It sounds like common sense: wear something on your head and it’ll stop you from getting hurt. But, as with so many other things, the value of bicycle helmets in the real world – and particularly whether people should be made to wear them – turns out to be more complicated than that.
Three different types of bicycle helmets. From left to right, an Australian Stackhat from the 1980s, a soft-shell helmet and a modern hard-shell helmet (from Expo 2010 Denmark). Images from Wikimedia Commons.
Can science help? Of course it can! There’s been so much research with so many different results that science can help support nearly any opinion you like. Consensus, not so easy.
However, from reading of some of the most-cited studies (see below), I’ve reached the following conclusions:
- Compulsory helmet legislation has not been proven to reduce the likelihood of head injuries. There are many theories why, but the important point is that there are lots of other road safety measures that are more effective.
- Even so, wearing a helmet still seems to reduce your own, individual chance of head injury. How much is hard to say, and it depends on you not riding more dangerously to compensate, but it’s still a good idea for you, yourself.
Also, there are indications that part of the problem is that the modern, soft foam helmets (the middle photo in the image above) are not as good as hard shell helmets, like the old Stackhats or the new hard ones that hipsters wear. So maybe get one of those too.
But also be aware that, helmet or no helmet, cycling is still a relatively safe activity.
In Victoria in 1989, the year before the helmet law was introduced, there were 8,502 fatalities and hospital admissions from non-cycling road injuries, but only 430 cyclists with head injuries admitted to hospital. Taking into account the number of cyclists, this works out to 1 head injury per 9,302 cyclist-years. Whereas for non-cycling injuries, the rate is 1 death or hospital admission per 517 person-years (Robinson 2007, ref. 4 below).
Click through after the jump for an overview of the actual research papers (listed from most to least recent), and see if you agree.
Where possible, I’ve linked to full text copies of the papers. But this itself has led to an interesting result. Cycling websites overwhelmingly post only the anti-helmet papers, whereas pro-helmet research is more likely to appear on government websites. I believe this phenomenon is known as cherry picking.
A re-analysis of published case control trials (you can’t do randomised controlled trials of bike accidents, so instead researchers use case control studies, which compare people who have a condition – in this case, a head injury, with people who don’t). By changing the criteria for which trials are included, and adding some more recent ones, the apparent benefit of helmets is less than found in other analyses, like 15 and 18 below. In particular, helmets don’t seem to prevent neck injuries.
Looked at the ratio of head to arm injuries in hospital admissions in New South Wales, expecting to see a reduction if helmet laws were effective. Although the ratio did reduce, it was already falling before the laws were introduced, so the effect was probably due to other factors.
A meta-analysis of the effect of compulsory bike helmets for children, which found they reduced deaths and head injuries, as well as the rate of helmet use (unsurprisingly). But they didn’t look at whether overall numbers of cyclists had declined, which may be a factor in injury reduction.
An Australian statistician, Dorothy Robinson is one of the key researchers in this field. One of the main concerns she raises in this paper is whether helmet legislation leads to a drop in people riding, especially because of the belief that increasing the number of cyclists is a good way to reduce injuries – the “safety in numbers” concept. She also discusses how the cost-benefit ratio of helmet laws is worse compared with other initiatives like speed or drink-driving reduction, or fixing accident blackspots.
Part of an ongoing exchange with the authors of the Cochrane review (reference 18 below), this paper argues that there is insufficient evidence to prove that helmets prevent brain injury. William Curnow’s chief concern, as we shall see, seems to be that rotational injuries are more dangerous than than the direct, head-on collisions that helmets are designed for; and that in fact, helmets may exacerbate the risk of rotational damage.
Another part of the exchange mentioned directly above, this paper argues that the case control studies used in analyses like the Cochrane review (reference 18) are appropriate for answering the question of whether helmets prevent head and brain injury.
In 2006 the British Medical Journal published two articles in favour of and against helmet laws. On the “against” side was Dorothy Robinson, who accepted that case control studies showed a benefit, but maintained that legislation did little else but discourage cycling, possibly due to a number of confounding factors. Instead, she argued, governments should focus on things like speeding, drink-driving, road design, adherence to road rules and cycling without lights at night.
The “for” side maintained that, despite all the confounding factors, statistics still showed that helmets reduced the proportion of head injuries, and that this conclusion has been supported by a number of independent studies. There was also a lively discussion in replies to the article.
Another article in the Curnow discussion, this one quite simply stating that it’s wrong to assume that all types of helmets are effective against brain injury.
