Ethical Considerations in Direct-to-Consumer Genetic Testing: An Expert Interview With Wylie Burke, MD, PhD

Shira Berman


September 09, 2010

Editor's Note:

A recent publication in PLoS Genetics by the direct-to-consumer genetic testing company 23andMe[1] described results of their analysis of data collected from consumers who purchased their testing kits. The accompanying editorial[2] raised several ethical concerns about the way the study was conducted and how it affected their decision to delay, but ultimately publish, the study.

On July 22, 2010, the Government Accountability Office issued a report[3] raising additional concerns about the use of direct-to-consumer genetic tests and their clinical utility.

In an interview with Medscape, Wylie Burke, MD, PhD, Professor and Chair of Bioethics and Humanities at the University of Washington in Seattle, discussed these and other ethical issues associated with direct-to-consumer genetic testing.

Medscape: Let's start with the design of the 23andMe study. Their model was to use self-reported phenotypic data collected via the company's Website and compare it with genotypic data submitted in saliva samples. Because the data are proprietary, they are not available to other researchers. What might be the impact of these factors for future studies down the road?

Wylie Burke, MD, PhD: The use of a Web-based infrastructure to generate large datasets addressing gene-trait associations is an interesting, innovative model. However, this model raises 2 significant questions. First, how reliable are self-reported phenotypes? The phenotypes reported in the paper are primarily those that most of us would agree can be reliably self-reported, such as curly vs straight hair. But, there are a relatively small number of such phenotypes. We know from the paper itself that self-reported phenotypes may be influenced by knowledge of one's own genetic results, and there may be other factors that influence reporting that we're unaware of.

Second, this type of study is tapping into a very nonrepresentative sample. It doesn't represent the full spectrum of socioeconomic status. So, to the extent that the study is assessing traits that are influenced by environmental as well as genetic factors, it's not really able to capture the diversity of environment. As innovative as this model is, it has very limited applications in its current iteration for producing high-quality research.

The question about data availability challenges us as to what we ask from research infrastructures that are going to give us high-quality data. Putting data into a federal data repository is not the only way to allow for adequate replication; presumably 23andMe or similar organizations could arrange for independent evaluation of the data. But, the fact that there is no such measure now is a limitation.

Medscape: One of the key points raised in the editorial was that the consent form did not specify that the data would be used in a research study. Can you elaborate on why this is such a critical issue?

Dr. Burke: The IRB that 23andMe ultimately consulted used the Office for Human Research Protections (OHRP) 2004 guidance to determine that this study was not human subjects research. The interesting question, then, is why they needed consent at all.

In truth, the OHRP guidance is actually very controversial. My understanding is that it was intended to allow for certain kinds of research to be declared "not human subjects research" even though they involved human samples.

It seems clear the guidance is referring to research in which samples are collected by one agency, then de-identified, and then made available for research in another setting such as a repository. For example, let's say a researcher conducts a study on cardiovascular disease and now has de-identified data that she is going to submit to a repository or share with another researcher who has no contact with the human subjects or any kind of identifying information. At that point, this guidance would say this is not human subjects research and therefore doesn't require Institutional Review Board (IRB) review.

Some people question whether it's a legitimate call that such data should ever be viewed as not requiring human subjects review. But, its application in the 23andMe study seems, to me, particularly questionable. The argument that the OHRP guidance applies to this particular study has to assume that there's one part of 23andMe that made the initial contact and obtained consent, and there's another part of 23andMe that is distinct enough to say it's a different entity that did the research, and therefore it was not human subjects research. And, that, I think, is a questionable judgment.

The other point about the OHRP guidance is that it clearly asks for a prospective determination. In other words, one needs to determine before the research is done whether the OHRP guidance applies. In this case, the record would suggest that the determination was done after the fact. So, to me, it is questionable whether the OHRP guidance applies.

Returning to the question of the consent form, what I'd look for in a consent process is one that prospectively informs people that they are participating in a research study.

From what I understand of the 23andMe consent and contractual agreement, people who read carefully do understand that their data are going to be used for research. But, as the editorial states, if the portions of the consent document that refer to research had been subjected to an IRB review (and we do expect the initial consent to undergo IRB review even if there's a subsequent determination about de-identified data being non-human subjects research), the consent would have probably been clearer and perhaps separated from the contract.

