The research began two decades ago, ostensibly to search for a genetic variant that might be contributing to the increasing rate of diabetes in the tribe. The diabetes research proved unfruitful, but the blood donated by the Havasupai tribe members, and the DNA extracted from it, led to a number of follow-on research projects, grants and publications. It was that research – including searching tribe members’ DNA for variants linked to schizophrenia, and inferring the likely ancestral origins of the tribe’s founders – that led to lawsuits, millions in legal fees and, ultimately, the settlement.

Implications of the Havasupai Settlement. Harmon’s article provides a concise background to the dispute, and briefly describes the $700,000 settlement between ASU and the tribe to “remedy the wrong that was done.” Harmon and unnamed “legal experts” suggest that the settlement is significant because “it implied that the rights of research subjects can be violated when they are not fully informed about how their DNA might be used.”

In some respects, this is a trivial conclusion. One of the most important and well-known elements of the Common Rule – the regulatory regime that governs federally-funded human subjects research – is that researchers must seek, and participants provide, informed consent. Participants that are uninformed cannot provide valid consent and, thus, their rights as subjects are violated. In that respect, at least, the Havasupai case tells us nothing new. (I have not seen the settlement, but I doubt that it will (a) be made public or (b) contain an express admission of guilt from ASU, both factors that will limit its relevance to future similar scenarios.)

But the Havasupai case and Harmon’s article shine light on an important, and difficult, problem that continues to face scientific researchers, particularly those exploring human genetic variation: what does it really mean to provide “fully informed” consent for genomic research?

Fully Informed Consent? Looking at the text of the Common Rule, there are requirements that the researchers describe the nature and purposes of the research (§ 46.116(a)(1)), as well as both reasonably foreseeable (§ 46.116(a)(2)) and unforeseeable (§ 46.116(b)(1)) risks of participation. But the standard of informed consent that must be achieved is not explicitly spelled out. Is it “reasonably informed,” “substantially informed,” “fully informed” or something else altogether? (Interestingly enough, the language that Harmon uses – “fully informed” – does appear, but only in sections regarding research on pregnant women and fetuses, which is not applicable in this case.) Even if a clearer standard were articulated, how would researchers demonstrate that it had been satisfied?

The requirement of informed consent is part of the bedrock of modern human subjects research in the United States, and it is not going anywhere. In fact, the story of the Havasupai, as well as the tale of Henrietta Lacks (told so remarkably well in Rebecca Skloot’s new book) and other past failures of informed consent, suggest the informed consent requirement is here to stay, as well it should be.

A Difficult Balance. Yet, as we push forward into an era of large-scale, personalized genomic research, it is impossible to ignore the difficulties – legal, ethical and practical – that informed consent requirements impose. For example, truly informed consent for genomic research might require participants to possess a deep – or at least working – understanding of the underlying science. That sets a very high bar, and finding sufficient numbers of participants capable of providing such consent could restrict important research.

Even more daunting, however, is the difficulty of fully informing participants of the benefits and risks of participation in genomic research – particularly where the resultant findings could conceivably be linked back to the individual or, as in the case of the Havasupai, the individual’s community – when the researchers themselves lack this understanding. What genetic information can tell us about disease and other traits, and how this information can be used or misused in the case of individuals, is an area of continuing uncertainty. With the publication of the draft human genome sequence a decade in the rearview mirror, we know more than ever before. But there is still much that we don’t know.

Thankfully, new research models and strategies for seeking informed consent are being developed and tested. For instance, the Personal Genome Project (PGP) – for which I am an advisor, including with respect to the informed consent protocol – employs a model of “open consent.” The PGP focuses on preemptive and extensive risk disclosure, along with rigorous participant pre-screening to ensure that the risks of participation are understood. More broadly, the difficulties of informed consent and genomic research is an issue that the NIH is studying on an ongoing basis, with multiple internal working groups looking at different dimensions of the problem. Nevertheless, informed consent for genomic research poses a considerable challenge for policymakers, funding bodies, researchers and participants, and it is unlikely that any of the existing models represent a perfect approach.

None of this should be taken to mean that informed consent for genomic research is impossible. To admit that would leave us with the unenviable choice of sacrificing either the informed consent of participants or the valuable scientific research they enable. What the case of the Havasupai tribe does underscore, however, is just how difficult a task this is. It is clear that the next generation of personal genomics research will require more than purely scientific breakthroughs. We also need to think creatively about the ethical and legal framework in which such research is conducted, to make sure that it continues to promote scientific progress while protecting the participants – no matter what their background – that make such research possible.

What makes the RPGEH proposal so exciting, from a research perspective, is not just the 700,000 SNPs that will be genotyped for 100,000 patients, although that alone would represent one of the largest genetic research databases currently in existence. The real value lies in the marrying of genetic information with robust medical, environmental and other phenotypic data that Kaiser already maintains as a health care provider. From the RPGEH’s official description:

Researchers are collecting medical, lifestyle, demographic, environmental and, in some cases, genetic information from up to 500,000 Northern California Kaiser Permanente members.

