Social media sites are full of misinformation. Not surprisingly, anti-abortion social media outlets, including twitter and blogs, abound with incorrect attributions of negative outcomes both to methods of contraception and to abortion.
First, some basics about research design. There are many different ways to study the effect one event (for instance, taking birth control pills or having an abortion) has on future events (for instance, getting cancer). Some of those methods are more reliable than others. Generally, the most precise way to study cause and effect is something called a “randomized controlled trial.” This is the best because it is possible to compare the effect a certain exposure (for instance, taking aspirin) has on a certain outcome (for instance, having a heart attack). Since there are two groups, half the participants are randomly assigned to take aspirin, the other half are assigned to take a sugar pill. Because the only difference between the two groups is whether they took the aspirin or the sugar pill, we can be reasonably certain that any difference in outcomes (number of heart attacks, for instance) is due to the one element we manipulated (whether someone to aspirin or a sugar pill). This is a simplistic explanation and there are always other factors at play, but that’s the basic idea.
Although a randomized controlled trial is considered the best method, in the real world it tends to be hard to carry out this kind of study. Let’s take the example of studying an association between abortion and breast cancer. First of all, in real life you can’t just take a group of 10,000 pregnant women and randomly tell half of them to have an abortion and half to continue their pregnancy. Second of all, although breast cancer is one of the most common cancers, the percentage of women who will ever get breast cancer is relatively small, and most will not get breast cancer until they are in their 50s or 60s. In order to have enough people with the outcome (breast cancer) to truly compare exposure (abortion versus continued pregnancy) we would have to monitor tens of thousands of women for 20, 30, even 40 years.
So we really can’t study the association between abortion and breast cancer using a randomized controlled trial. The next best thing is called a prospective observational study. What this means is that we collect information about a group of people, like what medicines they take, what they eat, how much exercise they do, and what illnesses they develop, over a period of years. Studies like this have given us some of the best information we have about what factors affect diseases. For instance, the Women’s Health Initiative is a groundbreaking prospective observational study that has dramatically changed our understanding of risk factors for heart disease and cancer. Unfortunately, we do not have a prospective observational study that reports on the association between abortion and breast cancer.
So although we don’t have a prospective observational study per se, we do have something close: a prospective study using hospital records in Denmark. This isn’t quite as good as an observational prospective study, because we don’t get the opportunity to ask everyone the same questions in the same way, ensuring that we have the best quality data possible. However, using hospital records for studies often allows us to look at large numbers of people (in the Denmark study, 1.5 million) that would be impossible to study in any other way. In the Denmark study, researchers reviewed the health registry kept by the government on all residents and found no link between history of abortion and breast cancer. This study is strong because there were so many women studied that any difference we find is unlikely to be due to random variation.
All of the methods I’ve discussed are difficult, time-consuming, and often impractical. As a result, scientists have come up with many methods to try to draw conclusions in other ways. A very common, and problematic, method is a case-control study. What this means is that the researcher identifies “cases” (that is, people with the disease or outcome being studied), and then looks for other people who seem quite similar to those people with the exception that they do not have the disease (the “controls”). The research team then proceeds to ask both groups the same questions about behaviors and other factors that might plausibly lead to the disease, and compares how often the groups cited those factors.
The majority of studies that have found a connection between abortion and breast cancer are of this type.
Case control studies are especially useful for rare diseases. If you’re trying to find out what factors lead to a disease that only happens to one in 10,000 people, there’s really no way to recruit the millions of people that would be needed to get a significant number of people in your sample who develop the disease. Case control studies also allow researchers to test hypotheses with far fewer resources than are necessary for large, population-based prospective studies.
Unfortunately, case control studies must be used with caution as they can easily lead us to draw unfounded conclusions. One of the principal problems with these studies is something called “recall bias.” The idea is that someone who suffers from breast cancer is more likely to remember and report that abortion that she had when she was 17 than someone without breast cancer, because our natural tendency is to try to put our finger on something that caused the cancer. In fact, a study in Sweden set out to look at just that phenomenon. Researchers compared women’s reports of having had an abortion in the past with records available through the country’s national medical records. Women with breast cancer were more likely to accurately report a history of abortion in the past than women without breast cancer.
Another big problem with case control methodology comes in choosing the “control” cases. How do you really know that the only difference between the two groups is the exposure (in our case, a history of abortion)? A very easy confounder to run across in any study examining induced abortion is race and socioeconomic status. We know that women of color and poor women are much more likely to get abortions. For many reasons, women who are poor are also more likely to be overweight or obese . Women who are overweight are at higher risk of developing breast cancer. So is the difference we see really due to a history of abortion, or does the difference have to do with obesity, poverty, or something else?
Returning to our example of case control studies examining a link between abortion and breast cancer, we see that those studies supporting a link use weaker methodology (i.e. case control), while those that do not find a link use a stronger methodology (i.e. observational). The studies that show no connection are the largest studies that are most reliable.
So the next time an anti points at a study claiming a connection between abortion and breast cancer (or pretty much anything else), take a look at how the study was done, and ask them what they think about the methodology. Let me know what they say.