The CycleChat Helmet Debate Thread

[Justinslow]

I see Robert Peston on the news last night had a bike accident, went over the handlebars, said live on the 10 o'clock news "thank God I was wearing my helmet". Can't argue with that.

 

[Justinslow]

Fair enough, I won't. Mostly because the weaknesses of such arguments have been explained to you many times over, so there really can't be much point in doing it again.

Lol, it was more of a discussion point, I was barely watching it untill I heard the words "bicycle accident" and it made my ears prick up.

 

[martint235]

I see Robert Peston on the news last night had a bike accident, went over the handlebars, said live on the 10 o'clock news "thank God I was wearing my helmet". Can't argue with that.

Can. But if 372 pages of arguing with that still doesn't stop you saying silly things I can't be a***d

 

[ufkacbln]

I see Robert Peston on the news last night had a bike accident, went over the handlebars, said live on the 10 o'clock news "thank God I was wearing my helmet". Can't argue with that.

Jeremy vine is on record for being thankful he wore a helmet cam

... Can't argue with that inconclusive and unequivocal argument for compulsory helmet cams

...

 

[swansonj]

While I share your suspicion that cyclehelmets.org is reliable, there is also an issue with the reporting of negative results in the scientific literature. Put simply, a study that fails to show any positive benefit is less likely to get published - due largely to the fact that the researchers are less likely to seek to publish. In this context, that means that a study that fails to show a benefit to wearing helmets is less likely to end up in the literature. How many - if any - studies that show no benefits have never seen the light of day I have no idea. Which is the problem: we don't know the complete picture.

Publication bias is one of the things that various meta-analyses of the case-control studies have tested for. The recent Olivier and Creighton study (the one from Australia that got publicised a month or two ago) tests for it using a statistical model, that I don't understand and certainly don't trust, and funnel plots, which I do vaguely understand and still don't trust terribly much. They also try to correct by trim-and-fill, which is sort of OK if you accept the premise of the funnel plot in the first place, but is subject to the same assumptions.

But this is one of the things where the different meta-analyses have concluded differently. TRT in their Cochrane review and Attewell didn't find it; Elvik did; now Olivier doesn't. I went to cyclehrlmets.org in the hope for finding some analysis of this; instead they basically just assume Elvik is right and Attewell wrong, and it's hard to avoid concluding that they are choosing which methodology they prefer on the basis of the result it gives .... which I don't need to tell you is bad science

 

[swansonj]

This is not the point I'm making. It is always a good idea to do some quick back of the envelope calculations to check whether the numbers seem reasonable. A sanity check, if you will. The Cheshire police number implies a 33 fold reduction in fatal injury in wearing a helmet. Or, considering a hypothetical population of 1000 cyclists of whom 100 have fatal accidents, 50% helmetted, the rest not, this 97% stat, to be consistent, must mean 97 of those weren't wearing helmets - that's a rate of 19.6% whilst the rate in the helmetted group is 0.6%. Which gives a 33 fold difference between groups. Of course, that differene becomes smaller as the nonhelmetted cohort increases. The rate is identical for both groups when the helmet wearers are 3% of the total cyclist population - and that's far below the actual percentage of those who wear helmets in the UK. (I am assuming that the accident rate per unit distance is identical for both cohorts.)

But even allowing for that, the TRT study reported less than a seven fold decrease - this is a very large discrepancy between the data. That alone ought to be enough to set alarm bells ringing. Were I, during the course of my work, to generate two data sets with such a large discrepancy between them, I'd immediately suspect both sets. In fact, I'd go back and check all the underlying assumptions and the model I was using - and check, if I could, by generating a third set of data by another way.

Here I've used the dataset that's already the most consistent with the Cheshire stat - a set we already know to be of dubious provenance. Even so, it still fails this most basic of sanity checks - badly. That can only lead to the conclusion that the 97% figure is very much suspect.

I think it’s common ground among anyone who has looked at any of these figures for more than 5 seconds that we intuitively don’t believe either the 97% Cheshire Police figure or the TRT 85% reduction figure.

But to compare the Cheshire Police figure (for the % of cyclists in fatal accidents not wearing helmets) and the TRT figure (for reduction in fatal accidents caused by wearing a helmet), you need one extra figure, the % of cyclists wearing helmets. You clearly understand that; apologies if that was implicit in your original post, but it did look a little as if you were directly comparing the two figures – my mistake.

Let:

h= proportion of cycle helmet wearing in general population, so (1-h) is proportion not wearing helmets

r=risk of fatal accident in helmeted cyclists

R=relative risk for fatal head accidents for non-helmeted cyclists, so the % reduction of fatal head accidents for helmets, as often quoted for studies like TRT, is 100(R-1)/R

(note: this treatment assumes that all fatal accidents are head injuries, which is clearly not true – but I’ve seen estimates that 75 or 80% are, so we’ll assume it is actually 100% so as to keep the algebra simple(r).)

