Lorazepam 1997b, USA.

Study characteristics
Methods	Allocation: randomised. Blinding: double blind. Design: not stated. Duration: 4 hours. Setting: Psychiatric Emergency Service, Eastern Pennsylvania Psychiatric Institute, US.
Participants	Diagnosis: schizophrenia (n = 13), bipolar (n = 13), schizoaffective (n = 4), psychotic disorder not otherwise specified (n = 7). n = 37. Age: 18‐61 years. Gender: 26 men, 11 women. Ethnicity: white (n = 21), African‐American (n = 14), Hispanic (n = 2). Consent: need for informed consent waived. History: drug abuse or dependence n = 10. Inclusion criteria: highly agitated people exhibiting psychotic symptoms, judged by emergency department staff to be an imminent danger to themselves, required restraint, scored ≥ 5 on ≥ 3 BPRS items and ≥ 4 on CGI. Exclusion criteria: not stated.
Interventions	Benzodiazepines vs antipsychotics. 1. Lorazepam 2 mg oral or IM (n = 17). 2. Haloperidol 5 mg oral or IM (n = 20). Medication administered every 30 minutes for 4 hours 'as needed' until the participant was sedated or no longer posed a danger to themselves or staff.
Outcomes	Global impression: sedation; mean endpoint score (CGI‐S). Mental state: mean endpoint score (BPRS). Adverse effects/events: EPS.
Notes	More bipolar participants received lorazepam (n = 10) than haloperidol (n = 3).
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomised; however, no further description of randomisation provided.
Allocation concealment (selection bias)	Unclear risk	Not stated.
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Double blind; however, no further description of blinding provided. Rating scales: symptom ratings were conducted by 1 of 4 raters (psychiatry residents or psychiatric nursing staff). Raters were trained by trial author SF.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Follow‐up: 100%.
Selective reporting (reporting bias)	Low risk	All measure outcomes were reported.
Other bias	Low risk	Funding: supported in part by a grant from the National Alliance for Research on Schizophrenia and Depression. We detect no other bias.