Schizophrenia and related psychotic disorders are associated with an increased risk of hostility and aggressive behaviour. Reference Whiting, Gulati, Geddes and Fazel1,Reference Fazel, Smith, Chang and Geddes2 The lifetime prevalence of violence – the most severe form of aggression – among affected individuals is about five times higher than in the general population. Reference Whiting, Gulati, Geddes and Fazel1 Estimates of the population-attributable fraction are as high as 10%. Reference Fazel, Smith, Chang and Geddes2 Aggressive behaviour, often preceded or accompanied by hostility, contributes not only to harm but also to stigma, poorer prognosis, criminal justice involvement and substantial economic costs. Reference Lamsma and Harte3
Antipsychotic medication is central to the management of hostility and aggression in schizophrenia. Reference Strassnig, Nascimento, Deckler and Harvey4 Its effects may arise from improvements in positive symptoms (delusions, hallucinations and disorganised thinking), Reference Lagerberg, Lambe, Paulino, Yu and Fazel5 modulation of serotonergic and dopaminergic systems involved in impulsivity and emotion regulation Reference Seo, Patrick and Kennealy6 and, in some cases, reductions in substance misuse. Reference Green, Noordsy, Brunette and O’Keefe7 However, the magnitude and specificity of these effects remain unclear.
Although numerous randomised controlled trials (RCTs) have examined the effects of antipsychotics on hostility and aggression, findings have been inconsistent. Differences in study design, outcome definitions and comparators (placebo or other antipsychotics) complicate interpretation. Post hoc analyses from two large, multi-arm trials – Clinical Antipsychotic Trials Intervention Effectiveness Reference Volavka, Czobor, Derks, Bitter, Libiger and Kahn8 and European First-Episode Schizophrenia Trial Reference Volavka, Czobor, Citrome and Van Dorn9 – reported reductions in hostility with certain antipsychotics. However, these analyses were limited to the small number of antipsychotics included in each trial. Findings from pairwise meta-analyses are similarly limited: one focused on risperidone, Reference Aleman and Kahn10 another compared first- and second-generation antipsychotics, Reference Faay, Czobor and Sommer11 a third included participants with diagnoses beyond schizophrenia, Reference van Schalkwyk, Beyer, Johnson, Deal and Bloch12 and a fourth, restricted to relapse prevention, assessed aggressive behaviour in only a small subset of trials (k = 5). Reference Leucht, Tardy, Komossa, Heres, Kissling and Salanti13 Moreover, reliance on direct comparisons, combined with the scarcity of head-to-head trials, limits statistical power to detect differences between individual antipsychotics.
Network meta-analysis (NMA) integrates direct and indirect evidence, providing more precise estimates of treatment effects and enabling comparisons between treatments that have not been directly compared within the same trial. Reference Chaimani, Caldwell, Li, Higgins, Salanti, Higgins, Thomas, Chandler, Cumpston, Li and Page14 To date, no NMAs have been published on antipsychotics for hostility or aggression in schizophrenia. A registered protocol has outlined an NMA of rapid tranquillisation (≤48 h) for aggressive behaviour. Reference Ostinelli, D’Agostino, Shokraneh, Salanti and Furukawa15 We therefore conducted an NMA of RCTs to compare the longer-term efficacy of individual antipsychotics for reducing hostility and aggression in schizophrenia.
Method
This systematic review and NMA followed the PRISMA 2020 guidelines Reference Page, McKenzie, Bossuyt, Boutron, Hoffmann and Mulrow16 and the PRISMA-NMA extension. Reference Hutton, Salanti, Caldwell, Chaimani, Schmid and Cameron17 The protocol was registered prospectively with PROSPERO (CRD42024572272; 3 August 2024). Reference Lamsma, Ostinelli, Cipriani and Fazel18 Throughout, we use the term ‘trial’ to refer to an RCT, ‘study’ to an analytic unit based on one or more trials and ‘article’ to a report describing one or more studies.
