Two review authors independently extracted the data and assessed the risk of bias. We contacted authors to obtain missing data where
needed. Where possible, results were presented in meta analyses using mean differences and standardized mean differences. Post-test
scores were used. In cases of significant baseline differences, we used change scores.