Background Past meta-analyses of studies assessing bicycle helmet efficacy have been criticised for poor methodology and the literature has not been systematically reviewed in over 15 years. The most recent meta-analysis reported time trend and publication biases, and found the summary odds ratio (OR) diminished when combining head, face and neck injuries. However, this study did not use standard methodology to identify biases, did not systematically review the literature, and the heterogeneity among studies reporting different injury outcomes was not assessed. The aim of this study is to systematically review and summarise results from studies assessing bicycle helmet efficacy to mitigate head, face and neck injury.

Methods Four electronic databases were searched for relevant, peer-reviewed articles in English. Included studies reported medically diagnosed head, face or neck injuries, other cycling injuries and helmet usage. Non-approved helmets were excluded where possible. Summary ORs were obtained using mixed effects models stratified by injury type and severity. Time trends were tested using cumulative models and mixed models with time as a moderator. Evidence of publication bias was assessed using funnel plot methods.

Results Study is ongoing with 53/70 studies assessed. Early results suggest bicycle helmets were associated with reduced odds of head and facial injuries, with the strength of association greater for more severe head injuries. Stratification by injury type and severity reduced heterogeneity. Early analyses do not suggest publication bias and no time effects were found from 1998 onwards.

Conclusions A systematic search of the literature is essential for meta-analysis, especially when assessing publication bias. Inadequate assessment of heterogeneity among included studies partly accounts for discrepancies in previously reported results. We found helmets were associated with significant reductions in head injury for cyclists injured in a crash.