Search strategy and selection criteria
Four electronic databases (ClinicalTrials.gov, the Cochrane Schizophrenia Group’s register, PubMed and PsycINFO) were searched from inception to 25 April 2025. The search strategy combined terms for study design, schizophrenia, antipsychotics and hostility and aggressive behaviour (Supplementary Table A1.1 available at https://doi.org/10.1192/bjp.2026.10800). Reference lists of the included articles and relevant previous reviews were screened for additional articles. No restrictions were applied regarding language, publication year, country or publication status. Two reviewers (L.V. and E.E.) independently screened titles and abstracts identified through the searches. The full texts of articles reporting potentially eligible trials were assessed independently by the same reviewers. Final inclusion decisions were verified by the lead author (J.L.) prior to data extraction. The study selection process is shown in Supplementary Fig. A1.1.
We included RCTs comparing one antipsychotic with another antipsychotic or placebo in adults (≥18 years) diagnosed with schizophrenia or related psychotic disorder (e.g. schizoaffective disorder, schizophreniform disorder, delusional disorder). Eligible trials were required to report at least one measure of hostility or aggressive behaviour. Because our aim was to examine effects beyond rapid tranquillisation, trials were required to have at least one follow-up assessment more than 48 h after treatment initiation. To maximise available evidence, we included trials regardless of blinding level (open-label, single-blind or double-blind) or design (parallel-group, cluster-randomised or cross-over). For cross-over trials, only the first randomisation phase was considered. Eligible interventions consisted of monotherapy with any antipsychotic classified under Anatomical Therapeutic Chemical code N05AXXX, 19 excluding lithium, administered at any dose and by any route.
Data extraction
Data extraction was conducted by the lead author (J.L.) from 1 May 2025. Extracted data were reviewed for completeness and consistency, and any uncertainties were resolved through discussion with the co-authors. Participant age and gender were obtained from the included articles or, if unavailable, from other articles describing the same studies. Gender was recorded as reported in the included studies, all of which categorised participants as male or female. Race and ethnicity were not extracted because reporting of these variables was insufficient for meaningful analysis. To allow comparison across antipsychotics, realised doses were converted to oral olanzapine equivalents based on established dose–response data for the treatment of acute psychosis Reference Leucht, Crippa, Siafis, Patel, Orsini and Davis20 and World Health Organization-defined daily doses 21 for maintenance treatment. For studies with multiple follow-up assessments, we selected the assessment closest to 14 weeks. This decision was made after protocol development because the most appropriate follow-up assessment could only be determined once eligible studies had been identified and their assessment schedules reviewed. Selecting the assessment closest to 14 weeks maximised comparability across studies while allowing treatment exposure up to the point at which response to antipsychotics generally plateaus. Reference Drosos, Johnsen, Bartz-Johannessen, Larsen, Reitan and Rettenbacher22 Most studies (k = 31) had a duration of 14 weeks or less, whereas a smaller number (k = 10) extended beyond 6 months. Five of the longer-term studies also included an assessment at approximately 14 weeks. In one study, an earlier assessment was selected to enable the inclusion of all treatment arms (Supplementary Table A2.1). We considered using the final assessment from each study but rejected this approach because the substantial variation in study duration would have introduced additional methodological and clinical heterogeneity. Stratification by study duration was also not undertaken, as the network of longer-term studies was too sparse to support meaningful synthesis. When multiple outcome measures were available, priority was given to instruments specifically designed to assess aggression and, secondarily, to those most commonly used across trials. When not explicitly reported, means and s.d. were extracted from figures using WebPlotDigitizer version 4.7 (Ankit Rohatgi, Pacifica, California, USA; https://automeris.io). Missing s.d. were estimated from test statistics or by using the mean s.d. of the remaining studies. Reference Higgins, Li, Deeks, Higgins, Thomas, Chandler, Cumpston, Li and Page23 Where available, we used results from intention-to-treat analyses. Results obtained with multiple imputation were preferred over those obtained with last observation carried forward or complete-case analyses. Risk of bias was assessed using the Cochrane Risk of Bias 2 (RoB 2) tool. Reference Sterne, Savović, Page, Elbers, Blencowe and Boutron24
Outcome
As is common in schizophrenia research, Reference Witt, Van Dorn and Fazel25 hostility and aggressive behaviour – two overlapping constructs – were analysed jointly. Hostility was conceptualised as encompassing emotional expressions and behaviours, such as irritability, anger and threats, that often precede and accompany aggression. Aggression was defined as behaviour intended to cause psychological or physical harm to another person. Reference Anderson and Bushman26 Outcome measures could include hostility items or related subscales from symptom rating scales (e.g. item 10 of the Brief Psychiatric Rating Scale Reference Overall and Gorham27 and the hostility–excitement factor of the Positive and Negative Syndrome Scale Reference Kay, Fiszbein and Opler28 ), aggression-specific instruments (e.g. the Buss–Perry Aggression Questionnaire Reference Buss and Perry29 and the Modified Overt Aggression Scale Reference Kay, Wolkenfeld and Murrill30 ) or documented incidents in clinical or criminal records.
