Comparative efficacy and tolerability of antidopaminergic and muscarinic

# Comparative efficacy and tolerability of antidopaminergic and muscarinic antipsychotics for acute schizophrenia: a network meta-analysis of randomised controlled trials indexed in international English and Chinese databases *The Lancet 2026* Articles Comparative efficacy and tolerability of antidopaminergic and muscarinic antipsychotics for acute schizophrenia: a network meta-analysis of randomised controlled trials indexed in international English and Chinese databases Johannes Schneider-Thoma*, Yikang Zhu*, Mengchang Qin, Yu Dong, Shiwei Guan, Jiaxi Wang, Jing Tian, Xiao Lin, Alessandro Rodolico, Spyridon Siafis, Irene Bighelli, Melanie Wehner, Christina Veith, Felix Krayer, Elfriede Scheuring, John M Davis, Josef Priller, Adriani Nikolakopoulou, Georgia Salanti, Chunbo Li†, Stefan Leucht† Summary Lancet 2026; 407: 876–91 Background Antipsychotic drugs are the established treatment for acute schizophrenia but dier in receptor-binding See Comment page 829 profiles. In 2024, a new-in-class muscarinic receptor agonist (xanomeline–trospium) was licenced, acting upstream of antidopaminergic agents, and providing hope to decrease the adverse eects burden of antipsychotics. We aimed to *Joint first authors compare the ecacy and tolerability of antipsychotics by performing network meta-analysis of randomised controlled †Joint last authors trials (RCTs). Technical University of Munich, TUM School of Medicine and Health, TUM University Methods This systematic review (PROSPERO, CRD42022380708) included blinded and open RCTs investigating Hospital Rechts der Isar, antipsychotic drugs in participants of any age with acute psychotic symptoms of schizophrenia over 3 weeks to Department of Psychiatry and 3 months. Included antipsychotics comprised 23 primarily dopamine-receptor blocking medications and the Psychotherapy, Munich, Germany muscarinic receptor agonist xanomeline–trospium in dierent applications. We searched Cochrane Schizophrenia (J Schneider-Thoma MD, M Qin, group’s register, previous reviews, and five Chinese databases for trials published from database inception until J Tian, X Lin, A Rodolico PhD, July 26, 2024 and contacted authors to assess trials’ methodological quality; only trials with appropriate randomisation S Siafis MD, I Bighelli PhD, indicated were included. The primary outcome was rating scale-measured overall symptoms of schizophrenia M Wehner, C Veith, F Krayer MSc, Prof J Priller MD, S Leucht MD); (ecacy) analysed with random-eects frequentist network meta-analysis. Secondary outcomes comprised 32 further German Center for Mental ecacy and tolerability outcomes. The confidence in the estimates was assessed using the Confidence in Network Health (DZPG), partner site Meta-Analysis approach. München/Augsburg, Munich, Germany (J Schneider-Thoma, A Rodolico, S Siafis, I Bighelli, Findings After screening 18 859 references and contacting authors of 5428 trials, we included 438 RCTs. Of those, Prof J Priller, S Leucht, M Qin, 388 RCTs with 78 193 participants (28 448 women and 49 745 men) provided usable data for at least one outcome. J Tian, X Lin); Shanghai Key 5117 Chinese trials were identified but most were excluded because authors did not reply or reported serious Laboratory of Psychotic methodological concerns. 256 double-blind studies with 58 948 participants provided usable data for the primary Disorders, Shanghai Mental Health Centre, Shanghai Jiao outcome. All antipsychotics reduced symptoms more than placebo with standardised mean dierences ranging Tong University School of from –0·90 (95% CI –1·03 to –0·77) to –0·23 (–0·39 to –0·06). Particularly clozapine, as well as amisulpride, Medicine, Shanghai, China olanzapine, and risperidone were more ecacious than at least three other antipsychotics (confidence in estimates (Y Zhu MD, Y Dong, S Guan, were low-to-moderate). Adverse eects varied across medications. J Wang, Prof C Li MD); BASTA - Bündnis für psychisch erkrankte Menschen, Munich, Interpretation This network meta-analysis provides evidence for small-to-medium clinically relevant dierences Germany (E Scheuring M. Ed.); between antipsychotics in ecacy; this finding warrants stronger and more specific emphasis in clinical guidelines. Psychiatric Institute, University of Illinois Chicago, Chicago, IL, Nonetheless, important dierences in tolerability need to be considered for individualised drug choice, with partial USA (Prof J M Davis MD); dopamine agonists having overall better tolerability and xanomeline–trospium lacking adverse eects of dopamine- Department of Psychiatry, John blocking agents but resulting in cholinergic and anticholinergic adverse events. Future research should directly Hopkins University, Baltimore, compare xanomeline–trospium with other antipsychotics to confirm its ecacy; modern trials using clozapine early MD, USA (Prof J M Davis); Neuropsychiatry and in schizophrenia are needed to establish whether it improves outcomes and prevents chronification. Laboratory of Molecular Psychiatry, Charité – Funding German Research Foundation, German Ministry of Research, Technology and Space, and National Natural Universitätsmedizin Berlin and Science Foundation of China. DZNE, Berlin, Germany (Prof J Priller); University of Edinburgh and UK DRI, Copyright © 2026 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 Edinburgh, UK (Prof J Priller); license. Department of Hygiene, Social- Preventive Medicine and Medical Statistics, School of Introduction eects on serotonin, acetylcholine, histamine, and Medicine, Aristotle University Multiple antipsychotics are available for schizophrenia norepinephrine receptors. Additionally, in 2024, the of Thessaloniki, Thessaloniki, which act by the common mechanism of blocking first non-dopamine-centric but muscarinic compound, Greece (A Nikolakopoulou PhD); dopamine D2 receptors. However, they dier in their xanomeline–trospium, was licensed. In contrast to 876 Articles Institute of Medical Biometry Research in context and Statistics, Faculty of Medicine and Medical Center- Evidence before this study xanomeline–trospium), sexual adverse events, heart rate University of Freiburg, Freiburg Multiple antipsychotic medications with different receptor- abnormalities, reductions in white blood cell counts, and im Breisgau, Germany binding profiles are licensed worldwide. As a particular novelty, seizures. (A Nikolakopoulou); Institute of xanomeline–trospium, the first antipsychotic acting primarily Social and Preventive For evidence-based medicine, incorporation of data from Medicine, University of Bern, as a muscarinic receptor agonist instead as a dopamine improperly randomised studies and unreliable data can lead to Bern, Switzerland antagonist has been licensed in 2024. The resulting differences misleading meta-analytic results. Building on previous surveys, (G Salanti PhD); School of in pharmacodynamic properties can lead to different efficacy Psychology, Shanghai Jiao Tong we therefore contacted the trial authors to request details on and side effects. University, Shanghai, China the randomisation procedure. Here, our Chinese–German (Prof C Li) We searched PubMed with the search term co-funded research collaboration had a particular focus on Correspondence to: “(((antipsychotic[Title/Abstract]) AND (schizophrenia[Title/ studies indexed in Chinese databases due to their high number Prof Stefan Leucht, Technical Abstract])) AND (network meta-analysis[Title/Abstract])) AND and potential impact on the meta-analysis. Only a small University of Munich, TUM School of Medicine and Health, ((“2019/06/11”[Date - Entry] : “2025/05/13”[Date - Entry]))” proportion of those authors that answered our request TUM University Hospital Rechts but found no comprehensive systematic review after a network confirmed appropriate randomisation and authenticity leading der Isar, Department of meta-analysis from 2019 which only searched for RCTs up to to the inclusion of only very few studies from China. Conversely, Psychiatry and Psychotherapy, Jan 8, 2019 and was restricted to adult participants with chronic this result indicates that caution should be exercised when Munich 81675, Germany Stefan.leucht@tum.de schizophrenia and oral drug formulations. Differences in side interpreting results of meta-analyses indiscriminately including effects were indicated, whereas differences in efficacy were studies indexed in Chinese databases. mostly gradual and did not allow for clear statements. Implications of all the available evidence Numerous Chinese studies could have substantially increased There are efficacy differences between antipsychotics. In the evidence base but were excluded in general for quality particular clozapine (as also suggested by observational data), concerns. Furthermore, the 2019 review had less strict inclusion followed by amisulpride, olanzapine, and risperidone, are more criteria regarding trials’ methodology and focused on statistical efficacious than at least three other antipsychotics. In addition significance rather than clinically meaningful effects. Other to efficacy, diverse side effect profiles need to be considered for reviews concentrated on specific subpopulations with limited individualised drug choice. Aripiprazole might be preferred statistical power, long-term studies (important for quality of among partial dopamine-agonists, which have generally milder life and functioning during maintenance treatment), side effects, due to its medium-ranked efficacy. Blonanserin or observational studies (with increased representativeness of (moderate efficacy) and lumateperone are also among those the population but potential confounding). drugs with fewer side effects, but lumateperone ranked last in Added value of this study efficacy; add-on of samidorphan has little effect on mitigating We extended the previous 2019 review by including additional olanzapine-induced weight gain but the average treatment antipsychotics (xanomeline–trospium, olanzapine– duration in our review is likely too short to observe benefits. samidorphan, lumateperone, blonanserin, and perospirone), Muscarinic agents appear as promising compounds as non-oral formulations (long-acting intramuscular or xanomeline–trospium ranked in the top third in efficacy and does subcutaneous and transdermal delivery), and studies from not cause the side effects associated with dopamine-blocking Chinese databases with confirmed randomisation after agents (weight gain, extrapyramidal motor disorders, and contacting authors. This review also includes studies with prolactin increase) but has cholinergic and anticholinergic children and adolescents (younger than 18 years), and elderly effects. However, only placebo-controlled trials of xanomeline– participants (older than 65 years) as well as individuals in their trospium are available; to confirm the ranking of xanomeline– first episode, not responsive or intolerable to previous trospium comparisons with other antipsychotics are needed. antipsychotics and with comorbid drug abuse disorder. Moreover, modern trials using clozapine first-line or second-line Furthermore, this review provides evidence for 33 outcomes (and not as currently third-line) are needed to find out whether (instead of 17), including additional clinically relevant its early use might improve outcomes and prevent persistence of outcomes, such as cognitive functioning, cholinergic adverse symptoms. events (especially important due to the new antipsychotic D2-targeting antipsychotics acting downstream of striatal might not be properly randomised or blinded.2,3 In this dopamine dysregulation in schizophrenia, xanomeline’s network meta-analysis (NMA) sponsored by the German M1 and M4 agonism presumably modulates the Research Foundation and the National Natural Science dopamine dysregulation. Foundation of China, we aimed to provide an up-to-date China produces 25% of schizophrenia trials1 but they review on the ecacy and tolerability of antipsychotics. are often not included in systematic reviews because they We expanded upon our latest review from 20194 by are predominantly listed in Chinese databases, and including five additional antipsychotics (xanomeline– because of previous observations that many of them trospium, olanzapine–samidorphan, lumateperone, Articles blonanserin, and perospirone) as well as long-acting on antipsychotic dosing.17 Lower doses were considered injectable administrations. We also increased the for people in their first episode, children and adolescents, number of outcomes to 33 (evaluating cognition and and older adults, and higher doses for people with multiple additional adverse eects), broadened the treatment-resistant symptoms. However, for tolerability inclusion criteria beyond