One review author (DB) extracted data using a standardised, piloted data extraction form. A second review author (AV) checked this process by performing data extraction (independently) on a random set of studies and comparing the results. Any disagreements were resolved by discussion. In cases of more than 5% of disagreements with the random sample, two review authors performed the data extraction of all studies. Data were extracted on patient population, experimental and control interventions, and outcomes. All original data on outcomes were converted into effect estimates, which were reported as: a mean difference (MD) when a continuous outcome was measured on comparable instruments in the included studies, a standardised mean difference (SMD) when a continuous outcome was measured on different instruments in the included studies, or relative risk (RR) when the outcome was dichotomous. Each of these estimates was reported with a 95% confidence interval (CI) whenever possible. An effect of 15% or more was considered to be clinically relevant. Statistical analyses were performed using Review Manager 5.2