Data analysis
We conducted a frequentist random-effects NMA using the netmeta package version 3.2-0 (https://cran.r-project.org/package=netmeta) in R version 4.4.1 for Windows (R Foundation for Statistical Computing, Vienna, Austria; https://www.r-project.org/). Effect sizes were expressed as standardised mean differences (SMDs). When means and s.d. were not reported, SMDs were derived from F-statistics using standard methods described in the Cochrane Handbook for two-group comparisons (numerator d.f. = 1). Reference Higgins, Li, Deeks, Higgins, Thomas, Chandler, Cumpston, Li and Page23 Odds ratios were converted to SMDs using the probit method, as implemented in the Practical Meta-Analysis Effect Size Calculator (Campbell Collaboration, Oslo, Norway; https://www.campbellcollaboration.org/calculator/). Analyses were adjusted for within-study correlations in multi-arm trials. Reference Rücker and Schwarzer31 We assessed the transitivity assumption by comparing the distribution of potential effect modifiers across treatments, including follow-up duration, gender, age, care setting and illness phase. Global consistency was assessed with the design-by-treatment interaction model Reference Higgins, Jackson, Barrett, Lu, Ades and White32 and local consistency with the node-splitting method (Separating Indirect from Direct Evidence, SIDE). Reference Dias, Welton, Caldwell and Ades33 The magnitude of between-study variance (τ 2) was classified as ‘low’ (<0.04), ‘moderate’ (0.04–0.35) or ‘high’ (>0.35) using the first and third quantiles of the empirical distributions reported by Turner et al. Reference Turner, Davey, Clarke, Thompson and Higgins34 We investigated small-study effects with a comparison-adjusted funnel plot. Confidence in each comparison was assessed using the Confidence in Network Meta-Analysis framework, which considers within-study bias, reporting bias, indirectness, imprecision, heterogeneity and incoherence. Reference Nikolakopoulou, Higgins, Papakonstantinou, Chaimani, Del Giovane and Egger35 Sources of residual heterogeneity were explored in sensitivity analyses and meta-regression analyses. Sensitivity analyses examined the influence of publication year, follow-up duration, patient population, dosage, baseline levels of hostility or aggression and outcome definition. Meta-regression analyses were performed to assess whether follow-up duration and care setting modified treatment effects.
Changes to the original protocol
Two additional databases (PubMed and PsycINFO) were searched to increase the yield of eligible studies. Four potential effect modifiers – mean age, route of administration (oral versus long-acting injectable), lifetime substance misuse and concurrent psychosocial interventions – were considered but not analysed. Mean age did not vary meaningfully between treatments (Supplementary Fig. A4.4), and the other variables were reported in too few trials for reliable analysis.