P=proportion of fatal accidents in which the cyclist was wearing a helmet

Then:

Frequency of fatal accidents in helmeted cyclists is proportional to hr

Frequency of fatal accidents in non-helmeted cyclists is proportional to (1-h)Rr

So:

P=h/(h+(1-h)R)

h=RP/((1+(R-1)P)

R=h(1-P)/P(1-h)

So if we have estimates of any two out of P, h and R, we can work out what the third must be in order to be consistent.

If P=0.03 (from Cheshire Police) and R=3 (from latest Olivier meta-analysis): h=0.08 – clearly far lower than the reality, confirming that both those figures cannot be right.

Even if we took the maximum figure ever offered for the benefit of helmets, R=7 or something like that from TRT, and combined it with the P=0.03 figure we get h=0.18. That may be in the ballpark for the UK but still strikes me as a bit low.

If we took R=3 from the latest Olivier work (which I don’t trust, but it is the latest published figure) and say h=0.3 as a generous estimate of UK helmet wearing, we’d deduce P actually = 0.13 rather than 0.03.

For me, there are three obvious take away messages:

The Cheshire Police 97% figure is highly suspect in the UK context (and a little googling suggests it is a historic figure for cyclists in ... wait for it … New York City….)

As long as helmet wearing is in a minority, so that the majority of cyclists aren’t wearing one, then the majority of cyclists in accidents won’t be wearing them either, so that figure doesn’t actually tell you very much

And, sadly, it’s far easier to quote a statistic than to understand it, something that applies to our society rather more widely than just cycle helmets and is rather more widespread than just Cheshire Police.

 

[lutonloony]

I was discussing a scheme where, as part of a "get back to work" initiative, there is a half day course where they will be given donated bikes and learn to maintain them, so that one obstacle to being able to get to work was removed. So far very admirable. They then said" we will need to provide helmets ". Playing devils advocate I asked why. I was then given a list of the usual suspects including the "my friends son went straight into a car door and his helmet saved his life" I did ask if that was based on her friend being a specialist neurologist ,but we just got into a loop. Not saying it did or didn't, but yet again the perception of street joes

 

[ufkacbln]

I was discussing a scheme where, as part of a "get back to work" initiative, there is a half day course where they will be given donated bikes and learn to maintain them, so that one obstacle to being able to get to work was removed. So far very admirable. They then said" we will need to provide helmets ". Playing devils advocate I asked why. I was then given a list of the usual suspects including the "my friends son went straight into a car door and his helmet saved his life" I did ask if that was based on her friend being a specialist neurologist ,but we just got into a loop. Not saying it did or didn't, but yet again the perception of street joes

I used to do a lot with the Scouts, both "Badgework" and Sponsored rides.

However once we had to exclude people from activities because of the imposition of helmets, I ceased.

The Group is several thousand pound a year worse off financially and also those who would have gained badges no longer do so

 

[theclaud]

Excellent Tweetifying from @beztweets.

 

[Scoosh]

Excellent Tweetifying from @beztweets.

Brilliant !

 

[mjr]

Excellent Tweetifying from @beztweets.

Local cyclenation group replies to the police commissioner at http://www.klwnbug.co.uk/2016/11/07/call-for-police-help-to-reduce-road-casualties/

 

[mjr]

Decathlon, who use a column of boxed helmets to make goalposts for the Sunday late shift warehouse footy game? I don't expect any shop staff on or close to minimum wage to treat the merch with kid gloves behind the scenes.

Has anyone investigated cycle crash helmet supply chains? Does damage in the supply chain explain some of the reason why real-world helmet use doesn't seem to reflect the increased physical impact protection?

Manufacturers acknowledge that users can't check helmets sufficiently to verify they still work with statements in the instructions like "There may be damage invisible to you, which may reduce the ability of the helmet to reduce the harmful effects of a blow to the head" (Specialized).

Helmet users are basically trusting the entire supply chain not to have damaged the helmet critically before they get it. Do we know how many crash helmets are DOA - Dud On Arrival?

 

[david k]

Has anyone investigated cycle crash helmet supply chains? Does damage in the supply chain explain some of the reason why real-world helmet use doesn't seem to reflect the increased physical impact protection?

Manufacturers acknowledge that users can't check helmets sufficiently to verify they still work with statements in the instructions like "There may be damage invisible to you, which may reduce the ability of the helmet to reduce the harmful effects of a blow to the head" (Specialized).

Helmet users are basically trusting the entire supply chain not to have damaged the helmet critically before they get it. Do we know how many crash helmets are DOA - Dud On Arrival?

Has this been identified as a problem in the past?

 

[mjr]

Has this been identified as a problem in the past?

I don't know. Has anyone verified reasons for the lack of real-world cycle helmet benefit other than risk compensation?

 

[greekonabike]

I personally wear a helmet 'most of the time'. I know that it's not likely to help me much in the event of a serious RTC but sometimes I find it incredibly hard to think of a reason not to wear one. If I was to slip on a trail and hit something with my head I'd rather be wearing a helmet than not, even though it's a cheap amalgamation of plastics. I don't ride any different when I'm wearing one and don't preach to anyone who never wears them. It's a choice and I think it should be left like that. If wearing a helmet was to become mandatory I don't think it would make much difference to peoples opinions.

GOAB