does clozapine work better than other antipsychotics in treatment-resistant schizophrenia?
“it tends to it turns out it tends to work better in treatment-resistant cases, meaning cases don't respond to the other um antipsychotics that we have”
a guest, a historian of psychiatry (Andrew Scull) said on Lex Fridman Podcast, 2026-09-20, at 16:25. hear it in context →
first, the part that holds
clozapine is the one antipsychotic with randomized evidence built specifically around people whose schizophrenia did not respond to other antipsychotics. in the trial behind its approval, 30% of treatment-resistant patients responded on clozapine against 4% on chlorpromazine, and it still beats first-generation drugs like haloperidol in pooled data.
what we found when we went and looked
against the older first-generation drugs, the record supports the speaker: a 268-patient double-blind trial from 1988 found a 30% response rate on clozapine against 4% on chlorpromazine, and a 2023 network meta-analysis (60 trials, 6,838 participants) still finds clozapine ahead of haloperidol. against the newer second-generation drugs the record is mixed. a 2016 network meta-analysis of 40 trials (5,172 patients) found "clozapine was not significantly better than most other drugs." a 2016 pairwise meta-analysis found clozapine better on positive symptoms and on short-term total symptoms (number needed to treat 9), but with no long-term difference in total symptoms or response. the 2023 network analysis put the clozapine-versus-olanzapine difference at "trivial and uncertain" (SMD −0.05). "tends to work better" is fair against older drugs. against "the other antipsychotics we have" as a whole, it states more certainty than the trials give.
assertion by assertion
| assertion | verdict | what the record says |
|---|---|---|
| clozapine works better than first-generation antipsychotics in treatment-resistant schizophrenia | confirmed | 1988 double-blind trial, 268 patients who had failed at least three antipsychotics and then a 6-week haloperidol run-in: 30% response on clozapine vs 4% on chlorpromazine, with greater improvement on BPRS, CGI and nurses' ratings. the 2023 network meta-analysis still has clozapine ahead of haloperidol (SMD −0.26). limitation: the 1988 trial ran only 6 weeks, and chlorpromazine was the only comparator. [1][4] |
| clozapine works better than the second-generation antipsychotics in treatment-resistant schizophrenia | overstated | 2016 network meta-analysis (40 RCTs, 5,172 patients): 'clozapine was not significantly better than most other drugs'; 'insufficient evidence exists on which antipsychotic is more efficacious.' 2023 network meta-analysis (60 RCTs, 6,838 participants): clozapine vs olanzapine 'trivial and uncertain (SMD −0.05, 95% CI −0.21 to 0.11)'. 2016 pairwise meta-analysis (21 RCTs) found clozapine better on positive symptoms at all timepoints and on short-term total symptoms and response (RR 1.17, NNT 9), but no long-term difference in total symptoms or response. this pairwise analysis pooled first- and second-generation comparators. [1][2][3][4] |
| 'treatment-resistant' means cases that don't respond to the other antipsychotics | confirmed | the US label restricts clozapine to 'severely ill patients with schizophrenia who fail to respond adequately to standard antipsychotic treatment'. the 1988 trial required failure on at least three prior antipsychotics. [1][5] |
| context: how many treatment-resistant patients respond to clozapine, and does it keep people out of hospital | confirmed | a meta-analytic summary reports that only about 40% of people with treatment-resistant schizophrenia respond to clozapine. it also reports a lower risk of hospitalization on clozapine in real-world data (RR 0.74, 95% CI 0.69–0.80). the hospitalization figure is observational, not randomized. [3] |
both directions
for the speaker: the original double-blind trial: 30% vs 4% response against chlorpromazine in 268 patients who had failed at least three antipsychotics. the 2016 pairwise meta-analysis found clozapine superior for positive symptoms at every timepoint, and for total symptoms and response in the short term (NNT 9). its authors concluded that clozapine "should remain the treatment of choice for refractory schizophrenia, at least in the short term." the 2023 network meta-analysis concluded "clozapine remains the gold standard" and found it ahead of haloperidol, sertindole and quetiapine. real-world data show fewer hospitalizations on clozapine (RR 0.74). the US label reserves clozapine for exactly the group the speaker describes.
against: the largest network meta-analysis of its time (2016, 40 trials) found clozapine "not significantly better than most other drugs" and judged the evidence insufficient to rank antipsychotics in this population. the 2023 network analysis found the clozapine-versus-olanzapine difference trivial and uncertain. the pairwise meta-analysis found no long-term difference in total symptoms or response, and its authors suggest considering lower-burden drugs if there is no response by 6 months. even on clozapine, only about 40% of treatment-resistant patients respond. so "works better" holds clearly against older drugs, holds partly against the newer ones (positive symptoms, short term), and does not hold against olanzapine on current data.