Results
Included studies
The database search identified 247 records, of which 217 remained after removal of duplicates and were screened by title and abstract. Forty-one full-text articles were assessed for eligibility, resulting in the exclusion of 16. An additional 14 articles were identified from reference lists, yielding a final set of 39 articles reporting on 60 trials that met the inclusion criteria (Supplementary Fig. A1.1). These articles contained 41 studies (Supplementary Table A2.1). The majority of studies (k = 37, 90%) were based on double-blind, parallel-group trials. The median follow-up duration was 8 weeks (interquartile range = 6), with 80% of studies (k = 34) providing outcome data collected between 4 and 14 weeks. In total, 21 025 participants were included. The mean age of participants was 38 years (s.d. = 3), and just over half were men (n = 12 603, 60%). A small number of studies focused exclusively on first-episode (k = 4, 10%) or treatment-resistant (k = 4, 10%) patients. Most studies evaluated oral formulations (k = 35, 85%). Four studies (10%) evaluated long-acting injectable formulations, while one study each (2%) evaluated short-acting intramuscular and transdermal formulations. Nearly all studies (k = 39, 95%) assessed hostility or aggression using standardised instruments (Supplementary Table A2.2). Thirty-five studies (85%) assessed hostility using items or related subscales from the Brief Psychiatric Rating Scale, Positive and Negative Syndrome Scale or Symptom Checklist-90-Revised. Four studies (10%) used the Modified Overt Aggression Scale. The remaining two studies assessed aggression as a binary outcome (i.e. whether or not an aggressive act occurred) rather than rating its severity on a scale. Thirty-four studies (83%) were rated as having some concerns regarding risk of bias, six (15%) were rated as high risk and one (2%) as low risk (Supplementary Fig. A3.1 and Supplementary Table A3.1). The most common sources of bias were post hoc analyses of larger trials and missing outcome data due to drop-out. Concerns relating to randomisation, deviations from intended interventions and outcome measurement were less common.
Network meta-analysis
The network of direct treatment comparisons included 18 antipsychotics and placebo (Supplementary Table A2.1; Fig. 1). About half of the studies included a placebo arm (k = 21, 51%). Among antipsychotics, haloperidol (k = 16, 39%), risperidone (k = 12, 29%), olanzapine (k = 11, 27%) and quetiapine (k = 8, 20%) had the highest degree of connectivity, reflecting their frequent inclusion in placebo-controlled or head-to-head trials.
Network of direct treatment comparisons among antipsychotics and placebo for aggression in schizophrenia. Each node represents a treatment, with its size proportional to the total number of participants assigned to that treatment. Lines connect treatments that have been directly compared in randomised controlled trials, and their thickness corresponds to the number of trials making each comparison.
Compared with placebo, several antipsychotics were associated with reductions in hostility and aggressive behaviour (Fig. 2). Clozapine (SMD, 95% CI = –0.75, –1.13; –0.38), paliperidone (–0.62, –0.88; –0.36) and olanzapine (–0.60, –0.86; –0.35) showed the largest estimated effects. Smaller but statistically significant effects were observed for asenapine, risperidone, aripiprazole and haloperidol (SMD range = –0.52 to –0.30). The effects of the remaining antipsychotics were non-significant and varied in precision. The league table (Fig. 3) summarises all pairwise comparisons among antipsychotics. Clozapine, olanzapine, paliperidone and risperidone were most often favoured, although there was considerable imprecision across comparisons.
Effects of antipsychotics on aggressive behaviour compared with placebo. SMD, standardised mean difference.
Comparative effects of antipsychotics on aggressive behaviour. Values are standardised mean differences with 95% confidence intervals. Results of the network meta-analysis are presented in the left lower half and results from pairwise meta-analyses in the upper right half. Negative values favour the treatment in the column, whereas positive values favour the treatment in the row. Bold indicates statistically significant differences. AMI, amisulpride; ARI, aripiprazole; ASE, asenapine; BRE, brexpiprazole; CAR, cariprazine; CHL, chlorpromazine; CLO, clozapine; FLU, fluphenazine; HAL, haloperidol; ILO, iloperidone; LEV, levosulpiride; LUR, lurasidone; OLA, olanzapine; PAL, paliperidone; PER, perphenazine; PLA, placebo; QUE, quetiapine; RIS, risperidone; ZIP, ziprasidone.
Assessment of transitivity revealed that participant demographics were broadly similar across treatment comparisons, though notable variation was present in methodological and clinical features such as care setting, illness phase, treatment resistance, baseline levels of hostility or aggression and outcome definition (Supplementary Figs A4.1–12). The design-by-treatment interaction model revealed significant inconsistency between direct and indirect evidence across the network (Q 34 = 53.40, p = 0.018). The SIDE analysis identified eight locally inconsistent comparisons (p < 0.05), all involving either amisulpride or olanzapine (Supplementary Table A5.1). Overall heterogeneity was moderate (τ 2 = 0.09). Visual inspection of the comparison-adjusted funnel plot (Supplementary Fig. A6.1) and Egger’s test (t 85 = –0.03, p = 0.970) suggested no evidence of substantial small-study effects. Confidence in the evidence was rated as ‘low’ or ‘very low’ for all comparisons, largely reflecting concerns about within-study bias, imprecision and incoherence (Supplementary Table A7.1).
Sensitivity and meta-regression analyses
Sensitivity (Supplementary Figs A.8.1–10) and meta-regression analyses (Supplementary Figs A.9.1–2) did not materially change the overall findings. Excluding studies not primarily focused on schizophrenia, those with atypical baseline levels of hostility or aggression, or those relying on hostility-related subscales produced slightly larger effects, as did adjustment for follow-up duration and care setting.
Discussion
This systematic review and NMA of 41 studies, comprising 60 RCTs and more than 21 000 participants, found that antipsychotic treatment reduces hostility and aggressive behaviour in schizophrenia, with moderate effects over the short to medium term. Because these findings are based on randomised trials, they strengthen support for a causal role of positive symptoms in hostility and aggression, previously inferred largely from observational studies prone to confounding by indication and time-varying factors. Reference Lamsma and Harte3 They are also consistent with pairwise meta-analyses of selected antipsychotics and antipsychotics as a class, Reference Aleman and Kahn10–Reference Leucht, Tardy, Komossa, Heres, Kissling and Salanti13 and with pharmacoepidemiological studies reporting lower rates of aggressive behaviour during antipsychotic treatment. Reference Fazel, Zetterqvist, Larsson, Långström and Lichtenstein36–Reference Swanson, Swartz and Elbogen38 The latter support their external validity in routine clinical practice. However, because most included studies assessed hostility, the findings may primarily reflect reductions in hostility rather than aggressive behaviour.
This study builds on previous meta-analyses by comparing individual antipsychotics within a network meta-analytic framework. Clozapine, paliperidone and olanzapine were associated with the largest reductions in hostility and aggressive behaviour. Smaller but statistically significant reductions were observed for asenapine, risperidone, aripiprazole and haloperidol. This pattern was broadly consistent across sensitivity and meta-regression analyses adjusting for follow-up duration, care setting, target population, baseline levels of hostility or aggression and outcome definition. However, confidence in the evidence was limited by the predominantly indirect nature of most comparisons, inconsistency between direct and indirect estimates and imprecision. In addition, confidence intervals for comparisons with placebo showed substantial overlap. Differences between antipsychotics should therefore be interpreted cautiously. The finding for clozapine aligns with its established role in treatment-resistant schizophrenia and the management of aggression. Reference Frogley, Taylor, Dickens and Picchioni39 The finding for paliperidone is novel and potentially clinically important. Although the comparison with placebo was supported by direct evidence, comparisons with other antipsychotics relied mostly on indirect evidence. Consequently, the apparent advantage of paliperidone over other antipsychotics requires confirmation in head-to-head trials. In contrast, quetiapine – widely prescribed for behavioural disturbance Reference Carney40,Reference Hardoon, Hayes, Viding, McCrory, Walters and Osborn41 – did not differ significantly from placebo.
