The testing process is similar at each company. After spitting into a tube or swabbing their cheeks for saliva, customers submit samples of their DNA. Within weeks, they receive an e-mail informing them that they can retrieve their information from a secure Web site. These are not diagnostic tests and their predictive value is subject to much debate. Many diseases involve the interaction of scores or even hundreds of genes. A SNP that shows a heightened risk for a particular condition almost always only tells part of the story, and some people worry that since the data is rarely definitive customers might be misled. “We are still too early in the cycle of discovery for most tests that are based on newly discovered associations to provide stable estimates of genetic risk for many diseases,” wrote Peter Kraft and David J. Hunter, both epidemiologists at the Harvard School of Public Health, in an article titled “Genetic Risk Prediction—Are We There Yet?” in the April 16, 2009, issue of the
New England Journal of Medicine.
They thought not. “Although the major findings are highly unlikely to be false positives, the identified variants do not contribute more than a small fraction of the inherited predisposition.”
None of the services pretend that genetic tests alone can explain complex health problems. On its Web site, 23andme states that “in order to make a diagnosis, your doctor considers not only your genetic information, but also your particular personal and family history and your physical condition, as well as any symptoms you are experiencing. Your genotype is only part of the equation.” Making a similar point, deCODE suggests that you explore your genetic risk factors and keep a vigilant eye on your prospects for prolonged health. Even Navigenics, the most clinically oriented of the three, tells prospective customers that there are no certain answers in the information they provide: “This level of personalization may help you take action to detect health conditions early, reduce their effects or prevent them entirely.”
Even knowledgeable consumers can struggle to put partial genomic data into perspective, particularly if a report indicates that they are at greatly increased risk of developing a serious illness. That information is based on
what is known—
which in most cases is only a fraction of what there is to learn. Three prominent health officials, including the editor of the
New England Journal of Medicine
and the director of the National Office of Public Health Genomics at the Centers for Disease Control, have suggested that until better data is widely available, a person would do more to improve his health outlook by “spending their money on a gym membership or a personal trainer.”
Caveats and caution are necessary because risk is relative and few people deal with abstract probabilities rationally. If, for instance, a person has four times the normal risk of developing a particular disease, should he worry? That is an extremely elevated figure. Without context, however, a number means nothing. Take the digestive disorder celiac disease; fewer than one in a hundred people develop celiac disease in the United States, so a relative risk figure four times the average would mean that you stand a slightly greater than 96 percent chance of avoiding it completely.
That doesn’t mean genomic tests aren’t useful. They can change (and save) your life. Jeff Gulcher, deCODE’s forty-eight-year-old chief scientific officer, wasn’t around the last time I was in Reykjavik. Stefansson and he have worked together for more than two decades, since the day that, as a graduate student, Gulcher walked into Stefansson’s laboratory at the University of Chicago.
When Gulcher took his deCODEme test, a month before I arrived, he learned that his relative risk of developing prostate cancer was 1.88. That meant he was almost twice as likely as the average person to get the disease. Gulcher took those results to his physician, who ordered a prostate-specific antigen test. PSA is a protein produced by cells of the prostate gland. The results fluctuate, but in general the higher they are, the more likely a man is to have prostate cancer. Gulcher’s test showed that he had 2.4 nanograms of PSA per millimeter of blood, well within the normal range. Those tests are routinely recommended for men fifty years of age and older. But because Gulcher was not yet fifty, most doctors would never have given his results a second thought.
With Gulcher’s genetic profile in hand, however, his physician scheduled an ultrasound, just to be sure. The films revealed an aggressive tumor, though it had not yet spread beyond the prostate. “Jeff was so young that nobody would have made anything of that kind of PSA score for ten years,” Stefansson said, staring into Gulcher’s empty office. “By which time he would surely have been long dead.” Gulcher had surgery, quickly recovered, and returned to work. His prognosis is excellent.