chronic adults with acute outcomes, studies with average doses above the exacerbations to better reflect the real-world population,5 consensus’ maximum were excluded. included Chinese studies after active confirmation of The NMA could be extended to subgroups of appropriate randomisation, and evaluated clinically schizophrenia and long-acting injectable formulations as relevant treatment eects, rather than simply focusing recent meta-analyses18,19 did not identify important eect on statistical significance. modification (see the Discussion). Suboptimal randomisation is a concern for RCTs Methods from China,2,3,20 but has also been raised for trials from Search strategy and selection criteria other countries.21,22 We therefore contacted authors of This systematic review and meta-analysis (registered all industry-independent trials and pharmaceutical For PROSPERO see https://www. with PROSPERO, CRD42022380708; protocol is provided companies by telephone, email, and letters to seek crd.york.ac.uk/prospero/ in the appendix pp 9–18) followed the PRISMA guideline confirmation of appropriate randomisation and blinding See Online for appendix for NMA (appendix pp 19–22).6 Changes to the protocol (appendix p 35). To standardise the request, and to avoid are documented in the appendix (pp 23–24); the most misunder standings, a specific online questionnaire was relevant is the inclusion of subpopulations with designed (appendix pp 38–75). Industry-independent schizophrenia. No ethics committee approval was needed trials from countries without long-standing tradition in for this systematic review. conducting clinical research, as described by Panagiotou We searched previous reviews,4,7–13 Cochrane and colleagues,23 were excluded if no confirmation of Schizophrenia Group’s register, and five Chinese appropriate randomisation was received; industry- databases (appendix pp 25–33) for articles published in independent trials from countries with long-standing any language from database inception to July 26, 2024. tradition in conducting clinical research23 and industry- We included blinded and open-label randomised funded trials were included in the primary analysis (as controlled trials (RCTs) comparing antipsychotics with no specific concerns were indicated), but those without each other or placebo in participants with acute psychotic active confirmation of appropriate randomisation were symptoms of schizophrenia (or schizoaective or excluded in sensitivity analyses. Moreover, we consulted schizophreniform disorder) with a duration of 3 weeks to Retraction Watch24 for retracted trials. 3 months. Study selection and data extraction was done in In addition to studies in adults with exacerbation of duplicate by two independent reviewers (out of YZ, YD, chronic schizophrenia,4 we included studies in children SG, JW, and MQ for Chinese studies, and out of JS-T, and adolescents (younger than 18 years), older adults MQ, JT, XL, CV, MW, and the authors of previous reviews (older than 65 years), and people experiencing their first for non-Chinese studies). Contacting of study authors episode, with comorbid substance abuse or with was done by YZ, YD, SG, JW, and MQ for Chinese resistance and/or intolerability to previous antipsychotics studies, and by JS-T, MQ, JT, XL, CV, MW, and authors of to increase real-life representativeness5 and statistical previous reviews for non-Chinese studies. power. We included all 19 second generation antipsychotics Data analysis (appendix p 34) licensed in Europe or the USA, We followed a pre-established hierarchy if the same including the most recently licensed combinations of outcomes were reported in dierent publications xanomeline–trospium (October, 2024) and olanzapine– (appendix p 76). samidorphan (April, 2022), as well as lumateperone The primary outcome was overall ecacy (overall (December, 2019). We also included the second- symptoms of schizophrenia; continuous outcome). generation antipsychotics blonanserin and perospirone Secondary continuous outcomes were positive, negative, licensed in Japan, and the first-generation antipsychotics and depressive symptoms, quality of life, functioning (as haloperidol and chlorpromazine (as prototypical first- measured by rating scales), cognition (composite score generation antipsychotics listed in WHO’s list of following Feber and colleagues25), weight, prolactin, and essential medicines14), perphenazine (investigated in the QTc interval. Secondary dichotomous outcomes were the US-based CATIE trial15), and sulpiride (frequently used number of participants with response, who discontinued in China and the UK-based CUtLASS trial15). The latter for inecacy and any reason, anti cholinergic were also reported as the four clinically most important (eg, constipation, hyposalivation, tachycardia, and first-generation antipsychotics.16 blurred vision), cholinergic (eg, nausea, vomiting, These drugs were included in any application (eg, oral, dyspepsia, and hypersalivation), extrap yramidal (as long-acting injectable, or transdermal) if the allowed indicated by antiparkinsonian medication use), and dose range overlapped with the international consensus sexual (eg, libido decrease, amenorrhea, ejaculation 878 Articles Search in international databases 661 studies in search of previous 2923 references from update search 3969 references from search in reviews for the previous reviews in Cochrane Schizophrenia Group‘s Cochrane Schizophrenia Group‘s register for Chinese trials register from Oct 14, 2016 published in international until July 26, 2024 journals from inception until Oct 13, 2016 2541 references excluded with basic criteria 3924 references excluded due to being published in 74 no randomised controlled trial journals not indexed in PubMed 635 Chinese study 9 naturalistic design (switch of drugs) 180 participants not schizophrenia-like 21 participants not acutely ill 64 comparator any antipsychotic 50 comparator antipsychotic combination 136 comparator same antipsychotic 1112 comparator not an included antipsychotic* 18 duration <3 weeks 241 duration >3 months 1 duration unclear Duplicate studies included in >1 review 382 references belonging to 27 studies with 45 references removed 128 studies included with basic published in journals indexed in criteria PubMed 15 studies (30 references) excluded with basic criteria 1 participant relative of persons with schizophrenia 3 comparator antipsychotic combination 3 comparator same antipsychotics 1 comparator not an included antipsychotic 3 duration <3 weeks 1 duplicate to study in specific search for studies published in China 3 duration >3 months 12 studies with 15 references included 582 studies included Search in Chinese databases 11306 references from search in Chinese electronic literature databases from 115 excluded because comparison not usable inception until May 6, 2024 467 with usable comparison 31 duplicates excluded 53 excluded after contacting for methods 7 from country without research tradition 5237 references excluded in title or abstract screening with confirmed inadequate methods 46 from country without research tradition without confirmed methods (no answer or 6038 references screened on full text level unclear) 921 references excluded with basic criteria 414 included 376 no randomised controlled trial 279 conducted or supported by pharmaceutical companies 21 wrong participant 20 from country with research tradition with confirmed adequate methods 37 intervention not an included antipsychotic 95 (54 published before 1994) from country with research tradition 473 comparator antipsychotic combination without confirmation 14 comparator same antipsychotic 20 from country without research tradition with confirmed methods 5117 references included (Figure 1 continues on next page) Articles 5089 studies excluded after contacting for methods 4769 no answer 83 disclosed that it was an observational study 125 confirmed suboptimal randomisation process 39 unclear randomisation process 15 disclosed that study results were not reliable 58 authors unsure or cannot remember if actually conducted 27 studies (28 references) with confirmed adequate methods included 3 studies excluded for this review 1 participant not acutely ill 1 participant with predominant negative symptoms 1 duration >3 months 24 studies included 438 studies included 388 studies with usable data Figure 1: Study selection *Among the references excluded because the comparator was not an included antipsychotic, there were no new studies evaluating first-generation antipsychotics. This number is relatively high because for some update searches, all references added to the Cochrane Schizophrenia Group’s register were assessed, irrespective of whether they were on studies investigating antipsychotic drugs or other interventions. Continuous outcomes were analysed with standardised Chlorpromazine (n=1288) Cariprazine (n=1293) mean dierence (SMD) or mean dierences (MD) if Clozapine (n=703) Brexpiprazole (n=1474) results were reported on dierent or the same scale, Blonanserin (n=1166) Haloperidol (n=5013) respectively. Dichotomous outcomes were analysed using Asenapine (n=1582) odds ratios (OR) but transformed to event rates for Iloperidone (n=2157) Aripiprazole (n=3753) interpretability (appendix p 79). Treatment eects on Lumateperone (n=386) rating scales, study discontinuation, and specific self- Amisulpride (n=797) reported adverse eects (appendix pp 76–78) were Lurasidone (n=2572) estimated using only double-blind studies because Zotepine (n=209) knowledge of the intervention received might influence Olanzapine−samidorphan participants’ or outcome assessors’ perception and (n=124) Ziprasidone (n=1585) interpretation of symptoms. Study arms for the same Olanzapine (n=6690) Xanomeline–trospium drug at dierent eligible doses or applications were Paliperidone (n=3090) (n=314) merged in the main analysis. Sulpiride (n=71) Study-specific eects were combined using NMA with Perospirone (n=87) Sertindole (n=1366) a common heterogeneity variance (τ²).26 Transitivity was Perphenazine (n=526) Risperidone (n=6956) assessed visually by comparing potential eect modifiers across comparisons, displayed by boxplots. Heterogeneity Placebo (n=11841) Quetiapine (n=3905) (ie, dierences in estimates between studies of the same Figure 2: Network plot of overall symptoms comparison, eg, because of dierent participant The lines link treatments with direct comparisons in trials with usable data for the primary outcome; the thickness characteristics) was assessed by comparing common-τ²27 of the lines corresponds to the number of trials evaluating the comparison; the size of the nodes corresponds to to empirical distributions,28,29 and presented in prediction the number of participants assigned to the treatment. intervals.30 Particularly, common-τ² in the primary NMAs dysfunction, and erectile dysfunction) adverse events, of each outcome were compared with those observed in weight gain, hyperprolactinaemia, QTc prolongation, NMAs excluding subgroups. NMA consistency was sedation, akathisia, heart rate abnormalities (and assessed using global and local methods.30 Heterogeneity separately for tachycardia, bradycardia, and arrythmia), and inconsistency were explored by network meta- reduction in white blood cells (and separately for regression for the primary outcome using the following leukopenia, neutropenia, and agranulocytosis), seizures, potential eect modifiers: participants’ severity of illness deep vein thrombosis, and death (appendix pp 76–78). at baseline, sponsorship by pharmaceutical companies, 880 Articles publication year, degree of placebo-response, sample size, implementation for computational ease as no major percentage of male (gender) participants, and mean age. dierences can be expected with non-informative or In sensitivity analyses for the primary outcome, we minimally-informative Bayesian priors.42 Network meta- excluded trials for the following reasons: trials not using regression analyses were conducted with Bayesian operationalised diagnostic criteria; trials using imputed random-eects models in R with the package standard deviations; trials reporting complete-case “multinma”,43 due to lack of frequentist network meta- analyses only; trials with overall high risk of bias; trials regression models implementation. published before 1994; trials shorter than 4 weeks or In reporting of the results, we define k as the number longer than 8 weeks; industry-independent trials from of studies with usable data and t as the number of countries with research tradition and industry-funded investigated antipsychotics. trials without active confirmation of appropriate People with lived experience provided constructive randomisation; trials in subgroup populations; trials with