Nonprofit, nonpartisan journalism. Supported by readers.

UCare generously supports MinnPost’s Second Opinion coverage; learn why.

Root canals cause cancer! (and other spurious correlations)

A delightful new website explores the (sometimes but not always funny) hazards of mistaking correlation and causation.

The divorce rate in Maine correlates with U.S. per capita consumption of margarine.
Spurious Correlations

Over the weekend, I came across an article via Facebook about the great “root canal cover-up.”

The article charges that dentists have known for almost a century that root canals can unleash “pathological bacteria” into people’s bodies, which can then trigger all sorts of diseases, including multiple sclerosis, amyotrophic lateral sclerosis (ALS) and cancer.

My health-scam antennae immediately went up, of course. And, indeed, the article was nothing more than a ridiculous collection of fearmongering misinformation that ended (as I suspected it would) with a sales pitch for some kind of dental herbal supplement.

Some “facts” in the piece were so ludicrous, however, that I find it difficult to understand how anybody would believe them. At one point for example, the author cites a German doctor who apparently claimed that “in his 40 years of treating ‘terminal’ cancer patients, 97 percent of his cancer patients had root canals.”

Article continues after advertisement

Well, gosh, I guess that proves it then. Root canals cause cancer.

What it actually proves, of course, is that people still don’t understand the difference between correlation and causation. Because even if (and it’s a huge, huge if) the doctor had kept meticulous records of the dental history of his cancer patients and 97 percent of them did have a history of getting a root canal, that would prove absolutely nothing.

Just because two things occur together does not mean they are related to one another.

Plotting uncanny correlations

That’s made delightfully clear in a great new website called Spurious Correlations, which was started earlier this month by Tyler Vigen, a first-year Harvard Law School student.

By plotting data from the U.S. Census and the Centers for Disease Control and Prevention side-by-side, Vigen has been able to “discover” uncanny correlations between, say,

  • the divorce rate in Maine and the per capita consumption of margarine (Eating margarine causes divorce!);
  • U.S. spending on science, space and technology and the number of suicides by hanging, strangulation and suffocation (Giving money to science causes people to kill themselves!), and
  • U.S. per capita consumption of cheese and the number of people who die from becoming tangled in their bed sheets (Eating cheese can cause you to suffocate in your sleep!).

Vigen also “found” an inverted correlation between the number of bee colonies in the U.S. and juvenile arrests for possession of marijuana (Teens smoking pot is behind mysterious disappearance of honey bees in U.S.!).

A start, not a conclusion

That’s not to say that uncovering correlations by good scientific research (and having those findings peer-reviewed and published in a journal for others to analyze and comment on) isn’t valuable. As Vigen explains in a delightful video that accompanies his website, finding a correlation can spur scientists to search for a cause.

Starting in 1939, he points out, scientists began publishing research showing a correlation between smoking and lung cancer. Those observational studies encouraged researchers to study the matter in the laboratory. Eventually, the correlation between smoking and lung cancer (and other diseases) was proven to be causative, which led, of course, to anti-smoking public health efforts that have saved countless lives.

But the medical field is also full of examples of correlations that ended up being false. Perhaps the most famous involves menopausal hormone therapy (HT). In the 1980s and 1990s, epidemiological research had found a correlation between the decrease in estrogen that occurs after menopause and an increased risk of cardiovascular disease. Other observational studies had also suggested that giving women HT at menopause lowered their risk of developing heart disease.

Article continues after advertisement

But when a large randomized controlled trial was conducted (the Women’s Health Initiative), researchers found (much to their surprise) that taking HT actually increased women’s risk of cardiovascular disease, including blood clots, heart attacks and stroke. Subsequent analyses of the data have added other risks, including breast cancer and dementia.

Needed: rational thinkers

“Statistical data can show correlations, and then it’s up to us, rational thinkers, to determine if there is an actual connection between the data or if it’s merely a coincidence,” says Vigen in his video.

First, we need to do the math to determine whether the correlation actually exists, he explains. And then we need to establish the why.

And that requires good scientific research.

Unfortunately, most people read about correlations — even spurious ones (Root canals cause cancer!) — and assume it means causation.

Maybe websites like Vigen’s, which is both amusing and educational, will help. (He received a million hits on the site within the first week alone.)

I hope so. But it’s been my experience that getting people to understand the difference between correlation and causation is like, well, pulling teeth.

You can watch Vigen’s energetic explanation of correlation vs. causation below.