Gulcher makes his living pondering the meaning of risk. Most people don’t. Critics of the tests say they are still too complex for an average consumer to fully understand. Kari Stefansson disagrees. “That is such bullshit,” he screeched. “We are actually criticized for revealing valuable information to unsuspecting citizens
at their request, people who paid for exactly that service
. If somebody does not want to know this information he should not have the test done. It’s not required. But it is extraordinarily patronizing to tell a person that he is not mature enough to learn about himself.
“By the way,” he continued, “in most American states, you can get in a car and use your driver’s license as identification to buy a gun. Then you can drive to a liquor store. You can have the bottle, the gun, and the car. That’s fine. But for heaven’s sake don’t learn anything important about yourself or your family. For some diseases there is no treatment or no useful response
yet
. But you have to remember that our ability to treat diseases was always preceded by our ability to diagnose them. So our ability to prevent diseases will surely be preceded by our ability to assess risk.”
That fact is easier to handle in theory than in practice, however. Throughout the early years of the AIDS epidemic many people who had reason to fear they might be infected nevertheless didn’t want to know. At the time, there was no treatment or cure. A positive test was a death sentence with no reprieve. “These decisions are never easy to make,” said Arthur Caplan, the director of the Center for Bioethics at the University of Pennsylvania. “That lag between knowledge and application can be excruciating. Maybe personal genomics will look different in ten years, but right now it’s a world of fortune-telling and bad news.”
That depends on what you learn. If, for example, you discover that you possess a greatly increased risk of developing type 2 diabetes or heart disease, there are changes in diet and lifestyle that can help. There are also numerous medications. Will they help enough? Nobody will know until more genetic information is available. The tests have already proven their value in other ways, though. Genome-wide association tests have revealed how abnormal control of inflammation lies behind one of the principal causes of age-related macular degeneration, which is a leading cause of vision loss in Americans sixty years of age and older. More than one promising drug is already under development. The tests have also discovered genes that reveal pathways of inflammation critical for the development of inflammatory bowel disease, as well as genetic pathways for heart disease, diabetes, and obesity.
A principal goal of this research is to provide doctors with information that will take the guesswork out of writing prescriptions. In the case of the blood thinner warfarin that has already begun to happen. Warfarin is prescribed to two million people each year in the United States. The proper dose can be difficult to determine, and until recently doctors simply had to make an educated guess. Too much of the drug will put a patient at high risk for bleeding; too little can cause blood clots that lead to heart attacks. The dose can depend on age, gender, weight, and medical history. But it also depends on genetics. Two versions of the CYP2C9 gene can retard the body’s ability to break down warfarin. This causes the drug’s concentration in the bloodstream to decrease more slowly, which means the patient would need a lower dose. Armed with that kind of information—which these tests now provide—a physician is far more likely to get the dose right the first time.
That is the essence of pharmacogenetics. If three people out of a thousand die during a clinical trial due to a drug reaction, that drug will never make it to the market in the United States, even though it would have worked without complications for more than 99 percent of patients. If we knew who those three people were likely to be, however, none of that would matter. Obviously, that kind of knowledge would have saved thousands of lives lost to Vioxx. And it would have permitted millions who were not at risk of heart attack or stroke to continue to take a drug that had helped them immensely.
“We are just starting all this,” George Church said. In addition to his academic and entrepreneurial commitments, Church advises several genomics companies, including 23andme. “But there is already great value to these tests. If you happen to have a SNP that leads to a disease that changing behavior will help, then it’s magnificent. So if you have a propensity to diabetes, you’re going to want to exercise, don’t eat certain things, etcetera. If you have a propensity to a certain type of heart disease, etcetera, etcetera. If you have a propensity towards Alzheimer’s, you might want to start on a statin early, you know?”
AFTER WALKING OUT of Church’s laboratory at Harvard, I took a cab to the airport and flew home. It wasn’t a pleasant flight because I couldn’t stop thinking about the terrifying phrase “propensity toward Alzheimer’s.” Who wouldn’t fear a disease that starts by making us forget much of what we would choose to remember and ends in feral despair? I have special reason to worry. A few years ago my father began to disappear into a cloud of dementia. His illness took the normal pattern—first forgetting keys (as we all do), then names, then simple directions, and eventually whatever you had told him five minutes before. Inevitably, he became incapable of fending for himself. I can think of no worse fate.
For most common diseases, the relative risks posed to individuals by specific genetic mutations remain unclear. There are just too many moving parts we have yet to analyze. Alzheimer’s is an exception. Genomic studies have provided compelling evidence that a variant of at least one protein, called APOE and found on chromosome 19, dramatically increases the risk of developing the disease. APOE contains the instructions necessary to make a protein called apolipoprotein, which plays a complicated role in moderating cholesterol and clearing fats from the blood. There are three common forms, or alleles—APOE2, 3, and 4. APOE4 is the time bomb.
People with two copies of APOE4 have fifteen times the risk of developing Alzheimer’s than a typical person of similar ethnic heritage. They are also at great risk of losing their memory far more rapidly than people without this allele, or those who have just one copy. The correlation between APOE4 and Alzheimer’s disease is so dramatic that when James Watson became the second person (Craig Venter was the first) to publish his entire genomic sequence in 2007, he chose, of all the billions of nucleotides that comprise his DNA, to block only that data. There is Alzheimer’s in Watson’s family, and despite his age—he was seventy-nine at the time—Watson said he didn’t wish to know the status of such a debilitating disease for which there is no cure.
Many, perhaps most, people would make the same decision, choosing to subscribe to that well-worn aphorism from Ecclesias tes: “With much wisdom comes much sorrow; the more knowledge the more grief.” Others adhere to a more radical, denialist vision: “Ignorance is bliss.” I prefer to see fate the way Lawrence of Arabia saw it after he managed to cross the Nefud desert. “Remember,” he said to a stunned Ali, who had warned that the trip would kill Lawrence, the camels, and all his men. “Nothing is written unless you write it.” It’s not as if I believed that knowledge would permit me to alter my prospects of developing Alzheimer’s, but it would surely permit me to alter everything else in my life.
“There’s almost nothing that you can’t act on in some way or another,” Church had told me at Harvard. “It’s probabilistic just like every decision you make in your life. What car you’ve got, whether to jog or not. You can always—if there’s no cure, you can make a cure. You can be Augusto Odone. You can do the next
Lorenzo’s Oil.
You may not be successful, but at least it will keep you busy while you’re dying, or somebody in your family is dying.
“And I would definitely prefer to be busy than to be ignorant,” he continued. “In other words, ‘Gee, I don’t know if I have the Huntington’s gene, so I don’t know if I should go out and raise money and get educated.’ I think an increasing number of people are going to be altruistic—or selfish, depending upon how you look at it—and say, ‘I want to know, so I can spend a maximum amount of time with my loved ones, fixing the family disease.’”
I had already signed up for the tests offered by Navigenics, deCODE, and 23andme. My APOE status was included on my Navigenics report, and it never occurred to me not to look.
A few days later, I poured myself a cup of coffee, sat down, and signed in to the Web site, where my data was waiting. Like the other companies, Navigenics issues a detailed guide, which it calls your Health Compass, that assesses the risks associated with many of the SNPs in your profile. (At the time it was the only company to provide customers with their APOE status, although at first it had done so by a complicated and misleading route that involved testing a different gene, one that is often inherited with APOE.)
I downloaded the 40,000-word report on my personal health. Each condition was described in three ways: as a percentile, which showed where my risks ranked compared to the sample population; as the likelihood that I would develop a given condition over my lifetime; and compared to the average person’s risk. I held my breath and turned to page six, where I discovered that my lifetime risk of developing Alzheimer’s—4.4 percent—was half that of the average man with my ethnic background. I don’t have either APOE4 allele, which is a great relief. “You dodged a bullet,” my extremely wise physician said when I told him the news. “But don’t forget they might be coming out of a machine gun.”