Confusing Correlation With Causality

I'm always amused by various attempts at linking birth control and abortion to cancers in women. Inevitably, some combination of statistics come out and and some people start trying to claim that there is a causal relationship, and therefore both should be outlawed.

The first article cited is this article which describes a statistical correlation between hormone-based birth control pills and cervical cancer. The correlation is interesting in itself, but of course no story is ever so simple:

For women in developed countries who took the pill from age 20 to 30, the number of cases of cervical cancer by age 50 rises to 4.5 per 1,000 women, from 3.8 cases per 1,000 in women who did not take the pill.

The incidence rises to 8.3 cases per 1,000 for pill users in less developed countries, compared with 7.3 cases per 1,000 for women who did not take the pill.
But this extra risk is outweighed by a reduction in the risk of ovarian and endometrial cancers, the researchers said in a statement.

In other words, one risk is traded for another risk. Unsurprising, as this is true of most medical treatments these days. However, it is important to note that the researchers are quite clear in pointing out the "other side" of the risk coin - a factoid that is conveniently ignored by the so-called "pro-life" advocates.

The second article is more egregious as it is a purely statistical analysis which suggests a relationship between abortion rates and breast cancer rates in women. The sad part about this is that the trends shown do not necessarily imply any kind of causal relationship. The trends are parallel, but there are also a myriad of other factors that could be at play.

I will grant that statistical correlation is a valid point from which to launch more in-depth investigations to establish whether there is any causal relationship. However, it is folly in the worst sense of the term to simply assume causality.

Additionally, when an article by one person is only cited by people with obvious "axes to grind" politically speaking (and Patrick Carroll is certainly in that category), it's worthwhile to start digging a bit to see what can be found about the researchers involved.

It's hard to tell what Mr. Carroll's credentials really are. He seems to have a fairly solid background in statistical analysis, which fits with references to actuarial research, but it's not clear if he holds an advanced degree in that field, or has any significant medical research background. He's clearly not exactly "widely published", and certainly not as a co-author of papers in journals of Oncology, which should give anyone reading his work pause before placing too much weight on the conclusions or inferences made.

However, I would be reluctant to make any kind of policy decision based simply on these limited points of data, without significantly more in-depth causal analysis being made.