Takahashi 1982.

Methods	Allocation: randomised. Blindness: double (implied). Duration: 12 weeks. Setting: inpatient, multi‐centre (31 psychiatric institutions), Japan.
Participants	Diagnosis: schizophrenia (n = 127 hebephrenic, n = 24 catatonic, n = 38 paranoid, n = 4 hebephrenic‐catatonic, n = 2 hebephrenic‐paranoid, n = 1 schizoaffective, n = 2 undifferentiated). N = 205*. Age: 13 ‐ 60 years. Sex: no details. History: inpatient. Included: not described. Excluded: early stages of gestation; severe liver, kidney, cardiovascular and blood diseases; idiosyncrasy to psychotropic drugs; consciousness disturbances or vascular collapse or conditions under strong effects of neuro‐depressants; organic brain disorders; those patients whose conditions in the judgement of the attending physician would make it unlikely that they could tolerate the study treatment. Consent: informed consent obtained from each participant and guardian prior to the trial.
Interventions	1. Perphenazine: 4 mg per tablet + placebo tablet matching to the timiperone tablet, initially two or three tablets daily, afterwards individualised according to symptoms and side effects, maximum daily dose was 12 tablets, n = 99. 2. Timiperone: 1 mg per tablet + placebo tablet matching to the perphenazine tablet, initially two or three tablets daily, afterwards individualised according to symptoms and side effects, maximum daily dose was 12 tablets, n = 99. Additional medication: mild hypnotics and anti‐parkinsonian drugs ‐ no further details.
Outcomes	Leaving the study early. Global state: no better or worse. Adverse events. Unable to use ‐ Mental state: Keio University Brief Psychiatric Rating Scale (unpublished scale).
Notes	*Only 198 participants were evaluated and described.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomised: quote, "participants were assigned randomly" (p258). 'Randomised code' referred to, which was broken only at the end of treatment period.
Allocation concealment (selection bias)	Unclear risk	No details.
Blinding (performance bias and detection bias) All outcomes	Low risk	Double‐blind: implied ‐ decoding of drugs at the end of week 12. Code broken at the end of treatment period. Double‐dummy design.
Incomplete outcome data (attrition bias) All outcomes	High risk	Of N = 205 randomised, n = 7 were excluded, either because they had been given other neuroleptics or they revealed physical disease during the trial prior to breaking the randomisation code. A total of n = 198 were included in the analysis.
Selective reporting (reporting bias)	Unclear risk	No means, SD or P values reported for scale data.
Other bias	Unclear risk	Funding: not stated ‐ study drugs were supplied by Daiichi Seiyaku Co, Ltd, Tokyo, Japan. Raters: not stated to be independent of treatment.