Hardly a day goes by without word that Gene X has been linked to Disease Y. (As in, people with the vanilla version of the ICECREAM gene are 20% more likely to develop Type 2 diabetes than people with the CHOCOLATE version of the gene.)
The results are usually based on a computer analysis of hundreds of different single-letter genetic variants in a population of thousands of people, some of whom developed Disease Y and some who remained healthy. They sure sound scientific. But are they meaningful?
I may sound a little skeptical, and I’m happy to report that I’m not alone. A group of researchers from Brigham and Women’s Hospital in Boston, Canada’s McMaster University and Amgen Inc. of Thousand Oaks wondered too. Specifically, they wanted to see whether the scores of genetic variants that have been linked to an increased risk of heart disease in women actually predicted which women have the greatest risk of developing heart disease.
The researchers focused on SNPs (short for single nucleotide polymorphisms) that were reported to have an effect on heart disease or an intermediate cardiovascular risk factor – such as high cholesterol or high blood pressure – in peer-reviewed journals. Combing through a list of SNPs maintained by the National Human Genome Research Institute, they came up with a dozen that purportedly boosted the risk of heart attacks, strokes, coronary disease or cardiovascular death. Those were used to calculate one version of a “genetic risk score.” They also added in 89 SNPs linked to intermediate risk factors to create a second genetic risk score.
Then they calculated both kinds of risk scores for 19,313 white women who participated in the Women’s Genome Health Study. (They concentrated on whites because most of the gene-disease linkages were found in studies limited to people of European descent.) The WGHS included data on each woman’s age, race, blood pressure, parental history of heart attacks and other background health information, along with genetic data.
Once all the scores were calculated, they divided the women into three groups according to their degree of risk. For the risk score based on 101 SNPs, the one-third of women with the highest scores had a 22% increased risk of actual cardiovascular disease compared to the one-third of women with the lowest scores.
Sounds like a big deal? Not so much. That means that the low-risk women had a 3% absolute risk of developing heart disease over 10 years, compared to a 3.7% risk for the high-risk women. The results were published Tuesday in the Journal of the American Medical Assn.
At this point, you might not be surprised to learn that none of the other comparisons – with either the 101-SNP risk score or the 12-SNP risk score – yielded a statistically significant difference in actual cardiovascular outcomes. You might not even be surprised to learn that 29 of the 101 SNPs on NHGRI’s list had no link to cardiovascular disease risk among these 19,313 women. (That’s another problem with these SNP studies – they often can’t be replicated in independent populations.)
So what’s a woman to do? The researchers found that parental history of heart attacks before age 60 was highly predictive – women in that cohort had a 57% increased risk of cardiovascular disease. Relying on standard risk factors like cholesterol and blood pressure also worked just fine.
The authors concluded that judging cardiovascular risk based on the “best available” genetic studies provides “no clinical utility.” Keep that in mind the next time you hear that a gene has been linked to a disease.
-- Karen Kaplan