October 09, 2006

Corresponding Editor Ricki Lewis
The Nutrigenetics War - Science vs Semantics

It sounds like a great idea. Extrapolate population-based gene/health biomarker links to the average consumer. Specifically, offer to match variants in DNA collected on a painless inside-the-cheek swab to “dietary and lifestyle” suggestions that just may include the purchase of personalized (and pricey) dietary supplements.

What’s the catch? After all, scientific support for gene-diet/lifestyle connections is robust. But what is curiously lacking, as far as I, the CDC, FTC, FDA, and who knows how many others can see, is investigation of the validity of applying pooled population health data to individuals. Consider an example: researchers genotyped 266 white women with breast cancer from western New York for the manganese superoxide dismutase gene, and compared them to age and neighborhood-matched healthy controls, also noting whether each woman ate a lot of fruits and vegetables or didn’t (1). For premenopausal women, having two copies of the “A” allele tracked with heightened breast cancer risk, but mostly only if they ate less than the median level of veggies. Does this mean that a black woman from Peoria with the same genotype should eat buckets of carrots to fight oxidation in her cells? That’s perhaps an exaggeration, but the gist of nutrigenetics.

Concerned that web-based marketing of such nutrigenetic tests to a less-than-science-savvy public is outpacing rigorous scrutiny of the premise, the Government Accountability Office (GAO) set up a sting operation. The results were reported at a hearing of the Senate Special Committee on Aging on July 27, and in a summary document [big PDF suitable for framing here].

The GAO Experiment

Here’s how the GAO targeted four nutrigenetics company websites. They took DNA from a 9-month-old girl and from a 48-year-old man. Then they concocted diet/lifestyle profiles for 14 “fictitious consumers,” 12 from the baby’s DNA, and 2 from the man’s. For example, the baby girl was transmogrified into a 6-foot-tall, 210-pound 45-year-old man who smokes and doesn’t exercise, drinks a lot of coffee, and eats a lot of fat. The baby also became a 72-year-old woman who weighs 100 pounds and stands 4’9”, regularly exercises, never smoked, doesn’t drink coffee, and eats fried foods. To make it interesting, the researchers also submitted cat and dog DNA but did not disclose whether the cat preferred Friskies to Fancy Feast or the dog Alpo to Purina so as not to tip their hand.

Results were not terribly surprising. Overall, the advice tended to track more with the made-up details involving exercise, smoking, and pigging out than to DNA variants, making one wonder why the tests are needed at all. But some results were disturbingly inconsistent. For example, the following fictitious consumers were all told that they were at increased risk of developing osteoporosis, high blood pressure, type 2 diabetes and heart disease: the male DNA/32-year-old male profile, the female DNA/33-year-old female profile, and female DNA masquerading as a 59-year-old male. The 59-year-old led a pretty healthy lifestyle, but the made-up 32-year-old man was a former smoker and the faux woman a smoker, neither of whom frequented the gym. All three were advised to purchase the same supplement package, tailored to their genes, to the tune of $1,200. The GAO subsequently found equivalents of these supplements at a regular store for about $35.

At least the fake trio who may face osteoporosis etc. were warned of familiar conditions. Another company used language so vague as to be meaningless. What do “faulty methylation patterns” mean to the average Joe or Jill? Is having methyls attached to your DNA a value judgment? What is “altered activity” in certain genes? “Brain aging”? Such pronouncements could be frightening or intimidating.

The GAO report is unfortunately a summary -- it is difficult to sort out which company actually did what, and my request for details has not yet been answered. Since multiple companies actually send the DNA to the same labs, assigning specific blame may be impossible anyway. But the overriding conclusion of the GAO report is clear: “Although these recommendations may be beneficial to consumers in that they constitute common sense health and dietary guidance, DNA analysis is not needed to generate this advice.” Case in point: suggesting that smokers stop smoking, and never-smokers never start. Ditto for “eat your vegetables” to up antioxidant intake, toxic spinach notwithstanding.

One Company’s Rebuttal

Not surprisingly, one of the skewered companies has written a very detailed and impressive rebuttal, taking on the GAO report point by point. This response starkly reveals the source of the controversy – and it is semantics, not science. (I won’t name the company because that isn’t really relevant. They have been helpful and forthcoming, and their website is a treasure trove of information, even if it doesn’t include any references that extrapolate the population studies to individuals. These might not exist.)

This nutrigenetics company’s website and others include disclaimers, such as “[This is] not a genetic test for disease or predisposition to disease, nor does it determine a medical condition,” followed by some variation of “if you’re sick, see a doctor”. Rather than diagnose, reads the rebuttal, the company “explains impacts on health parameters.” Like the confusingly positive labels on food supplements, nutrigenetics tests purport to assess such “parameters” as bone health, heart health, inflammation, antioxidant/detoxification, and insulin sensitivity.

Dubious Definitions

The disconnect between the GAO report and company websites stems from fuzzy interpretation of three definitions: a test can diagnose, predict (aka “detect predisposition”), or reveal elevated risk (aka “at-risk”). Some of the confusion stems from the fact that “at-risk” and “predictive” have very precise definitions in the human genetics field, with vaguer associations in plain English.