feedback regarding the study design, interpretation of potentially unfair dose comparisons (following Huhn and results, and writing of the report with a particular focus colleagues4); and trials with placebo. We also re-analysed regarding relevance for patients. the network without merging modes of administration (ie, treating oral, transdermal, and long-acting injectable Role of the funding source applications of the same drug as distinct interventions), The funders of the study had no role in study design, and using post-hoc adjusted data for clozapine trials from data collection, data analysis, data interpretation, or an individual patient data meta-analysis.31 Additionally, writing of the report. we conducted post-hoc sensitivity analyses restricted to single-blind and open-label trials, respectively, and Results combining double-blind, single-blind, and open-label From previous reviews (661 studies) and their trials. update-searches in international literature databases Risk of bias was assessed using Cochrane’s Risk-of- (6892 ref erences), we identified 467 eligible inter- Bias 2.32 Small-study eects and publication bias were nationally published RCTs. 414 were included after explored with comparison-adjusted33 and contour- contacting authors for confirmation of randomisation enhanced funnel plots;34 and overall reporting bias was (with 312 letters and emails sent); 53 studies from evaluated following ROB-MEN.35 Confidence in estimates countries without research tradition were excluded was assessed using the Confidence in Network Meta- because seven revealed suboptimal randomisation and Analysis (CINEMA) approach.36 for 46 no answers were received (figure 1). For CINEMA and for interpreting results, we considered From screening 11 306 references found in Chinese treatment dierences to be very small (and likely clinically databases, we identified 5117 Chinese RCTs. After calling insignificant) when SMDs were between –0·1 and 0·1. the authors of the 176 double-blind studies (ie, relevant This choice was informed by the fact that SMD 0·2 is for the primary outcome) by telephone (at least 3 times) typically considered a small eect according to Cohen’s and sending 698 emails and 7126 letters to the first and rule27,37 and thus SMD 0·1 might be considered a very second authors of all studies (facilitating answering by an small eect. Moreover, SMD 0·1 reflects a dierence in especially developed online questionnaire), answers for event rates of 30% (the lowest drug response rate observed 347 (7%) of 5117 studies were received. Of those, in our analysis) versus 35%,27 which we believe is an 27 (8%) of 347 had confirmed appropriate methods and appropriate threshold with larger or smaller dierences 24 investigating antipsychotics for acute psychosis were being likely clinically relevant or irrelevant, respectively. included. Other Chinese trials were not included because The same approach was used for other eect size they used suboptimal methods (125 [36%] of 347 with measures: ORs between 0·8 and 1·25 were deemed very answers), methods remained unclear (39 [11%] of 347), or small as these values can be transformed to SMD were observational studies although reported as –0·1 and 0·1,27 respectively. For MDs, margins of randomised (83 [24%] of 347). For another 73 (21%) of the equivalence approximately corresponding to SMD 347 studies, the authors did not explicitly confirm –0·1 and 0·1 were calculated using the outcome’s median authenticity. standard deviation—those were –0·5 kg to 0·5 kg for In total, 438 RCTs with 80 567 participants were weight, –5 ng/ml to 5 ng/ml for prolactin, and –5 ms included. Of those, 388 RCTs with 78 193 participants to 5 ms for QTc (appendix p 81). investigating 24 antipsychotics provided usable data for Network and pairwise meta-analyses were performed at least one outcome. 315 (81%) of these 388 were double- with frequentist random-eects models using R blind and 93 (24%) of 388 were conducted in (packages “meta”38 and “netmeta”39); outcomes with rare schizophrenia subgroups. Median study duration was events were (network) meta-analysed with Mantel– 6 weeks (IQR 6–8), participants’ median age was Haenszel models;40 and surface under the cumulative 37·28 years (IQR 33·58–40·50), 28 448 (36·4% of all ranking scores41 were calculated for ranking interventions. 78 193 participants) were women, and 49 745 (63·6%) of Frequentist were chosen over Bayesian NMA model 78 193 participants were men (appendix p 82). Articles A Overall symptoms B Positive symptoms C T SMD S|I C T SMD S|I (95% CI) (of 24) (95% CI) (of 23 ) Clozapine (n=703) 6 22 –0·90 (–1·03 to –0·77) 19| Clozapine (n=450) 4 12 –0·90 (–1·06 to –0·73) 20| Amisulpride (n=797) 4 12 –0·68 (–0·81 to –0·55) 11| Amisulpride (n=745) 4 9 –0·65 (–0·79 to –0·52) 8| Sulpiride (n=71) 1 2 –0·73 (–1·11 to –0·35) | Risperidone (n=6092) 18 59 –0·57 (–0·62 to –0·52) 5|1 Olanzapine (n=6690) 16 63 –0·57 (–0·62 to –0·52) 5|1 Xanomeline–trospium (n=314) 1 3 –0·59 (–0·78 to –0·40) 1| Zotepine (n=209) 4 6 –0·59 (–0·80 to –0·38) | Olanzapine (n=5258) 16 49 –0·53 (–0·59 to –0·47) 2|1 Xanomeline–trospium (n=314) 1 3 –0·57 (–0·76 to –0·37) | Paliperidone (n=2868) 4 16 –0·52 (–0·60 to –0·45) 1|1 Risperidone (n=6956) 18 72 –0·53 (–0·57 to –0·48) 3|1 Haloperidol (n=3846) 15 46 –0·45 (–0·51 to –0·39) 1|1 Perphenazine (n=526) 6 5 –0·51 (–0·65 to –0·36) |1 Olanzapine–samidorphan (n=132) 2 1 –0·46 (–0·73 to –0·20) |1 Paliperidone (n=3090) 4 19 –0·49 (–0·56 to –0·42) |1 Blonanserin (n=1169) 3 8 –0·45 (–0·56 to –0·33) |1 Blonanserin (n=1166) 3 8 –0·47 (–0·59 to –0·36) |1 Aripiprazole (n=3176) 11 22 –0·44 (–0·50 to –0·37) 1|1 Olanzapine–samidorphan (n=124)2 1 –0·49 (–0·77 to –0·22) |1 Asenapine (n=1217) 4 6 –0·43 (–0·55 to –0·32) |1 Aripiprazole (n=3753) 11 26 –0·45 (–0·52 to –0·39) |1 Lumateperone (n=222) 2 2 –0·43 (–0·65 to –0·22) |1 Haloperidol (n=5013) 16 72 –0·44 (–0·50 to –0·38) |2 Perphenazine (n=455) 5 3 –0·42 (–0·58 to –0·26) |1 Perospirone (n=87) 4 2 –0·42 (–0·73 to –0·10) |1 Chlorpromazine (n=772) 5 8 –0·39 (–0·53 to –0·25) |1 Quetiapine (n=3905) 11 26 –0·40 (–0·47 to –0·33) |2 Ziprasidone (n=1459) 9 13 –0·38 (–0·47 to –0·28) |2 Sertindole (n=1366) 4 8 –0·39 (–0·50 to –0·27) |2 Sertindole (n=1395) 4 8 –0·36 (–0·47 to –0·25) |2 Asenapine (n=1582) 4 8 –0·38 (–0·48 to –0·28) |2 Quetiapine (n=3568) 11 20 –0·35 (–0·42 to –0·28) |3 Ziprasidone (n=1585) 9 15 –0·38 (–0·47 to –0·29) |2 Lurasidone (n=2374) 5 11 –0·34 (–0·42 to –0·25) |3 Chlorpromazine (n=1288) 9 24 –0·36 (–0·47 to –0·26) |2 Cariprazine (n=1135) 3 4 –0·33 (–0·46 to –0·21) |3 Lurasidone (n=2572) 5 12 –0·34 (–0·42 to –0·26) |3 Zotepine (n=35) 1 1 –0·16 (–0·67 to 0·34) |1 Cariprazine (n=1293) 3 5 –0·32 (–0·44 to –0·21) |3 Iloperidone (n=918) 4 4 –0·28 (–0·41 to –0·16) |4 Iloperidone (n=2157) 4 7 –0·31 (–0·41 to –0·21) |4 Perospirone (n=87) 4 2 –0·17 (–0·49 to 0·15) |2 Brexpiprazole (n=1474) 3 7 –0·29 (–0·39 to –0·18) |4 Brexpiprazole (n=1474) 3 7 –0·21 (–0·32 to –0·11) |8 Lumateperone (n=386) 2 3 –0·23 (–0·39 to –0·06) |4 Placebo (n=10305) 18 101 Reference Placebo (n=11841) 22 129 Reference –1 –0·5 0 0·5 1 –1 –0·5 0 0·5 1 Favours comparator Favours reference Favours comparator Favours reference C Negative symptoms D Discontinuation for any reason C T SMD S|I C T OR S|I Events (95% CI) (of 23) (95% CI) (of 24) (95% CI) Perospirone (n=87) 4 2 –0·65 (–0·95 to –0·34) 2| Paliperidone (n=3106) 3 17 0·49 (0·42 to 0·56) 6| 28% (25% to 30%) Clozapine (n=525) 6 19 –0·55 (–0·69 to –0·41) 6| Olanzapine–samidorphan (n=345) 2 2 0·44 (0·29 to 0·69) 2| 26% (19% to 35%) Zotepine (n=174) 4 4 –0·51 (–0·73 to –0·29) 1| Olanzapine (n=6675) 14 56 0·50 (0·45 to 0·56) 6| 28% (26% to 30%) Amisulpride (n=735) 3 11 –0·47 (–0·59 to –0·34) 1| Clozapine (n=559) 6 19 0·50 (0·35 to 0·71) 1| 28% (22% to 36%) Blonanserin (n=1169) 3 8 –0·44 (–0·55 to –0·33) 1| Amisulpride (n=781) 4 10 0·51 (0·39 to 0·68) 3| 29% (23% to 35%) Olanzapine (n=5264) 16 50 –0·41 (–0·47 to –0·36) 1| Perphenazine (n=596) 9 7 0·58 (0·43 to 0·77) 1| 31% (25% to 38%) Perphenazine (n=455) 5 3 –0·40 (–0·54 to –0·25) | Risperidone (n=7276) 17 68 0·59 (0·53 to 0·65) 3| 32% (29% to 34%) Aripiprazole (n=2978) 10 21 –0·37 (–0·44 to –0·30) | Lumateperone (n=408) 2 3 0·59 (0·42 to 0·83) 1| 32% (25% to 39%) Paliperidone (n=2868) 4 16 –0·36 (–0·43 to –0·29) | Aripiprazole (n=3484) 10 24 0·63 (0·55 to 0·72) 1| 33% (30% to 36%) Sertindole (n=1395) 4 8 –0·35 (–0·45 to –0·24) | Quetiapine (n=3671) 10 21 0·64 (0·55 to 0·74) 1| 33% (30% to 37%) Risperidone (n=6174) 18 61 –0·35 (–0·39 to –0·30) | Zotepine (n=222) 5 5 0·63 (0·41 to 0·97) | 33% (24% to 43%) Asenapine (n=1216) 4 6 –0·34 (–0·45 to –0·23) | Perospirone (n=70) 1 1 0·66 (0·31 to 1·41) | 34% (20% to 52%) Olanzapine–samidorphan (n=132)2 1 –0·34 (–0·58 to –0·09) | Iloperidone (n=2209) 4 7 0·72 (0·59 to 0·88) | 36% (32% to 41%) Xanomeline–trospium (n=314) 1 3 –0·33 (–0·51 to –0·16) | Asenapine (n=1665) 4 8 0·73 (0·60 to 0·88) | 36% (32% to 41%) Cariprazine (n=1135) 3 4 –0·32 (–0·43 to –0·20) | Brexpiprazole (n=1495) 3 7 0·75 (0·59 to 0·94) | 37% (32% to 42%) Ziprasidone (n=1316) 8 11 –0·31 (–0·40 to –0·22) | Sulpiride (n=129) 3 4 0·77 (0·36 to 1·64) | 38% (22% to 56%) Lumateperone (n=222) 2 2 –0·27 (–0·47 to –0·07) | Lurasidone (n=2607) 5 12 0·77 (0·66 to 0·90) |2 38% (34% to 41%) Haloperidol (n=4078) 15 53 –0·29 (–0·34 to –0·23) |1 Chlorpromazine (n=1222) 9 23 0·78 (0·60 to 1·01) | 38% (32% to 44%) Quetiapine (n=3625) 11 21 –0·28 (–0·35 to –0·21) |1 Blonanserin (n=1206) 3 8 0·79 (0·63 to 1·00) | 38% (33% to 44%) Brexpiprazole (n=1474) 3 7 –0·26 (–0·35 to –0·16) |1 Ziprasidone (n=1789) 9 15 0·79 (0·67 to 0·94) |2 38% (34% to 42%) Lurasidone (n=2374) 5 11 –0·25 (–0·32 to –0·17) |1 Cariprazine (n=1143) 3 5 0·83 (0·67 to 1·03) |2 39% (34% to 45%) Iloperidone (n=918) 4 4 –0·21 (–0·32 to –0·10) |2 Haloperidol (n=4845) 17 66 0·85 (0·76 to 0·96) |4 40% (37% to 43%) Chlorpromazine (n=926) 7 14 –0·16 (–0·27 to –0·05) |6 Sertindole (n=1127) 4 6 0·95 (0·75 to 1·21) |5 43% (37% to 49%) Placebo (n=10398) 20 105 Reference Placebo (n=12410) 22 128 Reference 43·916% Xanomeline–trospium (n=341) 1 3 1·26 (0·85 to 1·85) |10 50% (40% to 59%) –1 –0·5 0 0·5 1 0·1 0·2 0·5 1 2 5 10 Favours comparator Favours reference Favours comparator Favours reference 882 Articles Figure 3: Panel of efficacy and acceptability outcomes 256 double-blind studies with 58 948 participants, The order of treatments is according to the SUCRA-ranking. The direction of the investigating 24 antipsychotics and placebo, provided effect is indicated below the x-axis. The column C indicates, for the treatment in usable data for overall symptoms of schizophrenia the line, how many other interventions it has been directly compared with in a study; the column T indicates the number of studies with data for this (figures 2–4 and appendix p 152). All 24 antipsychotics intervention. The column S|I indicates, for the treatment in the line, how many reduced symptoms compared with placebo with, on other antipsychotics the intervention is superior (before the |-symbol) or inferior average, more than very small eects (ie, SMDs less (after the |-symbol) to, defined as a point estimate above or below very small than –0·1) and all with 95% CIs excluding no dierence effects, respectively, and 95% CIs excluding very small differences (appendix p 80); in the parenthesis after S|I in the column headings is the total number of and all, except perospirone and lumateperone, with antipsychotics in this NMA. The colour of the lines of the CIs indicates the 95% CIs excluding very small eects. SMDs ranged CINEMA-confidence in estimates: green=high, blue=moderate, orange=low, and from –0·90 (95% CI –1·03 to –0·77) for clozapine to –0·23 red=very low. C=connectivity. CINEMA=Confidence in Network Meta-Analysis. (–0·39 to –0·06) for lumateperone. Clozapine, amisulpride, I=inferior. n=number of participants. OR=odds ratio. SMD=standardised mean difference. S=superior. SUCRA= surface under the cumulative ranking. olanzapine, and risperidone reduced symptoms more than NMA=network meta-analysis. T=trials. at least three other antipsychotics with 95% CIs excluding very small eects. Clozapine was also more ecacious than olanzapine and risperidone. Lumateperone, brexpiprazole, iloperidone, cariprazine, and lurasidone We first report, per outcome, the percentage of reduced symptoms less than at least three other antipsychotics with average eects versus placebo that antipsychotics with 95% CIs excluding very small eects. are not very small (see Methods for each outcome’s 23 (100%) of 23 antipsychotics reduced positive definition of very small): irrespective of the 95% CI, with symptoms, on average, more than placebo irrespective of 95% CIs excluding no dierence (SMD or MD 0, or the 95% CI (21 [91%] of 23 with 95% CIs excluding no OR 1), and with 95% CIs excluding very small eects. dierence; 21 [91%] of 23 with 95% CIs excluding very Then, to support drug choice, we report the specific small eects), with SMDs ranging from –0·90 (95% CI antipsychotics superior or inferior to at least three other –1·06 to –0·73) to –0·16 (–0·67 to 0·34; NMA based on antipsychotics with 95% CIs excluding very small eects 189 studies with usable data; with 49 466 participants; (ie, with a high probability of more than very small and investigating 23 antipsychotics; these numbers are thus likely clinically relevant dierences). denoted as k, n, and t, respectively, hereafter). Particularly Clozapine NA NA (–0·6 – 9 0 · t 3 o 1 0·06)(–0·6 – 6 0 · t 0 o 8 0·51) NA (–0·4 – 2 0 t · o 1 9 0·03) NA NA NA NA NA (–0·6 – 5 0 t · o 4 1 –0·18) NA NA NA NA NA (–0·9 – 7 0 t · o 7 6 –0·54) NA NA NA NA NA (–2·69 – 1 t · o 7 1 –0·73) (–0·4 – 0 0 · t 2 o 2 – 0·05)Amisulpride NA (–0·2 – 8 0 t ·0 o 8 0·13) NA NA (–0·35 –0 t · o 0 8 0·19) NA NA NA NA NA (–0·4 – 9 0 t · o 2 8 –0·07) NA NA NA NA NA NA NA NA NA NA NA (–1·37 – 0 to ·8 – 8 0·39) (–0·5 – 6 0 t ·1 o 7 0·22)(–0·35 0 t ·0 o 5 0·45) Sulpiride NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA (–0·73 –0 t · o 3 7 0·00) NA NA NA NA NA NA (–0·4 – 6 0 · t 3 o 3 –0·20)(–0·24 – 0 to ·1 0 1 ·02)(–0·5 – 4 0 t ·1 o 6 0·22) Olanzapine NA NA (–0·25 –0 t · o 1 3 –0·01)(–0·3 – 6 0 t ·1 o 1 0·14) (–0·1 – 8 0 · t 0 o 6 0·07) NA (–0·22 0 t ·1 o 0 0·41)(–0·31 –0 t · o 1 6 –0·01)(–0·25 –0 t · o 1 4 –0·02)(–1·2 – 9 0 t ·6 o 2 0·05)(–0·3 – 7 0 t ·1 o 6 0·04)(–0·5 – 1 0 t ·2 o 3 0·05)(–0·34 0 · t 1 o 4 0·06)(–0·38 –0 t · o 2 0 –0·02)(–0·6 – 9 0 t · o 22 0·26)(–0·4 – 7 0 t ·1 o 8 0·11) NA NA NA NA (–0·61 – 0 to ·5 – 3 0·45) (–0·5 – 4 0 t ·3 o 1 –0·08)(–0·33 – 0 to ·0 0 9 ·16)(–0·5 – 6 0 t ·1 o 4 0·29)(–0·1 0 9 · 0 to 2 0·24) Zotepine NA NA NA NA NA NA NA (–0·19 0 · t 0 o 8 0·35) NA NA NA NA NA (–1·17 – 0 to ·7 – 4 0·31) NA NA NA NA NA (–1·26 – 0 to ·8 – 2 0·39) (–0·5 – 7 0 t ·3 o 3 – 0·10)(–0·34 – 0 to ·1 0 1 ·12)(–0·5 – 9 0 t ·1 o 6 0·27)(–0·2 – 0 0 t ·0 o 0 0·20)(–0·3 – 1 0 t ·0 o 3 0·26) Xa t n ro o s m pi e u l m ine– NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA (–0·76 – 0 to ·5 – 7 0·37) (–0·5 – 0 0 t ·3 o 7 –0·25)(–0·28 – 0 to ·1 – 5 0·02)(–0·5 – 8 0 t ·2 o 0 0·18)(–0·1 – 0 0 t ·0 o 4 0·02)(–0·2 – 8 0 t ·0 o 7 0·15) (–0·2 – 4 0 · t 0 o 4 0·16) Risperidone (–0·1 0 6 · 0 to 6 0·27) (–0·1 0 6 ·0 t 6 o 0·27)(–0·30 –0 t · o 1 1 0·07) NA (–0·21 –0 t · o 0 3 0·14)(–0·16 –0 t · o 0 4 0·07) (–1·1 – 3 0 t · o 4 5 0·22) (–0·2 – 1 0 t ·0 o 4 0·12) (–0·3 – 5 0 t ·1 o 2 0·11) (–0·27 0 · t 1 o 4 0·56)(–0·2 – 3 0 t ·0 o 6 0·11) NA (–0·4 – 1 0 t ·2 o 0 0·01)(–0·59 –0 t · o 3 6 –0·13)(–0·46 –0 t · o 2 6 –0·06) NA (–0·4 – 7 0 t ·2 o 4 –0·01)(–0·62 – 0 to ·5 – 5 0·48) (–0·5 – 9 0 t ·4 o 0 –0·21)(–0·37 – 0 to ·1 0 7 ·02)(–0·6 – 3 0 t ·2 o 2 0 ·18)(–0·2 – 1 0 t ·0 o 6 0·08)(–0·3 – 4 0 t ·0 o 9 0·17) (–0·3 – 0 0 · t 0 o 6 0·18)(–0·17 –0 t · o 0 0 2 ·13) Perphenazine NA NA NA (–0·33 –0 t · o 0 3 0·26) NA NA (–0·2 – 7 0 t ·0 o 5 0·16) NA NA (–0·4 – 2 0 t ·1 o 6 0·11) NA NA NA NA NA NA (–1·51 – t 0 o ·8 – 9 0·27) (–0·5 – 5 0 t ·4 o 1 –0·27)(–0·33 – 0 to ·1 – 9 0·04)(–0·6 – 2 0 t ·2 o 4 0·15)(–0·1 – 6 0 t ·0 o 8 0·00)(–0·3 – 2 0 t ·1 o 0 0·12) (–0·2 – 8 0 · t 0 o 8 0·13)(–0·12 –0 t · o 0 3 0·05)(–0·1 – 7 0 t · o 0 1 0·15) Paliperidone NA NA (–0·29 0 · t 0 o 4 0·36) NA NA NA NA NA NA NA NA NA NA NA NA (–0·57 – 0 to ·4 – 9 0·40) (–0·5 – 9 0 t ·4 o 3 – 0·26)(–0·37 – 0 to ·2 – 0 0·03)(–0·6 – 5 0 t ·2 o 5 0 ·14)(–0·2 – 2 0 t ·1 o 0 0·03)(–0·3 – 5 0 t ·1 o 2 0·12) (–0·3 – 2 0 · t 0 o 9 0·13)(–0·17 –0 t · o 0 0 5 ·06)(–0·2 – 1 0 t ·0 o 3 0·15) (–0·1 – 5 0 · t 0 o 2 0·12) Blonanserin NA NA (–0·33 –0 t · o 1 3 0·06) NA NA NA NA NA NA NA NA NA NA NA (–0·66 –0 t · o 4 – 7 0·28) (–0·7 – 1 0 t · o 4 1 –0·11)(–0·48 –0 t · o 1 8 0·12)(–0·7 – 0 0 t ·2 o 3 0·23)(–0·3 – 5 0 t ·0 o 8 0·20)(–0·4 – 4 0 · t 1 o 0 0·25) (–0·4 – 1 0 · t 0 o 7 0·26)(–0·31 –0 t · o 0 3 0·25)(–0·3 – 2 0 t ·0 o 1 0·30)(–0·2 0 8 · 0 to 0 0·29)(–0·28 0 · t 0 o 2 0·32) O sa l m an i z d a o p r i p n h e a – n NA NA NA NA NA NA NA NA NA NA NA NA NA (–0·63 – 0 to ·3 0 1 ·00) (–0·5 – 9 0 t ·4 o 5 –0·31)(–0·37 – 0 to ·2 – 2 0·08)(–0·6 – 6 0 · t 2 o 7 0·11)(–0·1 – 9 0 t ·1 o 2 –0·04)(–0·3 – 6 0 t ·1 o 4 0·08)(–0·3 – 2 0 · t 1 o 1 0·09)(–0·15 –0 t · o 0 7 0·00)(–0·2 – 0 0 t ·0 o 5 0·10)(–0·1 – 3 0 t ·0 o 4 0·05)(–0·15 –0 t · o 0 2 0·11) (–0·3 – 2 0 · t 0 o 4 0·24) Aripiprazole(–0·28 –0 t · o 0 5 0·18) NA (–0·4 – 8 0 · t 0 o 7 0·35) NA NA (–0·4 – 7 0 t · o 1 6 0·16)(–0·6 – 4 0 t · o 2 9 0·05) NA (–0·25 0 t ·0 o 2 0·30) NA (–0·2 – 7 0 t ·0 o 8 0·10) NA (–0·54 – 0 to ·4 – 5 0·36) (–0·5 – 9 0 t ·4 o 6 –0·33)(–0·37 – 0 to ·2 – 4 0·11)(–0·6 – 7 0 t ·2 o 9 0·09)(–0·1 – 9 0 t ·1 o 3 –0·07)(–0·3 – 6 0 t ·1 o 5 0·05) (–0·3 – 3 0 · t 1 o 3 0·07)(–0·15 –0 t · o 0 9 –0·03)(–0·2 – 2 0 t ·0 o 6 0·09)(–0·1 – 4 0 t ·0 o 5 0·04)(–0·15 –0 t · o 0 3 0·08) (–0·3 – 3 0 · t 0 o 5 0·22)(–0·0 – 9 0 t · o 0 1 0·06) Haloperidol (–0·2 0 2 · 1 to 6 0·54)(–0·1 0 4 · 0 to 1 0·15)(–0·2 – 5 0 t ·0 o 9 0·08)(–0·28 0 · t 0 o 2 0·32)(–0·24 0 · t 0 o 2 0·27) (–0·5 0 2 · 2 to 9 1·11) (–0·6 – 2 0 t ·2 o 8 0·06) NA (–0·22 –0 t · o 0 8 0·06) NA NA (–0·59 – 0 to ·4 – 9 0·40) (–0·8 – 2 0 t ·4 o 8 –0·15)(–0·60 – 0 t · o 2 0 6 ·08)(–0·8 – 0 0 · t 3 o 1 0·18)(–0·4 – 7 0 t ·1 o 5 0·16)(–0·5 – 5 0 t ·1 o 8 0·20) (–0·5 – 2 0 · t 1 o 5 0·22)(–0·4 – 3 0 t · o 1 1 0·21)(–0·4 – 3 0 t ·0 o 9 0·26)(–0·4 – 0 0 · t 0 o 8 0·25)(–0·39 –0 t · o 0 6 0·27) (–0·5 – 0 0 · t 0 o 8 0·34)(–0·36 –0 t · o 0 4 0·28)(–0·3 – 4 0 t ·0 o 2 0·29) Perospirone (–0·3 0 3 · 3 to 4 1·01) NA NA NA NA NA NA NA NA NA NA (–0·6 – 4 0 · t 5 o 0 –0·37)(–0·42 – 0 to ·2 – 8 0·14)(–0·7 – 1 0 t · o 33 0·05)(–0·2 – 5 0 t ·1 o 7 –0·09)(–0·4 – 1 0 t ·2 o 0 0·02)(–0·3 – 8 0 t ·1 o 7 0·04)(–0·21 – 0 to ·1 – 3 0·05)(–0·2 – 6 0 t ·1 o 1 0·05)(–0·1 – 9 0 t ·0 o 9 0·00)(–0·21 –0 t · o 0 8 0·05) (–0·3 – 8 0 · t 1 o 0 0·19)(–0·15 –0 t · o 0 6 0·03)(–0·1 – 2 0 t ·0 o 4 0·04)(–0·3 – 4 0 t ·0 o 2 0·30) Quetiapine NA NA (–0·34 –0 t · o 0 8 0·19)(–0·0 0 4 · t 1 o 0 0·25)(–0·4 – 7 0 t ·1 o 8 0·12) NA NA (–0·5 – 1 0 t ·2 o 2 0·08) NA (–0·52 – 0 to ·4 – 1 0·31) (–0·6 – 8 0 · t 5 o 2 –0·35)(–0·46 –0 t · o 2 9 –0·13)(–0·7 – 4 0 t ·3 o 4 0·05)(–0·3 – 0 0 t ·1 o 8 –0·07)(–0·4 – 4 0 · t 2 o 1 0·03)(–0·4 – 1 0 t ·1 o 8 0·04)(–0·26 – 0 to ·1 – 4 0·03)(–0·3 – 0 0 t ·1 o 2 0·06)(–0·2 – 4 0 t ·1 o 1 0·03)(–0·25 –0 t · o 0 9 0·07) (–0·4 – 1 0 · t 1 o 1 0·19)(–0·20 –0 t · o 0 7 0·06)(–0·1 – 7 0 t · o 0 5 0·06)(–0·3 – 6 0 t ·0 o 3 0·30)(–0·1 – 4 0 t ·0 o 1 0·12) Sertindole NA NA NA NA NA NA NA NA (–0·63 – 0 to ·4 – 3 0·23) (–0·6 – 8 0 · t 5 o 2 –0·36)(–0·46 –0 t · o 3 0 –0·13)(–0·7 – 4 0 t ·3 o 5 0·05)(–0·3 – 0 0 t ·1 o 9 –0·08)(–0·4 – 4 0 · t 2 o 1 0·02)(–0·4 – 0 0 · t 1 o 9 0·03)(–0·25 – 0 to ·1 – 5 0·04)(–0·3 – 0 0 t ·1 o 2 0·05)(–0·2 – 3 0 t ·1 o 1 0·01)(–0·25 –0 t · o 0 9 0·06) (–0·4 – 1 0 · t 1 o 1 0·18)(–0·19 –0 t · o 0 7 0·05)(–0·1 – 7 0 t · o 0 6 0·05)(–0·3 – 7 0 t ·0 o 4 0·30)(–0·1 – 4 0 t ·0 o 2 0·10)(–0·1 – 5 0 t ·0 o 0 0·15) Asenapine NA NA NA NA NA NA NA (–0·43 – 0 to ·3 – 2 0·21) (–0·6 – 7 0 t ·5 o 2 –0·37)(–0·45 – 0 to ·3 – 0 0·14)(–0·7 – 3 0 t ·3 o 5 0·04)(–0·2 – 8 0 t ·1 o 9 –0·09)(–0·4 – 4 0 · t 2 o 1 0·01)(–0·4 – 0 0 · t 1 o 9 0·03)(–0·24 – 0 to ·1 – 5 0·05)(–0·2 – 9 0 t ·1 o 2 0·04)(–0·2 – 2 0 t ·1 o 1 0·00)(–0·24 –0 t · o 0 9 0·05) (–0·4 – 0 0 · t 1 o 1 0·17)(–0·18 –0 t · o 0 7 0·03)(–0·1 – 6 0 t ·0 o 6 0·04)(–0·3 – 6 0 t ·0 o 4 0·29)(–0·1 – 2 0 t ·0 o 2 0·09)(–0·1 – 5 0 t ·0 o 0 0·14)(–0·1 – 3 0 t ·0 o 0 0·13) Ziprasidone (–0·3 – 7 0 t ·0 o 8 0·22) NA NA (–0·29 – 0 to ·0 0 2 ·26) NA NA (–0·46 –0 t · o 3 0 –0·15) (–0·6 – 8 0 t · o 54 –0·40)(–0·48 –0 t · o 3 2 –0·15)(–0·7 – 3 0 t ·3 o 7 0·00)(–0·3 – 2 0 t ·2 o 1 –0·10)(–0·4 – 5 0 t ·2 o 3 –0·01)(–0·4 – 2 0 t ·2 o 0 0·02)(–0·27 – 0 to ·1 – 6 0·05)(–0·3 – 2 0 t ·1 o 4 0·03)(–0·2 – 5 0 t ·1 o 3 0·00)(–0·26 –0 t · o 1 1 0·04)(–0·4 – 2 0 · t 1 o 3 0·16)(–0·21 –0 t · o 0 9 0·03)(–0·1 – 9 0 t ·0 o 8 0·03)(–0·3 – 8 0 t ·0 o 5 0·28)(–0·1 – 4 0 t ·0 o 3 0·07)(–0·1 – 7 0 t · o 0 2 0·13) (–0·1 – 6 0 t ·0 o 2 0·13)(–0·14 –0 t · o 0 2 0·11) Chl a o z r i p n r e om- NA NA NA NA NA (–0·73 – 0 to ·5 – 0 0·26) (–0·71 –0 t · o 5 6 –0·41)(–0·49 –0 t · o 3 4 –0·18)(–0·7 – 8 0 t ·3 o 9 0·00)(–0·3 – 2 0 t ·2 o 3 –0·14)(–0·4 – 7 0 t ·2 o 5 –0·03)(–0·4 – 4 0 · t 2 o 3 –0·02)(–0·28 – 0 to ·1 – 8 0·09)(–0·3 – 3 0 t ·1 o 6 0·00)(–0·2 – 6 0 t ·1 o 5 –0·04)(–0·27 –0 t · o 1 3 0·01) (–0·4 – 4 0· t 1 o 5 0·13)(–0·22 –0 t · o 1 1 –0·01)(–0·1 – 9 0 t ·1 o 0 0·00)(–0·4 – 0 0 · t 0 o 7 0·25)(–0·1 – 6 0 t ·0 o 6 0·05)(–0·1 – 8 0 t ·0 o 4 0·09)(–0·1 – 7 0 t · o 0 4 0·09)(–0·16 –0 t · o 0 4 0·08) (–0·1 – 5 0 t ·0 o 2 0·11) Lurasidone NA NA NA NA (–0·45 – 0 to ·3 – 6 0·27) (–0·75 –0 t · o 5 8 –0·41)(–0·53 – 0 to ·3 – 5 0·18)(–0·8 – 0 0 · t 4 o 1 –0·01)(–0·3 – 7 0 t ·2 o 5 –0·12)(–0·5 – 1 0 t ·2 o 7 –0·03)(–0·4 – 7 0 t ·2 o 4 –0·02)(–0·33 – 0 to ·2 – 0 0·08)(–0·3 – 7 0 t ·1 o 8 0·00)(–0·3 – 0 0 t ·1 o 7 –0·03)(–0·31 –0 t · o 1 5 0·01) (–0·4 – 7 0 · t 1 o 7 0·13)(–0·26 –0 t · o 1 3 0·00)(–0·2 – 4 0 t ·1 o 2 0·01)(–0·4 – 3 0 t ·0 o 9 0·24)(–0·2 – 1 0 t ·0 o 7 0·06)(–0·2 – 2 0 t ·0 o 6 0·10)(–0·2 – 1 0 t ·0 o 6 0·10)(–0·20 –0 t · o 0 6 0·09)(–0·20 –0 t · o 0 4 0·12) (–0·1 – 6 0 · t 0 o 2 0·12) Cariprazine NA NA NA (–0·47 – 0 to ·3 – 4 0·21) (–0·75 –0 t · o 5 9 –0·43)(–0·53 – 0 to ·3 – 7 0·21)(–0·8 – 1 0 t ·4 o 2 –0·02)(–0·3 – 7 0 t ·2 o 6 –0·15)(–0·5 – 1 0 t ·2 o 8 –0·05)(–0·4 – 7 0 t ·2 o 6 –0·04)(–0·32 – 0 to ·2 – 1 0·11)(–0·3 – 7 0 t ·1 o 9 –0·02)(–0·3 – 0 0 t ·1 o 8 –0·06)(–0·31 –0 t · o 1 6 –0·01)(–0·4 – 7 0 · t 1 o 8 0·11)(–0·26 –0 t · o 1 4 –0·03)(–0·23 –0 t · o 1 3 –0·03)(–0·4 – 3 0 t ·1 o 0 0·22)(–0·2 – 0 0 t ·0 o 9 0·03)(–0·2 – 2 0 t ·0 o 7 0·07)(–0·2 – 1 0 t ·0 o 7 0·07)(–0·19 –0 t · o 0 7 0·06)(–0·19 –0 t · o 0 5 0·09)(–0·16 –0 t · o 0 3 0·10) (–0·1 – 6 0 · t 0 o 1 0·14) Iloperidone NA NA (–0·39 – 0 to ·2 – 4 0·10) (–0·78 –0 t · o 6 1 –0·45)(–0·56 – 0 to ·3 – 9 0·22)(–0·8 – 3 0 t ·4 o 4 –0·05)(–0·4 – 0 0 · t 2 o 8 –0·17)(–0·5 – 4 0 t ·3 o 0 –0·07)(–0·5 – 0 0 t ·2 o 8 –0·06)(–0·35 – 0 to ·2 – 4 0·12)(–0·40 – 0 to ·2 – 2 0·04)(–0·3 – 3 0 t ·2 o 0 –0·08)(–0·34 –0 t · o 1 9 –0·03)(–0·5 – 0 0 · t 2 o 1 0·09)(–0·28 –0 t · o 1 7 –0·05)(–0·27 –0 t · o 1 5 –0·03)(–0·4 – 6 0 · t 1 o 3 0·20)(–0·2 – 3 0 t ·1 o 1 0·01)(–0·2 – 5 0 t ·1 o 0 0·06)(–0·2 – 4 0 t ·0 o 9 0·05)(–0·23 –0 t · o 0 9 0·04)(–0·22 –0 t · o 0 7 0·07)(–0·1 – 9 0 t ·0 o 5 0·08) (–0·1 – 9 0 · t 0 o 4 0·12)(–0·17 –0 t · o 0 0 2 ·12) Brexpiprazole NA (–0·39 – 0 to ·2 – 7 0·15) (–0·8 – 8 0 t · o 6 8 –0·47)(–0·66 – 0 to ·4 – 5 0·24)(–0·9 – 2 0 t · o 5 0 –0·09)(–0·5 – 2 0 t ·3 o 5 –0·17)(–0·6 – 3 0 t ·3 o 7 –0·10)(–0·6 – 0 0 · t 3 o 4 –0·09)(–0·47 – 0 to ·3 – 0 0·13)(–0·50 – 0 to ·2 – 8 0·06)(–0·4 – 5 0 t ·2 o 7 –0·09)(–0·4 – 5 0 t · o 2 5 –0·05)(–0·5 – 9 0 · t 2 o 7 0·05)(–0·4 – 1 0 t · o 2 3 –0·05)(–0·39 –0 t · o 2 2 –0·04)(–0·5 – 5 0 t ·1 o 9 0·16)(–0·3 – 5 0 t ·1 o 7 0·01)(–0·3 – 6 0 t ·1 o 6 0·04)(–0·3 – 5 0 t ·1 o 6 0·04)(–0·34 –0 t · o 1 6 0·03)(–0·33 –0 t · o 1 4 0·06)(–0·30 –0 t · o 1 2 0·07) (–0·3 – 0 0 · t 1 o 0 0·10)(–0·28 –0 t · o 0 9 0·11) (–0·2 – 6 0 t ·0 o 6 0·13)Lumateperone(–0·36 – 0 to ·1 0 8 ·00) (–1·03 –0 t · o 9 – 0 0·77)(–0·81 – t 0 o ·6 – 8 0·55)(–1·11 – 0 to ·7 – 3 0·35)(–0·6 – 2 0 t ·5 o 7 –0·52)(–0·8 – 0 0 · t 5 o 9 –0·38)(–0·7 – 6 0 t ·5 o 7 –0·37)(–0·57 – 0 to ·5 – 3 0·48)(–0·65 – 0 to ·5 – 1 0·36)(–0·5 – 6 0 t ·4 o 9 –0·42)(–0·59 –0 t · o 4 7 –0·36)(–0·7 – 7 0 · t 4 o 9 –0·22)(–0·52 –0 t · o 4 5 –0·39)(–0·50 –0 t · o 4 4 –0·38)(–0·7 – 3 0 t ·4 o 2 –0·10)(–0·4 – 7 0 t ·4 o 0 –0·33)(–0·5 – 0 0 t ·3 o 9 –0·27)(–0·48 –0 t · o 3 8 –0·28)(–0·4 – 7 0 t · o 3 8 –0·29)(–0·4 – 7 0 t · o 3 6 –0·26)(–0·4 – 2 0 t ·3 o 4 –0·26)(–0·4 – 4 0 t · o 3 2 –0·21)(–0·4 – 1 0 t · o 3 1 –0·21)(–0·39 –0 t · o 2 9 –0·18)(–0·3 – 9 0 t ·2 o 3 –0·06) Placebo Figure 4: League table of overall symptoms Order of treatments is according to SUCRA. Results of the network-meta-analysis are presented in the left lower half and results of pairwise meta-analyses in the right upper half. Each cell provides the effect estimate and the corresponding 95% CI of a comparison (lower left half: treatment in column versus treatment in row; right upper half: treatment in row versus treatment in column). The type of effect size measure is SMD. Light yellow indicates that the point estimate is outside the range of very small differences but the 95% CI includes the value of no difference (SMD=0). Yellow indicates that the point estimate is outside the range of very small effects and the 95% CI does not include the value of no difference but very small differences. Orange indicates that the point estimate and the 95% CI exclude very small effects. Blue indicates that the point estimate and the 95% confidence interval are within the range of very small effects. NA=not