Thinking about causation


German Upsidedown by Jan-Willem Reusink, Attribution License

A colleague directed me to a recently published study which examined the relationship between neuter status and joint disease in German Shepherds. We had an interesting discussion about the merits of the research methods, however the most significant flaw I saw in the study was in the conclusions the authors drew about their results.

The researchers found a statistical relationship between neutering dogs before one year of age and the occurrence of some types of joint disease. In their discussion they make it clear that they conclude that this association is causal, and that its direction is from neutering to joint disease. In other words they conclude that early neutering causes an increased risk of joint disease and they go on to propose possible mechanisms.

However, in concluding this the researchers fall into a trap that is easy to fall into. Firstly they assumed that the association they found was causal and secondly, that it was in one particular direction. Their study was designed to reduce the false detection of chance associations, however they did not discuss the (very real) possibility of other reasons for the association. One possibility is a relationships between whether owners are likely to neuter early and whether they are likely to do things which increase the chances of joint disease occurring or being detected. As an example, dogs selected for work, may be neutered because it is the organisation’s policy to do so, and the fact they are in work may increase the risk of joint disease or its detection, compared to pet dogs. This then means that there is a confounding variable—whether the dog belongs to this organisation—which links neutering to joint disease statistically, but with no direct association between neutering and joint disease. Such confounders can be accounted for in study designs, but they were not in this case, and therefore it is premature to conclude the relationship is causal.

Even if there was a reason to suspect a causal relationship, we need to be careful not to assume it is in one particular direction. This paper provides a good example of this. Although one could postulate (as the authors did) reasons why neutering might increase the risk of joint disease, one could also postulate reasons why joint disease might increase the risk of neutering. For example dogs with poor conformation, visible as pups, may be neutered because they are not good for showing or breeding, and then go on to develop detectable joint disease later.

For a previous post about the difference between correlation and causation see here . And here is the reference and link to the article about neutered German Shepherds and joint disease:

Hart, B. L., Hart, L. A., Thigpen, A. P. and Willits, N. H. (2016). Neutering of German Shepherd Dogs: Associated joint disorders, cancers and urinary incontinence. Veterinary Medicine and Science 2: 191–199. doi:10.1002/vms3.34



Leave a comment

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

You are commenting using your 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