An analogy may clarify matters.

Imagine visiting a fortuneteller. If said seer diagnoses, she might say, “I see you are with a tall, dark, stranger.” Should she predict, she might say, “You will meet a tall, dark stranger.” But if her information is less precise, she’d say, “You are at increased risk of meeting a tall, dark, stranger.” This last option is what the companies claim they do; the GAO report claims it is the middle option. Figures in the government report, for example, use a label of “medical predictions” above lists of “increased risks”, as if the terms are synonymous. They’re not.

With this ambiguous language, what’s a health care consumer to do? Possibly panic. If a company reports that a cheek-swabber is at increased risk of developing, say, beriberi, might he or she interpret the suggestion as a diagnosis? One company’s language of “your genetic profile indicates you have variations in the B vitamin metabolizing genes … “ could be upsetting to someone not realizing that any copy of a gene is, technically, a variant. We all have them. (At least the websites use “variant’ and not “mutant”.) Yes, the websites provide excellent background material on genetics, with great references. But I doubt most consumers actually read all of it, let alone understand it.

The company rebuttal wisely steers clear of comparing their services to tests for well-studied single gene disorders -- such as for Huntington disease or BRCA breast cancer, -- likely because these tests underwent years of testing to see how people react to and use the resulting information. As far as I (and others) can tell, that simply isn’t known for nutrigenetics tests. Instead, the company uses as their model FDA’s quasi-guidelines for health claims on certain food products, which, like labeling on food supplements, steadfastly avoid mention of disease to the point of maddening obfuscation. The rebuttal offers an example from FDAspeak that must have won a contest for maximizing qualification: “While many factors affect heart disease, diets low in saturated fat and cholesterol may reduce the risk of this disease.” Similarly, a nutrigenetics company offers a “biological description of the effect of a genetic variation on cholesterol metabolism” – not a prediction that a person will meet a tall, dark stranger late at night in a garage and have a heart attack from fright.

Using the intentionally loose FDA food labeling guidelines as the model, I think, falls into the category of “two wrongs don’t make a right.” But it serves a purpose – it absolves the nutrigenetics companies, at least for now, from entering the fracas over genetic test regulation. That is an entire, huge area I’ll reserve for a future essay. In a nutshell, of the 900+ genetic tests, FDA regulates only about two dozen, as medical devices; the rest are dubbed “home brews”. The 1988 Clinical Laboratory Improvement Amendments of 1988 (CLIA) do not cover validity of genetic tests, despite many requests for their revision. Nor need labs publicly disclose the scientific basis for their tests, according to testimony of Kathy Hudson, director of the Genetics and Public Policy Center at Johns Hopkins University.

If the rebuttal company wins the contest for delivering exactly what they promise, they lose ground when they attempt to justify the tests. They seemingly want to have it both ways, admitting that “… the presence of the particular variation does not necessarily mean that each individual will respond in the same way.” This is directly counter to their website’s “Why settle for a generic diet or lifestyle plan when you can have a personalized program based on your unique genetic make-up?” Homocysteine level is offered as an example. Ranges of levels of this amino acid track with genotype of the MTHFR gene, that’s true. But the rebuttal points out that high homocysteine is a risk factor, not a diagnosis, and can be measured without a genetic test. Again, why do the genetic test?

(Interestingly, this particular company’s website prominently sites their beginning as June 8, 2000, to coincide with the day that Francis Collins and J. Craig Venter announced the sequencing of the human genome. It’s common knowledge that the date was selected because it was the only opening on the White House calendar. The website claims that the company was founded on a “simple premise … to utilize the scientific information uncovered in this landmark study…” But many of the 19 genes that the company tests have been studied since as long ago as the early 1980s, when no one had even dreamed of a human genome project. An unfortunate bit of historical hype.)

First, Do No Harm

For most people, the advice offered by nutrigenetics companies probably can’t do much harm. But the company behind the rebuttal, at least, does not ask for information on health history or current drug usage. And so a problem could arise, given that we have no idea how most genes and environmental factors interact. The GAO report offers the example of a person with a retinoic-acid sensitive form of lung cancer being told to up his veggie intake. Extra beta-carotene could hasten the growth of the cancer.

I’m not too worried about the danger of web-based nutrigenetics tests, though, because consumers are savvy. I think the media reports on the Congressional hearing will go a long way to counter the positive spin that health and fitness magazines and the like have given these types of tests. Economics could be an issue. The genetic tests aren’t all that costly – 19 tests for about $250 isn’t unreasonable. Hopefully consumers, after they recover from the pricetag of the suggested supplements (assuming they do not have gene variants that compromise their heart health), will figure out cheaper ways to get what they think they need.

At least the dog and cat DNA samples were outed, returned because they “couldn’t be processed”.

The most recent addition to the team at The American journal of Bioethics, new Corresponding Editor Ricki Lewis is a regular contributor to many major science journals, author of several textbooks, and just generally insightful. Welcome.

View blog reactions

| More