Far from clinical reality…

February 28, 2011 at 15:29 | Posted in bicycles | 3 Comments

In the Jan / Feb edition of ‘Australian Cyclist‘, Dr Michael Dinh sets out the case for mandatory helmet legislation. In previous posts, I have discussed how the article acts as little more than anti-cycling propaganda by painting cycling as an extremely dangerous sport, and how his premise that ‘helmets prevent injury’ simply does not lead to the conclusion that they should be mandatory. In this post, I shall look at the research data Dr Dinh presents to support his case – research that I presume underlines the article subheading – ‘the hard clinical reality’.

I apologise in advance, but this post gets a bit technical with a discussion of research methodology. However, once again it’s worth pointing out that this is not a discussion about whether helmets ‘work’. Indeed, the conclusions I reach once again remain valid irrespective of their efficacy, and as such should be embraced by everyone whatever their views on the wearing of helmets.

However, before we start talking about Dr Dinh’s research, I’d like to take you on a bit of a detour into something rather different – hormone replacement therapy. HRT, as most people know, involves giving women hormones as they enter the menopause, to help relieve the symptoms. It was quickly established that HRT reduced the risk of osteoporosis (brittle bones), and about twenty years ago doctors became interested in whether taking HRT had any other protective effects, for example on heart disease.

A number of studies were undertaken, and it appeared that HRT  had a significantly protective effect on cardiac disease. For example, one study notes: ‘four cross-sectional studies … have provided some of the most convincing evidence that estrogen replacement reduces cardiovascular risk in postmenopausal women‘. This led to the development of clinical practice where doctors routinely prescribed HRT to their patients in order to protect them against heart disease.

It is instructive to take a look at how these studies were undertaken. The methodology was quite straightforward; take a group of patients and divide them into two groups – those taking HRT, and those not. Then look at their medical records and compare the outcomes for both groups. This study, for example, looks at the survival rate following heart surgery. When the HRT (also called ERT; ‘estrogen replacement therapy’) and the non-HRT groups are compared, the group taking HRT live on average much longer; as the study notes: ‘These data suggest that ERT improves survival significantly‘.

All very compelling stuff, and the basis for clinical practice for many years. It was accepted by doctors that HRT protected against cardiac disease – this is what they were told when they trained, and this is what they told their patients.

Unfortunately, it was completely wrong. In fact, taking HRT actually increases the risk of heart disease in women. Women who may be taking HRT to prevent heart disease are now urged to see their doctor to see if they should stop.

So what happened? How did those studies get it so wrong? The answer is quite simple, and bedevils all attempts to use medical case data to compare different groups. The studies assumed that the groups of women were similar in terms of cardiac health risks, aside from their HRT usage. This, as it turned out, was a very flawed assumption. The problem is that people choose to take HRT; it is not something that is randomly applied to some people and not others. This means that, along with that choice, go other related factors. Women taking HRT are more likely to be of a higher socio-economic background, for example. They are more likely to eat healthily and do exercise. And so on. And his is what the studies I discussed before had actually discovered – that people who choose to take HRT are also more likely to make healthy lifestyle choices. This means they do tend to live longer – but despite, rather than because of, the HRT!

(As an aside, the studies that finally showed that HRT was linked to an increased risk or heart disease were based on randomised studies; groups of similar women were split in half, with half given HRT and the other half a placebo. This is a much more powerful research technique, as it is much less prone to ‘confounding variables’ (things not under test that skew the result). It is also a useful cautionary tale about choice; if your data rely on people consciously choosing one course of action over another, then you need to be alert to the possibility that along with the specific choice you are looking at there may be other correlated factors that can affect the result.)

So what does this tell us about bicycle helmets? Well, many studies looking at the effectiveness of bicycle helmets are undertaken by comparing the post-accident medical outcomes of riders who choose to wear helmets and those who chose not to.

Let us pause there for a moment, and reflect on the HRT example again. There are lots of reasons why people might make a particular choice, and it is likely that that those reasons will also influence other (perhaps unrelated) aspects of their behaviour. Could it be that people who choose not to wear helmets might also behave differently in other ways too? And is it possible that those behavioural differences might also affect the risks they run whilst riding a bicycle?

