Twenty-seven placebo-controlled RCTs (5,313 participants reported and 5,423 calculated, range 21 to 705) and four active-controlled RCTs (210 participants) were included in the review. Jadad scores were reported to range from 2 to 8, but this did not seem to tally with the Cochrane Risk of Bias criteria. Drop-out rates ranged from 0% to 44%.
Compared to placebo, botulinum toxin A injections were associated with a statistically significant reduction in the frequency of chronic daily headaches (WMD -2.06 per month, 95% CI -3.56 to -0.56; Ι²=28.2%, three RCTs) and chronic migraine (WMD -2.30 per month, 95% CI -3.66 to -0.94; Ι²=32.2%, five RCTs). There were no differences compared to placebo in the number of episodic migraines per month and/or chronic tension headaches (16 RCTs) and there was some evidence of statistical heterogeneity.
Compared to active comparators, botulinum toxin A was associated with a significantly greater reduction in average headache severity in patients with chronic tension-type headaches (one RCT). No other outcomes were statistically different compared to active comparators.
Botulinum toxin A was statistically significantly more effective than placebo in achieving 50% improvement in chronic migraine headaches (RR 2.21, 95% CI 1.30 to 3.78; Ι²=0%, two RCTs). There were no differences for other types of headache.
Sensitivity analyses did not significantly alter the results. Reanalysis using the Higgins t distribution and a non-informative Bayesian model showed that none of the results were statistically significant.
Patients who received botulinum toxin A were more likely to experience adverse events. There was a placebo effect (placebo patients reported substantial improvements in headaches over time).
There was no evidence of publication bias.