The methodology and outcome measures of the included studies were similar and this allowed for pooling of results. All data were analyzed on an intention-to-treat basis. A random effect model was used as it provides a more conservative estimate of effect size.