DNA from participants’ saliva and/or blood samples will be used to obtain information about genes and exposures to environmental contaminants. This information will be entered into the new, secure database that contains participants’ survey responses and medical record information. Once we have all this information in one secure place, we’ll have one of the largest databanks of its kind in the United States.

The combination of genomic and phenomic information is widely recognized as a key step in constructing the large, rich datasets that will enable researchers to continue to explore the complex pathways by which a genome is ultimately manifest as an unique human being.

Not long ago, the rate-limiting step in building comprehensive genomic research databases was the generation of genotypic data. After all, sequencing the first composite human genome took over a decade and cost several billion dollars. In recent years, however, the cost of genomic sequencing has declined precipitously. Continued scientific and technological advances seem certain to render genomic sequencing a commodity in short order, perhaps within as little as 3-5 years.

As the generation of genomic information is rendered faster, easier and less expensive, the rate-limiting step is increasingly at the other end: the accumulation of accurate and comprehensive phenomic data. Detailed information on factors like environmental exposures, medical histories and other physical and behavioral characteristics is necessary to construct a complete picture of the process by which genes are converted into traits in individuals. And that is where health care providers such as Kaiser, who are in the business of collecting detailed phenotypic information about their patients, have the potential to provide an invaluable resource.

As Cathy Shaefer, a Kaiser research scientist, puts it in the Technology Review, “The importance of this project is that it will, almost overnight–well, in two years–produce a very large amount of genetic and phenotypic data that a large number of investigators and scientists can begin asking questions of, rather than having to gather data first.”

But what do the 100,000 to 500,000 Kaiser patients receive in return for their participation in the RPGEH? The big picture answer is that those patients will contribute to the creation of a scientific database of indisputable value for the study of human health and disease, unquestionably a sufficient reason for any individual to choose to participate.

But at an individual level, there’s not much else to look forward to. Although patients’ health records will be combined with newly generated genotype data for purposes of the RPGEH database, that genetic information will not be made available to the patients themselves, or to their doctors, except at the discretion of RPGEH scientists. From the RPGEH’s FAQ:

Will I receive any personal benefit from giving a saliva sample that may impact or improve my own health?

This research is not intended to benefit individual participants directly. However, we hope that results from the research will improve how clinicians diagnose, treat and maybe even prevent major illnesses in the future.

Will I be getting any test results?

You will not receive personal health or medical results from taking part in the RPGEH. We do not expect that results from the RPGEH will be the kind of information that can be used by you or your health care provider to make decisions about your current health care. However, if scientists discover information as a result of RPGEH research that we believe is of substantial medical importance to you, we may contact you and ask if you want to learn the results.

As Catherine McCarty touched on in her ELSI piece last week here at the GLR (“To Share or Not to Share: That is the Question”), the logistics of returning genetic results to research participants is a key issue facing the fields of genomics and personalized medicine. As McCarty notes, “research findings . . . that were initially tentative may become truly significant and ideally clinically relevant. Researchers will then need to consider the ethics of withholding information known to be clinically relevant . . . .” Although it depends on the exact SNP data generated by the RPGEH project, with 700,000 SNPs per participant and 100,000 participants, it seems a statistical certainty that some of the genotype data will be of clinical relevance.

The norm in the field of genetic research continues to be the RPGEH approach, with research projects generally declining to return information to participants or patients except in cases of “substantial medical importance,” the threshold for which is unlikely to be clearly defined. New projects such as the Coriell Personalized Medicine Collaborative also restrict the information that is returned to participants, but are more specific about what information should be returned, focusing on genetic variants that are deemed potentially medically actionable.

Still other projects, most notably the Personal Genome Project, are testing traditional genomic research norms by returning all research data directly to participants, a formula that has required the PGP to take steps (pdf) to ensure that participants are given some guidance as to the information content of that data and reminded, repeatedly, of the importance of confirming any research findings with a clinical care provider before taking (or avoiding) any particular course of clinical action.

There is little question that as the cost of genomic sequencing falls, large-scale research databases such as Kaiser’s will continue to proliferate, helping researchers and scientists push the boundaries of genomic understanding. These datasets cannot exist without patients and participants willing to contribute their genomic and phenomic information. Whether we have now reached the point that McCarty and others have envisioned, at which a compelling argument can be made that research data is of such individual value that it should be returned to participants, the very size and promise of the RPGEH project signal that point may not be far off.

The Importance of Open Consent

In considering the use of DNA samples and phenotypic data provided by children to biobanks, Gurwitz et al. argue that the traditional notion of confidentiality or anonymity, at least when it comes to genomic data, is an illusory one:

DNA remains unique as a permanent identifier throughout an individual’s life… As sequencing of entire genomes becomes a routine procedure, DNA donors’ privacy can never be completely ensured within biobanks. Individuals can be traced even in very large aggregate data sets spanning thousands of donors. As a consequence, there is no ‘opting out’ from biobanks once DNA sequences have been published and deposited with public databases.

Along with one of the co-authors of the Science piece (Lunshof), I’ve written previously about the inability to promise privacy in the genomic context (pdf). That premise, coupled with the determination that informed consent requires open and complete disclosure of the risks of participation in genomics research, has served as the basis for of the Personal Genome Project’s (PGP) informed consent protocol (pdf):

If you are enrolled in the PGP, your genetic and trait information will not be maintained or made available in a confidential or anonymous fashion. Your genetic and trait information will be made available via a publicly accessible website and database….

But the PGP’s “open consent” model, which works in the case of adults, who are capable of providing informed consent to serve as “health information altruists,” does not work for children.  Very simply, children are unable to provide informed consent at the time of participation.

For population biobanks (which the authors treat separately from disease-specific pediatric research), Gurwitz et al. propose a regime of tightly controlled access to the DNA samples and data supplied by children, at least “until donors are recontacted as adults and give their own informed consent.” They propose policies that would “minimize the risks of revealing children’s identifying genetic data, thus protecting their privacy, while still allowing the advancement of pediatric research.”

While this is a laudable goal—protect privacy while still allowing pediatric biobanks and research to thrive—it is an impossible one, for the very reasons that Gurwitz et al. pointed out only a few paragraphs before. “DNA donors’ privacy can never be completely ensured within biobanks.” Despite the robust privacy apparatus proposed by Gurwitz et al.—which includes restricted access by third parties, publication of aggregate instead of specific data, and procedures for re-consent upon adulthood—sharing genomic data will always subject participants to both known and unknown risks, including re-identification, as demonstrated by the NHGRI’s recent experience with publicly available dbGaP data.

To their considerable credit, Gurwitz et al. admit that “there are no perfect solutions.” What they propose, ultimately, looks something like this:

• We recognize that genomic privacy can never be fully ensured in the biobank setting.
• Although with adults this risk may be addressed via a robust informed consent process, children are incapable of providing informed consent, at least until they reach adulthood.
• However, pediatric biobanking research is of crucial importance to scientific research and cannot be prohibited entirely.
• Therefore, we will protect privacy as best we can—realizing that this protection will be necessarily imperfect—and determine that the remaining risk to children is outweighed by the benefits that accrue to scientific research and to society at large.

Although this is a chain of reasoning that, ultimately, I agree with, the importance of the issue necessitates that the policy be made more explicit than Gurwitz et al. manage (although, in their defense, they were given a scant two pages to make their case).

Not made explicit enough is which party will weigh the benefits of submitting a child’s genotypic and phenotypic data to a biobanks against the risks imposed by the possibility of a breach of privacy. Although researchers may view the risks of re-identification as overwhelmingly outweighed by the benefits, in the absence of the child’s ability to provide his or her own consent the decision must fall to the parents of that child, and not to the researchers. Whether with children or adults, the solution remains complete openness and transparency as to the nature of those privacy risks, thereby satisfying the principles of veracity and voluntariness that underlie truly informed consent.

I’ll close with two final points raised by Gurwitz et al. and the surrounding commentary (see the Times Online, GenomeWeb and Nature).

Research vs. Medical Use of Pediatric Genomic Data

Mark Henderson, the Science Editor of the Times, has responded to Gurwitz et al. with an excellent commentary of his own in which he concludes that “as and when knowledge of a person’s genome sequence has clear medical benefits, it would be wrong to deny these to children purely because they cannot consent to access their DNA data.” Henderson’s point is an important one, but it risks confusing two distinct issues. Whether or not it is appropriate to use the DNA of children for research purposes has no bearing on whether or not the same DNA could and should be used for medical purposes.

It is non-controversial for parents and doctors to collaborate to make medical decisions that they conclude are in the best interest of a child, quite often in the face of the child’s vociferous dissent (as The Onion humorously notes, the majority of children are opposed to children’s healthcare). But a doctor’s examination of a child’s DNA for an identified medical purpose—even if it is then stored in the child’s medical record (which, as I discussed earlier, may not be a foregone conclusion)—is a far different proposition from submitting the same DNA to a biobank for use by the scientific research community.

DTC Pediatric Genomic Research

Gurwitz et al. hint briefly at what may be a coming paradigm shift in genomic research: the development of for-profit, direct-to-consumer (DTC) genomics companies as legitimate and important sources of genomic research. Gurwitz et al. write that “apart from biobanks, some direct-to-consumer personal genomics providers…are already analyzing children’s DNA samples, creating another avenue by which personal privacy might be compromised.”

I covered this topic last month (see “Genomic Research Goes DTC”), and the fundamental point remains the same: most, if not all, of the human subject protections that apply to genomic research, including in the case of children, are unlikely to cover much of the activity undertaken by the DTC industry. While it’s possible that companies such as 23andMe—which appears to be leading the DTC genomic research race—can self-regulate effectively, in order for any changes to the current practices of pediatric genomic research to apply to DTC genomics companies, it is likely that considerable structural changes would be required to the system of human subjects research protections generally, which would be a decidedly non-trivial task. But as Gurwitz et al. write in conclusion, “the long-term benefits of maintaining public trust in biomedical research” may “justify extra governance efforts and added costs.”