Broad genetic score adds nothing above traditional CV risk factors February 16, 2010 16:00 | Shelley Wood Chicago, IL - A new study trying to improve cardiovascular risk prediction by combining as many known genetic polymorphisms as possible has found that a genetic risk score made up of more than 100 single nucleotide polymorphisms (SNPs) offers nothing over and above traditional risk factors [1].
In an interview with heartwire, lead author Dr Nina Paynter explained that the current study, published in the February 17, 2010 issue of the Journal of the American Medical Association, was an "outgrowth" of previous research by her and her colleagues that compared the predictive powers of traditional risk factors with presence of a SNP in chromosome 9p21, one of the strongest genetic predictors of early MI discovered to date. That study found no added benefits of knowing 9p21 status.
"The big question after that was, would you do better if you didn't look at just one SNP? If you added a bunch together and used all of the information you have, could you have done better?"
Two scores, no winner
To answer this question, Paynter and colleagues developed two risk scores, one combining 12 SNPs, all of which are markers for cardiovascular disease, and a second using these 12 SNPs, plus any other genetic markers for intermediate CVD phenotypes—a total of 101 SNPs. They then applied these scores to a prospective cohort of 19 313 initially healthy women enrolled in the Women's Genome Health Study, followed for 12 years.
They found that, after adjusting for age, both the 12-SNP genetic risk score and the 101-SNP risk score were associated with a very slight increase in risk of cardiovascular events (of which 777 occurred over follow-up). But after adjustment for other traditional risk factors, neither score remained a significant predictor of events, and neither could help discriminate between women at risk or not or reclassify their risk category.
Speaking with heartwire, Paynter called the findings "disappointing."
"The 9p21 study sort of got close," she said, in terms of establishing a meaningful role for genetic screening in risk prediction over and above traditional risk factors. Intuitively, boosting that information with additional genetic markers seemed like it might tip the balance: ultimately something not seen with this study.
"Where we are right now for risk prediction, with the genetic information we have, taking blood pressure or a lipid value is more useful at predicting risk," she said. "That kind of implies that we certainly don't have the whole picture yet. The genetic information isn't complete enough to use as a global predictor in this way."
Scratching the surface
Commenting on the study for heartwire, however, Dr Eric Topol (Scripps Translational Science Institute, La Jolla, CA) said there were "lots of problems with this study," chief among them the fact that it looked at only 101 SNPs and, critically, did not include two recently identified SNPs for LP(a).
"This is not by any means a comprehensive study. This is just scratching the surface of the common variants that are related to CV risk," Topol said.
He also faulted the study for focusing on common SNPs, found in approximately 5% of the population, whereas there's recently been "a big shift in the field."
"It was thought that common SNPs were going to be a large explanation for common diseases like heart attack, but it turns out that the low-frequency SNPs, below the 5% level, are really where the action is at."
The paper does not address low-frequency SNPs, Topol continued, or structural variants—copy number variants, insertions, deletions, and block substitutions—all of which may also be important genetic markers for CVD risk.
To heartwire, Paynter acknowledged that she collected SNPs for her study until June 2009 but said that while research linking newly identified SNPs to specific risk factors has emerged in the interim, she's seen nothing that she thinks would "tip the scale" and convince her to redo her study any time soon.
"I don't think that there's been one huge SNP [emerge] that I definitely wish I'd included. . . . I think that this was, at least for me, a bit discouraging. So I might wait a little while before looking at this again."
Topol, for his part, pointed out that the field is moving so quickly, studies of this kind quickly become "obsolete."
"This can lead to the misimpression that we're not making progress in the field, when indeed I think we are," Topol said.
The road ahead
In fact, Paynter says her study points in two directions, both of which "give us hope" for future genetic advances in CVD research.
"My paper is a fairly simple, brute-force kind of approach, and that didn't seem to work. However there are options for it to work in the future," she told heartwire. "The first would be to find even more genes, maybe still with weak individual effects, and keep adding them together until they are strong enough in aggregate. The other would be to get narrow and nuanced, perhaps to find genes that predict a specific type of disease in a specific population when you add them together in a specific way under specific conditions."
And, while not a novel finding, Paynter et al's paper highlights the fact that one of the most potent predictors of future CVD remains family history, which combines shared environment and genetics. "You'd think that family history would be kind of a messy indicator of genetic susceptibility [and that] a genetic score could definitely be a better way to get at that," she said. "That's what keeps us hopeful, that one day we'll be able to explain the genetic part and find something useful. But for now, family history remains very robust, and it's consistent across studies and seems the best kind of summary marker of your genetic risk."
