Zimbroff 2007.

Methods	Allocation: random, computer‐generated, centre blocked Blindness: double, dummy administration. Duration: four weeks. Location: multicentre.
Participants	Diagnosis: schizophrenia or schizoaffective disorder (DSM‐IV). N = 256. Age: 18‐70 years. Gender: 169 M, 57 F. History: duration of illness not reported, age at onset not reported, hospitalised > 14 consecutive days prior to screening, PANSS ≥ 80, 4/more on 2+ PANSS positive items, score of ≥ 4 on CGI‐S Setting: inpatient.
Interventions	1. Aripiprazole: dose 15 mg/day first 14 days, up to 30 mg/day thereafter, mean 21 mg/day (SD 7). N = 129. 2. Ziprasidone: dose 80 mg/day on first 14 days, up to 160 mg/day thereafter mean 149 mg/day ( SD 25): N = 127.
Outcomes	Leaving the study early. Global state: CGI‐S. Mental state: BPRS, PANSS. Adverse effects: system‐specific effects, AIMS, BAS, SAS. Unable to use ‐ Global state: ORDQ. Mental state: CDSS (no numerical data). Adverse effects: SAS (reported as zero difference with P value ‐ not credible data), weight, glucose, blood lipids (median reported with no other data). Adverse effects: AIMS, BAS (P values, and numbers that did allow calculation of SD).*
Notes	* These data, although possible to report, carried assumptions on numbers of people from which they were derived, identical SDs per group, normality of data on top of other assumptions usually associated with scale‐derived numbers. We feel that these are assumptions too far.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Random, computer‐generated, blocked by centre.
Allocation concealment (selection bias)	Unclear risk	No further details.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Double, double dummy administration. Whether blinding was successful has not been examined, but both compounds have similar adverse effects.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Objective outcomes such as laboratory measures or death are unlikely to have been much affected by lack of blinding.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	LOCF method used to account for people leaving the study early.
Selective reporting (reporting bias)	High risk	Only adverse events with an incidence of more than 5% were reported. This procedure may have missed important adverse events. CDSS and ORDQ ‐ no numerical data reported because"no statistically significant differences were observed. . ." AIMS, BAS, SAS were portrayed very differently to other scales ‐ allowing computation of data but not giving clear person‐based numbers for the ratings.
Other bias	High risk	Random sequence generation (selection bias)