applicable. SMD=standardised mean difference. SUCRA=surface under the cumulative ranking. Articles clozapine, along with amisulpride and risperidone, reduced symptoms more than at least three other A Cholinergic events antipsychotics, where clozapine was more ecacious C T OR S|I Events than risperidone. Brexpiprazole, iloperidone, cariprazine, (95% CI) (of 22) (95% CI) and lurasidone reduced symptoms less than at least Perphenazine (n=543) 8 6 0·62 (0·37 to 1·04) 6| 4% (2% to 6%) three other antipsychotics (figure 3, appendix p 168). Quetiapine (n=3315) 9 17 0·75 (0·59 to 0·95) 6| 4% (4% to 6%) 100% of antipsychotics reduced negative symptoms Olanzapine (n=3973) 11 33 0·78 (0·62 to 0·97) 6| 5% (4% to 6%) Perospirone (n=70) 1 1 0·76 (0·24 to 2·40) 2| 5% (1% to 13%) (100% excluding no dierence; 21 [83%] of 23 excluding Placebo (n=10550) 19 110 Reference 5·848% very small eects) with SMDs ranging from –0·65 Paliperidone (n=2356) 2 14 1·02 (0·77 to 1·36) 2| 6% (5% to 8%) (95% CI –0·95 to –0·34) to –0·16 (–0·27 to –0·05; k=203; Sertindole (n=649) 3 4 1·02 (0·61 to 1·70) 2| 6% (4% to 10%) n=49 952; t=23). Clozapine reduced symptoms more, and Ziprasidone (n=1513) 8 14 1·08 (0·82 to 1·43) 2| 6% (5% to 8%) chlorpromazine less, than at least three other Risperidone (n=5110) 15 47 1·14 (0·96 to 1·35) 2| 7% (6% to 8%) antipsychotics (figure 3, appendix p 181). Blonanserin (n=1206) 3 8 1·18 (0·86 to 1·63) 2| 7% (5% to 9%) Asenapine (n=1366) 4 6 1·19 (0·75 to 1·89) 2| 7% (4% to 11%) 100% of antipsychotics reduced depressive symptoms Sulpiride (n=129) 3 4 1·27 (0·40 to 4·02) | 7% (2% to 20%) (19 [79%] of 19 excluding no dierence; 11 [58%] of 19 Zotepine (n=197) 4 4 1·29 (0·57 to 2·91) | 7% (3% to 15%) excluding very small eects) with SMDs ranging from Brexpiprazole (n=1160) 2 5 1·32 (0·84 to 2·05) 2| 8% (5% to 11%) –0·91 (95% CI –1·37 to –0·44) to –0·11 (–0·29 to 0·06; Cariprazine (n=1143) 3 5 1·34 (0·93 to 1·93) 2| 8% (5% to 11%) k=111; n=28 536; t=19). Sulpiride reduced symptoms more, Aripiprazole (n=2697) 10 20 1·33 (1·09 to 1·64) 2|3 8% (6% to 9%) and haloperidol less, than at least three other Amisulpride (n=430) 3 5 1·43 (0·77 to 2·66) | 8% (5% to 14%) antipsychotics, but the evidence for sulpiride is based on Lumateperone (n=408) 2 3 1·46 (0·91 to 2·32) 2| 8% (5% to 13%) Haloperidol (n=3461) 16 38 1·53 (1·25 to 1·86) 2|3 9% (7% to 10%) only two studies versus chlorpromazine with Chlorpromazine (n=977) 8 18 1·59 (1·09 to 2·33) 2|3 9% (6% to 13%) 52 participants on sulpiride, and thus without indirect Lurasidone (n=2360) 5 11 1·72 (1·33 to 2·24) |3 10% (8% to 12%) evidence (appendix p 193). Clozapine (n=422) 4 13 3·46 (2·23 to 5·35) |16 18% (12% to 25%) Patterns for dichotomous treatment response and Xanomeline–trospium (n=341) 1 3 4·11 (2·27 to 7·43) |16 20% (12% to 32%) discontinuation due to inecacy were similar to the 0·1 0·2 0·5 1 2 5 10 primary outcome (see appendix pp 201–236). For cognition, SMDs versus placebo ranged from –1·04 Favours comparator Favours reference (95% CI –2·06 to –0·03) to 0·29 (–0·25 to 0·83) but B Anticholinergic events 95% CIs included no dierence for all drugs, except C T OR S|I Events sertindole (k=25; n=3397; t=12). Sertindole improved (95% CI) (of 24) (95% CI) cognition more than at least three other antipsychotics Brexpiprazole (n=653) 3 4 0·61 (0·31 to 1·22) 5| 3% (2% to 6%) but this was based on one study with 16 participants Lurasidone (n=2235) 5 11 0·92 (0·63 to 1·34) 5| 5% (3% to 7%) taking sertindole only (appendix p 237). Placebo (n=9620) 21 97 Reference 5·177% 12 (55%) of 22 antipsychotics had higher odds of Asenapine (n=942) 3 4 1·00 (0·54 to 1·84) 4| 5% (3% to 9%) cholinergic adverse events than placebo (six [27%] of Blonanserin (n=1038) 3 6 1·14 (0·71 to 1·82) 3| 6% (4% to 9%) Aripiprazole (n=2624) 9 20 1·26 (0·91 to 1·72) 3| 6% (5% to 9%) 22 excluding no dierence; three [14%] of 22 excluding Risperidone (n=6197) 15 53 1·28 (1·03 to 1·59) 3| 7% (5% to 8%) very small eects) with ORs above very small eects Amisulpride (n=453) 3 6 1·24 (0·58 to 2·67) | 6% (3% to 13%) ranging from 4·11 (95% CI 2·27 to 7·43) to 1·27 Ziprasidone (n=1404) 7 14 1·37 (0·93 to 2·03) 2| 7% (5% to 10%) (0·40 to 4·02; event rates 20–7% versus 6%; k=178; Cariprazine (n=1143) 3 5 1·40 (0·84 to 2·32) 1| 7% (4% to 11%) n=44 376; t=22). Perphenazine, olanzapine, and Paliperidone (n=2781) 3 15 1·38 (0·98 to 1·95) 2| 7% (5% to 10%) quetiapine had lower odds than at least three other Sertindole (n=631) 4 4 1·40 (0·80 to 2·47) 1| 7% (4% to 12%) Lumateperone (n=408) 2 3 1·41 (0·68 to 2·96) | 7% (4% to 14%) antispychotics, whereas lurasidone, chlorpromazine, Perospirone (n=70) 1 1 1·39 (0·34 to 5·65) | 7% (2% to 24%) haloperidol, aripiprazole, and in particular xanomeline– Haloperidol (n=2867) 15 32 1·53 (1·14 to 2·05) 2| 8% (6% to 10%) trospium and clozapine, had higher odds than at least Sulpiride (n=129) 3 4 1·65 (0·66 to 4·10) | 8% (3% to 18%) three other antipsychotics (figure 5, appendix p 242). Olanzapine (n=3952) 12 32 1·60 (1·22 to 2·09) 1| 8% (6% to 10%) 18 (75%) of 24 antipsychotics had higher odds of Perphenazine (n=135) 4 4 1·96 (0·83 to 4·62) | 10% (4% to 20%) anticholinergic adverse events than placebo (eight [33%] Iloperidone (n=303) 2 1 2·39 (0·83 to 6·89) | 12% (4% to 27%) of 24 excluding no dierence; five [21%] of 24 excluding Olanzapine–samidorphan (n=134) 2 1 3·01 (0·90 to 10·10) | 14% (5% to 36%) Quetiapine (n=3211) 9 19 2·67 (1·94 to 3·68) |6 13% (10% to 17%) very small eects), with ORs ranging from 3·55 (95% CI Xanomeline–trospium (n=341) 1 3 3·35 (1·65 to 6·80) |3 15% (8% to 27%) 1·31 to 9·66) to 1·28 (1·03 to 1·59) (event rates Zotepine (n=151) 3 3 3·55 (1·31 to 9·66) |2 16% (7% to 35%) 16–7% versus 5%; k=170; n=42 841; t=24). Brexpiprazole, Clozapine (n=344) 3 9 3·41 (1·94 to 5·99) |9 16% (10% to 25%) lurasidone, asenapine, blonanserin, aripiprazole, and Chlorpromazine (n=1075) 8 18 3·51 (2·31 to 5·33) |12 16% (11% to 23%) risperidone had lower odds than at least three other 0·1 0·2 0·5 1 2 5 10 antispychotics, whereas chlorpromazine, clozapine, xanomeline–trospium, and quetiapine had higher odds Favours comparator Favours reference than at least three other antipsychotics (figure 5, appendix (Figure 5 continues on next page) p 260). 884 Articles 21 (88%) of 24 antipsychotics had higher odds of C Sedation sedation than placebo (17 [71%] of 24 excluding no C T OR S|I Events dierence; 15 [62%] of 24 excluding very small eects), (95% CI) (of 24) (95% CI) with ORs ranging from 6·37 (95% CI 4·18 to 9·69) Placebo (n=10255) 21 108 Reference 4·087% to 1·38 (0·74 to 2·58; event rates 21–6% versus 4%; Brexpiprazole (n=1082) 3 5 1·05 (0·59 to 1·89) 10| 4% (2% to 7%) k=180; n=44 846; t=24). Brexpiprazole, cariprazine, Cariprazine (n=680) 2 3 1·14 (0·66 to 1·98) 10| 5% (3% to 8%) iloperidone, sertindole, paliperidone, blonanserin, Iloperidone (n=303) 2 1 1·38 (0·74 to 2·58) 5| 6% (3% to 10%) aripiprazole, and lurasidone had lower odds than at least Xanomeline–trospium (n=90) 1 1 1·02 (0·13 to 7·78) | 4% (1% to 25%) three other antipsychotics, whereas particularly Sertindole (n=989) 4 5 1·48 (0·95 to 2·31) 7| 6% (4% to 9%) clozapine, along with zotepine, sulpiride, chlorpro- Paliperidone (n=2730) 3 15 1·59 (1·17 to 2·17) 7| 6% (5% to 8%) Blonanserin (n=1103) 3 7 1·72 (1·09 to 2·73) 4| 7% (4% to 10%) mazine, quetiapine, ziprasidone, and olanzapine had Aripiprazole (n=2496) 9 18 1·87 (1·44 to 2·44) 5| 7% (6% to 9%) higher odds of sedation than at least three other Perospirone (n=70) 1 1 1·81 (0·45 to 7·31) | 7% (2% to 24%) antipsychotics (figure 5, appendix p 279). Amisulpride (n=315) 3 4 1·90 (0·94 to 3·87) 1| 7% (4% to 14%) 17 (74%) of 23 antipsychotics were associated with Lurasidone (n=2413) 5 11 2·14 (1·52 to 3·00) 3| 8% (6% to 11%) higher odds of using antiparkinsonian drugs for Lumateperone (n=408) 2 3 2·33 (1·51 to 3·60) 1| 9% (6% to 13%) extrapyramidal symptoms than placebo (12 [52%] of 23 Risperidone (n=5351) 15 45 2·41 (1·99 to 2·91) 2| 9% (8% to 11%) Haloperidol (n=3428) 15 40 2·60 (2·10 to 3·20) 1|2 10% (8% to 12%) excluding no dierence; four [17%] of 23 excluding very Olanzapine–samidorphan (n=345) 2 2 2·92 (1·50 to 5·68) | 11% (6% to 19%) small eects), with ORs ranging from 4·38 (95% CI Olanzapine (n=4461) 12 37 3·15 (2·57 to 3·86) 1|4 12% (10% to 14%) 3·54 to 5·43) to 1·28 (0·71 to 2·28; event rates 31–12% Asenapine (n=1366) 4 6 3·21 (1·99 to 5·17) |2 12% (8% to 18%) versus 10%; k=132; n=34 221; t=23). Clozapine, Ziprasidone (n=1597) 7 15 3·25 (2·47 to 4·28) |5 12% (10% to 15%) quetiapine, sertindole, and olanzapine were associated Quetiapine (n=3217) 9 19 3·35 (2·65 to 4·24) |5 12% (10% to 15%) with lower odds than at least three other antipsy- Perphenazine (n=218) 3 3 4·82 (1·91 to 12·16) |2 17% (8% to 34%) Chlorpromazine (n=1215) 7 21 4·61 (3·26 to 6·52) |9 16% (12% to 22%) chotics, whereas particularly haloperidol, along with Sulpiride (n=97) 2 3 5·84 (2·44 to 13·98) |7 20% (9% to 37%) chlorpromazine, zotepine, lurasidone, and risperidone Zotepine (n=165) 3 4 5·84 (2·87 to 11·86) |8 20% (11% to 34%) were associated with higher odds than at least three other Clozapine (n=452) 4 14 6·37 (4·18 to 9·69) |13 21% (15% to 29%) antipsychotics (figure 5, appendix p 296). For 0·1 0·2 0·5 1 2 5 10 xanomeline–trospium see the appendix (p 567). 17 (71%) of 24 antipsychotics had higher odds of Favours comparator Favours reference akathisia than placebo (13 [54%] of 24 excluding no D Use of antiparkinsonian drugs dierence; 12 [50%] of 24 excluding very small eects), C T OR S|I Events with ORs ranging from 4·32 (95% CI 3·52 to 5·30) (95% CI) (of 23) (95% CI) to 1·32 (0·92 to 1·88; event rates of 15–5% versus 4%; k=152; n=41 649; t=24). Iloperidone, sertindole, Clozapine (n=175) 3 7 0·68 (0·31 to 1·50) 6| 7% (3% to 14%) Lumateperone (n=234) 2 2 0·69 (0·21 to 2·29) 1| 7% (2% to 19%) olanzapine–samidorphan, olanzapine, quetiapine, Quetiapine (n=2742) 7 15 0·94 (0·67 to 1·31) 4| 9% (7% to 12%) brexpiprazole, and paliperidone had lower odds than at Xanomeline–trospium (n=126) 1 1 0·33 (0·01 to 8·90) | 3% (0% to 48%) least three other antipsychotics, whereas particularly Placebo (n=7214) 18 76 Reference 3| 9·489% haloperidol along with perphenazine, chlorpromazine, Sertindole (n=1121) 3 5 0·99 (0·65 to 1·53) 3| 9% (6% to 14%) blonanserin, sulpiride, cariprazine, perospirone, Olanzapine (n=3176) 10 23 1·06 (0·81 to 1·39) 4| 10% (8% to 13%) lurasidone, risperidone, aripiprazole, and ziprasidone Asenapine (n=484) 3 3 1·28 (0·71 to 2·28) 1| 12% (7% to 19%) Amisulpride (n=391) 2 6 1·45 (0·86 to 2·45) 1| 13% (8% to 20%) had higher odds than at least three other antipsychotics Iloperidone (n=427) 3 2 1·48 (0·67 to 3·28) 1| 13% (7% to 26%) (appendix p 311). Paliperidone (n=2962) 3 15 1·56 (1·20 to 2·04) 1| 14% (11% to 18%) Nine (41%) of 22 antipsychotics led to higher prolactin Brexpiprazole (n=598) 1 2 1·58 (0·73 to 3·40) | 14% (7% to 26%) than placebo (nine [41%] of 22 excluding no dierence; Aripiprazole (n=1566) 6 12 1·64 (1·16 to 2·31) 1| 15% (11% to 19%) six [27%] of 22 excluding very small eects), with MDs Ziprasidone (n=948) 5 9 1·65 (1·14 to 2·38) 1| 15% (11% to 20%) ranging from 58·23 ng/ml (95% CI 39·00 to 77·45) to Sulpiride (n=97) 2 3 1·99 (0·81 to 4·86) | 17% (8% to 34%) 7·42 ng/ml (4·36 to 10·48; k=114; n=28 642; t=22). Blonanserin (n=389) 3 3 1·92 (1·05 to 3·52) 1| 17% (10% to 27%) Risperidone (n=4369) 13 36 1·89 (1·52 to 2·35) 1|3 17% (14% to 20%) Particularly amisulpride, paliperidone, and risperidone, Lurasidone (n=2413) 5 11 2·20 (1·60 to 3·03) 1|4 19% (14% to 24%) along with zotepine, perphenazine, and haloperidol, Zotepine (n=176) 3 3 2·40 (1·22 to 4·72) |1 20% (11% to 33%) increased prolactin more than at least three other Cariprazine (n=279) 1 2 2·60 (1·19 to 5·65) |1 21% (11% to 37%) antipsychotics, whereas aripiprazole, cariprazine, Chlorpromazine (n=358) 5 8 2·63 (1·41 to 4·88) |4 22% (13% to 34%) quetiapine, lumateperone, clozapine, and some other Perospirone (n=70) 1 1 2·91 (1·07 to 7·94) | 23% (10% to 45%) antipsychotics increased prolactin less than at least Perphenazine (n=146) 1 1 2·91 (1·09 to 7·73) | 23% (10% to 45%) Haloperidol (n=3760) 17 42 4·38 (3·54 to 5·43) |14 31% (27% to 36%) three other antipsychotics. Aripiprazole resulted in lower prolactin levels than placebo with a 95% CI 0·1 0·2 0·5 1 2 5 10 excluding no dierence (figure 5, appendix p 326). Favours comparator Favours reference Five (26%) of 19 antipsychotics increased QTc (five [26%] of 19 excluding no dierence; three [16%] of (Figure 5 continues on next page) Articles 19 excluding very small eects) with MDs ranging sertindole, along with amisulpride and ziprasidone from 22·15 (95% CI 18·06 to 26·24) to 6·29 ms increased QTc more than at least three other (3·96 to 8·62; k=66; n=21 022; t=19). Particularly antipsychotics, whereas brexpiprazole, aripiprazole, and lurasidone increased QTc less than at least three other antipsychotics (figure 5, appendix p 356). 