If you have followed our Isala project for a while, you know that we are very enthusiastic and keen to study a lot, especially when it comes to advancing the health of women (and their partners and children). This ranges from the effect of certain vitamins on women’s health to the effect of underwear on menstrual cycles. Some of these questions can be answered by studies in which we ask people to do something and then we investigate the effects. This type of study is called an intervention study. For example: we might ask a group of women to wear only cotton or polyester underwear for a month to see if there are changes in the composition of their vaginal microbiome. A similar small-scale study was previously organised by our Isala colleagues (check our GeneDoe blog), and we are currently planning a new study on menstrual hygiene in collaboration with the Isala sisterhood in Peru, Switzerland, and Cameroon. But, what if we, for example, want to better understand the effect of pregnancy or the effect of smoking? We cannot ask people to get pregnant or engage in unhealthy behaviours for a study. So, we have to look for other solutions and that is where my freshly started PhD will contribute.
One might assume that we could just compare a group of smokers with a group of non-smokers, see what the difference is and be done with it. Unfortunately, it is not as easy as it seems. Smokers probably have some common characteristics that non-smokers do not, and vice versa. For example, it is possible that smokers generally have a less healthy lifestyle than non-smokers. So, we cannot be sure that the observed difference is caused by smoking and not by any of the other common characteristics. This comes down to the difference between association on the one hand and causation on the other.
Associawhat?
Association means that there is a relationship between two factors. This is what classical statistical methods aim to show and why you can theoretically use statistics to draw connections between anything. Causation, on the other hand, means that one factor is the cause of another factor. This is way more challenging to demonstrate. To illustrate this point, consider a scenario where classical statistics reveal a correlation between the sale of ice cream and the number of shark attacks on a summer day. As a result, people might advise the city council to curb the sale of ice cream to prevent accidents with sharks. This is of course complete nonsense.
We naturally feel that something is wrong here but why did they find a correlation anyway? Because while there is an association between ice cream sales and the number of shark attacks, there is no causation. In other words: selling ice cream does not cause a massive bloodlust in sharks, but both ice cream sales and the number of sunbathers rise when the weather is nice. A lot of people are tremendously good at sensing these kinds of connections. They immediately know that sharks and ice cream have little to do with each other and also that the rooster crows because the sun rises and not the other way around. But in a lot of scientific research, it is not at all clear what exactly causes what. For example, we know that women who smoke are more likely to have vaginal infections. What we don’t know (yet) is whether smoking has a direct impact on the vaginal microbiome. Or are women who smoke generally less concerned about their health, so they take more risks? Or is there something else?
To be able to say with more certainty that smoking is the cause of a change in the vaginal microbiome (or, in other words, that there is a causal relationship), we asked our amazing Isala participants whether they smoke or not. By using a special statistical technique, we can then look for an answer to our question without asking the women to change their behaviour. This technique is called “causal interference” which is exactly what I am going to apply during my PhD in the coming years. In other words: we will be looking for the real cause of the shark attacks in the story above.
And now in practice
One of the things where we are going to use causal interference is to look at the effect of medication on the vaginal microbiome composition, more specifically pain medication available without prescription. Many women carry painkillers in their bag by default. However, most of these medications have mainly been tested on healthy, young men. So, we would like to investigate the effects of these medications on the vaginal microbiome, especially during or around menstruation.
In addition, we want to investigate potential differences in the vaginal microbiome within a population. By this we mainly refer to the influence of the environment surrounding their home, how much is spent on greenery in the streets there, the proximity to major roads, etc. While these factors may have a relatively small effect on one person, combined for the whole population they can ultimately make a big difference. Think of it this way: if your weight drops by 1 kg, you probably won’t even notice. But if the average weight of the whole country suddenly drops by 1 kg, it becomes a reason to look into it.
As you can see, there is still a lot to discover in the huge collection of Isala data. Thanks to the contribution of our amazing participants, we can answer a large variety of questions. I am already very curious to see what will come out of the analyses!
Who am I?
My name is Kato Michiels. I graduated last year with a master’s degree in epidemiology at the University of Antwerp after a bachelor’s degree in biochemistry and biotechnology. Now I am working on a PhD in biostatistics at Hasselt University. As a statistician, you rarely collect your own data, so I wanted to collaborate with an interesting research group that already has datasets but has too many questions to answer them all by themselves. In the Isala lab there is a wealth of challenging data, and the collaboration became reality soon. I was immediately enthusiastic because microbiome data is super interesting, and I am a big fan of research where women get a central role.
“A PhD in statistics? But why?”. It’s a question I answered so many times. Statistics is not many people’s favourite subject, nor has it a sexy reputation. But if I am honest, I have lost my heart to it. Statistics is everywhere. It is a kind of toolbox that allows you to turn data into information. You can use it to answer questions like “What can we consider the norm?”, “In what sense do groups differ from each other?”, “Can you make a prediction?”, and so on. Also, to do your job well as a statistician, you always engage with other specialists outside your own field. To arrive at the best solution, you have to step into the other(s)’ world. This automatically leads to fascinating conversations and collaborations. So, Isala team: here I come!