Ten years ago the Human Genome Project reported the complete sequencing of human DNA, one of the greatest accomplishments of modern science. Having a map of our genome opens the door to exciting new possibilities for medical research and personalized medicine. Futurist Ray Kurzweil predicted “Genomics testing may soon be able to predict precisely what foods are best for us, prescribe individualized exercise and other lifestyle prescriptions, and recommend a personalized list of supplements, neutraceuticals [sic], and prescription drugs for maximum health and disease avoidance.”
How far have we progressed towards those goals after a decade? People keep asking, like the kids in the back seat, “Are we there yet?” Many enthusiasts think we are, and direct-to-consumer genomic testing is becoming increasingly popular.
In The Language of Life: DNA and the Revolution in Personalized Medicine, Francis S. Collins describes his experience with personal DNA testing. He submitted samples to all three of the major labs (23andMe, Navigenics, and deCODE) to see if their results would agree. Each lab tested for a different panel of gene variants although there was a lot of overlap.
To assess prostate cancer risk the different companies tested for 5, 13, and 9 variants respectively, but no company tested for all 16 variants known to be related to increased risk. One company told Collins his risk of prostate cancer was average; another told him his risk was 40% higher than average.
Even when testing the same variants, their interpretations varied. Variants associated with sensitivity to the drug Coumadin were rated by one company as indicating average sensitivity and by another as showing usually high sensitivity.
One result highlighted the unreliability of these tests: Collins’ genes predicted that his eyes were brown. They are blue.
These tests only sample less than 1/10th of 1 percent of your DNA. They don’t test for some of the most significant variants like the BRCA1 gene of familial breast cancer. They don’t ask about family history or test specifically for diseases that run in your family. They don’t really know whether the high risk gene variants cause disease: all they know is that a variant was statistically associated with a disease in a population they tested to provide a basis for comparison. Correlation doesn’t always mean causation. They don’t know whether the risk from one gene variant is modified by the effects of other genes. The testing was done in northern European populations, so we don’t know if the results are valid for other ethnic groups. And any predictions the tests currently make will be altered by new data that are constantly coming in.
So yes, the labs were accurate in identifying genetic variants, but we don’t know how accurately the variants predict risk, and the risk information may not be useful. Does it do any good to know you have an increased risk of Parkinson’s disease if there is no action you can take to reduce your risk?
Even if preventive actions are possible, do we really need the results of individual tests to persuade us to take those actions? Collins’ tests showed that his risk of diabetes was 29% versus the average risk of 23%. He claims that was enough to change his motivation and persuade him to adopt a healthier lifestyle. One could argue that the 23% average risk should be enough to encourage all of us to adopt a healthier lifestyle: even those who might be at lower than average risk are still at some risk and should consider prevention.
Nearly all human illnesses may be due to a combination of hereditary and environmental factors. It will be very difficult to tease out the various contributions. Parkinson’s disease can be entirely genetic in the case of certain familial mutations, or entirely environmental as in the case of poisoning with MPTP. Or it can be anywhere in between.
We tend to think of one gene/one effect, but the reality is far more complex. We have only 20,000 genes. They code for an incredibly complex human body. One gene in the brain is known to make 38,000 different proteins. Genes interact to enhance or suppress each other’s expressions in very complex ways.
Diabetes is 50% hereditary but only 10% of the inherited genes have been discovered. There could be many other variants each contributing only a small amount of risk, there could be rare variants with large effects, and it could be a matter of copy number variants where a section of DNA is repeated.
Doctors get frustrated when patients bring in printouts of genome-based risk estimates. Relative risks are frequently on the order of 1.5 or lower, with little power to predict who will actually develop the disease and with no real implications for management. Patients who test negative may be falsely reassured and thus less motivated to comply with preventive recommendations. Doctors have to tell patients the printouts don’t mean much: family history and general preventive advice are probably more important.
The laboratories offering these tests are not regulated by the FDA, but come under state jurisdiction. They have faced legal challenges. The California Department of Public Health sent “cease and desist” letters to 13 companies in 2008. New York State sent 31 letters to companies, saying they will require licenses to solicit DNA specimens from state residents. The FTC is investigating possibly deceptive advertising and marketing. Legal experts are arguing about whether giving patients information about their risks for certain diseases constitutes “medical testing” and whether the tests should be ordered only by a doctor.
Some of the less reputable companies make promises that go way beyond present capabilities. If they offer personalized advice about diet, lifestyle, and supplements based on DNA profiles, it’s probably a scam to sell supplements. There is no scientific evidence to support those practices.
Pharmacogenetics promises to individualize prescriptions based on genetics, and it has already had some successes. A young patient with acute lymphocytic leukemia (ALL) might have died from her treatment if she’d been given the standard dose. Genetic testing showed she was unusually sensitive to 6-MP and her doctors knew to give her 1/5 of the standard dose.
The label for Abacavir (an HIV drug) now recommends that patients be tested for a gene that causes severe hypersensitivity reactions. Detecting genes associated with rare drug reactions can be very difficult, requiring studies with huge numbers of subjects.
A clinical trial of a new cancer drug Iressa showed little effect overall, and the FDA withdrew approval, but genetic testing determined that it worked very well for the 10% of patients who had a specific mutation. The current drug approval process doesn’t take into account that a drug that fails for 90% of cancer patients might be lifesaving for 10%.
Some pharmacogenetic findings are not really very helpful. One gene raises the risk of developing myopathy when taking statin drugs, but that complication is uncommon, most cases are mild, and many patients with the incriminating genotype don’t have the complication.
The greatest promise of genome analysis is not in personal testing, but in research. Genetic analysis is redefining our concepts of illness. Psychiatric diagnoses have been based on subjective evaluations; genomics may make molecular classifications possible and totally change the approach to mental illness. Cancers that were lumped together are being found to differ significantly at the DNA level. A single gene has been linked to multiple diseases as different as diabetes, rheumatoid arthritis and Crohn’s disease.
In addition to our genome, we have a microbiome: the genes of the microbes that live on us and in us and greatly outnumber our own cells. These microbes contribute to illness, for instance the H. pylori bacteria that cause ulcers. We haven’t even scratched the surface of what we may someday learn from sequencing the genes of our passengers.
Can bad genes be fixed? Experimental gene therapy is already being tested, but it won’t be easy. Getting the gene to its desired location can be difficult, and then you have to get it to function in that new location and make sure it doesn’t have adverse consequences (in some early trials, patients developed leukemia). But the potential is exciting: in one recent trial, colorblind monkeys were given color vision.
People with the CCR5Δ32 mutation are unusually resistant to HIV infection. An AIDS patient in Germany who developed leukemia was given a stem cell transplant to treat the leukemia. They deliberately chose a donor with the favorable mutation, and the recipient no longer has any signs of HIV infection.
Should you be tested?
There are legal and social considerations. Genomic information might be used to deny employment. If a hereditary disease is identified, must other family members be notified? What if they would rather not know? If children are tested, should information be withheld until the child is old enough to give informed consent? Will people reject a potential marriage partner or choose adoption to avoid the risk of a hereditary disease? Testing can lead to rude surprises. Collins’ book tells of an African American man who wanted to learn what part of Africa his ancestors came from. His DNA showed that he was 57% Indo-European, 39% Native American, 4% East Asian and 0% African!
Doctors order specific genetic tests for valid medical reasons. But direct-to-consumer testing is less justified. Should you get your DNA tested? Maybe, maybe not. You might learn some interesting-to-know facts, but it probably won’t have any real impact on your health. We’re not there yet. We’ve embarked on the journey but we’re nowhere near as close to useful results as the tests’ marketers would like us to believe.
This article was originally published as a SkepDoc column in Skeptic magazine.