People often talk about genes as if they were directly related to specific diseases, as in “the breast cancer gene” or “the gene for alcoholism.” But the fact of the matter is that genes aren’t so simple.
They tend to work together in intricate networks, so any one gene’s function only contributes a small part of a person’s total inherited risk for any given disease. So when newspapers report the discovery of a “new gene for” some disease, what are we to make of it?
In their new book “The Tiger That Isn’t: Seeing Through a World of Numbers” (Profile Books, 2007), journalist Michael Blastland and economist Andrew Dilnot give us tools for making sense of such situations.
Chapter 7 opens with a real lemon, reported by the BBC: “For every alcoholic drink a woman consumes, her risk of breast cancer rises by 6 per cent.”
If every drink increased cancer risk by 6 percent, the disease would be almost universal. The more accurate summary (from a Cancer Research UK press release) is that “a woman’s risk of breast cancer increases by 6 per cent for every extra alcoholic drink consumed on a daily basis” (italics added).
But as Blastland and Dilnot point out, we’re still left in the dark: 6 percent of what? Is that a lot?
There are two problems here. The first is due to the fact that a woman’s risk of getting breast cancer is already a percentage. American women, for example, have a lifetime breast cancer risk of about 13 percent. So “increases by 6 per cent” can mean two different things:
1) It could mean an absolute difference of six percentage points. In that case you would add 13 percent and 6 percent to get a total of 19 percent risk.
2) It could also mean 6 percent of the original 13 percent risk. In that case, you would add 13 percent and 6 percent of 13 percent to get a much less dramatic 13.8 percent risk.
The convention in reporting risk is the latter. But the media often create a second problem by reporting the change but leaving out the baseline risk.
Take, for example, this passage from the New York Times:
“Carriers of a single copy have a 15 to 20 percent greater risk of heart disease, while owners of two copies are up to 60 percent more likely to develop heart disease than people who have none.” (New York Times: May 3, 2007).
But the article never mentions what the baseline risk for heart disease is. If only 1 in 1,000 people with no copies of the higher-risk version of the SNP get heart disease–i.e., an individual’s chance is 0.1 percent–then having two copies increases one’s chance to 0.16 percent, or 1.6 out of 1000. That’s still a pretty scant risk.
But if heart disease were so common that even for people with no copies of the risky version, 60 out of 100 get heart disease, then those with two copies would have a whopping 96% chance (60 percent of 60 is 36) of getting heart disease.
That’s a significant risk.
That’s why Blastland and Dilnot recommend that news stories report both baseline rate of disease and the rate after accounting for the risk factor.
They also suggest converting percentages into a format that appears more personal:
“In every 100 women, about nine will typically get breast cancer in a lifetime. If they all had two extra drinks every day, about ten would.”
We do something like this in 23andMe’s Gene Journal (now called Health and Traits), where we compare the average incidence of a disease to that for people with a specific genotype. Along with the additional context we provide, this allows our customers to better understand their data.