Nevertheless, although a prospective human subjects review process might have led to a better consent document, that doesn't necessarily mean that the consent document was insufficient.

The bigger problem is that consent for research was bundled with a contractual agreement to buy a service. In general, this is a problematic situation from a human subjects research prospective. We'd really like an agreement to participate in research to be a freestanding process -- clear and separate from any other kind of agreement that a person is entering into. There should be a clear, separate decision being made: Do I, or do I not, participate in research. And, it doesn't appear that the research opportunity was offered in this way by 23andMe, or that 23andMe offered the opportunity to sign up for the product without signing up for the research.

Medscape: What are the lessons that can be learned for other companies who might be considering conducting these types of studies down the road?

Dr. Burke: If I were advising these companies, I would encourage them to have a contractual agreement that describes the product, and then to have a separate section that would allow a person to decide whether to sign up for research. In other words, the research becomes a separate, voluntary choice. And if it's not a separate choice, if buying the product also requires a person to sign up for the research; that also should be made crystal clear.

Medscape: One of the concerns raised in the Government Accountability Office report is that inconsistent results were reported by different companies. What responsibility does a company -- or a physician -- have in ensuring that the patient understands the results? What expectations should a patient have from any direct-to-consumer (DTC) product?

Dr. Burke: We expect truth in advertising from all marketers of consumer products. We expect them to be clear about the product they're selling, and about what consumers will get for their money.

From the Government Accountability Office report, it seems that this standard of truth in advertising is not being achieved. They found some evidence of truly fraudulent claims, where some companies claim that a consumer's DNA could be used to create personalized supplements to cure diseases. Two of these companies further stated that their supplements could "repair damaged DNA" or cure disease. When companies make such claims, which have no factual basis, they are going well beyond the failure of truth in advertising.

But outside of this, the question of what a patient is getting, or what a doctor should expect, really goes to the heart of the issue for DTC genetic testing -- whether there is any medical value to the information that's being provided. Let's assume we're talking about a reputable company that's trying its best to report its results accurately. Their test may include single nucleotide polymorphisms shown in peer-reviewed literature to be related to age-related macular degeneration or diabetes or heart disease. We could imagine a situation in which the company could truthfully say that they're testing for these risks, describe what the data will look like, and confirm that the information they provide is accurate in terms of what we know from the scientific literature.

The problem is that these single nucleotide polymorphisms have no medical value. The vast majority of reported health-related single nucleotide polymorphisms are not tests that meet a clinical utility standard and are not currently in use in clinical care. They haven't passed the sequential process of review that occurs as a product becomes available, and then ultimately gets adopted into healthcare, with experts doing a careful, evidence-based review and generating practice guidelines.

I would argue that truth in advertising should clarify that point. It should make very clear that these tests are not currently recommended in clinical use and don't have any established clinical utility.

A separate complicating issue is that a small subset of tests provided in DTC genetic testing is, in fact, used in clinical care. For example, some DTC products offer tests for mutations in the BRCA1/BRCA2 genes that are associated with an increased risk of breast cancer. In this case, the concern is whether individuals who receive that information are getting other components of care at an adequate clinical standard. Are they getting good counseling? Do they understand what the risk data mean? Are they getting good advice about appropriate clinical responses to that information?

Medscape: Where does the responsibility lie for ensuring that appropriate follow-up is done? With the company, the physician, or the patient who bought the product?

Dr. Burke: I would say it's unfair to ask an overworked healthcare system to take on the general responsibility of educating people about the lack of clinical utility generated from these kinds of tests. What I think we really need is a more aggressive effort to define what "truth in advertising" really means in this kind of situation.

I suspect it would be very useful it the tests that had no established clinical utility were so labeled and clearly separated from those that do have clinical utility. For the latter, the labeling would allow people to better understand the test they were purchasing, the nature of the clinical use of the test, and what kind of services might be appropriate for them to seek based on their results.

The fundamental point is that we still have work to do in defining what is considered an adequate amount of information in order to feel confident that the consumer has the information needed before purchasing a DTC genetic test.