Data collection and analysis
Data were extracted independently from included papers by at least two reviewers. Authors of trials were contacted for additional and missing data. Relative risks (RR) and 95% confidence intervals (CI) of homogeneous dichotomous data were calculated. A random effects model was used for heterogeneous dichotomous data. Where possible the numbers needed to treat (NNT) were also calculated. Weighted or standardised means were calculated for continuous data.