STEPHEN STRAUSS: SCIENCE FRICTION
How much stock should we put in genetic testing?
November 16, 2006
Here's a proposition to make you sit up and go slack-jawed: "There is no such thing as a healthy person, merely an incomplete diagnosis."
This startling pronouncement was recently put forward by University of Toronto health policy professor Raisa Deber in an effort explain to journalists how difficult it sometimes is to determine the benefits of genetic testing.
To get the exact, subversive flavour of her contention, you have to have a sense of the volcanic eruption that has gone on around genetic testing. Twenty years ago, there were effectively no tests. Now, genetic tests can be done for 1,317 diseases. For most of those, hundreds — and in the case of cystic fibrosis, more than 1,000 — mutations of a gene have been associated with a disease state.
Alongside hospital-ordered tests, a consumer-testing industry has arisen that promises, in the words of the Genovations company, "to create precise, powerful, and personalized health care designed just for you."
More specifically, for prices ranging from $200 US to $3,500, you can learn which sports your DNA says your body is best suited for, whether you are going to go blind, whether stress is going to lead you into depression, where your genes say your ancestors came from, and more.
False positives could outnumber true ones
With this at least quasi-revolution as the context, Deber gave her audience a mini-seminar on the fallibility of diagnostic statistics.
Let's assume, she said, that the gene tests are quite good — say 95 per cent accurate. Well, that would be fine if you had a genetic condition that afflicted half the population. That would mean that if you tested 10,000 people you would end up with 4,750 true positives and 250 false positives. On the other side you would end up with 4,750 true negatives, and 250 false negatives.
But wait.
What if you are looking for the genetic marker of something rare, such as sickle cell anemia, which is only found in one in 500 African North Americans? Statistically speaking, a 95 per cent effective test would churn up 19 true positives, one false negative and 499 false positives among every 10,000 people.
These false positives are going to needlessly scare the snot out of 499 people. Even if you tested all the true and false positives again, you would churn up more false positives than true positives. Not to mention the mountainous costs which testing, retesting and testing again.
Bogus labs could fake tests
The false positives are so pervasive that Deber and some colleagues have amused themselves by sardonically discussing the virtue of setting up "Bob's Genetic Testing Laboratory."
"You would go to some place without an extradition treaty, set up a lab and have people fearing they had rare conditions send you blood or saliva or some other DNA-containing samples," she said.
"Only instead of actually conducting a test, if the condition was rare enough you could send people back the answer that they didn't have the condition, and you would usually be right."
This sounds horribly cynical, and maybe is. From published reports, it looks as if raw genetic testing is getting more accurate, and accounts of accuracy of 98 to 100 per cent are frequent in more common tests.
How accurate are they?
However, I subsequently had a long chat with Ingeborg Blanquaert, a scientist working for the Quebec government agency responsible for health services and technology assessment, the Agence d'évaluation des technologies et des modes d'intervention en santé.
"Popular belief has been that genetic tests have a higher analytical validity than many other biomedical tests," she said.
"Unfortunately, relatively few molecular tests have undergone rigorous evaluation of either analytical or clinical validity and that this may bring down performance estimates, when evaluation is done in real world situations."
That is to say, despite claims by testing companies, we don't really know how good tests are.
Gene often doesn't mean disease inevitable
But that's not all.
It's one thing to be tested for a mutation that, say in the case of Huntington's disease, tells you that you will absolutely come down with an illness. But most illnesses involve the interactions of many genes with the environment. Carrying one mutation simply indicates a propensity for something occurring.
For example, mutations in the BRCA2 gene are found in less than 2.5 per cent of women in the United States who come down with breast cancer. Moreover, even if they carry the mutation, only about 45 per cent of women actually come down with breast cancer by the time they reach 70.
And all of this skims over the uncomfortable fact that all mutations are probably a load of pluses or minuses. Thus, one of our genes which slightly increases our propensity for, say, diabetes might at the same time decrease the risk of pancreatic cancer, or make us slightly smarter, or add a tiny joy bubble to a generally airless personality.
So, should I be tested?
So what's a cautious consumer to make of this Himalayan heap of incomplete genetic diagnostics?
My recommendation is to view all claims in this area as if you were Doubt's own skeptic.
In the short term, the principle thing that genetic tests have done is increase our awareness of how incomplete our knowledge of genetics is.
But beyond embracing doubt, I suggest that we demand legislation. In Canada, no regulations govern the claims and practices of the genetic testing industry, nor do governments get any guidance on whether a specific gene test is worth funding. As the ironic possibility of Bob's Genetic Testing Laboratory indicates, this is worse than unhealthy: it's nuts.