In addition we found that efficacy of fasudil was best with doses of 10 mg/kg; however in the clinical trial of fasudil the dose used was 60 mg which equates to roughly 1 mg/kg and this was administered intravenously whereas the most frequent route of delivery was intraperitoneal in the preclinical literature. meta-analysis on publications describing the efficacy of RhoA and ROCK inhibitors in animal models of focal cerebral ischaemia where outcome was assessed as a change in lesion size or neurobehavioural score, or both. Results We identified 25 published papers which met our inclusion criteria. RhoA and ROCK inhibitors reduced lesion size by 37.3% in models of focal cerebral ischaemia (95% CI, 28.6% to 46.0%, 41 comparisons), and reduced neurobehavioural data by 40.5% (33.4% to 47.7%, 30 comparisons). Overall study quality was AZD9898 low (median=4, interquartile range 3C5) and steps to reduce AZD9898 bias were seldom reported. Publication bias was prevalent and associated with a substantial overstatement of efficacy for lesion size. Conclusions RhoA and ROCK inhibitors appear to be effective in animal models of stroke. However the low quality score, publication bias and limited number of studies are areas which need attention prior to conducting clinical trials. animal model of focal cerebral ischemia. We AZD9898 did not include studies which reported the effects of drugs known to inhibit molecules in the Rho pathway upstream of RhoA and ROCK. We included studies that reported the number of animals per group, AZD9898 outcome as a lesion size (infarct volume or infarct area; primary outcome) or a neurobehavioural score (secondary outcome) or both, and the mean and its variance (standard error of the mean (SEM) or standard deviation (SD)). Experiments with co-treatments were excluded. Data were extracted to the CAMARADES data manager. Quality assessment We assessed studies against the CAMARADES 10-item quality check list . One point was awarded for each of: (1) publication in a peer-reviewed journal; and reporting of: (2) control of heat, (3) random allocation to groups, (4) allocation concealment, (5) blinded assessment of outcome, (6) use of an anaesthetic without intrinsic neuroprotective activity, (7) the use of co-morbid animals, (8) performing a sample size calculation, (9) compliance with animal welfare regulations, (10) a statement of potential conflicts of interest. Data extraction We extracted data on study design including the time, route and dose of the drug administration, the species, sex and strain of the animal, the type of ischaemia (permanent, temporary or thrombotic), the anaesthetic and ventilation method used during the induction of injury and the method of quantification of lesion size. For each comparison on drug efficacy we extracted data on the number of animals per group, the mean outcome and the variance for both the control and treatment group. When a single control group was used for multiple treatment groups this was adjusted by dividing by the number of treatment groups served. Where data were not reported we made efforts to contact authors. Where data were reported graphically we used digital ruler software (Universal Desktop Ruler) and where data were expressed serially we extracted the final time point. Where it was not clear if the measure of variance was SD or SEM we extracted data as SEM, as for the purpose of meta-analysis this is a more conservative estimate. All data were extracted by a single, AZD9898 non-blinded, reviewer. Data analysis We deemed infarct volume and area to be sufficiently similar to be grouped into the same meta-analysis which we refer to as lesion size. We calculated a normalised mean difference effect size for each comparison (Vesterinen 0.004 for each of infarct volume and neurobehavioural scores. Publication bias was assessed using funnel plotting , Egger regression  COL3A1 and trim and fill . Results We identified 3,286 publications in our electronic search of which 3,237 were excluded in the first instance (513 duplicates and 2,724 publications which did not meet our inclusion criteria). We screened 49 publications in detail from which we excluded a further 24 publications (16 had no relevant outcome measures; four only reported outcomes measured outside the brain; two were abstracts later published in full; one did not use a relevant intervention; and one was a review). Our systematic review therefore included 25 articles published between 1992 and 2011 which met our inclusion criteria (24 full publications and one conference abstract; Additional file 1). We extracted data for 41 comparisons describing infarct.