Data extraction
Three reviewers (LOB, SA, AOS) independently extracted data using a standardised collection form for all eligible studies, including study characteristics and reported outcomes for analysis. Where multiple publications using the same data existed, we included the most recent study, provided the earlier publication did not contain reported information not available in the most recent study. In studies in which adequate data was not reported, an effort was made to contact authors to provide us with additional information to allow us to compute effect estimates. Risk of bias assessment was undertaken by three reviewers (LOB, SA, AOS) using an appropriate quality assessment tool described by McDonald et al15, with all included studies classified as low to moderate risk of bias.