BACKGROUND: It has been estimated that violence accounts for more than 1.6 million deaths worldwide each year and these fatal assaults represent only a fraction of all assaults that actually occur. The problem has widespread consequences for the individual and for the wider society in physical, psychological, social and economic terms. A wide range of pharmacological, psychosocial and organisational interventions have been developed with the aim of addressing the problem. This review was designed to examine the effectiveness of these interventions when they are developed in mental health and criminal justice populations. OBJECTIVE: To update a previous review that examined the evidence base up to 2002 for a wide range of pharmacological, psychosocial and organisational interventions aimed at reducing violence, and to identify the key variables associated with a significant reduction in violence. DATA SOURCES: Nineteen bibliographic databases were searched from January 2002 to April 2008, including PsycINFO (CSA) MEDLINE (Ovid), Cumulative Index to Nursing and Allied Health Literature (CINAHL), Allied and Complementary Medicine Database (AMED), British Nursing Index/Royal College of Nursing, International Bibliography of the Social Sciences (IBSS), Education Resources Information Center (ERIC)/International ERIC, The Cochrane Library (Cochrane reviews, other reviews, clinical trials, methods studies, technology assessments, economic evaluations), Web of Science [Science Citation Index Expanded (SCIE), Social Sciences Citation Index (SSCI), Arts & Humanities Citation Index (A&HCI)]. REVIEW METHODS: The assessment was carried out according to accepted procedures for conducting and reporting systematic reviews, including identification of studies, application of inclusion criteria, data extraction and appropriate analysis. Studies were included in meta-analyses (MAs) if they followed a randomised control trial (RCT) design and reported data that could be converted into odds ratios (ORs). For each MA, both a fixed-effects model and a random-effects model were fitted, and both Q statistic and I2 estimates of heterogeneity were performed. RESULTS: A total of 198 studies were identified as meeting the inclusion criteria; of these, 51 (26%) were RCTs. Bivariate analyses exploring possible sources of variance in whether a study reported a statistically significant result or not, identified six variables with a significant association. An outcome was less likely to be positive if the primary intervention was something other than a psychological or pharmacological intervention, the study was conducted in an penal institution, the comparator was another active treatment or treatment as usual and if a between-groups design had been used. An outcome was more likely to be positive if it was conducted with people with a mental disorder. The variation attributable to these variables when added to a binary logistic regression was not large (Cox and Snell R(2) = 0.12), but not insignificant given the small number of variables included. The pooled results of all included RCTs suggested a statistically significant advantage for interventions over the various comparators [OR 0.59, 95% confidence interval (CI) 0.53 to 0.65, fixed effects; OR 0.35, 95% CI 0.26 to 0.49 random effects, 40 studies]. However, there was high heterogeneity {I(2) = 86, Q = 279 [degrees of freedom (df) = 39], p < 0.0001}, indicating the need for caution in interpreting the observed effect. Analysis by subgroups showed that most results followed a similar pattern, with statistically significant advantages of treatments over comparators being suggested in fixed- and/or random-effects models but in the context of large heterogeneity. Three exceptions were atypical antipsychotic drugs [OR 0.21, 95% CI 0.16 to 0.27, fixed effects; OR 0.24, 95% CI 0.14 to 0.43, random effects; 10 studies, I(2) = 72.2, Q = 32.4 (df = 9), p < 0.0001], psychological interventions [OR 0.63, 95% CI 0.48 to 0.83, fixed effects; OR 0.53, 95% CI 0.31 to 0.93, random effects; nine studies, I(2) = 62.1, Q = 21.1 (df = 8), p = 0.007] and cognitive behavioural therapy (CBT) as a primary intervention [OR 0.61, 95% CI 0.42 to 0.88, fixed effects; OR 0.61, 95% CI 0.37 to 0.99, random effects; seven studies, I(2) = 21.6, Q = 7.65 (df = 6), p = 0.26]. LIMITATIONS: The heterogenity of the included studies inhibits both robust MA and the clear application of findings to establishing improvements in clinical practice. CONCLUSIONS: Results from this review show small-to-moderate effects for CBT, for all psychological interventions combined, and larger effects for atypical antipsychotic drugs, with relatively low heterogeneity. There is also evidence that interventions targeted at mental health populations, and particularly male groups in community settings, are well supported, as they are more likely to achieve stronger effects than interventions with the other groups. Future work should focus on improving the quality of evidence available and should address the issue of heterogenity in the literature. FUNDING: The National Institute for Health Research Health Technology Assessment programme and the Research for Patient Benefit programme.