Heterogeneity (I 2) and publication bias (Harbord's small-study effects) were formally tested. Results 25 trials, including 26 610 participants with 3464 injuries, were analysed. The overall effect ...