By Anni Stonebridge
This morning I was waiting for the kettle to boil and I noticed an FB post which referenced an article from the EVJ from some Swedish researchers, conducted in 1999. The title caught my eye because it specifically included measures comparing shod and unshod conditions. If you are as familiar with scouring research papers on hoof function as I am, you will realise that this is pretty unusual. So while I appreciated the first tea of the day (Lady Grey, little bit of milk, big cup), I took a scooby at the study itself. I opened up Deep Dyve, which is a way to access full text articles for free, but you have to read quickly as its time limited.
I love research analysis, yes it’s nerdy but picking out the bias in how research is written up and how these findings travel within the wider world is fascinating. Analysis is a hobby I have, I loved it at uni and it preps me for starting with as much of an open mind as I can. First of all I ask myself the question ‘what am I likely to fix on in this paper?’ This flags up my personal biases – if an article is hoof related, as a barefoot hoof care professional I am definitely likely to focus on information that supports my views. We will always be biased, this is something we need to accept and deal with but that’s interesting in itself! Then I look for the features of the study that challenge my assumptions and I hope, add stuff to my ‘mind workshop’ (shameless ref to Sherlock there, apologies).
Here is the paper we are looking at today:
Roepstorff, L., Johnston, C and Drevemo, S. (1999). The effect of shoeing on kinetics and kinematics during the stance phase. Equine Veterinary Journal 31, suppl. 30, 279-285.
How are we likely to be biased?
Firstly the word ‘shoeing’ in the title will stimulate an emotional reaction – we already do not support the findings should they not support our world view.
Secondly the word ‘unshod’ in the abstract will prime us to look for, and place more value on, evidence supporting the barefoot condition. We will pick up little pieces of evidence and inflate their significance, and we will decrease the value of evidence that supports shoeing.
Thirdly, if we come across elements within the study that are complex, or that we don’t understand, we are likely to avoid them or diminish their importance. Human brains are inherently lazy, but if you know that you can pick out when your own is trying to duck out of some work!
How are the authors likely to be biased?
The authors are also going to have their perspectives, which will affect how they design and conduct the study. This is going to be influenced by their education, life experiences and beliefs, and if they are a team, even the team dynamics, funding or time pressures. The historical focus prior research in the area will play a strong role. We often believe researchers are inherently objective but far from it, there is a fair amount of research in this area which finds inherent bias rather than objectivity in research practice.
So lets get into the nitty gritty. The OP who highlighted the study highlighted it as evidence that “[i]t doesn’t really matter if your horse is shod or barefoot, trotting on roads will lead to a similar degree of concussive damage within the hoof and joints within the hoof and fetlock.”
Does it? The aim of the study was the describe changes in ground reaction forces and locomotion patterns during the stance phase of the stride, due to the application of a standard iron shoe. Their hypothesis was that shoeing would cause changes mainly in the initial phase of contact between ground and hoof.
Comments on the thread quickly highlighted that the horses used were not barefoot adapted but were normally shod, which buys into the belief that shoeing-adapted horses initially struggle without shoes. In real life some do and some don’t, but we will tend when looking at the paper, to assume that all of the sample horses do. The researchers removed their shoes, ran them through the study for the unshod condition, and shod them again for the shod condition. Convenient, but questionable, our barefoot bias will suggest. The small sample size also came up (common for equine studies – this one started with ten warmbloods, but four of them were evaluated as unsound when their shoes were removed so were excluded).
Reading the methodology, the horses were trotted along a asphalt (tarmac) road at a ‘comfortable speed’, then they went over three rails and a force plate buried in 1-2 cm of sand. At the point the measurements were taken they were trotting on a different surface – sand. This may have been because it was more convenient to bury the force plate than to dig up the road and embed it, that the plate was slippery, or that it was impossible to set up a scenario where the surface was continuous, but it is questionable whether the findings have any relevance to trotting on tarmac at all.
This post isn’t really about the findings from the study, but out of interest the researchers concluded that when unshod, the horses showed a decrease in initial horizontal loading. It appeared that there were significant differences in the alteration of concussion-dampening mechanisms in the distal limb. These may have been related to heel loading, quicker joint rotation and transmission of forces more quickly up the limb. It is likely that from our barefoot world view we will pick up on those findings as ‘better’, confirming our biases, but it is also possible that the newly unshod horses were trying to unload the front of their feet, which were suddenly exposed to the ground, or that the suddenly unshod horses trotting on tarmac and experiencing a surface change, altered their stride more quickly because they were closer to it or sank into it more quickly than the shod horses.
Let’s think, in an ideal world if we were going to improve the design and re-run the study we would:
- Increase the sample size
- Use barefoot adapted horses as an additional comparison condition
- Use a continuous surface
- I find myself querying whether using the same sample of horses and running them through three different conditions of shod, newly unshod and barefoot adapted, (including time for barefoot adaptation) would ever be possible but hey, we’re in an ideal world so why not. If money is not object and the sample size is big enough it could be possible. In that case we could add fourth condition ‘never shod barefoot managed’, and even a fifth of ‘never shod regular domestic management’. In fact you could go on forever adding conditions, but it would be hellishly complex to draw useful conclusions.
Well, enough distraction and time for another cup of tea I think. I strongly recommend reading Daniel Kahneman’s writing on bias in his book ‘Thinking, fast and slow’ if you are interested in how our thinking is affected by bias and representation.
Look out for our new book ‘Barefoot Horse Keeping – The integrated horse’, published by Crowood in May this year. An extensive collection on barefoot and evidence based horse keeping, where myself and Jane Cumberlidge demonstrate precisely how biased towards barefoot we actually are!