Krakowski 2006.

Methods	Allocation: random; block randomisation with no baseline stratification; parallel group. Blinding: double‐blind. Duration: 1 to 2 weeks' pre‐trial baseline screening; 12 weeks' trial ‐ 6 weeks' escalation and fixed dose, 6 weeks' variable dose.
Participants	Diagnosis: schizophrenia/schizoaffective disorder (DSM‐IV). N=110. Age: 18 to 60 years. History: psychiatric inpatient at 1 of 2 New York state hospitals, hospitalised within the last year. Inclusion: clearly confirmed episode of physical assault during this hospitalisation, some persistence of aggression i.e. evidence of other aggressive event (physical or verbal or against property). Exclusion: people hospitalised for over 1 year; history of non‐response or intolerance to interventions; people with medical conditions adversely affected by interventions; people who received a depot antipsychotic drug within 30 days before randomisation.
Interventions	1. Haloperidol: first 6 weeks, target endpoint 20 mg/day, mean dose 19.6 mg/day; last 6 weeks, range 10 mg/day to 30 mg/day, mean dose 23.3 mg/day. N=36. 2. Olanzapine: first 6 weeks, target endpoint 20 mg/day, mean dose 19.8 mg/day; last 6 weeks, range 10 mg/day to 35 mg/day, mean dose 24.7 mg/day. N=37. 3. Clozapine: first 6 weeks, target endpoint 500 mg/day, mean dose 457.1 mg/day; last 6 weeks, range 200 mg/day to 800 mg/day, mean dose 565.5 mg/day. N=37.
Outcomes	Leaving study early. Unable to use ‐ Aggression: MOAS (skewed data), subgroup: MOAS (randomisation status unclear). Mental state: PANSS (skewed data). Adverse effects: ESRS and checklist (no data reported).
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"Block randomisation" ‐ no further details.
Allocation concealment (selection bias)	Unclear risk	Not stated.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	"Double blind" but no description of how blinding was ensured.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Raters, blind to treatment group, performed all clinical assessments.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Study attrition clearly reported ‐ table presented and intention‐to‐treat analysis performed.
Selective reporting (reporting bias)	High risk	Adverse effects data not fully reported; numerical data for additional medications not fully reported.
Other bias	Unclear risk	Supplemental funding from pharmaceuticals for "encapsulation of the medications".

N: number of participants.

Rating scale abbreviations: ESRS: Extrapyramidal Symptom Rating Scale; MOAS: Modified Overt Aggression Scale; PANSS: Positive and Negative Syndrome Scale.

Diagnostic standards abbreviations: DSM‐IV: Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition.