Genomewide Single-Nucleotide Polymorphism Analyses: Boon or Bane for Clinical Medicine?

Jacquelyn K. Beals
January 9, 2008 — The role of personal genome analysis in current medical practice is the topic of a Perspective article in the January 10 issue of the New England Journal of Medicine. Considering the clinical value of direct-to-consumer genomics services, the authors conclude that the tests may have high analytic validity but are "not ready for prime time."
Coauthor Muin J. Khoury, MD, PhD, director of the National Office of Public Health Genomics, Centers for Disease Control and Prevention, Atlanta, Georgia, was interviewed for an online NEJM supplement by Rachel Gotbaum, an independent producer based in Boston, Massachusetts.
"These are exciting methodologies for research," Dr. Khoury said, "because they're essentially beginning to narrow down the list of genes that have something to do with...common diseases. And once you know what the genes are, and you can study their gene products and interactions, you might be able to develop interventions...that essentially begin to counteract the abnormal gene functions related to these genes. But that's not where we are today."
Today, however, there are 2 companies (soon to be 3) offering personal genome analyses directly to consumers. To receive a personal genome analysis, a client submits a DNA sample to the company, which analyzes arrays of 500,000 to 1 million single-nucleotide polymorphisms (SNPs), using tools employed by researchers to scan the whole genome. The client then receives a printout indicating his or her risk for certain medical conditions.
The article identifies 3 major areas of concern with this direct-to-consumer process. First, there is the question of the test's analytic validity, which is rarely addressed, although researchers generally consider the validity to be quite high. Errors still may occur in handling samples, however, and even low error rates could have an appreciable effect when applied across the 500,000 to 1 million SNPs analyzed.
Second, when determining the test's validity for clinical detection or for prediction of a given medical problem, one must weigh its sensitivity, specificity, and both positive and negative predictive capacity.
"Most of the diseases that we get in adults...are very complicated," said Dr. Khoury. Many gene variants exist and interact with each other, as well as with environmental exposures. "When you take 1 gene at a time and measure its relationship or association with that disease, you get a very incomplete picture of the complex array of associations...the amount of information or extra risk that you measure is very weak at best. And therefore, by itself, it has no clinical value for the prediction of that disease."
Finally, the test's clinical utility must consider the advantages — and disadvantages — of introducing the test into medical practice. A fundamental question is, What can be done for a patient identified as being at risk for a given disease? The article states that few data are available on the value (or risk) of screening for gene variants. One basic, but unverified, assumption is that patients who are considered to be genetically at risk will respond to interventions proven effective in the broader population.
Other potential disadvantages are that patients who test negative for disease risk may be less likely to follow preventive health measures. Dr. Khoury also noted that the current lack of legal protection for genetic information could lead to "potential psycho-social and ethical issues...and potentially discrimination.... So we may end up with more harm at this point, given that the science is incomplete."
Steven L. Salzberg, PhD, director, Center for Bioinformatics and Computational Biology, and Horvitz Professor of Computer Science, University of Maryland, College Park, told Medscape Pathology via email: "Personal genomics has tremendous potential, but I think we're not there yet.... [T]o make useful clinical predictions, we need to collect SNP data from thousands of individuals." Further, to identify " 'weak' correlations between SNPs and disease, we might need tens of thousands of individual samples," Dr. Salzberg said.
"It is far too early to start telling people that they have SNPs that indicate a tendency to get one or another genetic disease — we just don't have that data, except for a very few markers such as BRCA1," observed Dr. Salzberg. "This being said, I think it's a positive sign that scientists are trying to commercialize genotyping technology."
Wylie Burke, MD, PhD, professor and chair, Department of Medical History and Ethics, University of Washington, Seattle, also commented by email to Medscape Pathology: "Does it help to know that your risk for diabetes is 1.5 times higher than the population risk, for example? Not necessarily, particularly if the proper steps to reduce risk are to improve dietary and exercise habits."
Testing is only beneficial if it motivates behavioral change or leads to proper treatment. "Clinical studies are needed to see what happens when people get their test results...whether [or not] the test in fact leads to healthy change. Or...to determine whether health outcomes are truly improved after testing and appropriate changes in treatment," said Dr. Burke. The health benefits of these tests remain to be proven.
For the present, when patients walk in with genetic printouts, Dr. Khoury advises physicians to tell them politely that "the information is not predictive...[and] not of sufficient magnitude to warrant concern.... Then proceed to advise them about health promotion and disease-prevention messages related to those diseases on the printout, regardless of whether the person is positive or negative for these genes."
He concluded, "Family history is probably a much more useful tool for clinical practice today than most of these genes on the chip that people are selling right now."
Dr. Khoury, Dr. Salzberg, and Dr. Burke have disclosed no relevant financial relationships.
N Engl J Med. 2008;358(2):105–107.