A similar pattern of findings has been reported in NMAs of overall symptom severity in schizophrenia, Reference Leucht, Cipriani, Spineli, Mavridis, Örey and Richter42–Reference Dong, Schneider-Thoma, Bighelli, Siafis, Wang and Burschinski44 suggesting improvements in positive symptoms are likely to account for a substantial proportion of the reductions in hostility and aggression. Nevertheless, additional pharmacological mechanisms may be involved. Clozapine’s distinctive profile – characterised by potent 5-HT2A antagonism, high D4 receptor affinity and modulation of glutamatergic signalling Reference Meltzer and Massey45 – may underlie its relatively strong effects through reduced impulsivity and enhanced affect regulation. Reference Frogley, Taylor, Dickens and Picchioni39,Reference Krakowski46 Olanzapine, paliperidone and risperidone all share combined 5-HT2A and D2 receptor antagonism, Reference Mauri, Paletta, Maffini, Colasanti, Dragogna and Di Pace47 a profile that may improve inhibitory control. Reference Fink, Anastasio, Fox, Rice, Moeller and Cunningham48,Reference Pattij, Janssen, Vanderschuren, Schoffelmeer and van Gaalen49 The mood-stabilising and sedative properties of clozapine Reference Hopkins and Miller50–Reference Li, Tang, Wang and de Leon52 and olanzapine Reference Kishi, Ikuta, Matsuda, Sakuma, Okuya and Mishima53,Reference Fang, Sun, Wang, Ren, Calabrese and Gao54 may also contribute. In contrast, the modest effects observed for first-generation antipsychotics suggest that dopamine D2-receptor blockade alone is insufficient. Moreover, extrapyramidal side-effects – particularly akathisia – can heighten agitation and irritability in some patients, thereby increasing the risk of hostility and aggressive behaviour. Reference Wu, Siafis, Wang, Burschinski, Schneider-Thoma and Priller55 A recent data-driven classification based on receptor affinity identified several clusters of antipsychotics with distinct efficacy and side-effect profiles. Reference McCutcheon, Harrison, Howes, McGuire, Taylor and Pillinger56 Clozapine and olanzapine fall within the muscarinic cluster, while paliperidone and risperidone fall within the dopamine–serotonin antagonist cluster. This may point to the involvement of muscarinic and serotonergic pathways in impulse and emotion regulation. Reference Seo, Patrick and Kennealy6,Reference Sarter and Paolone57,Reference Dulawa and Janowsky58
A number of important limitations should be acknowledged. First, most included studies were judged to carry at least a moderate risk of bias due to post hoc analyses and incomplete outcome data. Second, several potential effect modifiers, including illness phase, treatment resistance, care setting and baseline levels of hostility or aggression, were unevenly distributed across treatment comparisons. Treatment effects may be larger during first-episode psychosis or acute exacerbations, when hostility is more closely linked to active psychotic symptoms. Reference Hodgins and Klein59 Clozapine has established superiority in treatment-resistant schizophrenia. Reference Frogley, Taylor, Dickens and Picchioni39 Higher baseline levels of hostility or aggression, as well as the greater clinical acuity typically seen in in-patient settings, may provide more scope for improvement. Their uneven distribution raises concerns about the transitivity assumption, which underpins the validity of indirect comparisons. However, sensitivity and meta-regression analyses examining selected potential effect modifiers produced results similar to those of the main analyses. Third, follow-up durations in most studies ranged from 4 to 14 weeks, limiting conclusions regarding the longer-term effects of antipsychotics on hostility and aggressive behaviour. Fourth, because most included studies evaluated oral formulations, the generalisability of the findings to other formulations – particularly long-acting injectables – remains uncertain. In addition, there were too few studies of non-oral formulations to determine whether treatment effects differ according to formulation. Fifth, we combined hostility and aggressive behaviour because of their conceptual overlap and to maximise the available evidence. As a result, it remains uncertain whether antipsychotics differ in their effects on hostility and aggressive behaviour. Furthermore, because most studies assessed hostility using items embedded within symptom rating scales, shared variance with improvements in overall psychopathology may partly account for the observed effect estimates. Finally, first-episode and forensic populations were underrepresented, and substance misuse – common in schizophrenia Reference Hunt, Large, Cleary, Lai and Saunders60 and a risk factor for aggression Reference Lamsma and Harte3 – was often an exclusion criterion, limiting the generalisability of the findings across the full range of patients seen in routine clinical practice.
