BACKGROUND: Proponents of early intervention have argued that outcome might be improved if more therapeutic efforts were focused on the early stages of schizophrenia or on people with prodromal symptoms. Early intervention in schizophrenia has two elements that are distinct from standard care: early detection and phase-specific treatment. Both elements may be offered as supplements to standard care, or may be provided through a specialised early intervention team. Early intervention is now well established as a therapeutic approach in America, Europe and Australasia, but it is unclear how far early detection, phase-specific treatments, and the use of early intervention teams are underpinned by evidence of effectiveness. OBJECTIVES: To evaluate the effects of: (a) early detection; (b) phase-specific treatments; and (c) specialised early intervention teams in the treatment of people with prodromal symptoms or first episode psychosis. SEARCH STRATEGY: We searched CINAHL (1982-2002), The Cochrane Controlled Trials Register (November 2001), The Cochrane Schizophrenia Group Register (July 2003), EMBASE (1980-2002), MEDLINE (1966-2002), PsycINFO (1967-2002), reference lists and contacted the European First Episode Network (2003). For the 2006 update we searched the Cochrane Schizophrenia Group's register. SELECTION CRITERIA: We included all randomised controlled trials designed to prevent progression to psychosis in people showing prodromal symptoms, or to improve outcome for people with first episode psychosis. Eligible interventions, alone and in combination, included early detection, phase-specific treatments, and care from specialised early intervention teams. We accepted cluster-randomised trials but excluded non-randomised trials. DATA COLLECTION AND ANALYSIS: We reliably selected studies, quality rated them and extracted data. For dichotomous data, we estimated relative risks (RR), with the 95% confidence intervals (CI). Where possible, we calculated the number needed to treat/harm statistic (NNT/H) and used intention-to-treat analysis (ITT). MAIN RESULTS: We included seven studies with a total of 941 participants. Six studies were small with numbers of participants ranging between 56 and 83, and one study randomised 547 people. None of the studies had similar interventions and therefore they were analysed separately. One small Australian trial (n=59) was concerned with a phase-specific intervention (low dose risperidone and cognitive behavioural therapy) for people with prodromal symptoms. This group were significantly less likely to develop psychosis at a six month follow up than people who only received care from a specialised team which did not involve phase-specific treatment (n=59, RR 0.27 CI 0.1 to 0.9, NNT 4 CI 2 to 20). This effect was not significant at 12 month follow up (n=59, 1 RCT, RR 0.54 CI 0.2 to 1.3). A UK-based study (EDIE) randomised 60 people with prodromal symptoms, to cognitive behavioural therapy (CBT) or a monitoring group. Only two outcomes were reported: leaving the study early and transition to psychosis, both sets of data were non-significant. A Chinese trial used a phase-specific intervention (family therapy) plus out patient care trial for people in their first episode of psychosis and found reduced admission rates care compared with those who received only outpatient care (n=83, RR 0.28 CI 0.1 to 0.6, NNT 3 CI 2 to 6). The applicability of this finding was, however, questionable. One Dutch study (n=76) comparing phase-specific intervention (family therapy) plus specialised team with specialised team for people in their first episode of schizophrenia found no difference between intervention and control groups at 12 months for the outcome of relapse (n=76, RR 1.05 CI 0.4 to 3.0). The large Scandinavian study (n=547) allocated people with first episode schizophrenia to integrated treatment (assertive community treatment plus family therapy, social skills training and a modified medication regime) or standard care. Global state outcome GAF significantly favoured integrated treatment (n=419, WMD -3.71 CI -6.7 to -0.7) by one year, but by two years data were non-significant. Rates of attrition were significantly lower (n=547, RR 0.59 CI 0.4 to 0.8, NNT 9 CI 6 to 18) for integrated treatment by one and two year follow-up. PRIME (USA) was the only double blind study and allocated people with prodromal symptoms to olanzapine or placebo. No significant differences were found between olanzapine and placebo in preventing conversion to psychosis by about 12 months (n=60, RR 0.58 CI 0.3 to 1.2). Clinical Global Impression change scores 'severity of illness' were equivocal by 12 months. Scale of Prodromal Symptoms (SOPS) scores were also equivocal and the PANSS, total, positive and negative outcomes were non-significant. There were no significant differences between the olanzapine and placebo group on adverse effects rating scales - SAS, BAS and AIMS scores; Weight gain was significantly higher in the olanzapine group (n=59, WMD 7.63 CI 4.0 to 11.2) by 12 months. Finally one more Australian study included people in their first episode of psychosis who were acutely suicidal and allocated people to phase-specific cognitively orientated therapy or standard care. Outcome data for leaving the study early and suicide were equivocal. AUTHORS' CONCLUSIONS: We identified insufficient trials to draw any definitive conclusions. The substantial international interest in early intervention offers an opportunity to make major positive changes in psychiatric practice, but making the most of this opportunity requires a concerted international programme of research to address key unanswered questions.