What the systematic reviews say about bicycle helmets.

What these numbers measure

The reviews below compare cyclists who were injured in a crash while wearing a helmet with those who were not. The Cochrane review and the Attewell meta-analysis pooled case-control studies, and the Cochrane authors found no randomised trial. Their results are usually given as an odds ratio (OR): a figure below 1 means lower odds of that injury among helmet wearers, and the further below 1, the larger the association. An OR of 0.49, for example, is often expressed as a reduction of about 51 percent.

Two limits apply to all of them. They measure injury given that a crash happened, not the chance of crashing in the first place. And because they compare wearers with non-wearers, they depend on how alike those groups are, which is one of the points disputed below.

The reviews, one by one

Thompson, Rivara and Thompson: the Cochrane review (1999)

D. C. Thompson, F. Rivara and R. Thompson's Cochrane review was first published on 25 October 1999; its literature search was last updated in November 2006, and a search-result summary of the Cochrane Library record says it was edited without changing its conclusions and republished in 2010. The authors found no randomised trials and included five case-control studies. They reported that helmets reduced the risk of head, brain and severe brain injury by 63 to 88 percent for cyclists of all ages. Protection was similar in crashes with motor vehicles (69 percent) and in other crashes (68 percent). Injuries to the upper and middle face fell by 65 percent, but helmets did not prevent lower-face injuries.

Attewell, Glase and McFadden (2001)

This meta-analysis in Accident Analysis and Prevention pooled 16 case-control studies with data from 1987 to 1998. As tabulated in the 2023 review by Büth and colleagues, it reported odds ratios of 0.40 for head injury, 0.42 for brain injury, 0.27 for fatal head injury and 0.53 for face injury. For neck injury it reported a raised odds ratio of 1.36, with a 95 percent confidence interval of 1.00 to 1.86, a borderline result.

Olivier and Creighton (2017)

Jake Olivier and Prudence Creighton's review in the International Journal of Epidemiology (online 2016, in print 2017) pooled 40 studies. They found helmet use associated with lower odds of head injury (OR 0.49), serious head injury (OR 0.31), fatal head injury (OR 0.35) and face injury (OR 0.67), and no association with neck injury (OR 0.96). UNSW summaries express these as reductions of about 51, 69, 65 and 33 percent. According to Büth and colleagues, they found no strong evidence of publication bias and, after adjusting for possible risk-taking by helmet wearers, got nearly identical results.

Høye (2018)

Alena Høye's meta-analysis in Accident Analysis and Prevention pooled 55 studies. As tabulated by Büth and colleagues, it reported odds ratios of 0.47 for head injury, 0.40 for serious head injury, 0.29 for fatal head injury and 0.77 for face injury, with no clear effect on cervical spine injury (OR 1.09, confidence interval 0.90 to 1.32). Høye applied a statistical adjustment (trim and fill) where the funnel plots were asymmetric, and stated that it remains unclear whether helmet wearing changes behaviour for better or worse.

Büth, Barbour and Abdel-Aty (2023): a review of the reviews

Carlson Moses Büth, Natalia Barbour and Mohamed Abdel-Aty's systematic review of meta-analyses (Scientific Reports, 26 May 2023) concluded that wearing a helmet while cycling is beneficial regardless of age, crash severity or crash type, with the greatest benefit for severe head injuries and in high-risk settings such as shared roads. It did not examine the risk-compensation question itself, pointing instead to a separate review by Esmaeilikia and colleagues that found little or no support for the idea that helmet wearers take more risks. It also described Olivier and Radun's 2017 paper, "Bicycle helmet effectiveness is not overstated", as a methodological re-analysis arguing that case-control results are the best evidence available.

The dispute over the size of the effect

Elvik (2011, corrected 2013). Rune Elvik re-analysed the Attewell meta-analysis in Accident Analysis and Prevention in 2011 and concluded it was affected by publication bias and by time-trend bias, so that it overstated the benefit of helmets; his adjusted results showed smaller benefits. In 2013 he published a corrigendum correcting errors in his data and analysis, and kept the conclusion that the 2001 figures were overstated. We read Elvik's conclusions from summaries of the papers, not the full texts.

The other view. Olivier and Creighton's later review found no strong sign of publication bias, and Olivier and Radun argued in 2017 that helmet effectiveness is not overstated. Büth and colleagues left Elvik's re-analysis out of their 2023 review, saying it added only a few studies to Attewell's and that Olivier and Creighton had since examined those studies in more detail.

Clarke (2017). Colin Clarke, writing in World Transport Policy and Practice, argued that the headline figures are unreliable because of weaknesses in the evidence and method, especially the comparison of wearers with non-wearers whose crash risks may differ. We read this from a summary of the paper.

Neck injury. Cycling UK's June 2019 evidence briefing cites evidence that helmets may raise the risk of neck injury. Attewell (2001) reported a borderline raised odds ratio; Olivier and Creighton (2017) and Høye (2018) found no association.

In short, the reviews agree that helmet wearers who crash have lower odds of head injury. How large that reduction is remains disputed. Whether laws or campaigns that raise helmet wearing produce a benefit for a whole population is a separate question, covered on the helmet law debate page.

What official bodies say

• Highway Code, Rule 59 (Great Britain, page updated 22 October 2025). The Code advises, but does not require, a helmet that meets current regulations, is the right size and is securely fastened, and says evidence suggests a correctly fitted helmet will reduce the risk of head injury in certain circumstances.
• NHTSA (July 2026). In its 2024 data fact sheet, NHTSA states that all bicyclists should wear properly fitted helmets every time they ride, and calls a helmet the single most effective way to prevent head injury in a bicycle crash.

Rotational-impact systems such as MIPS are makers' technologies, and any benefit stated for them is the maker's claim; the Insurance Institute for Highway Safety described MIPS in 2018 as a low-friction layer that lets the head slide relative to the helmet. What helmets are tested to is on the helmet standards page, and independent laboratory ratings, with their method and year, are on the independent helmet tests page.