Cumulative evidence of helmet effects on bicycle injuries

Background: With widely-used use of bicycles, studies of helmet effects on bicycle injuries were intensively conducted after the problem of bicycle injuries emerged. This study aims to justify whether current evidence is sufficient to manifest the effects of helmets. MethodsWe exhaustively searched the articles in the databases of Medline, Scopus, and Embase by the term of (helmet* AND (cycl* OR bicycle* OR bike*)) AND injur* by the time of April 10, 2019. The meta-analysis and SSA (study sequential analysis) were conducted. ResultsA total of 55 studies are eligible for meta-analysis. The OR (odds ratio) of helmet effect on head injuries compared with other injuries is 0·50 (0·43, 0·59) and effect of helmets on serious head injuries compared with other injuries are protective with OR of 0·34 (0·28, 0·43). Compared with control injuries, the OR of helmet effect on facial injuries is 0·63 (0·45, 0·88). Helmets is not associated with protective effects of neck injuries and OR is 0·98 (0·82, 1·17). SSA results of head injuries, serious head injuries, and face injuries showed, the cumulative Z-curve crossed both the conventional and the trial sequential monitoring boundary. SSA results of neck injuries showed the cumulative Z-curve does not cross both the conventional and the trial sequential monitoring boundary. ConclusionsThe helmet has protection effects on head injuries, serious head injuries, and face injuries. The SSA showed the current evidence was sufficient to support the results. More studies of helmet promotion are warranted in the future.

Meanwhile, the concomitant bicycle-related injuries have become a global public health concern. In the US, bicycle usage conferred 1000 deaths and 467000 injuries in the latest report in 2015. 2 China, with the largest number of bicycle users, had more than 2·5 million injuries and 50000 fatalities related to bicycles from 2004 to 2010. 3 In Netherland where the per capital bicycle ownership is as high as 1, 456 bicycle-related injuries per 100 thousand were reported in 2012. 4 To prevent bicycle-related injuries, helmet wearing is considered an effective measure.
Research has been conducted to investigate its protective effects on various injury types including head, brain, face, and neck injuries. In 1989, Thompson et al. conducted a casecontrol study, reporting that helmets reduced 85% of head injury risk and 88% of brain injury risk. 5 In 2000, Thompson et al. initiated the first systematic review of helmet effects with inclusion of only prospective studies and medical certificates of injuries, and found that the risk reduction provided by helmets on head, brain, and severe brain injuries ranged from 63-88%. 6 The first meta-analysis including 16 studies by Attewell et al. in 2001 showed that helmets reduced the risk of head injury, brain injury, facial injury, and fatal injury by 60%, 58%, 47%, and 73%. 7 Later, Elvik et al. re-analyzed the work of Attewell et al. through evaluating publication by trim-and-fill method, estimating time trend bias, correcting the zero-count cell, and updating the meta-analysis with new publications in 2011. 8,9 With the concerns of usage of trim-and-fill method and the heterogeneity of different injuries which should be evaluated separately, Olivier et al. performed a meta-analysis including 40 studies in order to take into account sources of bias in 2017. It is found the efficacy of helmets reduce the risk of head injury by 51%, of serious head injury by 69%, of fatal head injury by 65%, and of face injury by 33%, The evidence did not support the helmet protection effects on neck injury. 10 Høye et al replicated the results of previous studies with inclusion of newly published data in 2018. The article investigated multiple confounding variables such as influence of alcohol, cyclists' age, and speed limit. 11 It also analyzed the influence of data resources and the moderators such a helmet usage rate and crash type.
The helmet effects of reduction of head injuries and face injuries are 48% and 23%, respectively, which showed significant effects. The reduction of neck injuries was not statistically significant.
Though the helmet protection effects on head, brain, and face injuries are conclusive, it is unclear that whether the evidence is sufficient to support current effects of helmet in the addition of new publications and whether we should draw more attention from manifesting effects of helmets on other aspects to improve bicycle safety. 8,9 With the questions for previous literature gap, this meta-analysis aimed to evaluate the effects of helmet on risk of bicycle injuries, and explore whether the evidence of current studies are sufficient and conclusive to support the effects of helmets.

Methods
The checklist PRISMA for systematic reviews and meta-analysis was followed during the process of implementation and report. The protocol was registered in Prospero (www.crd.york.ac.uk/PROSPERO/, ID: CRD42019131751).

Study Eligibility, Search Strategy, And Selection Criteria
The study inclusion criteria were as the following: the studies evaluated or compared bicycle-related injuries; the studies reporting individual cyclists' injuries of head, brain, face, and neck with medical diagnosis (studies with self-report injuries were excluded); the studies with details to complete 2 × 2 table of injury comparison by presence or absence of helmet; and studies were English language publications in peer review journals. Three electronic databases (Medline, Scopus, and Embase) were exhaustively searched for the articles and reports on April 10, 2019. Previous review article references were served as an additional source. Search terms were used (helmet* AND (cycl* OR bicycle* OR bicycle*)) AND injur*, to include as many studies as it could. Only the articles published in English were included. Two authors independently reviewed the titles and abstracts of the obtained studies and made first evaluation of inclusion or exclusion. Next, the studies with discrepancies at the first step were retrieved with full-text review in order to judge whether they met eligibility criteria. Disagreements between the two authors were resolved through discussion with the third author. Figure 1 presents the study selection process. One author gleaned information (PMID, author name, country, study design, sample size, 2 × 2 table of injuries of different parts, the ORs if provided, the type of helmet, and other risk behavior factors) from the included articles and summary the information into Table 1. A second author checked the quality and accuracy of the data.

Study Sequential Analysis
Study sequential analysis (SSA) was conducted to determine whether the sample size included in the meta-analysis was sufficient for manifesting the effect size of helmets. For the SSA, when the Z-curve crosses the conventional boundary, a significant difference is considered which means the effect size of helmets is significant. Moreover, if the Z-curve passes through the trial sequential monitoring boundary or required information size (RIS) boundary, it indicates the evidence of meta-analysis is sufficient and conclusive to support the effect size of helmets. Otherwise, the evidence is rendered inconclusive and more studies were warranted to verify the effect size.
In our study, we also conducted subgroup analysis about the helmet effects under different legislation status and different helmet wearing rates. The heterogeneity did not decrease after subgroup analysis. The source of heterogeneity could be the selection of controls which are undefined other kinds of injuries rather than the use status of helmet.
For the head injuries and serious head injuries, in different situations of helmet wearing rates and mandatory helmet legislation, the protection effects remain significant. For these two types of injuries, with the increase of helmet wearing rate, the effect sizes of helmet protection consistently increase. From the situation of legislation absence to mandatory legislation, the effect sizes of helmet protection decrease. Currently, we could not confirm the association between helmet use status and effect size of helmet protection and we could not observe the similar situation in face injuries and neck injuries.
However, future studies could further investigate whether helmet wearing rate and legislation will influence helmet protection effect size and the underlying mechanisms.
Moreover, in subgroup analysis, we also conducted SSA analysis which indicated the evidence was sufficient to support the results in subgroup analysis.
Other factors will also influence the effects of helmets such the texture of the helmets. It is reported that the soft texture and hard texture of helmets may have different effects in protection. The specific design of the helmet such as chin bar will also influence the function. 65 The magnitude of protection effects of hard helmets is larger than that of soft helmets, even some studies showed soft helmets may increase the incidence of neck injuries. 8 However, the application of soft helmets become more and more widely. 66 Other studies believed the variation in the effects of soft helmets and hard helmets is because of different definitions of injuries in different meta-analysis. 27

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.