19 (83%) of 23 antipsychotics increased weight E Change in prolactin (ng/mL) (16 [70%] of 23 excluding no dierence; 11 [48%] of C T MD S|I (95% CI) (of 22) 23 excluding very small eects), with MDs ranging from 3·21 kg (95% CI 2·21 to 4·22) to 0·51 kg (–1·36 to 2·39; Aripiprazole (n=1834) 11 18 –6·52 (–10·53 to –2·50) 10| k=171; n=46 720; t=23). Xanomeline–trospium, Cariprazine (n=714) 3 3 –2·73 (–9·95 to 4·49) 6| ziprasidone, lurasidone, lumateperone, aripiprazole, Quetiapine (n=2608) 10 18 –1·85 (–5·68 to 1·97) 6| Placebo (n=6374) 14 66 Reference haloperidol, blonanserin, cariprazine, amisulpride, Lumateperone (n=344) 2 3 –0·18 (–6·95 to 6·58) 6| brexpiprazole, and asenapine resulted in less weight gain Clozapine (n=37) 2 2 –1·41 (–21·13 to 18·30) 5| than at least three other antipsychotics, whereas Xanomeline–trospium (n=126) 1 1 0·20 (–12·19 to 12·59) 5| particularly zotepine, olanzapine, and sertindole, along Brexpiprazole (n=1070) 2 5 1·29 (–4·43 to 7·01) 5| with olanzapine–samidorphan, clozapine, iloperidone, Perospirone (n=51) 1 1 0·58 (–17·91 to 19·08) 3| chlorpromazine, quetiapine, risperidone, and Ziprasidone (n=1303) 9 10 1·18 (–4·01 to 6·37) 5| Blonanserin (n=357) 3 4 1·97 (–7·14 to 11·08) 5| paliperidone showed more weight gain than at least Iloperidone (n=451) 4 3 4·18 (–2·90 to 11·26) 5| three other antipsychotics (figure 5, appendix p 385). Lurasidone (n=2130) 5 11 4·42 (0·41 to 8·44) 5|1 The analyses using SMD and on participants with Asenapine (n=910) 4 5 4·87 (–1·17 to 10·91) 5| hyperprolactinaemia, QTc prolongation, and weight gain Olanzapine (n=3105) 12 27 7·42 (4·36 to 10·48) 5|1 yielded results similar to the MD results for prolactin, Chlorpromazine (n=244) 2 2 13·40 (1·48 to 25·32) 3|1 QTc, and weight (appendix pp 366–419). Sertindole (n=129) 1 1 15·58 (1·47 to 29·69) 3|1 Six (43%) of 14 antipsychotics had higher odds of sexual Haloperidol (n=1960) 12 24 18·40 (14·89 to 21·91) 3|12 Perphenazine (n=243) 6 2 26·63 (15·69 to 37·57) 2|13 adverse events than placebo (one [7%] of 14 excluding no Zotepine (n=28) 1 1 34·03 (6·53 to 61·52) |5 dierence; one [7%] of 14 excluding very small eects), Risperidone (n=3299) 14 36 37·09 (33·99 to 40·18) |17 with ORs ranging from 8·35 (95% CI 4·07 to 17·14) Paliperidone (n=1227) 4 10 44·52 (38·94 to 50·11) |18 to 1·26 (0·62 to 2·57; event rates 6–1% versus 1%; k=48; Amisulpride (n=98) 3 3 58·23 (39·00 to 77·45) |18 n=18 281; t=14). Sertindole had higher odds than most –50 –30–10010 30 50 other antipsychotics; however, low event rates per study make estimates uncertain (appendix p 420). Favours comparator Favours reference 14 (82%) of 17 antipsychotics had higher odds of heart F Change in QTc (ms) rate abnormalities than placebo (seven [41%] of C T MD S|I 17 excluding no dierence; six [35%] of 17 excluding very (95% CI) (of 19) small eects), with ORs ranging from 13·56 (95% CI Cariprazine (n=293) 2 1 –2·35 (–7·15 to 2·46) 2| 4·23 to 43·44) to 1·34 (0·78 to 2·29; event rates 22–3% Brexpiprazole (n=695) 1 2 –1·48 (–4·57 to 1·62) 3| versus 2%; k=81; n=23 198; t=17). Iloperidone, clozapine, Lurasidone (n=1585) 4 8 –1·18 (–3·08 to 0·71) 3| and chlorpromazine had higher odds than at least Aripiprazole (n=1127) 8 7 –1·21 (–3·96 to 1·54) 3| three other antipsychotics. The increased odds were Placebo (n=3556) 14 38 Reference mainly due to tachycardia and arrythmias and less due to Paliperidone (n=222) 2 1 0·93 (–2·94 to 4·81) 1| Haloperidol (n=1904) 10 19 0·98 (–0·87 to 2·82) 2| bradycardia (appendix pp 434–472). Xanomeline–trospium (n=89) 1 1 1·10 (–5·42 to 7·62) 1| Due to the low number of events per study, estimates Perospirone (n=51) 1 1 1·49 (–4·51 to 7·48) 1| were very uncertain for reductions in white blood cell Perphenazine (n=238) 6 1 1·65 (–3·06 to 6·37) 1| counts, seizures, deep vein thrombosis, and death Blonanserin (n=154) 1 1 2·41 (–5·06 to 9·89) 1| (appendix pp 473–533). Risperidone (n=2482) 12 19 3·61 (1·87 to 5·36) 1| All antipsychotics except haloperidol were associated Olanzapine (n=2631) 11 18 3·70 (1·69 to 5·70) 1| Asenapine (n=57) 2 1 4·95 (–1·41 to 11·30) 1| with higher quality of life (four [36%] of 11 excluding no Quetiapine (n=1670) 8 8 4·58 (2·42 to 6·75) 1| dierence; two [18%] of 11 excluding very small eects) Iloperidone (n=869) 4 4 6·29 (3·96 to 8·62) 1| with SMDs from –0·42 (95% CI –1·02 to 0·18) to –0·16 Ziprasidone (n=2374) 9 17 7·73 (5·76 to 9·70) 1|3 (–0·63 to 0·31; k=17; n=5336; t=11). There were no clear Chlorpromazine (n=142) 1 1 11·83 (1·27 to 22·40) | dierences between antipsychotics (appendix p 534). Amisulpride (n=178) 1 2 11·91 (6·53 to 17·30) |5 All antipsychotics improved social functioning Sertindole (n=705) 4 4 22·15 (18·06 to 26·24) |16 (13 [93%] of 14 excluding no dierence; 12 [86%] of –20 –10 0 10 20 14 excluding very small eects), with SMDs ranging from –1·05 (95% CI –1·63 to –0·48) to –0·15 Favours comparator Favours reference (–0·45 to 0·15; k=37; n=10 675; t=14). Clozapine improved (Figure 5 continues on next page) functioning more than at least three other antipsychotics 886 Articles (but based on 28 participants in two studies only; G Change in weight (kg) appendix p 541). C T MD S|I 20 (83%) of 24 antipsychotics had lower odds of discon- (95% CI) (of 23) tinuation for any reason than placebo (16 [67%] of 24 Xanomeline–trospium (n=340) 1 3 –0·37 (–1·27 to 0·53) 11| excluding no dierence; nine [38%] of 24 excluding very Ziprasidone (n=2299) 9 16 –0·12 (–0·53 to 0·29) 11| small eects), with ORs ranging from 0·44 (95% CI Placebo (n=9477) 20 94 Reference 0·29 to 0·69) to 0·79 (0·67 to 0·94; event rates of 38–26% Lurasidone (n=2160) 4 11 0·40 (–0·01 to 0·81) 10| versus 44%; k=241; n=59 381; t=24). Paliperidone, Lumateperone (n=372) 2 3 0·45 (–0·33 to 1·23) 6| olanzapine, amisulpride, and risperidone had lower odds Aripiprazole (n=3429) 11 25 0·55 (0·24 to 0·87) 9| Haloperidol (n=3330) 15 31 0·56 (0·24 to 0·89) 9| whereas xanomeline–trospium, sertindole, and Blonanserin (n=1028) 5 8 0·58 (–0·09 to 1·26) 6| haloperidol had higher odds than at least three other Cariprazine (n=877) 3 4 0·61 (–0·02 to 1·24) 6| antipsychotics (figure 3, appendix p 548). Perospirone (n=75) 2 2 0·51 (–1·36 to 2·39) 2| We did not find clear evidence against the transitivity Amisulpride (n=539) 4 6 0·97 (0·26 to 1·69) 3| assumption (appendix p 568). Heterogeneity was low Brexpiprazole (n=1216) 2 6 1·02 (0·44 to 1·60) 3| or low-moderate compared with empirical distributions Perphenazine (n=431) 7 3 1·16 (0·36 to 1·95) 2| Asenapine (n=1343) 4 7 1·24 (0·73 to 1·74) 3|2 in the primary outcome (common-τ=0·097) and all Paliperidone (n=2658) 5 19 1·45 (1·11 to 1·79) 2|3 secondary outcomes, except prolactin, use of Risperidone (n=5027) 17 46 1·55 (1·27 to 1·82) 2|5 antiparkinsonian medication, and anticholinergic Quetiapine (n=3123) 9 18 1·94 (1·53 to 2·34) 1|8 adverse event (all moderate-high). Uncertainty due to Chlorpromazine (n=384) 6 6 1·98 (1·21 to 2·74) |5 heterogeneity, as represented by prediction intervals, Iloperidone (n=939) 4 4 2·12 (1·50 to 2·73) |8 challenged the interpretation only for very few Clozapine (n=152) 5 8 2·22 (1·14 to 3·29) |5 Olanzapine–samidorphan (n=336) 2 2 2·44 (1·48 to 3·39) |8 comparisons in most outcomes (1·33% for the primary Sertindole (n=1019) 4 5 2·59 (1·94 to 3·24) |11 outcome). Heterogeneity in network meta-analyses Olanzapine (n=5980) 15 54 2·91 (2·64 to 3·17) |16 excluding subgroup populations was similar to the Zotepine (n=186) 4 4 3·21 (2·21 to 4·22) |15 primary analyses (appendix p 571). Statistical indications of inconsistency were observed for the primary outcome –5 –4 –3–2–10 1 2 3 4 5 and several secondary outcomes. However, our detailed Favours comparator Favours reference exploration revealed no critical concerns regarding the Figure 5: Panel of tolerability outcomes validity of NMA results (appendix p 578). The order of treatments is according to the SUCRA-ranking (or the mean effect for rare binary outcomes because Sensitivity and network meta-regression analyses SUCRA-ranking technically not possible). The direction of the effect is indicated below the x-axis. The column C yielded no materially dierent results (appendix indicates, for the treatment in the line, how many other interventions it has been directly compared with in a pp 626–709) and there were no indications for clinically study; the column T indicates the number of studies with data for this intervention. The column S|I indicates, for the treatment in the line, how many other antipsychotics the intervention is superior (before the |-symbol) relevant dierences in ecacy between oral, transdermal, or inferior (after the |-symbol) to, defined as a point estimates above or below very small effects, respectively, and and long-acting-injectable applications (appendix p 670). 95% CIs excluding very small differences (an information only available in the league tables in the appendix For the primary outcome, 1% of studies were rated as otherwise); in parenthesis the total number of antipsychotics in this NMA. The colour of the lines of the CIs low risk of bias, 63% as some concerns of bias, and indicates the CINEMA-confidence in estimates: green=high, blue=moderate, orange=low, and red=very low. C=connectivity. CINEMA=Confidence in Network Meta-Analysis. I=inferior. MD=mean difference. n=number of 36% as high risk of bias (details, other outcomes, and participants. OR=odds ratio. S=superior. SUCRA=surface under the cumulative ranking. NMA=network meta- subgroup analysis by bias risk are provided in the analysis. T=trials. appendix p 710). There were some indications for potential small study publication bias in placebo- olanzapine–samidorphan, lumateperone, blonanserin, and controlled trials in funnel plots (appendix p 781). perospirone—absent in 2019—alongside other CINEMA-ratings were mostly moderate-low and low-very antipsychotics, and includes 16 additional outcomes low for antipsychotics versus placebo (figures 3 and 5) (eg, cognition, cholinergic and sexual adverse eects, and and antipsychotics versus antipsychotics (appendix seizures). The relative ecacy and safety of xanomeline– pp 152–567, 797–817), respectively. trospium is particularly important given its muscarinic mechanism, as it is the first antipsychotic not acting Discussion primarily on dopamine receptors.44 To our knowledge, this is the most up to date comprehensive Clozapine, along with amisulpride, olanzapine, and NMA on antipsychotic treatment for acute schizophrenia, risperidone, was more ecacious than at least three other which provides evidence for potentially guideline-changing antipsychotics with small to medium eect sizes. clinically relevant dierences between antipsychotics in Currently, no major guidelines state specific ecacy ecacy. Compared with our 2019 review,4 this study dierences between antipsychotics (except for treatment includes a 1·5-times larger dataset (26 802 participants resistance). Our findings support guideline updates in more; 170 newly included studies), focuses on studies this regard, reinforced by the largest antipsychotics RCT with appropriate randomisation indicated, and it to date to our knowledge (3067 participants),45 in which emphasises clinical relevance rather than statistical olanzapine and risperidone were superior to aripiprazole, significance alone. It also includes xanomeline–trospium, ziprasidone, and quetiapine. Articles Xanomeline–trospium ranked in the top third for online questionnaire. Of these, only 27 described ecacy in the current review, avoids dopamine-blocking appropriate methods, whereas the majority indicated adverse events, but causes cholinergic and anticholinergic suboptimal methods, use of observational instead of adverse events. Its comparison with other antipsychotics randomised data (as clinicians chose the antipsychotics), was indirect, based only on three placebo-controlled or unreliable data. Given related concerns for other trials, so uncertainty remains despite no clear bias or countries without long-standing research tradition transitivity issues, and head-to-head studies are needed. (eg, India and Iran),23,22,53 we approached those countries Patients’ acceptance might also be problematic, as it had alike and only included trials with actively confirmed the highest discontinuation risk across antipsychotics, appropriate methods. although only one participant discontinued for inecacy. It could be considered a limitation of this study that Clozapine had high odds of cholinergic adverse we included trials conducted under the auspices of events similar to xanomeline–trospium, supporting a pharmaceutical companies and industry-independent speculative idea that muscarinic