The findings of this NMA suggest that antipsychotics are effective for managing hostility and aggressive behaviour in schizophrenia over the short to medium term. Evidence was strongest for clozapine and several other second-generation antipsychotics, particularly paliperidone, olanzapine and risperidone. However, uncertainty regarding comparative efficacy makes clinical considerations – including adverse effects, tolerability, adherence, comorbidities and patient preferences – particularly important when selecting between antipsychotics. Previous studies suggest that long-acting injectable formulations of paliperidone and risperidone improve adherence, which may help sustain treatment response. Reference Okoli, Kappi, Wang, Makowski and Cooley61 Combining antipsychotic medication with targeted psychosocial interventions – such as cognitive–behavioural therapy, anger management and social skills training – may produce additional and longer-term reductions in hostility and aggression. Future research should prioritise pre-registered RCTs in which aggressive behaviour is the primary outcome, assessed using standardised and validated instruments designed for this purpose, and conducted over longer follow-up periods to determine the specificity and durability of treatment effects. Further trials are needed in first-episode psychosis, treatment-resistant schizophrenia and forensic populations, alongside studies examining the impact of comorbid substance misuse on treatment response. Biomarker approaches, such as neuroimaging and electrophysiology, together with neuropsychological testing, may clarify the mechanisms through which antipsychotics reduce hostility and aggression.
This NMA suggests that antipsychotic treatment reduces hostility and aggressive behaviour in schizophrenia over the short to medium term. Clozapine, paliperidone and olanzapine showed the largest estimated effects, although low confidence in the evidence limits conclusions about differences between antipsychotics. Further trials with aggression as the primary outcome and longer follow-up are needed.
Supplementary material
The supplementary material for this article is available online at https://doi.org/10.1192/bjp.2026.10800
Data availability
The data that support the findings of this study are available from the corresponding author, S.F., upon reasonable request.
Acknowledgements
We thank Larissa van Vlodorp (L.V.) and Esmee Ellens (E.E.) for their assistance with the literature search. During the preparation of this manuscript, the authors used ChatGPT (versions 5.0 and 5.5) to assist with language editing and improving clarity. The authors reviewed and edited the manuscript as needed and take full responsibility for its content.
Author contributions
J.L. conceived the study, designed the methods, verified the data, conducted the analyses, and wrote the first draft of the manuscript. E.G.O., A.C. and S.F. provided input on the analysis plan, contributed to the interpretation of the results and critically reviewed the manuscript. All authors approved the final manuscript and agreed to its submission for publication.
Funding
This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.
Declaration of interest
A.C. is supported by the National Institute for Health Research (NIHR) Oxford Cognitive Health Clinical Research Facility, an NIHR Research Professorship (grant RP-2017-08-ST2-006), the NIHR Oxford and Thames Valley Applied Research Collaboration, the NIHR Oxford Health Biomedical Research Centre (grant NIHR203316) and the Wellcome Trust (GALENOS Project and PRADA Trial). A.C. has received research, educational and consultancy fees from the Italian Network for Paediatric Trials, the Cariplo Foundation, Lundbeck, Teva and Angelini Pharma, outside the submitted work. S.F. is supported by the NIHR Oxford Health Biomedical Research Centre (grant NIHR203316). S.F. has received a speaker fee from Angelini Pharma, outside the submitted work. E.G.O. is a member of the BJPsych editorial board and did not take part in the review or decision-making process for this manuscript. J.L. declares no competing interests. The views expressed are those of the authors and not necessarily those of the UK National Health Service, the NIHR or the UK Department of Health and Social Care.
Open access
eLetters
No eLetters have been published for this article.