One way to see if this could be true is to look at the overall injury severity of cyclists who choose to wear helmets compared to those who choose not to. This has not often been done; however one notable study from the US did look at this. Interestingly they found that riders who chose not to wear a helmet were more likely to have serious injuries, even if head injuries were excluded. In other words, for some reason non-helmeted cyclists were having more serious crashes. This is not the only study to have seen this effect; for example this Australian study also found that unhelmeted cyclists suffered more severe injuries to their extremities and pelvis. Several studies (including the two cited) have also found that unhelmeted riders suffer worse facial injuries, even though bicycle helmets do not cover the face.

These data seem to show that, along with the choice not to wear a helmet, other choices are also made that tend to increase the severity of cyclist injuries. None of the studies can tell us why that should be so, but we could make some educated hypotheses. For example, perhaps people who choose not to wear helmets like taking risks; choosing not to wear a helmet is one ‘risk’ which perhaps goes along with others such as riding faster or more aggressively.

Such hypotheses are not important, however, to the overall conclusion. Bearing in mind the cautionary tale of HRT, when we compare groups of people making an active choice and see differences in outcomes between those groups we should be cautious about drawing too firm a conclusion lest there are confounding variables. When we have further data that to show that such confounding variables definitely exist, then we should toss out this type of study design and instead look to use a more robust, randomised methodology that eliminates this error.

So what does this have to do with Dr Dinh’s article in Australian Cyclist? Quite simply, Dr Dinh references a study he undertook with a colleague that compared the head injury outcomes of helmeted and unhelmeted cyclists admitted to his hospital. He notes that helmeted cyclists have less severe head injuries, and uses this as the ‘hard clinical reality’ that underlines the need for mandatory helmet laws.

However, Dr Dinh’s paper is flawed. In it, he does not even discuss the possibility of confounding variables. He is apparently unaware of the dangers of this experimental design. He does not reference studies that illustrate that there could be a problem. Interestingly he does note that helmeted cyclists also had less severe facial injuries, but even this apparently did not set off any alarm bells, nor apparently prompt Dr Dinh to look to see if there were any other differences, for example in the injury severity of other parts of the body as found in other similar studies (although it should be noted that the absence of such differences does not remove the possibility of confounding).

One rather gets the impression that Dr Dinh already believed that helmets ‘work’; indeed he actually admits that in the article when he notes ‘our distant medical training reminds us that helmets do protect from head injuries‘. It seems that this belief is blinding Dr Dinh to undertaking proper sceptical enquiry and making it difficult for him to see the potential problems with his research approach.

It should be noted that the poor quality of Dr Dinh’s research in no way adds weight to the argument that helmets ‘do not work’. Dr Dihn fails to provide compelling evidence that they do, but this is not the same thing as saying such evidence does not exist. If there is one thing to take from this, it is that much research in this field is of very low quality. (Indeed, many people will remember another Sydney academic, Dr Chris Rissel, recently withdrew a paper he had published about bicycle helmets because of data errors – a paper that (perhaps ironically) claimed that increased rates of helmet wearing had not improved cyclist head injury rates in Australia). This makes it very difficult for anyone to properly weigh up the arguments without spending a lot of time researching the issue and reading endless rather dry papers and critiques. Perhaps because of this, may people short-cut proper sceptical enquiry into this topic, and instead rely on intuition – which ultimately makes for a very low quality of debate on this issue. Something that is unfortunate for everyone.

(note: Australian Cyclist articles are usually available online around two months after the magazine is published; the article referred to should be available here in a few weeks time)

Advertisements

3 Comments »

RSS feed for comments on this post. TrackBack URI

  1. The major confounding factor for TRT (which I think you are referring to) appears to be socioeconomic class. Helmeted cyclists were predominantly white middle class children cycling on quiet back streets. Unhelmeted were typically poor black children cycling on busier roads. In fact I believe there is a reanalysis of the TRT data using the same methodology but a different injury that shows that helmets are actually better at preventing knee injuries than they are at preventing head injuries.

    There’s even research to indicate that wearing a helmet might put you at more risk (even if risk compensation by the cyclist is discounted): http://www.bath.ac.uk/news/articles/archive/overtaking110906.html. Passing drivers passed a helmeted cyclist several inches closer.

  2. Great stuff! – love your writing style – am hooked on your fabulous blog!

  3. This is the absolute classic:

    “our distant medical training reminds us that helmets do protect from head injuries”

    Ah, I see – so Dr Dinh went to a lecture 20 years ago when he was told that “helmets were good”. That’s like saying “My mother told me so it must be true”.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Blog at WordPress.com.
Entries and comments feeds.

%d bloggers like this: