The Wild Genetic Goose Case: An Interview With Neil Holtzman
by Brandon Keim

A renowned researcher, scholar, and activist, Neil Holtzman has studied for decades the medical use of genetic technologies. He has largely focused on the policy implications of the technologies, genetic screening, and the use of genetic tests to determine susceptibility to common diseases or the possibility of altered reactions to drug treatments. From 1995 to 1997, he was Chair of the Task Force on Genetic Testing of the National Institutes of Health - Department of Energy Working Group on Ethical, Legal, and Social Implications of Human Genome Research.

For years you've denounced 'genohype', a term which seems two have two separate but related targets. First is genetic reductionism, the idea that genes are responsible for more than they are, and second is the promotion of treatments and therapies based on genetic information. What do you think is driving this genohype?

I don't think all genetic research, or all the attempts to associate genes with disease, is misplaced. There are, as you know, Mendelian single-gene disorders, and the genome project has accelerated discovery of the genes that are implicated in them. Although I never like to use the term 'determinism', one has to recognize that, for those rare disorders, the genome project has paid off handsomely. But they account for a very small portion of all human disease -- well less than five percent, even as we discover more and more of them.

It's possible, one might say -- perhaps naively -- that success in finding genes involved in Mendelian diseases, which extends much farther back than the genome project, stimulated people to search for the role of genes in much more common diseases. I think there was some naivete there. Many of the people who were making discoveries in the genome project were not geneticists, but molecular biologists, and had not really studied the complexity of what have now come to be called 'common complex diseases,' which I think geneticists were aware of. But even those who were aware lent themselves to pursuing common complex diseases.

There were a few early successes that further contributed to this search. The discovery of BRCA-1 and BRCA-2 was felt to be a signal that we could continue to find genes for a small proportion of the incidence of a common complex disease -- breast cancer in that case, and then hereditary nonpolyposis colorectal cancer -- and that we could nibble away at common complex diseases which might turn out to be nothing but an aggregation of fairly 'high-penetrant' single genes.

Could you define 'high-penetrant' genes?

A penetrant gene is one that is almost always expressed. Rather than use the term 'gene', you should really say 'allele.' So if someone had an allele for another gene like BRCA-1, it would mean that person had a very high risk of developing whatever the disease was.

That's where the genetic epidemiologists came in. If one looks at populations which have a common complex disease, if there were highly penetrant alleles that accounted for those diseases, the pattern within the family that possessed the particular genotype should be very close to Mendelian; if it's highly penetrant, then every family member, or almost every family member, who possessed that particular genotype would get the disease. But we weren't seeing those patterns. So it was clear, quite early on, that we were on a wild goose chase, and the molecular biologists would have seen that if they'd gone back and looked at the genetic research.

So, to answer your question, I think there was genuine naivete, and optimism that we would find more and more genes for common complex diseases. Now, as the search progressed, and announcements were made -- and then retracted -- that associations had been made with disorders like schizophrenia, bipolar affective disorder, and asthma, it became very apparent that we were not covering a large number of highly penetrant genotypes that were accounting for those disease. Instead, two things were happening. One, within a small number of people for some of these diseases, we were discovering rare highly penetrant diseases. So, in addition to colon cancer, breast cancer, Parkinson's, Alzheimer's, a rare form of macular degeneration, and several other diseases, we were finding a very small proportion of affected people who had a high-penetrant genotype. But the cases were often very early-onset; so, where we found high-penetrant genotypes, they were usually in atypical cases.

Among the more typical cases, what were reported, and what were difficult to corroborate, were low-penetrant genotypes that appeared in some populations to increase risk -- but only to a very small extent. Here I think there is still a lot of misunderstanding and ignorance about what these genes mean, and the mass media played a role, right up until about 1999 or 2000, in giving discoveries of low-penetrant genotypes an inordinate amount of publicity.

These low-penetrant genotypes confer such a low risk on an individual that they're hardly worth pursuing. For instance, if you take something that sounds on the surface as though it ought to be quite important, such as a genotype which confers a relative risk of disease of two, so that a person who possesses this genotype would have twice the chance of getting a disease as someone who didn't, it sounds quite striking. Almost all of the associations that have been found, although they're not often reported that way, report odds ratios that are slightly higher than one.

