Tests for heart-disease risk could be misleading
Genetic variants may not really be linked to heart troubles after all.
A study looking at 85 genetic variations thought to be linked to heart disease — some of which are already used in clinical tests — has been unable to confirm that any of these links are real.
The findings, published this week in the Journal of the American Medical Association1, highlight a recurring problem as geneticists struggle to extend their reach to complex diseases such as diabetes, obesity and heart disease. These diseases cannot be attributed to mutations in a single gene or genetic pathway, and teasing each genetic contribution from the tangled knot of environmental and genetic factors is so challenging that it can sometimes go wrong.
Often, researchers have found, preliminary results from relatively small studies turn out to be false positives, and are overturned when followed up in larger populations. This survey of heart disease genes may be such a case.
At the same time, the push for personalized medicine — the long-promised ability to tailor medical care to an individual's genetic make-up — has created an incentive for companies to rush from preliminary results to clinical tests, researchers say. The result can be misleading tests, they warn.
The weakest link
Thomas Morgan of the Washington University School of Medicine in St Louis, Missouri, and his colleagues culled through the literature to find published accounts of genetic risk factors for acute heart disease. They then tested 1,461 Caucasian patients from Kansas City, Missouri, hospitals for these genetic quirks, and compared those results to the incidence of heart disease.
They found no significant association between any of the specific gene variants and development of acute coronary syndrome. "We were unable to unequivocally validate even a single one," says Morgan. "This raises serious questions." At least 7 of the 85 gene variants looked at by Morgan are used in commercial genetic tests, which may be ordered and used by physicians on their patients as they see fit. These specific tests are not typically available for direct purchase by individuals.
Morgan thinks his survey is more complete than previous assessments, thanks to the larger sample size. It is possible that other researchers have also found no link between these genetic factors and disease, he adds, but they may have found it hard to publish negative results.
Pressure to produce a positive association can encourage some researchers to skew their data, he notes, analysing just the right subgroups that will give them the association they are looking for. "Even though the plan may be to analyse all cases and all controls, if nothing comes up with that analysis, someone will say, 'let's look at what happens in all women or all men'," agrees John Ioannidis, an epidemiologist at the University of Ioannina, Greece. That decreases the effective size of small studies even further, leaving them more vulnerable to random chance.
Robert Zee, a geneticist at Harvard Medical School in Boston, Massachusetts, who published the initial work on three of the gene variants that Morgan studied, argues that Morgan's results are just as open to fault. "One needs to interpret all of these genetic findings — including this particular paper — with caution," he says. "We still cannot yet claim that none of these would be a suitable risk factor."
Morgan's study, argues Zee, is itself quite small by today's standards, and it evaluated markers that were discovered more than two years ago. Since then, the field has moved on to studying larger populations of patients — sometimes on the order of tens of thousands — using genome-wide screens for subtle genetic differences.
But despite the larger data sets, Ioannidis points out that some of these studies have also been difficult to reproduce. Possible genetic risk factors for Parkinson's disease, for example, were later knocked down by follow-up studies.
All of this highlights the importance of spending the time and money to validate initial results before they are transported to the clinic, says Morgan. "We need to value validation as much as we value the discovery process," he says. "I was told by some that I was wasting my time going back over things that were already known."
Meanwhile, the availability of genetic tests that have not been clinically verified could harm consumers, says Ioannidis. "If you are given the wrong information, you can really go down the wrong path," he says. "You can get deeper into having medical tests and medical care and perhaps even treatment that you might not need."
- Morgan T. M., Krumholz H. M., Lifton R. P. & Spertus J. A. J. Am. Med. Assoc., 297 . 1551 - 1561 (2007).