Getting closer to the start of this discussion, this is a response to Curnow’s initial criticism of the Cochrane review (reference 18), questioning some of his assertions.
Curnow’s first, direct attack on the Cochrane review (reference 18), basically claiming that because we don’t fully understand the theory of how helmets protect the brain – especially with regard to rotational injuries -we can’t conclude that they work no matter what case control studies show. The Cochrane authors, on the other hand, believe that their studies of real incidents speak for themselves.
Another paper of Curnow’s, also examining the theoretical mechanisms of brain injury, going through a number of situations where he believes they’re not effective. He particularly criticises the soft foam helmets, designed to disintegrate on impact.
Another paper critical of soft-shell helmets. The authors performed a case control study in Norway, with 991 injured patients. They too found that hard shell helmets reduced the risk of injuries to the head, while foam helmets seemed to increase the risk of face injuries.
An interesting paper looking at the cost to society of purchasing helmets, compared to the healthcare savings of any drop in head injuries. They basically found that, at the level of the whole country, the costs aren’t worth it (at least for adults). Although, you have to wonder how well that translates to risk and choice at the individual level.
This study simply looked at whether head injuries in children had reduced in the Canadian provinces that had introduced helmet laws. They found there was a reduction, but as far as I know they didn’t address the oft-repeated question of whether that was due to a reduction in cycling.
A meta-analysis of case control studies, re-examined in our first reference by Rune Elvik. This study claimed to find clear evidence that helmets prevent serious injury and death, and even suggested that lighter helmets are better for neck injuries. They went on to recommend that helmets should be used as much as seat belts in cars.
Another study that found that helmets laws had led to a decrease in head injuries – in this case, a 19% drop in 3 years. Again though, it’s not clear whether this is because fewer people are cycling.
Another meta-analysis, similar to reference 15 above, probably because it’s by the same authors. But at least the full text of this one is available for free.
At last, the Cochrane review that sparked all the controversy with William Curnow and others above. The Cochrane Collaboration, as discussed before on Lost in Science, is an independent group that performs systematic reviews or meta-analyses of medical research. But this particular review has been criticised heavily, partly because of its trial criteria (remembering that unlike most Cochrane reviews, in this case it’s not possible to use randomised controlled trials), but also because the authors of the review have themselves performed some of the included studies.
Despite those concerns, this is a very thorough review that makes its analysis very clear, and demonstrates how they’ve reached conclusions of the protective benefit of helmets. They also have an extensive discussion section, in which the authors respond convincingly to the criticism.
But it’s also important to remember that this study, and those that it references, only address the question of whether helmets protect individuals against head injury. It can’t really prove that helmet legislation is effective as a population safety measure, although the authors do draw their own conclusions.
Another of Dorothy Robinson’s statistical analyses, where she claims that there is “little or no obvious effect in hospital data” of a benefit from wearing helmets. She still doesn’t declare they are useless, but as usual she points out that other road safety campaigns can be more effective.
Similar research, published around the same time as reference 19. This claims that the most notable effect of helmet laws is to reduce cycling – here she uses cyclist counts from New South Wales and Victoria. But even for this claim there is very little data, and complicating factors that mean adjustments have to be made to see a trend. Unfortunately, although today there are better statistics for cycling numbers, we can’t go back in time to re-do the figures from when legislation was introduced. (Or can we?)
Another fascinating study, this time from the renowned Monash University Accident Research Centre. Using a lot of statistical adjustments, researchers claimed to find Victorian helmet laws had caused a 40% decrease in the rate of head injuries, compared with what they would have been without the laws. Even so, the numbers considered are very small: 342 injuries in 1993/94, down from 490 in 1987/88.
Finally, another study of Victorian helmet laws, also finding a significant drop in injuries. This study did also consider the changes in overall numbers of cyclists: their figures showed a 36% drop in children riding bikes in the first year of the law, but a 44% increase in adult cyclists. So even as straightforward a figure as that cannot be agreed in this very contentious argument.
Thanks for fab and well informed discussion of the issue. Although I am probably just as unsure of whether helmets are a good idea, it has confirmed that my crappy helmet (yes, the one pictured in the middle) should be upgraded… seeing as it’s the law to wear a helmet, might as well wear one that’s going to be the most effective. (Though I worry I am not cool (or rich) enough to buy one of the ‘hipster’ ones. May have to source a Stackhat.)