Now, if you turn that around, recognizing that many of these associations are with very common genotypes, and you calculate the chance of somebody who has such a genotype actually getting the disease, it turns out often to be less than one in four. It's very doubtful, particularly when dealing with common genotypes, that people will act -- or should act -- on a relative risk of twenty-five percent or less. Nor do we always know that these genotypes are conveying information about the mechanism of diseases. I think these are increasingly recognized as being not-very-fruitful areas of research.

A geneticist was describing to me what drives the continuation of genohype in his field. The younger generation of scientists largely recognizes the complexity of genes in relation to development, he said, but funding is typically available for those who say, 'X implies Y, it's very straightforward and simple' -- so they end up translating their own work into reductionist language, and pursuing reductionist projects, because that's what they can get the money for.

I would hope that people reviewing research grants these days don't fall as often as they used to for deterministic, reductionist claims made in grant applications.

Theresa Marteau and I wrote a paper that the New England Journal of Medicine, much to my astonishment, rapidly accepted, called 'Will Genetics Revolutionize Medicine?' In the article we pointed out what I described in response to the previous question. Other geneticists, particularly Ken Weiss, Joe Terwilliger, and several others, have really raised the issue of complexity to the point where funding agencies in the government should recognize that this is a low-yield area of research. I've said before that I think the government, and the NIH in particular, should get out of the business of funding searches for genes or genotypes for common complex diseases because it is so low-yield. If private industry in this country wants to do it, so be it.

There's another path that's been taken. It's best characterized by DeCode Genetics, a private company which succeeded in getting the Icelandic government to grant them access to medical records of all the people of Iceland. The idea was that the Icelandic population had a common origin, so it would be easier to find associations between genes and disease. There are a number of other variants on this process: Estonia is one, there is another in the United Kingdom, and there is even talk at NIH of trying to gather a large cohort of people who could be genotyped and then followed for the appearance of common complex disease. I think this is a waste of resources. It's very clear that if you want to find these small relative risks and have them corroborated, you need enormous populations, because the risks are so small and the predictive value so tiny to begin with.

I've looked at a number of papers written by Kari Stefansson, DeCode's president, and the odds he's published of associations in the Icelandic population between particular genotypes and diseases are very low. It's not persuasive to me that this is a fruitful line of research, or one that should be supported by public funds.

So far you have discussed predictive applications of genetics. Do you think there are any more useful avenues of research in pharmacogenomics -- in tailoring treatments to an individual's genotype?

It's actually quite analogous to disease. There are a few genotypes that do alter a person's reaction to particular drugs, and we've known these for years. One very striking example is between glucose 6 phosphate dehydronyase deficiency -- G6DD -- and hemolytic anemia following the ingestion of one category of malaria drugs. It turns out that the forms of G6DD deficiency are present in fairly large frequencies among Black and Asian populations. Even though the predictive values may not be very large, they are larger than those between genotypes and common complex diseases, and the reactions are severe and quite predictable. There is also a rare genetic predisposition for adverse reactions to a category of drugs used to treat one form of leukemia, and people who have it cannot metabolize these drugs and often have severe, and sometimes lethal, reactions. That genotype occurs in about one in three hundred people, which is considered to be relatively rare, though I would argue that for this particular form of leukemia people should be tested for this rare genotype before the drug is administered.

Having said that, there has been a lot of interest by the pharmaceutical industry and government in looking for other genotype-related reactions. In 2001, I published a chapter in a book by Mark Rothstein on pharmacogenetics where I looked at the claims of pharmacogenomicists, and when one did the arithmetic, and translated the associations and the odds ratios and the relative risks and the predictive values, they weren't very high. The chance that someone who ingested one of the drugs they discussed would have a bad reaction or would not respond to the drug was only about one in three or one in four, and very often there was no alternative drug to give people. Many physicians, even if aware of this, would have no choice but to use these drugs. It would be good medical practice to simply follow patients closely, starting on low doses and titrating them, and not relying only on the report of a pharmacogenetic test, which says that a patient has a greater chance of having a severe reaction. The tests simply aren't good enough.

Why do these tests, whether for diseases or for drug reactions, only discover genotypes of low penetrance? It's not a fault of the tests. When the laboratories which perform them are of high quality, the tests can be quite accurate. We're not talking about test mistakes; we're talking about the fact that genotypes themselves do not always result in disease. The reason is that there are many other factors: genes that we haven't yet identified that also have low penetrance, or may even have no penetrance by themselves at all but are expressed only in reaction to a disease when other genotypes at independent loci are present, or when there are environmental factors. Clearly this is the case with drug reactions, but it may be the case with diseases, too.

So, in the area of pharmacogenetics, there is some promise, but not very much. The pharmaceutical industry is probably split on this. The people who are working on pharmacogenomics think it's terrific, but other people are aware of the relatively low success rate, and recognize that many of the promised tests will have a lot of false positives: people who would not have a reaction to the drug, but look like they might -- and it's impossible to predict who they will be. So there would be a smaller market open to the pharmaceutical industry, which would have to try to develop an alternative for those people. That brings up a tricky issue. There has been the assocation, which I think is largely spurious, between drug reactions and so-called race categories. Depending on circumstances, the drug companies may say, 'Well, these people aren't going to pay for the drug anyway, so why should we bother with an alternative?'

I'm not deprecating all of genetic research, though I've sometimes been accused of that. The yields for single-gene diseases have been very impressive; treatments have lagged behind, but as we learn more about those diseases, treatments will come. From a clinical point of view, continued research into single-gene diseases which follow a Mendelian pattern of inheritance is certainly important -- but, beyond that, the area of basic research, of trying to understand how genes function, of things like gene-protein relations, is extremely important. Further research into evolution, including primate and human evolution in terms of genetic differences, is also important in telling us much more about our origins. Another vital area is the unquestionable fact that in many common diseases, somatic genetics plays a role. As a disease unfolds in an individual, for reasons we don't understand, but where there is no clear pattern of inheritance, genes in certain tissues subjected to environmental stress will undergo reactions and will mutate in one particular cell and all the daughter cells derived from that cell. These mutations are not transmitted from one generation of individuals to the next, but occur within a particular cell or tissue of one person. That's how most cancers develop: not because of inherited mutations, but because of acquired mutations in particular cells in, say, a breast or lung, after which further somatic or acquired mutations transform the tissue from healthy to malignant. This area is receiving some attention, and should receive much more.

Stepping back from the utility of medical genetics, what sort of regulatory system does the United States have in place? What are its strengths and shortcomings? Are there regulatory changes which need to be made -- and, if so, where does the impetus need to come from?

As you probably know, I chaired the NIH/Department of Energy task force on genetic testing. We met from about 1995 to 1997, and produced a report called Promoting Safe and Effective Genetic Testing in the United States. I'm quite proud of that, for it laid out what should be done to validate a genetic test. We introduced terms that are now widely accepted: analytical validity, clinical validity, and clinical utility, and defined each of them. They had been used, one or the other, before, but in a fairly loose sense. We said there had to be some demonstration of these three characteristics before a test became routinely used in clinical practice.

When we looked at the regulatory framework at that time, which is still the framework today, we found that it was particularly lacking in the area of tests which are marketed as services. For instance, laboratories today do not sell kits for testing BRCA-1 and BRCA-2, but market a service. If a doctor collects a specimen from a patient for whom he thinks BRCA testing might be useful, he must send it for analysis to the laboratory of Myriad Genetics, a private company which owns the patent on that test. For years Myriad has successfully claimed that they are not under FDA purview, since they provide their tests as a service rather than as a kit, and to this day this company has never made available data on the clinical validity of their BRCA-1 and BRCA-2 tests. By contrast, companies that manufacture kits which are then marketed and sold to physicians for use in their office laboratories, or to hospitals or other large clinical laboratories, must go through strict FDA study. The FDA requires that they demonstrate clinical validity. They must establish the sensitivity, specificity, and predictive value of the test.

When the task force met, we asked the FDA whether they had the legal authority to regulate tests marketed as services. They said yes, but claimed that because of resource limitations they could never do so. It also turns out that the medical device industry lobbied very hard against this sort of regulation.

We started talking with the FDA about how they could regulate tests marketed as services. We began to develop with them algorithms for doing so. And then, when the task force disbanded, the Secretary's Advisory Committee on Genetic Testing was developed, partly at our recommendation, by Donna Shalala, the Clinton administration's Secretary of the Department of Health and Human Services. They did what we couldn't do: they recommended in their report that the FDA should regulate all genetic tests, with the clear implication that they regulate tests marketed as services as well as test kits. We then had a change in administration; the committee was disbanded, and a number of its more regulation-prone members were removed when a new committee was created. Looking at the agenda of this committee, it's not at all clear that they will again tackle the question of FDA regulation, though there is as great a need for it as ever.
There is an important regulatory role to be played, and it has been dropped.

How do, and should, physicians approach medical genetic technologies?