modulation might trials from countries categorised as countries with long- underlie its ecacy.46 Although its high ecacy aligns standing research tradition according to Panagiotou and with observational data,47 uncertainty remains in RCT colleagues (see Methods)23 without explicit confirmation data because estimates stem mainly from small, relatively of methods. However, our experience from contacting old (often published before 1990) trials. Moreover, authors, in which none of them revealed inappropriate individual patient data meta-analysis in treatment- methods and many confirmed appropriate methods, as resistant patients did not confirm superiority over well as our sensitivity analyses, support this decision. olanzapine and risperidone.31 Although we included far Another potential limitation of this study is the broad more participants taking clozapine than in 20194 inclusion criteria, but these were based on a large meta- (eg, 450 compared with 31 for positive symptoms), data analysis18 (537 RCTs with 76 382 participants) that found on QTc and prolactin remain insucient. no major ecacy and tolerability dierences across Partial dopamine-agonists (aripiprazole, brexpiprazole, patient populations; thus, no key eect modifiers and cariprazine) were not among the top drugs regarding emerged, and transitivity was unlikely to have been ecacy but overall showed better tolerability than compromised. Reported higher response rates in first- dopamine-antagonists.48 Specifically, aripiprazole might episode patients52 are likely confounded, as they come be considered when tolerability is a primary concern, as solely from head-to-head trials, which systematically it also ranked in the middle of all antipsychotics in show higher response than placebo-controlled trials.54 ecacy (rank 12 of 24). Our sensitivity analyses excluding subgroups yielded Lumateperone ranked low in ecacy even after similar results and heterogeneity, indicating that average adjustment for placebo-response,49 similar to cariprazine relative eects are applicable across patient groups. Still, and brexpiprazole, which is relevant for cost-eectiveness research in subgroups remains needed because their calculations. Samidorphan reduced olanzapine-related trial networks are sparse. weight gain only slightly (0·47 kg less), this was possibly Another limitation is that for sexual adverse eects, because the included trials were short as longer studies, low and probably under-reported event rates impeded which were ineligible for our review,50,51 indicate greater the analysis (eg, for the strongly prolactin-inducing but not neutralising eects (1·9 kg less51). amisulpride) and we refer to an ongoing larger review.55 Blonanserin, which is only licensed in some Asian Only 25 included trials investigated cognition; however, countries, showed moderate ecacy and tolerability, another broader review25 with 68 trials similarly found no supporting its use outside Asia (for perospirone more clear dierences between antipsychotics. research is needed as data are from 87 participants only). Extensive sensitivity and meta-regression analyses We excluded several first generation antipsychotics confirmed our initial findings. However, residual bias from our review protocol because second generation persists due to factors inherent in schizophrenia drug antipsychotics are first-line treatments in middle-to-high- research, such as participants’ premature study discon- income countries and most first generation antipsychotic tinuation, compromised blinding by adverse eects or trials were old, small, methodologically outdated, and suboptimal over-encapsulation, and potential reporting produced inconclusive NMA results in 2019.4 Since no bias, which prevented high certainty CINEMA new trials on the excluded first generation antipsychotics judgements. Additionally, certainty for estimates exist (figure 1), findings were unlikely to dier from indicating small or very small dierences is reduced, as earlier reviews.4,9,10,12,13,52 only extremely large sample sizes would ensure precision. Of 5117 Chinese trials labelled as RCTs, only 24 were Conversely, the review’s conclusions, based on estimates included. Following our protocol we only selected RCTs whose CIs excluded very small dierences, were not for which authors confirmed appropriate randomisation, downgraded for imprecision and achieved better certainty. because of raised concerns.2,3,20 However, authors of only Furthermore, in principle, chance findings across the 347 RCTs responded to our requests sent in Chinese NMAs for multiple outcomes can occur. Therefore, the from Shanghai and facilitated by a specifically designed overall repeated patterns across outcomes or reviews are 888 Articles important for interpretation and not only the singular but placebo response or spontaneous remissions could statistically significant results (particularly for also be more frequent among them.54,65 interventions informed by only few participants or Regarding methods of evidence-based medicine, studies and loosely connected to the network as those systematic reviewers, guideline makers, and guideline are more prone to outlying results due to chance users should be aware that many RCTs listed in Chinese findings, imperceptible intransitivity, or bias, as control databases likely have methodological shortcomings, through other studies or network-pathways is missing ranging from suboptimal randomisation to the use (see figures 3, 5, and appendix pp 152–567). of observational data and publication of unreliable Overall, this NMA documents dierences in ecacy results. Unfortunately, this also aects recent trials, and tolerability between antipsychotics. Pragmatic drug despite government actions in China.66,67 In this review choice depends on individual patient characteristics. For we only addressed antipsychotics, but other studies have example, antipsychotics with fewer adverse eects and flagged similar problems with trials from other medical adapted doses might be advisable for the still maturating disciplines.20–22 adolescents and for older adults with reduced drug Contributors metabolism and age-related brain loss as, for example, SL and CL obtained the funding and supervised the study. SL, CL, YZ, older people have a higher risk of tardive dyskinesia JS-T, SS, IB, JMD, AN, and GS designed the systematic review and associated with extrapyramidal symptoms. Prolactin- meta-analysis. JS-T, MQ, JT, XL, CV, and MW (for non-Chinese studies) and YZ, YD, SG, JW, and MQ (for Chinese studies) screened the increasing drugs have been associated with breast cancer literature searches and contacted study authors. JS-T, AR, MQ, XL, JT, in women.56 Drug choice should also consider MW, CV, and SS (for non-Chinese studies) and YZ, YD, SG, JW, and MQ comorbidities (eg, avoid olanzapine in diabetes and QTc (for Chinese studies) extracted data from included studies and assessed risk of bias. FK optimised and maintained the database for extraction, prolonging amisulpride in cardiac disorders), target engaging in quality assurance. JS-T conducted the network meta- symptoms (eg, the D3-partial agonist cariprazine and analyses under the supervision of AN and GS and with support by low-dose amisulpride have the best evidence in persistent SS and AR. AN conducted the network meta-regression analyses. negative symptoms57), preferred administration route ES provided the patients’ perspective when designing the study and interpreting the results. JS-T, YZ, and SL accessed and verified the study (non-adherence is more common in substance abuse), data. JS-T, YZ, SL, GS, AN, SS, AR, IB, CL, JP, and JMD interpreted the and particularly patients’ values. The importance patients results. JS-T, GS, and SL drafted the manuscript. All authors had full give to individual adverse eects varies, and some access to all the data in the study and had final responsibility for the prioritise symptom control over adverse eects. Finally, decision to submit for publication. there is high inter-individual variability in drug response Declaration of interests and not all patients develop all adverse eects. Electronic SL has received honoraria for consulting from Angelini, Bristol-Myers- Squibb, Boehringer Ingelheim, Johnson and Johnson, Karuna, Kynexis, devices58,59 (such as those listed on Illuminatum) can Neurotorium, NovoNordisk, Otsuka, ROVI, and TEVA, and has received For Illuminatum see support this highly individualised shared decision- payment for lectures from Angelini, Apsen, Boehringer Ingelheim, www.illuminatum.com making process (appendix p 818). Gedeon Richter, Johnson and Johnson, Mitsubishi, Orionpharma, Future research should identify predictors of response Otsuka, Lundbeck, ROVI, Sun Pharma, and TEVA. All other authors declare no competing interests. and adverse eects for even more individualised Data sharing treatment; therefore, individual patient data are needed, Aggregate data (ie, summary results per study arm) and a data dictionary but currently these are only available for a few drugs at can be made available for methodological purposes and validation of portals such as Vivli. results on written request to the corresponding author including a For Vivli see www.vivli.org Head-to-head trials are needed to confirm the relative detailed study plan. ecacy of xanomeline–trospium (to date, none are Acknowledgments registered at ClinicalTrials.gov or WHO-ICRTP) and This study was funded by the German Research Foundation (Deutsche it would be important to explore whether certain Forschungsgemeinschaft [DFG], grant number 468853597), the German Center for Mental Health (funded by the German Ministry of Research, individuals respond more favourably to muscarinic Technology and Space, Bundesministerium für Bildung und Forschung, rather than antidopaminergic antipsychotics. grant number 01EE2303B) and the National Natural Science Foundation RCTs following current standards, comparing of China (NSFC, grant number 82161138021). AN was supported by the clozapine to amisulpride, olanzapine, and risperidone in DFG Project-ID 499552394–SFB 1597. We specifically thank all the 493 authors who answered to our requests regarding study methods and the general population (ie, not limited to treatment- missing outcome data. Listing all respondents by name would not be resistance) are needed to confirm clozapine’s superior feasible within the constraints of this manuscript and would raise data ecacy (some trials are underway60,61). If RCT results are protection concerns. Additionally, we thank Miao Zhang, Juanjuan Ren, Yue Hu, Mengyi Gao, Yuanmeng Li, Jiayun Qiu, Huiqi Li, and positive, clozapine could be recommended earlier, not Shuoyuan Zhang for their help in contacting Chinese trial authors; only in cases of resistance to two other antipsychotics, Nobuyuki Nomura for contacting Japanese authors; Filip Milosavljević despite its adverse eects. for contacting Serbian authors; Xiaochun Qiu of Library of Shanghai Moreover, placebo-controlled studies in antipsychotic- Jiao Tong University School of Medicine and Farhad Shokraneh of Systematic Review Consultants (https://systematicreview.info) for naive (never treated) individuals do not exist62 (except for designing and conducting the professional literature searches; and subgroup analyses in adolescents63,64) but would be Orestis Efthimiou for his advice in performing network meta-analysis of important to clarify the true absolute eects of anti- rare event data. Moreover, we explicitly thank the authors of the psychotic treatment. Drug-response might be higher,13,52 previous reviews on which this work was built (first authors Articles Maximilian Huhn, Marc Krause, Dongfang Wang, Myrto Samara, 18 Leucht S, Chaimani A, Krause M, et al. The response of subgroups Shimeng Dong) and Hui Wu, Angelika Burschinski, Lena Feber, of patients with schizophrenia to dierent antipsychotic drugs: Natalie Peter, Yuki Furukawa, and Yaohui Wei for their help in study a systematic review and meta-analysis. Lancet Psychiatry 2022; selection and data extraction and David Kim for advising on the 9: 884–93. language of the manuscript as a native speaker. Wulf-Peter Hansen, 19 Wang D, Schneider-Thoma J, Siafis S, et al. Long-acting injectable who collaborated as a person with lived-experience in this project and second-generation antipsychotics vs placebo and their oral provided the patients’ perspective when designing the study and formulations in acute schizophrenia: a systematic review and meta- analysis of randomized-controlled-trials. Schizophr Bull 2024; interpreting the results, passed away in April, 2025. We thank him for 50: 132–44. his contributions to this review and his lasting commitment to patient 20 Wu T, Li Y, Bian Z, Liu G, Moher D. Randomized trials published in welfare and research. some Chinese journals: how many are randomized? Trials 2009; References 10: 46. 