Please note that reference 2 listed above (Voukelatos A & Rissel C, “The effects of bicycle helmet legislation on cycling-related injury: The ratio of head to arm injuries over time”, Journal of the Australasian College of Road Safety, August 2010) has been found to contain serious data and arithmetic errors and was formally withdrawn in Feb 2011 by the journal which published it.
See
http://www.smh.com.au/nsw/authors-admit-errors-in-study-on-bike-helmets-and-head-injuries-20101229-19a9x.html
and the formal retraction of the paper on page 39 of http://www.acrs.org.au/srcfiles/ACRS-Journal-22No1WebLR2.pdf
Tim Churches failed to tell you that even when the error was corrected, it didn’t effect the outcome of the research by Voukelatos A & Rissel C. Tim is a helmet zealot that clings to this error in order to cast doubt on the paper. Nice try TIm. Do you have anything new these days? Choice works everywhere else in the world, 98% of it in fact. Why not here? Hope you wear a helmet when you drive a car or a put on a lifejacket when you go to the beach… or do you enjoy the fact that you have the freedom to choose?
As far as I am aware, Voukelatos and Rissel have never corrected the errors in their paper. Here are the details from the editorial note announncing the retraction of the paper (p39 at
http://acrs.org.au/wp-content/uploads/ACRS-Journal-22_No1WebLR1.pdf ):
“After much deliberation, the journal editors
have decided to formally retract the publication by A
Voukelatos and C Rissel, ‘The effects of bicycle helmet
legislation on cycling-related injury: The ratio of head to arm
injuries over time’, published in the August 2010 issue of the
journal. This decision was made in compliance with the
guidelines provided by the Committee on Publication Ethics
(COPE) as ratified by the ACRS Executive Committee on 18
November 2010
(http://www.acrs.org.au/publications/journalauthorguidelines.html)
Retraction of the paper is made for the following reasons:
a) The authors had been given the opportunity to provide a
response to the Tim Churches letter, and had done so.
b) The authors’ response was sent out for peer review to five
independent reviewers: three Australians, one American and
one international reviewer from Germany. The reviewers’
qualifications range across the professions of psychology,
engineering, medicine and science, while their extensive
expertise ranges across the areas of epidemiology, biostatistics,
cycling safety, transport engineering, hospital and crash
databases, and crash investigations. As a result of the review the
authors were asked to further revise their response.
c) This revised response was again sent to the peer reviewers,
but was found to still contain serious errors: it contained data
errors (incorrect ICD-9-CM codes used); it excluded the first
year of data from the original paper without good reason; it
still had graphing errors (RTA survey data still shown in wrong
place on graph); it failed to implement simple but essential
adjustments (sample weighting and exclusion of hospital
transfers), which are routinely done for analysis of such data;
and it introduced new data (on cycling fatalities), which was
not in the original paper and which was inappropriate to
include in such a correction.
In retracting this paper, the journal is not trying to stifle
scientific debate; however, in the absence of a response from the
authors that addresses reviewers’ concerns – in effect, that is free
of data errors and that has no basic methodological flaws – the
journal has no choice but to retract the paper and apologises for
any inconvenience this has caused.
The authors have been offered the opportunity to submit a new
paper on this topic for consideration for publication by the
journal.”
As far as I know, Voukelatos and Rissel have not published a corrected version of their paper in any other peer-reviewed journal (or anywhere, in fact).
As for “clinging to this error in order to cast doubt on the paper”, well, they used the wrong ICD codes to select the data for their analysis, the numbers in their main data table literally didn’t add up (there were pervasive arithmetic errors), and they drew the graph which was the basis of their analysis wrongly. In other words, the paper was riddled with errors and, as such, its conclusions cannot in any way be relied upon.
However, a subsequent analysis by different researchers (including me) of the same data sources (but this time using correct ICD codes for data selection) has been published, and it shows that not only were Voukelatos’ and Rissel’s data and analysis wrong, but their conclusions based on their faulty analysis were incorrect as well – see http://handle.unsw.edu.au/1959.4/50858
There’s a follow-up paper looking at long-term trends too: http://handle.unsw.edu.au/1959.4/52030
The first reference you give above (Rune Elvik, “Publication bias and time-trend bias in meta-analysis of bicycle helmet efficacy: A re-analysis of Attewell, Glase and McFadden, 2001”, Accident Analysis and Prevention, January 2011) is also looking more than a little flaky – see https://github.com/timchurches/meta-analyses/blob/master/benefits-of-reproducible-research/benefits-of-reproducible-research.md