Nine studies were included in the review (n=1,924): three double-blinded randomised controlled trials (RCTs) (n=285) and six observational studies (n=1,639)
Treatment response: All analyses significantly favoured the intervention over controls for this outcome, but with substantial heterogeneity in most cases. When all nine studies were pooled, the odds ratio was 18.8 (95% CI 5.2 to 68.3, I2=76%; nine studies). The most marked effect was in RCTs (OR 70.8, 95% CI 16 to 314, I2=36%; three RCTs); restricting this analysis to RCTs of children reduced heterogeneity (OR 38.4, 95% CI 10.7 to 137.0, I2=8%). The odds ratio for observational studies was 7.8 (95% CI 2.1 to 28.9, I2=59%; six studies); restricting analysis to observational studies of adults reduced the odds ratio to 6.3 (95% CI 1.6 to 24.7, I2=60%).
Assessment for publication bias indicated that 223 statistically insignificant unpublished studies would be required to negate the statistical significance of these findings.
Clinical response: The intervention significantly lowered the risk of a clinical event (heart disease or mortality) (OR 0.29, 95% CI 0.16 to 0.53, I2=0%; two observational studies)
Adverse reactions: A median of 10% of participants (range 1% to 18%) discontinued treatment due to toxicity (most commonly cutaneous or gastrointestinal). Adverse reactions were less common in children than adults. Reporting was poor for this outcome.