1 Miyar J, Adams CE. Content and quality of 10,000 controlled 21 Van Noorden R, Thompson B. Audio long read: Medicine is trials in schizophrenia over 60 years. Schizophr Bull 2013; plagued by untrustworthy clinical trials. How many studies are 39: 226–29. faked or flawed? Nature 2023; 619: 454–58. 2 Leucht S, Li C, Davis JM, Bighelli I, Zhu Y, Furukawa TA. About the 22 Zhang D, Freemantle N, Cheng KK. Are randomized trials issue of including or excluding studies from China in systematic conducted in China or India biased? A comparative empirical reviews. Schizophr Res 2022; 240: 162–63. analysis. J Clin Epidemiol 2011; 64: 90–95. 3 Tong Z, Li F, Ogawa Y, Watanabe N, Furukawa TA. Quality of 23 Panagiotou OA, Contopoulos-Ioannidis DG, Ioannidis JPA. randomized controlled trials of new generation antidepressants and Comparative eect sizes in randomised trials from less developed antipsychotics identified in the China National Knowledge and more developed countries: meta-epidemiological assessment. Infrastructure (CNKI): a literature and telephone interview study. BMJ 2013; 346: f707. BMC Med Res Methodol 2018; 18: 96. 24 The Center for Scientific Integrity. The Retraction Watch Database. 4 Huhn M, Nikolakopoulou A, Schneider-Thoma J, et al. http://retractiondatabase.org/ (accessed July 27, 2024). Comparative ecacy and tolerability of 32 oral antipsychotics for 25 Feber L, Peter NL, Chiocchia V, et al. Antipsychotic drugs and the acute treatment of adults with multi-episode schizophrenia: cognitive function: a systematic review and pairwise network meta- a systematic review and network meta-analysis. Lancet 2019; analysis. JAMA Psychiatry 2024; 82: 1–10. 394: 939–51. 26 Salanti G. Indirect and mixed-treatment comparison, network, 5 Taipale H, Schneider-Thoma J, Pinzón-Espinosa J, et al. or multiple-treatments meta-analysis: many names, many benefits, Representation and outcomes of individuals with schizophrenia many concerns for the next generation evidence synthesis tool. seen in everyday practice who are ineligible for randomized clinical Res Synth Methods 2012; 3: 80–97. trials. JAMA Psychiatry 2022; 79: 210–18. 27 Higgins JPT. Cochrane Handbook for Systematic Reviews of 6 Hutton B, Salanti G, Caldwell DM, et al. The PRISMA extension Interventions. Newark: John Wiley & Sons Incorporated, 2019. statement for reporting of systematic reviews incorporating network 28 Rhodes KM, Turner RM, Higgins JPT. Predictive distributions were meta-analyses of health care interventions: checklist and developed for the extent of heterogeneity in meta-analyses of explanations. Ann Intern Med 2015; 162: 777–84. continuous outcome data. J Clin Epidemiol 2015; 68: 52–60. 7 Wang D, Schneider-Thoma J, Siafis S, et al. Ecacy, acceptability 29 Turner RM, Jackson D, Wei Y, Thompson SG, Higgins JPT. and side-eects of oral versus long-acting- injectables Predictive distributions for between-study heterogeneity and simple antipsychotics: systematic review and network meta-analysis. methods for their application in Bayesian meta-analysis. Stat Med Eur Neuropsychopharmacol 2024; 83: 11–18. 2015; 34: 984–98. 8 Zhu Y, Krause M, Huhn M, et al. Antipsychotic drugs for the acute 30 Salanti G, Higgins JP. Network meta-analysis. In: Egger M, treatment of patients with a first episode of schizophrenia: Higgins JP, Davey Smith G, eds. Systematic reviews in health a systematic review with pairwise and network meta-analyses. research. Wiley, 2022: 238–57. Lancet Psychiatry 2017; 4: 694–705. 31 Schneider-Thoma J, Hamza T, Chalkou K, et al. Ecacy of clozapine 9 Krause M, Zhu Y, Huhn M, et al. Ecacy, acceptability, and versus second-generation antipsychotics in people with treatment- tolerability of antipsychotics in children and adolescents with resistant schizophrenia: a systematic review and individual patient schizophrenia: a network meta-analysis. Eur Neuropsychopharmacol data meta-analysis. Lancet Psychiatry 2025; 12: 254–65. 2018; 28: 659–74. 32 Sterne JAC, Savović J, Page MJ, et al. RoB 2: a revised tool for 10 Krause M, Huhn M, Schneider-Thoma J, Rothe P, Smith RC, assessing risk of bias in randomised trials. BMJ 2019; 366: l4898. Leucht S. Antipsychotic drugs for elderly patients with 33 Chaimani A, Salanti G. Using network meta-analysis to evaluate the schizophrenia: a systematic review and meta-analysis. Eur Neuropsychopharmacol 2018; 28: 1360–70. existence of small-study eects in a network of interventions. Res Synth Methods 2012; 3: 161–76. 11 Krause M, Huhn M, Schneider-Thoma J, Bighelli I, Gutsmiedl K, Leucht S. Ecacy, acceptability and tolerability of antipsychotics in 34 Peters JL, Sutton AJ, Jones DR, Abrams KR, Rushton L. Contour- enhanced meta-analysis funnel plots help distinguish publication patients with schizophrenia and comorbid substance use. bias from other causes of asymmetry. J Clin Epidemiol 2008; A systematic review and meta-analysis. Eur Neuropsychopharmacol 61: 991–96. 2019; 29: 32–45. 35 Chiocchia V, Nikolakopoulou A, Higgins JPT, et al. ROB-MEN: 12 Samara MT, Dold M, Gianatsi M, et al. Ecacy, acceptability, and a tool to assess risk of bias due to missing evidence in network tolerability of antipsychotics in treatment-resistant schizophrenia: a network meta-analysis. JAMA Psychiatry 2016; 73: 199–210. meta-analysis. BMC Med 2021; 19: 304. 36 Nikolakopoulou A, Higgins JPT, Papakonstantinou T, et al. 13 Dong S, Schneider-Thoma J, Bighelli I, et al. A network meta- CINEMA: an approach for assessing confidence in the results of a analysis of ecacy, acceptability, and tolerability of antipsychotics in treatment-resistant schizophrenia. Eur Arch Psychiatry Clin Neurosci network meta-analysis. PLoS Med 2020; 17: e1003082. 2024; 274: 917–28. 37 Cohen J. Statistical power analysis for the behavioral sciences. 14 WHO. eEML – Model List of Essential Medicines. World Health Hillsdale, New Jersey: L. Erlbaum Associates, 1988. Organization. https://list.essentialmeds.org/?section=451&indicatio 38 Balduzzi S, Rücker G, Schwarzer G. How to perform a meta-analysis n=243&year=&age=&sex= (accessed Sept 4, 2025). with R: a practical tutorial. Evid Based Ment Health 2019; 22: 153–60. 15 Lewis S, Lieberman J. CATIE and CUtLASS: can we handle the 39 Balduzzi S, Rücker G, Nikolakopoulou A, et al. Netmeta: truth? Br J Psychiatry 2008; 192: 161–63. an R package for network meta-analysis using frequentist methods. J Stat Soft 2023; 106: 1–40. 16 Leucht S, Davis JM. Which first-generation antipsychotics should be “repurposed” for the treatment of schizophrenia. 40 Efthimiou O, Rücker G, Schwarzer G, Higgins JPT, Egger M, Eur Arch Psychiatry Clin Neurosci 2022; 272: 1–3. Salanti G. Network meta-analysis of rare events using the Mantel- Haenszel method. Stat Med 2019; 38: 2992–3012. 17 McAdam MK, Baldessarini RJ, Murphy AL, Gardner DM. Second International Consensus Study of Antipsychotic Dosing (ICSAD-2). J Psychopharmacol 2023; 37: 982–91. 890 Articles 41 Salanti G, Ades AE, Ioannidis JPA. Graphical methods and 55 Schneider-Thoma J, Dong S, Kim DD, et al. Sexual dysfunction and numerical summaries for presenting results from multiple- other prolactin-related side eects of antipsychotic drugs in treatment meta-analysis: an overview and tutorial. J Clin Epidemiol schizophrenia: protocol for a systematic review with single-arm, 2011; 64: 163–71. pairwise, and network meta-analyses of randomized controlled 42 Chaimani A, Caldwell DM, Li T, Higgins JPT, Salanti G. Additional trials and non-randomized studies. F1000 Res 2025; 13: 973. considerations are required when preparing a protocol for a 56 Taipale H, Solmi M, Lähteenvuo M, Tanskanen A, Correll CU, systematic review with multiple interventions. J Clin Epidemiol 2017; Tiihonen J. Antipsychotic use and risk of breast cancer in women 83: 65–74. with schizophrenia: a nationwide nested case-control study in 43 Phillippo DM, Dias S, Ades AE, et al. Multilevel network meta- Finland. Lancet Psychiatry 2021; 8: 883–91. regression for population-adjusted treatment comparisons. 57 Krause M, Zhu Y, Huhn M, et al. Antipsychotic drugs for patients J R Stat Soc Ser A Stat Soc 2020; 183: 1189–210. with schizophrenia and predominant or prominent negative 44 Dolgin E. Revolutionary drug for schizophrenia wins US approval. symptoms: a systematic review and meta-analysis. Nature 2024; 634: 276–77. Eur Arch Psychiatry Clin Neurosci 2018; 268: 625–39. 45 Zhao G, Sun Y, Zhang Y, et al. Ecacy and tolerability of 58 Leucht S, Siafis S, Rodolico A, et al. Shared decision making seven antipsychotic drugs in acutely ill patients with schizophrenia: assistant (SDMA) and other digital tools for choosing antipsychotics a randomised, multicenter, assessor-blinded trial. Am J Psychiatry in schizophrenia treatment. Eur Arch Psychiatry Clin Neurosci 2023; 2026; 183: 112–23. 273: 1629–31. 46 Murray RM, Egerton A, Gao Y, et al. Why is clozapine uniquely 59 Pillinger T, Howes OD, Correll CU, et al. Antidepressant and eective in treatment-resistant schizophrenia? A review and antipsychotic side-eects and personalised prescribing: a systematic proposal. Biol Psychiatry 2025; published online June 23. https:// review and digital tool development. Lancet Psychiatry 2023; doi.org/10.1016/j.biopsych.2025.06.011. 10: 860–76. 47 Tiihonen J, Mittendorfer-Rutz E, Majak M, et al. Real-world 60 Li X, Lu C, Zhai Z, et al. Clozapine after one failed antipsychotic eectiveness of antipsychotic treatments in a nationwide cohort of drug trial in first-episode psychosis: a randomised, single-blind, 29 823 patients with schizophrenia. JAMA Psychiatry 2017; multi-center study. JAMA Psychiatry (in press). 74: 686–93. 61 Wagner E, Strube W, Görlitz T, et al. Eects of early clozapine 48 Yu H, Yan H, Wang L, et al, and the Chinese Antipsychotics treatment on remission rates in acute schizophrenia (the EARLY Pharmacogenomics Consortium. Five novel loci associated with trial): protocol of a randomized-controlled multicentric trial. antipsychotic treatment response in patients with schizophrenia: Pharmacopsychiatry 2023; 56: 169–81. a genome-wide association study. Lancet Psychiatry 2018; 5: 327–38. 62 Danborg PB, Gøtzsche PC. Benefits and harms of antipsychotic 49 Leucht S, Leucht C, Huhn M, et al. Sixty years of placebo-controlled drugs in drug-naïve patients with psychosis: a systematic review. antipsychotic drug trials in acute schizophrenia: systematic review, Int J Risk Saf Med 2019; 30: 193–201. bayesian meta-analysis, and meta-regression of ecacy predictors. 63 Correll CU, Tocco M, Hsu J, Goldman R, Pikalov A. Short-term Am J Psychiatry 2017; 174: 927–42. ecacy and safety of lurasidone versus placebo in antipsychotic- 50 Martin WF, Correll CU, Weiden PJ, et al. Mitigation of olanzapine- naive vs previously treated adolescents with an acute exacerbation induced weight gain with samidorphan, an opioid antagonist: of schizophrenia. Eur Psychiatry 2022; 65: 1–35. a randomized double-blind phase 2 study in patients with 64 Kryzhanovskaya L, Schulz SC, McDougle C, et al. Olanzapine schizophrenia. Am J Psychiatry 2019; 176: 457–67. versus placebo in adolescents with schizophrenia: a 6-week, 51 Correll CU, Newcomer JW, Silverman B, et al. Eects of randomized, double-blind, placebo-controlled trial. olanzapine combined with samidorphan on weight gain in J Am Acad Child Adolesc Psychiatry 2009; 48: 60–70. schizophrenia: a 24-week phase 3 study. Am J Psychiatry 2020; 65 Brandt L, Efthimiou O, Siafis S, et al. Antipsychotic acute-phase 177: 1168–78. treatment in individuals with and without recent treatment: 52 Zhu Y, Li C, Huhn M, et al. How well do patients with a first episode an individual participant data meta-analysis. JAMA Psychiatry 2025; of schizophrenia respond to antipsychotics: a systematic review and 82: 692–700. meta-analysis. Eur Neuropsychopharmacol 2017; 27: 835–44. 66 国家药监局 国家卫生健康委关于发布药物临床试验质量管理规范 53 Carlisle JB. False individual patient data and zombie randomised 的公告 (2020年第57号). https://www.nmpa.gov.cn/yaopin/ controlled trials submitted to Anaesthesia. Anaesthesia 2021; ypggtg/20200426162401243.html (accessed Aug 12, 2025). 76: 472–79. 67 国家食品药品监督管理总局关于开展药物临床试验数据自查核查 54 Dong S, Schneider-Thoma J, Siafis S, Peter N, Leucht S. Single-arm 工作的公告(2015年第117号). https://www.nmpa.gov.cn/xxgk/ggtg/ meta-analysis of drug response in placebo-controlled versus active- ypggtg/ypqtggtg/20150722173601172.html (accessed Aug 12, 2025). controlled antipsychotic drug trials in schizophrenia. Eur Neuropsychopharmacol 2024; 84: 21–26. --- [PDF原文](https://sci-net.xyz/storage/7932541/5589c478902eecf0dcb7cb1b47245dc548732028f82e1c3628567ff447bfebec/Comparative-efficacy-and-tolerability-of-antidopaminergic-and-muscarinic-antipsychotics-for.pdf) DOI: 10.1016/S0140-6736(25)02365-7