BACKGROUND: Migraine occurs in around 15% of adults and is ranked as the seventh most disabling disease amongst all diseases globally. Despite the available treatments many people suffer prolonged and frequent attacks which have a major impact on their quality of life. Chronic migraine is defined as 15 or more days of headache per month, at least eight of those days being migraine. People with episodic migraine have fewer than 15 headache days per month. Botulinum toxin type A has been licensed in some countries for chronic migraine treatment, due to the results of just two trials. OBJECTIVES: To assess the effects of botulinum toxins versus placebo or active treatment for the prevention or reduction in frequency of chronic or episodic migraine in adults. SEARCH METHODS: We searched CENTRAL, MEDLINE & MEDLINE in Process, Embase, ClinicalTrials.gov and World Health Organization International Clinical Trials Registry (to December 2017). We examined reference lists and carried out citation searches on key publications. We sent correspondence to major manufacturers of botulinum toxin. SELECTION CRITERIA: Randomised, double-blind, controlled trials of botulinum toxin (any sero-type) injections into the head and neck for prophylaxis of chronic or episodic migraine in adults. Eligible comparators were placebo, alternative prophylactic agent or different dose of botulinum toxin. DATA COLLECTION AND ANALYSIS: Two review authors independently selected trials and extracted data. For continuous outcomes we used mean change data when available. For dichotomous data we calculated risk ratios (RRs). We used data from the 12-week post-treatment follow-up time point. We assessed the evidence using GRADE and created two 'Summary of findings' tables. MAIN RESULTS: Description of trialsWe found 90 articles describing 28 trials (4190 participants), which were eligible for inclusion. The longest treatment duration was three rounds of injections with three months between treatments, so we could not analyse long-term effects. For the primary analyses, we pooled data from both chronic and episodic participant populations. Where possible, we also separated data into chronic migraine, episodic migraine and 'mixed group' classification subgroups. Most trials (21 out of 28) were small (fewer than 50 participants per trial arm). The risk of bias for included trials was low or unclear across most domains, with some trials reporting a high risk of bias for incomplete outcome data and selective outcome reporting.Botulinum toxin versus placeboTwenty-three trials compared botulinum toxin with placebo. Botulinum toxin may reduce the number of migraine days per month in the chronic migraine population by 3.1 days (95% confidence interval (CI) -4.7 to -1.4, 4 trials, 1497 participants, low-quality evidence). This was reduced to -2 days (95% CI -2.8 to -1.1, 2 trials, 1384 participants; moderate-quality evidence) when we removed small trials.A single trial of people with episodic migraine (N = 418) showed no difference between groups for this outcome measure (P = 0.49).In the chronic migraine population, botulinum toxin reduces the number of headache days per month by 1.9 days (95% CI -2.7 to -1.0, 2 trials, 1384 participants, high-quality evidence). We did not find evidence of a difference in the number of migraine attacks for both chronic and episodic migraine participants (6 trials, N = 2004, P = 0.30, low-quality evidence). For the population of both chronic and episodic migraine participants a reduction in severity of migraine rated during clinical visits, on a 10 cm visual analogue scale (VAS) of 3.3 cm (95% CI -4.2 to -2.5, very low-quality evidence) in favour of botulinum toxin treatment came from four small trials (N = 209); better reporting of this outcome measure from the additional eight trials that recorded it may have improved our confidence in the pooled estimate. Global assessment and quality-of-life measures were poorly reported and it was not possible to carry out statistical analysis of these outcome measures. Analysis of adverse events showed an increase in the risk ratio with treatment with botulinum toxin over placebo 30% (RR 1.28, 95% CI 1.12 to 1.47, moderate-quality evidence). For every 100 participants 60 experienced an adverse event in the botulinum toxin group compared with 47 in the placebo group.Botulinum toxin versus other prophylactic agentThree trials studied comparisons with alternative oral prophylactic medications. Meta-analyses were not possible for number of migraine days, number of headache days or number of migraine attacks due to insufficient data, but individually trials reported no differences between groups for a variety of efficacy measures in the population of both chronic and episodic migraine participants. The global impression of disease measured using Migraine Disability Assessment (MIDAS) scores were reported from two trials that showed no difference between groups. Compared with oral treatments, botulinum toxin showed no between-group difference in the risk of adverse events (2 trials, N = 114, very low-quality evidence). The relative risk reduction (RRR) for withdrawing from botulinum toxin due to adverse events compared with the alternative prophylactic agent was 72% (P = 0.02, 2 trials, N = 119).Dosing trialsThere were insufficient data available for the comparison of different doses.Quality of the evidenceThe quality of the evidence assessed using GRADE methods was varied but mostly very low; the quality of the evidence for the placebo and active control comparisons was low and very low, respectively for the primary outcome measure. Small trial size, high risk of bias and unexplained heterogeneity were common reasons for downgrading the quality of the evidence. AUTHORS' CONCLUSIONS: In chronic migraine, botulinum toxin type A may reduce the number of migraine days per month by 2 days compared with placebo treatment. Non-serious adverse events were probably experienced by 60/100 participants in the treated group compared with 47/100 in the placebo group. For people with episodic migraine, we remain uncertain whether or not this treatment is effective because the quality of this limited evidence is very low. Better reporting of outcome measures in published trials would provide a more complete evidence base on which to draw conclusions.