." It adds that trials sufficient to show effectiveness "may not include a sufficient number of participants or a sufficient treatment duration to conclude that the drug is safe," so a larger, longer or extra trial may be needed. This is a draft, but it describes long-standing safety-database expectations (ICH E1A). [2] |
| two positive trials are enough regardless of how many other trials failed | overstated | The direction is right: approvals do happen with many negative trials on file. For four newer antidepressants, the FDA judged 15 of 30 trials positive; for twelve older ones, about half of 74 trials. But the FDA's stated method is not "regardless." The guidance says "all relevant trials should be considered." It also says a well-controlled trial showing no effect, with confidence intervals that rule out a meaningful effect, "could call into question the results from additional adequate and well-controlled trials" unless the difference can be explained. What the record does not show is how hard the agency applies that principle case by case. [2][4] |
| two is the minimum number of trials needed | undersold | The real minimum is lower. The 1997 Modernization Act amended section 505(d) so the FDA may accept "data from one adequate and well-controlled clinical investigation and confirmatory evidence." Of 206 indications for new drugs approved from 2005 to 2012, 74 (36.8%) rested on a single pivotal trial; the median was two. The June 2026 draft guidance is built around this one-trial route. [1][2][3] |
| a drug could be approved with 2 successes out of 15 trials (13 failures) | unverifiable | The speaker gave this as an illustration, not a named case. The antidepressant groups we read show about a 50% positive rate, not 13%. We found no document showing a 2-of-15 approval, and the per-drug breakdown in the newer-drug paper was not readable in what we fetched. [4] |
both directions
for the speaker: The FDA's own 1998 guidance confirms that the agency's working rule was two trials, each convincing on its own, which matches the past-tense "you needed two trials." FDA review documents analyzed in a peer-reviewed paper show that four newer antidepressants were approved even though half their 30 placebo-controlled trials were not positive. The older group of twelve antidepressants had about the same split across 74 trials. So the core point, that approval can sit on top of a pile of failed or negative trials, holds up for this drug class.
against: The two-trial rule never covered safety. The FDA treats safety as a separate determination, which may need more participants or a longer trial than the efficacy trials provide. "Regardless" is contradicted by the FDA's written standard: all relevant trials count, and a well-run trial showing no effect can undercut the positive ones. Two trials is also not the floor. Since 1997 the law has allowed one trial plus confirmatory evidence, and more than a third of new-drug indications from 2005 to 2012 were approved on a single pivotal trial. On that point the real bar is lower than the speaker said. The 13-of-15 figure has no document behind it, and the failure rates we found are much lower.
what this cannot say
Guidance documents describe how the FDA says it weighs evidence, not how hard it applied that standard in each decision. The 2026 document is a draft. The antidepressant figures cover 16 drugs in one class and use the FDA's own positive/negative calls. They do not separate "failed" trials (where even a proven comparator drug did not beat placebo) from truly negative ones, and the FDA discounts the former differently. We could not read the per-drug table, so we cannot say whether any single drug came close to the 2-of-15 example. The single-trial figure covers novel drugs from 2005 to 2012, not psychiatric drugs specifically or approvals before 1997. We did not check the speaker's own written sources.
sources, each one fetched and read
- U.S. Food and Drug Administration, Guidance for Industry: Providing Clinical Evidence of Effectiveness for Human Drug and Biological Products, May 1998 What we read there: "it has been FDA's position that Congress generally intended to require at least two adequate and well-controlled studies, each convincing on its own, to establish effectiveness"; section 115(a) of the Modernization Act lets the agency consider "data from one adequate and well-controlled clinical investigation and confirmatory evidence". Fetched 2026-10-21 (HTTP 200). https://www.fda.gov/media/71655/download
- U.S. Food and Drug Administration (CDER/CBER/OCE), Demonstrating Substantial Evidence of Effectiveness for Human Drug and Biological Products, draft guidance, Revision 1, June 2026 What we read there: Substantial evidence of effectiveness is "necessary but not sufficient" and "there must be sufficient safety data"; "All relevant trials should be considered in evaluating a drug's effectiveness"; a trial showing no effect "could call into question the results from additional adequate and well-controlled trials unless there is a clear and compelling explanation"; trials sufficient for effectiveness "may not include a sufficient number of participants" to show safety. Fetched 2026-10-21 (HTTP 200). https://www.fda.gov/media/133660/download
- Clinical trial evidence supporting FDA approval of novel therapeutic agents, 2005-2012. JAMA, 2014 (PMC4144867) What we read there: 188 novel therapeutics for 206 indications on 448 pivotal trials; "74 (36.8%) indications were approved on the basis of a single pivotal trial"; the median number of pivotal trials per indication was two. Fetched 2026-10-21 (HTTP 200). https://pmc.ncbi.nlm.nih.gov/articles/PMC4144867/
- Selective publication of antidepressant trials and its influence on apparent efficacy: updated comparisons and meta-analyses of newer versus older trials. PLOS Medicine, 2022, doi:10.1371/journal.pmed.1003886 What we read there: For 4 newer antidepressants (desvenlafaxine, vilazodone, levomilnacipran, vortioxetine), "of these 30 trials, the FDA deemed 15 (50%) to be positive," "consistent with the proportion previously reported for the older cohort" of 74 trials of 12 antidepressants. Fetched 2026-10-21 (HTTP 200). https://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.1003886
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