People presenting with agitated or violent behaviour thought to be due to severe mental illness may require urgent pharmacological tranquillisation. Several preparations of olanzapine, an antipsychotic drug, are now being used for management of such agitation. To estimate the effects of intramuscular, oral-velotab, or standard oral olanzapine compared with other treatments for controlling aggressive behaviour or agitation thought to be due to severe mental illness. We searched the Cochrane Controlled Trials Register (Issue 1, 2002), The Cochrane Schizophrenia Group's Register (November 2004) and reference lists. We contacted authors of trials and the manufacturers of olanzapine. Randomised clinical trials comparing oral-velotab or intramuscular, or standard oral olanzapine to any treatment, for agitated or aggressive people with severe mental illnesses. We reliably selected, quality assessed and data extracted studies. For binary outcomes we calculated a fixed effects Risk Ratio (RR) and its 95% Confidence Interval (CI) with a weighted Number Needed to Treat/Harm statistic (NNT/H). For continuous outcomes, we preferred endpoint data to change data and synthesised non-skewed data from valid scales using a weighted mean difference (WMD). Four trials compared olanzapine IM with IM placebo (total n=769, 217 allocated to placebo). Fewer people given olanzapine IM had 'no important response' by 2 hours compared with placebo (4 RCTs, n=769, RR 0.49 CI 0.42 to 0.59, NNT 4 CI 3 to 5) and olanzapine IM was as acceptable as placebo (2 RCTs, n=354, RR leaving the study early 0.31 CI 0.06 to 1.55). When compared with placebo, people given olanzapine IM required substantially fewer additional injections following the initial dose (4 RCTs, n=774, RR 0.48 CI 0.40 to 0.58, NNT 4 CI 4 to 5). Olanzapine IM did not seem associated with extrapyramidal effects (4 RCT, n=570, RR experiencing any adverse event requiring anticholinergic medication in first 24 hours 1.27 CI 0.49 to 3.26). Two trials compared olanzapine IM with haloperidol IM (total n=482, 166 allocated to haloperidol). Studies found no differences between olanzapine IM and haloperidol by 2 hours for the outcome of 'no important clinical response' (2 RCTs, n= 482, RR 1.00 CI 0.73 to 1.38) neither was there a difference for needing repeat IM injections (2 RCTs, n=482, RR 0.99 CI 0.71 to 1.38). More people on haloperidol experienced akathisia over the five day oral period compared with olanzapine IM (1 RCT, n=257, RR 0.51 CI 0.32 to 0.80, NNT 6 CI 5 to 15) and fewer people allocated to olanzapine IM required anticholinergic medication by 24 hours compared with those given haloperidol IM (2 RCTs, n= 432, RR 0.20 CI 0.09 to 0.44, NNT 8 CI 7 to 11). Two trials compared olanzapine IM with lorazepam IM (total n=355, 119 allocated to lorazepam). For the outcome of 'no important clinical response' , there was no difference between people given olanzapine IM and those allocated to lorazepam at 2 hours (2 RCTs, n=355, RR 92 CI 0.66 to 1.30) but fewer people allocated to olanzapine IM required additional injections by 24 hours compared with those on lorazepam IM (2 RCTs, n=355, RR 0.68 CI 0.49 to 0.95, NNT 10 CI 6 to 59). People receiving IM olanzapine were less likely to experience any treatment emergent adverse events, than those on lorazepam (1 RCT, n=150, RR at 24 hours 0.62 CI 0.43 to 0.89, NNT 5 CI 4 to 17) and over the same time period there were no clear differences in the use of anticholinergic medication between groups (1 RCT, n=150, RR 1.16 CI 0.38 to 3.58).No studies reported outcomes related to hospital and service use. Nor did any report on issues of satisfaction with care or suicide, self-harm or harm to others. No studies evaluated the oro-dispersable form of olanzapine. Data relevant to the effects of olanzapine IM are taken from some studies that may not be considered ethical in many places, all are funded by a company with a pecuniary interest in the result. These studies often poorly report outcomes that are difficult to interpret for routine care. Other important outcomes are not recorded at all. Nevertheless, olanzapine IM probably has some value in helping manage acute aggression or agitation, especially where it is necessary to avoid some of the older, better, known treatments. Olanzapine causes fewer movement disorders than haloperidol and more than lorazepam. The value of the oro-dipersable velotab preparation is untested in trials. There is a need for well designed, conducted and reported randomised studies in this area. Such studies are possible and, we argue, should be designed with the patient groups and clinicians in mind. They should report outcomes of relevance to the management of people at this difficult point in their illness.