what this cannot say
the trials comparing clozapine with second-generation drugs are mostly small and short, and many were industry-funded. a meta-analytic review notes that funding source and sensitivity analyses moved the results, and that recent non-industry trials are lacking. some comparator trials may have underdosed clozapine. pairwise and network methods give different answers from largely the same trials. the hospitalization advantage comes from observational data and may reflect who gets prescribed clozapine. we read abstracts and summaries, not every trial's full text. we could not open the 2016 network meta-analysis page on the journal site, so we read its abstract through the PubMed record. the speaker hedged ("tends to"), and our verdict weighs the claim sentence as published.
sources, each one fetched and read
- PubMed abstracts: 'Efficacy, Acceptability, and Tolerability of Antipsychotics in Treatment-Resistant Schizophrenia: A Network Meta-analysis', JAMA Psychiatry 2016;73(3):199-210, doi:10.1001/jamapsychiatry.2015.2955 (PMID 26842482); and 'Clozapine for the treatment-resistant schizophrenic. A double-blind comparison with chlorpromazine', Arch Gen Psychiatry 1988;45(9):789-96, doi:10.1001/archpsyc.1988.01800330013001 (PMID 3046553) What we read there: 2016: 40 RCTs, 5,172 participants; 'clozapine was not significantly better than most other drugs'; 'insufficient evidence exists on which antipsychotic is more efficacious.' 1988: 268 patients; '30% of the clozapine-treated patients were categorized as responders compared with 4% of chlorpromazine-treated patients.' Fetched 2026-10-21 (HTTP 203). https://pubmed.ncbi.nlm.nih.gov/26842482/
- 'Clozapine v. first- and second-generation antipsychotics in treatment-refractory schizophrenia: systematic review and meta-analysis', British Journal of Psychiatry 2016;209(5):385-392 What we read there: 21 RCTs (25 comparisons); NNT 9; clozapine superior for positive symptoms at all timepoints; short-term response RR 1.17 (8 studies, n=1,218); no significant long-term difference in total symptoms or response; 'Clozapine should remain the treatment of choice for refractory schizophrenia, at least in the short term.' Fetched 2026-10-21 (HTTP 200). https://www.cambridge.org/core/journals/the-british-journal-of-psychiatry/article/clozapine-v-first-and-secondgeneration-antipsychotics-in-treatmentrefractory-schizophrenia-systematic-review-and-metaanalysis/2CD17754D62A1044C6089EEDD005F0C5
- 'What do meta-analyses tell us about clozapine's efficacy and effectiveness for treatment refractory schizophrenia?', Schizophrenia Bulletin 2018, conference abstract 41.1 (PMC5888558) What we read there: only 40% of people with TRS responded to clozapine; network and pairwise meta-analyses equivocal for total symptoms, pairwise finds clozapine superior for positive symptoms; hospitalization RR 0.74 (95% CI 0.69–0.80); 'a lack of recent non-industry funded randomised control trials.' Fetched 2026-10-21 (HTTP 200). https://pmc.ncbi.nlm.nih.gov/articles/PMC5888558/
- 'A network meta-analysis of efficacy, acceptability, and tolerability of antipsychotics in treatment-resistant schizophrenia', European Archives of Psychiatry and Clinical Neuroscience 2024;274(4):917-928 (online 2023), doi:10.1007/s00406-023-01654-2 What we read there: 60 RCTs, 6,838 participants; clozapine vs olanzapine 'trivial and uncertain (SMD −0.05, 95% CI, −0.21 to 0.11)'; clozapine vs haloperidol SMD −0.26; 'Clozapine remains the gold standard for patients with treatment-resistant schizophrenia. Olanzapine seems to be second-best.' Fetched 2026-10-21 (HTTP 200). https://pmc.ncbi.nlm.nih.gov/articles/PMC11127860/
- US prescribing information, clozapine tablets, DailyMed (National Library of Medicine) What we read there: 'indicated for the treatment of severely ill patients with schizophrenia who fail to respond adequately to standard antipsychotic treatment'; effectiveness 'demonstrated in a 6-week, randomized, double-blind, active-controlled study comparing clozapine tablets and chlorpromazine in patients who had failed other antipsychotics.' Fetched 2026-10-21 (HTTP 200). https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=883b5d43-0339-7dc1-f775-93791fb9b978
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