"Misleading and of little or no practical use to consumers" is the way that the Government Accountability Office (GAO) described the results of direct-to-consumer genetic screening tests in a report unveiled last week. To reach this damning conclusion, the GAO sent in genetic samples from five people for testing by four leading direct-to-consumer testing companies. For each donor the GAO sent two DNA samples, one sample using the person’s actual profile and one using a fictitious profile. Although the testing companies were not identified in the report, it’s pretty clear that they are 23andMe, deCodeMe, Navigenics, and Pathway Genomics.
However, being a customer of two of the four companies, I was puzzled by the GAO’s claim that I had been duped. I found the information obtained from my two genetic profiles neither misleading nor useless. (There is one criticism from the GAO that is right on, however. Most genetic studies so far have been done on ethnically European populations, and the testing companies have been remiss in failing to warn customers who are not ethnically European that many results may not apply to them.)
The GAO reached its conclusions, which were presented at a Capitol Hill hearing before the House of Representatives Subcommittee on Oversight and Investigations, by comparing the test results from the four companies for 15 common diseases and conditions: Alzheimer’s disease, atrial fibrillation (a type of irregular heart beat), breast cancer, celiac disease (a chronic digestive problem caused by an inability to process gluten), colon cancer, heart attack, hypertension, leukemia, multiple sclerosis, obesity, prostate cancer, restless leg syndrome, rheumatoid arthritis, type 1 diabetes, and type 2 diabetes. Not all four companies tested for all 15 disease risks and conditions. The GAO found that “each donor’s factual profile received disease risk predictions that varied across all four companies, indicating that identical DNA can yield contradictory results depending solely on the company it was sent to for analysis.” In addition, the risk predictions “often conflicted with the donors’ factual illnesses and family medical histories.”
To illustrate the “contradictory” risk predictions it found, the GAO reports that one 48-year-old male donor received conflicting information “that he was at below-average, average, and above average risk for prostate cancer and hypertension.” Another donor, a 61-year-old male who had had a pacemaker installed 13 years ago, was told that he was at below-average risk for an irregular heartbeat by two companies and at average risk by two others. Different results from the same DNA and results that diverge from the donors’ actual medical conditions—must be pure snake oil, right? Well, maybe not.
The GAO acknowledged that the “contradictory” results it uncovered “can be attributed in part to the fact that the companies analyzed different genetic ‘markers’ in assessing the donors’ risk for disease.” Generally, these “markers” consist of slight differences in genes called single nucleotide polymorphisms (SNPs). SNPs are relatively common genetic variations in which a single DNA base pair differs among individuals. Some SNPs affect the susceptibility of people to certain diseases or influence their response to certain drugs. That’s what the companies are testing for.
Different companies select the markers they think most relevant and the studies they believe are most accurate. For example, 23andMe offers customers results based on “established research” which is defined as being based on studies with a minimum sample size of 750 cases with appropriate controls and which has been independently replicated by at least one other study. The company also offers results based on less stringent “preliminary research” which is generally based on fewer cases and has not been replicated independently. Pathway Genomics reports results based on studies using 1,000 cases and 1,000 controls.
Many of the "contradictory" findings are actually artifacts of the agency's ranking of results as "below-average, average, or above-average." Comparing my 23andMe and Pathway results, one finds a continuum of information. Part of the "differential" in reported results (as opposed to "contradictions") occurs because the two companies test and report different alleles associated with each condition. In general, Pathway reports on more markers than 23andMe does.
That being said, scoring my personal results using the GAO's rather crude tripartite system I find that the two companies agree on 12 out of 20 disease risks they both report. Confining the results to just the ones compared by the GAO, the agreement comes to 8 out of 11 disease risks they both test. To get a sense of why results differ let’s compare the risk predictions for my chances of colorectal cancer, melanoma, and heart attack.
With regard to colorectal cancer, both 23andMe and Pathway Genomics find that my risk is about average. 23andMe tests for three different variants and calculates that 6.2 out of 100 men of European ethnicity who share my genome will get colorectal cancer between the ages of 15 and 79. The average risk for this population is 6.1 out of 100. Pathway Genomics screens for the same three gene variants plus seven others and calculates that my genetic profile is “typical” of the general population for colorectal cancer risk. However, I had a small polyp excised a few years back suggesting that I might be at a slightly higher risk for colorectal cancer. I plan to keep those colonoscopies coming.
The companies disagree about my risk for melanoma. Pathway Genomics reports that my genetic profile for the disease risk is typical, whereas 23andMe reports that my genetic data suggests that my risk is slightly below average at 2.2 out 100 versus 2.9 out of 100. Considering that at age 10 during my first visit ever to an ocean beach, I suffered a second degree sunburn over my shoulders and upper back, I’m not going to be too sanguine about my melanoma risk.
Finally, the companies also disagree about my risk for a heart attack. 23andMe tests for only one variant which indicates that my risk of heart attack is slightly below average at 20.9 out 100 Caucasian males, whereas the average is 21.2 out 100. The GAO report worried that “a consumer with a strong family history may be falsely reassured by below-average risk predictions related to heart attacks and consequently make poor health choices.” Does the GAO really think that a difference of three-tenths of a percent in risk will encourage me switch to a diet of Double Bypass Burgers? Besides, Pathway Genomics tested for 11 variants associated with heart attacks and reports that I am “somewhat more susceptible” than the average to a heart attack. Given my family history—my father died of a heart attack at age 70 and my mother died of stroke brought on by atrial fibrillation at the same age—I’ll go with Pathway Genomics’ analysis. By the way, both 23andMe and Pathway find that I am at greater risk than average for experiencing atrial fibrillation. Note to self: Eat more green leafy vegetables and get to the gym!
Given personal and family history of polyps, severe sunburn, and heart disease, do the genetic tests add anything? Considering that I am likely to suffer heart disease, it’s a good thing to know that people with my “genetic markers receive significantly greater benefit from intensive statin therapy (such as atorvastatin—Lipitor) than people who do not have these markers,” according to Pathway Genomics. Pathway also reports that I am unlikely to have a bad reaction to taking statins. (I’ll be talking with my doctor about this soon.) And if I do survive a heart attack, I will definitely mention to my cardiologist that if she’s considering treatment with warfarin, my genetic tests suggest that I should start out with a lower dose of that blood thinner.
The differential tests results do not bother me, and I would be surprised if many gene testing pioneer customers find the information they receive all that confusing. The results are probabilistic calculations based on a selection of low risk susceptibility alleles. The right way to think about the current direct-to-consumer genotype screening tests is that they are a preliminary technology. They offer supplementary, not dispositive information about various health risks. The tests are not perfect, but they are the beginning of the process through which consumers, physicians, and purveyors will learn how to better interpret and use genetic information over time.
We are in the Apple II era of genetic testing. It would have been silly to ban the Apple II just because it was not as easy to use or immediately comprehensible as the MacBook Air. Standardization of test results will come as more information about the interaction between genetic variants and environmental influences accumulates. The current tests function as training wheels for curious consumers who will be using the whole genome and epigenetic screening tests that will be widely and cheaply available by the end of this decade. As one of those curious consumers, I don't want or need federal regulators to protect me from my own test results.
Ronald Bailey is Reason's science correspondent. His book Liberation Biology: The Scientific and Moral Case for the Biotech Revolution is now available from Prometheus Books. This column first appeared at Reason.com.