Kane 2003.

Methods	Allocation: randomised. Blindness: double. Duration: 6 weeks (preceded by 2 weeks screening period; 2 days wash‐out; 4‐6 weeks open‐label treatment [olanzapine vs risperidone] to confirm treatment‐resistant; 2‐10 days wash‐out). Setting: multi‐centre, 59 centres in US & Canada (2000‐2002).
Participants	Diagnosis: schizophrenia (DSM‐IV) classified as ‘treatment resistant*’ (n = 215 paranoid; n = 14 disorganised; n = 71 undifferentiated). N = 300. Age: mean 42.1 (Perphenazine mean 41.6; aripiprazole mean 42.6). Sex: F 92; M 208, (Perphenazine: F 52, M 94; aripiprazole: F 40, M 114). History: mean age at time of first hospitalisation 22.8 (Perphenazine mean 22.9; aripiprazole mean 22.6). Included: > 18 years of age; schizophrenia DSM‐IV; classified as ‘treatment resistant*’; PANSS total score ≥ 75 and a score of ≥ 4 on at least 2 items of conceptual disorganization; suspiciousness, hallucinatory behaviour, or delusions; Clinical Global Impression‐Severity of Illness (CGI‐S) score ≥ 4; treated as outpatient for at least 1 continuous 3‐month period during 2‐years prior to study. Excluded: DSM‐IV schizoaffective disorder, residual schizophrenia, bipolar disorder, clinical presentation or history consistent with delirium, dementia, amnesic or other cognitive disorders; refractory response to prior clozapine treatment administered at therapeutic doses for 6 weeks; previous unsatisfactory response to perphenazine; likelihood to require prohibited concomitant therapy; history of or current drug or alcohol abuse or dependence; history of suicide attempts/ serious suicidal thoughts; known allergy/ hypersensitivity to study drugs; treatment with an investigational drug within 4 weeks of wash‐out phase; previous enrolment in aripiprazole study; acute or unstable medical condition; pregnancy or lactating. Consent: written informed consent required; institutional review board approval obtained.
Interventions	1. Perphenazine: range 8 to 64 mg/day, mean 39.1 mg/day; started at 8 mg/day and could be increased to 16 mg/day on day 4 if needed; max 64 mg/day, doses greater than 8 mg/day administered twice daily, n = 146. 2. Aripiprazole: 15 to 30 mg/day, mean 30 mg/day; started at 15 mg/day, dose adjustments made to maximum 30 mg/day after week 1, n = 154. (Dose reductions also permitted provided participants remain in permitted dose range).
Outcomes	Primary outcome: PANSS change from baseline. Secondary outcomes: CGI‐I; CGI‐S; BPRS; QoL using QLS (clinically important improvement defined as > 20% improvement in total score from baseline); response to treatment (> 30% decrease in PANSS total score or CGI‐I score of 1 or 2); adverse events; lab data; EPS (SAS, AIMS and BAS).
Notes	*treatment resistant defined as “failure to experience satisfactory symptom relief despite at least 2 periods of treatment (lasting at least 6 weeks) with adequate doses of antipsychotic agents (one of which had to be a typical antipsychotic) during the 2 years prior to the study. Nor should participants have experienced satisfactory symptom relief with their most recent course of antipsychotic therapy (p214).
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomised – no further details; emailed authors 15/11/13.
Allocation concealment (selection bias)	Unclear risk	No details.
Blinding (performance bias and detection bias) All outcomes	Low risk	Double‐blind treatment phase: treatment‐resistant participants, as previously diagnosed, entered a 6‐week, open‐label treatment phase (olanzapine vs risperidone) to confirm treatment‐resistance. If completed, entered 2‐ to 10‐day single‐blind placebo wash‐out phase; afterwards randomised to double‐blind treatment with perphenazine vs aripiprazole.
Incomplete outcome data (attrition bias) All outcomes	Low risk	N = 512 screened; n = 416 entered open‐label treatment phase with olanzapine and Risperidone to confirm treatment resistant; n = 334 completed; of these, n = 34 discontinued due to reasons including adverse events, withdrawal of consent, lost to follow‐up; lack of efficacy, patient unreliability, and other causes. N = 300 entered double‐blind randomised study phase; of these, n = 225 (75%) completed the full 6‐week study (n = 110 receiving aripiprazole; n = 115 receiving perphenazine); of the n = 75 who discontinued, n = 33 was due to adverse events; n = 18 was due to lack of efficacy; n = 11 was due to consent withdrawal; n = 13 unknown cause. N = 3 participants did not receive study medication (n = 1 aripiprazole; n = 2 perphenazine). A further n = 3 (all aripiprazole) were excluded due to lacking post‐randomisation efficacy rating. QLS analysis included n = 207 participants (n = 104 aripiprazole; n = 103 perphenazine), missing participants were excluded due to lack of baseline/ post‐randomisation data. LOCF used; observed case analyses performed to corroborate data.
Selective reporting (reporting bias)	Low risk	None detected from cross‐checking published protocol and review.
Other bias	High risk	Funding: supported by Bristol‐Myers Squibb Co. and Otsuka Pharmaceutical Development & Commercialization, Inc. PI is consultant to Abbott, Bristol‐Myers Squibb, Janssen, Eli Lilly, Pfizer, Wyeth and AstraZeneca; received honoraria from Bristol Myers‐Squibb and Janssen. All other co‐authors bar‐one are either consultants, shareholders, have received grant support from or are employees of ACADIA, AstraZeneca, Bristol‐Myers Squibb, Eli Lilly, GlaxoSmithKline, Janssen, Pfizer, Sanofi and Solvay, or Otsuka. Rating scales: raters not stated to be independent of treatment.