In the late 1990's, one of my doctoral students did a study in which she surveyed physicians' use of genetic tests, particularly for breast cancer. What she found in that area, and to a lesser extent with cystic fibrosis, was that physicians ordering the tests did not have a good knowledge of either the disease or the characteristics of the test they were ordering.

Clearly there is a need for educating physicians. How that need is being fulfilled I don't know. One of the problems is that, largely under the instigation of Francis Collins -- Director of the National Human Genome Research Institute and a leader in the Human Genome Project -- there is something called the National Coalition for Health Professional Education in Genetics, which has as its main task the education of healthcare providers about genetic tests. I think it's partly promotional, designed to make sure that, when there are tests available, physicians will use them. Physicians also receive a great deal of promotional literature from companies like Myriad. The organizations responsible for education need to be regulated. Which brings me to another point: direct-to-consumer marketing. Myriad last year took out television commercials for their tests in Denver and Atlanta. How much that stimulated business I don't know, but the probability is that many women who had their physicians order tests were so low-risk that the test was of no benefit. In fact, the genetics community has now developed a good protocol for ordering genetic tests like those involving BRCA-1 and BRCA-2. The risks must be fairly significant for them to be ordered.

There are now a number of websites, documented in the peer-reviewed literature, that offer genetic tests. In some cases, these sites offer tests that I never heard of, such as 'neutragenomics.' They offer tests for which there is no evidence at all that they predict anything to any degree.

In another example, a genotype was found and thought to be highly predictive of hemochromatosis, a disease which is usually adult-onset and leads to severe iron overload and eventually liver, kidney, and heart failure. It turns out that many people who have the genotype never develop severe iron overload, or any overload at all -- yet there is now a consumer organization that promotes genetic screening for the genotype. These are indications of inappropriate use of genetic tests and the failure of regulatory agencies to control them.

Do you think that advances in medical genetics will be widely accessible, or will they be affordable by only a few?

In countries with some form of universal health insurance, there's a greater chance that, if these tests prove to have clinical validity and utility, they will be equitably distributed. That's not to say that countries with national health insurance will decide to cover them -- but at least there's a format and protocols for looking at new medical interventions and deciding whether they are appropriate for coverage.

In this country, of course, it's a completely different situation. I think that, as is the case with something as expensive as in vitro fertilization, it's not going to be equitably distributed. That's compounded by the fact that many genetic tests are patented. That is the case with Myriad's tests, about whom there has been a great deal of international controversy. The company has a monopoly, and has optimized its price to maximize its profits. Provinces in Canada, particularly Ontario, and in the United Kingdom and France, have really acted very vigorously not to adhere to Myriad's claims of exclusivity. Ontario, for instance, provides a test much more cheaply than Myriad, making it available to all women in its health system. If they had to adhere to Myriad's pricing, they couldn't do it. Patenting claims are an important complicating factor.

There is one area of genetics in which equity, even in the United States, has been practiced, and that is newborn screening. Every infant in the United States is screened for a number of disorders, and the number of disorders which we can detect has increased in the last several years because of a new technology called tandem mass spectrometry. But it turns out that this technology is patented, with the rights held by a company called Pediatrix. So far, Pediatrix has not asserted its claim, but it could do so. If Pediatrix did, it could undermine statewide screening programs and the universal availability of newborn screening.

Looking into the future, what do you see as the role of genetics in medicine in, say, ten years? A generation from now? Do you see points of possible divergence in these outcomes, where one set of key decisions, or a certain social climate, will lead to one particular future, and another set to a different one?

I think a worst-case scenario is that weakly predictive tests will get out into the marketplace and that their usefulness will be exaggerated. If that happens, the notion of genetic determinism -- of blaming the victim, of holding his or her genotype responsible -- is likely to diminish efforts to improve the environment and social situations. That would be disastrous.

We are not going to learn much or accomplish much by the advancement of tests for common complex diseases, and they will deter us from other avenues of research, particularly with respect to environmental factors and genetic-environmental interactions, that would have a much higher yield. I hope there will be a growing realization that this is not fruitful, and that attention given to the genetics of common complex diseases and pharmacogenetics dwindles.

On the other hand, I think we will make, with sufficient funding, advances in basic genetic research, as was the case in 1953 with the discovery of the double helix. I don't think it's possible to predict where that basic research will lead us. It may be that scientists will be clever enough to be able to identify what I call 'constellations' of genotypes -- that is, particular alleles at independently segregated loci, which are inherited independently of each other -- that will have a significant amount of predictive value. But I don't see that happening in the foreseeable future.