Beware the Epidemiologic Dragon

A recent online health service noted that “40% of cancers are due to avoidable life choices….Tobacco causes 23% of [cancer] cases in men and 15.6% of cases in women. The next largest cause of cancer in men was lack of fruits and vegetables in their diets…”

Hold it right there! When someone says that a certain activity “causes” another event to occur, it should be backed up by some very good studies in one of two ways. One method is for a group of people to be given activity A and then monitored for specific result B, and then the same group of people not to be allowed activity A and again monitored for result B.

Alternatively, two groups of people, as alike as possible (to rule out other causes), are selected. One group is given activity A and the other group is not, and both groups are monitored for result B. This approach is applicable to the case of smoking or eating fruits and vegetables, where the time is too long for the appearance of cancer and the appearance of cancer cannot be taken away to complete the second phase of the study.

As you can imagine, it is impossible and probably unethical to have one group of people specifically smoke or specifically not eat fruits and vegetables just to see if they get cancer. The next best thing is to find two groups of people as alike as possible (to rule out other interfering variables), one of which already does activity A, and the other of which does not do activity A, and then watch them over time to look for result B.

The less desirable study is to find people who have suffered the result B and then compare them to people who have not suffered result B, trying to isolate their different activities. Comparing groups of people to each other or to information gathered on the population in general is the stuff of epidemiology, an attempt to ferret out possible relationships between known activities and observed results. The epidemiologist is looking for probable “cause.”

However, these kinds of studies to not prove “cause.” Just because a person does activity A and gets result B does not mean that A caused B. There might be another activity C (which might or might not be related to activity A) that caused result B. While statisticians try to eliminate these possibilities with statistics, epidemiologic studies cannot prove that activity A “causes” result B.

Don’t dismiss these studies, but be careful. These studies tell us where to go looking for probable cause. While in some cases (like making people smoke to see if they get cancer) these studies cannot be done with people, they can be done with animals that closely approximate people. The results can then be theorized to have a similar effect in people.

Here’s another study: “Over 6 years, Spanish researchers found that those who ate the most fast foods and processed pastries were 37% more likely to develop depression.” Does eating fast food “cause” depression? People who have a stressed lifestyle, who are pressed for time and demand perfection in the workplace, who have no time to prepare healthy food, might also be prime targets for depression, especially if they don’t succeed or burn out prematurely. However, there is no direct causal link between fast foods and depression, as the authors of the article accurately pointed out.

Consider another study: “Each 4 point increase in BMI (body mass index) increases the risk of heart disease by 52%.” The researchers went on to say that “higher BMI influences well-known risk factors like hypertension and type-2 diabetes.” The researchers noted that BMI (activity A) influenced activities C and D, which are known to “cause” heart disease (result B).

When you read these studies or hear them on the news, ask yourself – “does this make sense?” Did they prove that activity A actually “causes” result B? Or is there just a suspicion and more well-controlled studies are called for?

Read carefully and thoughtfully. Don’t assume causality, but don’t ignore the warning either.