Various drugs 1993.

Study characteristics
Methods	Randomisation: centrally randomised by a specialised unit using an "adaptive randomisation method”. Allocation: procedure not described. Blinding: open, only key rating scales were additionally rated by a second blind assessor. Duration: 2 years. Design: parallel. Location: multi‐centre. Setting: outpatient.
Participants	Diagnosis: schizophrenia or schizoaffective disorder (ICD‐9 and Research Diagnostic Criteria). N = 237. Gender: 124 women, 113 men. Age: mean 34.6 years. History: duration stable‐ at least 3 months in addition titrated to minimally effective dose which was maintained for at least 4 weeks, duration ill‐ mean 7.3 years, number of previous hospitalisations‐ n.i., age at onset‐ mean 27.3 years, severity of illness‐ mean CGI 3.8; mean BPRS total score 28.5, baseline antipsychotic dose‐ n.i..
Interventions	1. Drug: various antipsychotic drugs. Flexible dose, minimum 100 mg/day chlorpromazine equivalent. Allowed dose range: 100 mg ‐ unlimited chlorpromazine equivalents/day. Mean dose: 201 mg/day. N = 122. 2. No treatment (= crisis management, medication was only given in case of a full relapse). Duration of taper: 50% every two weeks, thus after 6 weeks only 12.5% of initial dose left, thus 42 days. Note that participants were not withdrawn after they had received crisis intervention. N = 115. Rescue medication: in the no treatment group additional antipsychotic medication could only be given in case of relapse.
Outcomes	Examined Relapse: BPRS total score ‐ >10 increase, GAS < 20 reduction, deterioration Clinical Global Impression Scale CGI >7. Leaving the study early. Service use: number of participants hospitalised. Unable to use/Not included Global state: CGI (no usable data ). Mental state: BPRS, AMDP system, Paranoid Depression Scale (all no means, no SDs / no predefined outcome of interest). Functioning: Strauss and Carpenter scale, another scale validated by the study group (no usable data) Subjective well‐being (own scale ‐ no mean, no SD/no predefined outcome of interest). Adverse effects: extrapyramidal side‐effects (AIMS ‐ no SD, SAS, Dosage Record and Treatment Emergent Symptoms Scale ‐ all no means, no SDs/continuous side‐effect results were not among the predefined outcomes of interest). Concept of illness (concept of illness scale ‐ no mean, no SD). Compliance: doctors’ assessment (no predefined outcome of interest). Physiological measures: routine laboratory, ECG, EEG (no data/no predefined outcome of interest).
Notes	There was a third group using intermittent treatment which was not of interest for this review.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Centrally randomised by a specialised unit using an "adaptive randomisation method”.
Allocation concealment (selection bias)	Unclear risk	Procedure not described.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Open, only key rating scales were additionally rated by a second blind assessor.
Blinding (performance bias and detection bias) Subjective outcomes	Unclear risk	Open, only key rating scales were additionally rated by a second blind assessor.
Blinding (performance bias and detection bias) Objective outcomes	Low risk	Open, only key rating scales were additionally rated by a second blind assessor.
Incomplete outcome data (attrition bias) All outcomes	High risk	High two‐year discontinuation rate of 43.7%. Analysis was ITT based on Kaplan‐Meier survival curve analysis, completer analyses were presented in addition if different. A risk of bias can not be excluded given the high discontinuation rate.
Selective reporting (reporting bias)	Low risk	No evidence for selective reporting.
Other bias	Low risk	No clear evidence for other bias.