Nipples, fashion and guns

• A study has shown that nipple piercing can lead to breast abscesses.
• No study has been conducted to establish whether nipple piercings are fashionable.
• I threw guns in the title to attract a broader segment of readers. You are not going to find anything on guns in this blog post. Nor on fashion. But some do believe that nipple piercing is a fashion statement.

I found this NY Times article that summarizes what turned out to be a part of a larger study. The study examined which risk factors are connected to the development of primary breast abscesses and subsequent recurrence. The NYT article did not mention these other factors and focused on nipple piercing, however this was the biggest contribution of this study to the world of medical science anyway. 

The matter sounded interesting so I googled the study (Google Scholar is great for searching, but to actually get my hands  mouse cursor on the study, I had to use the MU Library access to the world of article databases) and got this nicely formatted document that was easy to read. Now, sometimes you get a report that has no structure (or even worse, no pictures!) and those are not nice to read. This is not a scientific criterion, but I decided to like that study based on that :)

The second thing I liked about it was the phrase “case control study”. I did not mention this, but I was looking for a case control study because I wanted to describe the possibilities and challenges of this study design to my faithful readers. I personally like it because it is rather simple and fairly reliable.This makes it a very common study design (based on my experience...which, I admit, is not that extensive). It is especially useful when we are dealing with rare diseases and/or when diseases occur a long time after exposure - the latter part being more relevant to the study I am presenting here.

Now, many of you do not know (and quite as likely do not care) that in science, there are many kinds of studies with various pros and cons. The biggest issue in social sciences is that the most reliable study designs are often unethical to conduct on people. For example, you cannot conduct a study where you observe what happens if you deprive people of food (or access to Facebook) – it is not ethical. Yes, there a big ethical difference if you compare depriving people of food and Facebook, but the ethical dilemma is real when you try to think how far you can go. In healthcare, the matter is much worse because you may be jeopardizing your research subjects’ health (or lives!).

OK, what then are these so called “case-control studies”? As the name suggests, there are two kinds of people involved – cases (those who have the disease) and controls (those who don’t). Gordis explains that when conducting a case-control study, we start by selecting the cases and then we go about finding suitable controls. What we compare among these two groups it the exposure – did they come in contact with the risk factor. Usually, the risk factor is defined, but in the case of the breast abscess study, the researchers were looking for ANY risk factor significantly associated with the development and recurrence of primary breast abscesses (the disease). The most significant risk factors were smoking and nipple piercing (I will not go into the details about the other discoveries they made). While smoking was an expected risk factor, nipple piercing was not.

Now, I already talked about risk factors a few weeks ago, so I will just briefly explain significant associations at this point. First, significance is a statistical term which, in essence, means that something is highly likely. Not 100% sure (reality check: science is NOT 100% sure most of the time), just pretty sure. And then when I say association, I mean that there is some kind of connection between two things – it is just not sure if we can say that one thing caused the other. This is exactly the argument people that defended smoking used against smoking-restricting policies – there is a connection (association) between smoking and lung cancer, but that alone is not a 100% proof. Now, over the years, science provided good biological explanations on how smoking causes cancer (for example, the chemical compounds inside cigarettes), so many of the skeptics were silenced; but when you do not have a logical explanation for the causation, you cannot (or better said “should not”) jump to any conclusions based on case-control studies.

Soo…. Where was I? Oh, yeah, nipples. OK, how did our researchers choose the subjects for the study? The easy part was getting the cases – women with breast abscesses. Obviously they have read a book or two about epidemiology, like the one from Gordis, because they were aware of some problems associated with selection criteria. One of the biggest problems that Gordis mentions is when all the cases come from one hospital – what if the risk factor was significantly associated with that hospital? Maybe we cannot generalize our findings to other hospitals/cities/states/countries. This was exactly the case in our breast abscess study – all the cases came from the University of Iowa Hospital. However, based on the description of what precaution measures they took, I can conclude that they did a good job anyway. Out of the initial pool of 773 patients with breast abscesses, they selected the best 68 examples that were suitable for the research.

To select the controls, they chose a matching sample. Gordis calls it individual matching. It means, based on our example, that for every case they found a control that was of the same age, race and puerperal status. Not that this was the best possible scenario, but due to practical reasons, it was probably the best (they chose the controls from the same hospital as the cases). Paraphrasing Gordis: the downside is that after making these matches, they could not investigate whether age, race and puerperal status were related to breast abscesses.

After collecting data from the hospital records, they did various analyses. This also seems like a good place to mention the disadvantages of the case-control study design. I already mentioned the problems of selecting the cases and controls - we call that selection bias. There is another kind of bias called recall bias - it is when you don't really remember how often you did something (Like how many times did you drink coffee this week? How many times did you exercise this month and how intensive? Our memory is just fuzzy sometimes). I mentioned that for the breast abscess study, they collected data from medical records. You'd think that this is reliable information, right? But what if there is something missing from the records? Maybe not all doctors paid attention whether their patients had nipple piercing or not... It is really important to have these challenges in mind from the time you start designing your study. You can improve the reliability of your study a lot if you control for these disadvantages, like matching the cases and controls.

I will now summarize their findings about how smoking and nipple piercing are associated with abscesses. The univariate analysis showed that the odds ratio for smokers getting the abscess was 8. In English, this means that the odds for smokers getting an abscess are 8 times higher than the odds for non-smokers. For nipple piercing, the odds ratio was 10.2.

When I reached this point in the study, I began thinking about the habits of people who are inclined to get their nipples pierced. I may sound judgmental, but don’t you find it that someone who get’s their nipples pierced (you know the rebellious type) is also more likely to smoke? What if nipple piercing is associated with smoking (or, better said, both can be related to rebellious behavior), wouldn’t that discredit the idea that nipple piercing causes abscesses? This is a good example of a confounding factor – a factor that may cause an association that is not meaningful.

In order to control for these confounding factors between the risk factors themselves, they performed a multivariate analysis. Now, this does not mean that there can be no confounding factors, but it does give more reliable results. After this analysis, the results were a bit different – while smoking was still significantly associated with all kinds of abscesses (odds ratio in this case lowered to 6.15), nipple piercing only remained linked to subareolar abscesses (odds ratio: 20.26; smoking odds ratio for subareolar abscesses was 11.49).

What this study successfully accomplished was that it confirmed what other studies have found – that smoking is associated with breast abscesses; but more importantly, it discovered that so is nipple piercing. They have found numerous biological explanations that support the causation, however they do not go on to claim that there is one. Rather, they suggest further research be done, but also that doctors might want to start warning their patients about the possible threats of nipple piercing. The researchers also reflect on the study design, explaining that there are limitations connected with a retrospective case-control study and that their matching was not perfect.

All this makes me believe that this study was a good one and more importantly, a responsible one when it came to making conclusions. I also find the NY Times article to responsibly point out the new discovery – the danger of nipple piercing; however they could have mentioned at least the smoking part too.

Now, these findings will not affect me much – I never considered nipple piercing and I don’t think I’m facing the risk of breast abscesses anyway, but I think it is important to raise awareness of such issues, especially among the female population. In moments of youthful indiscretion, an argument like this one could prevent a girl or two from doing something she would eventually regret.