In the final session of the workshop, a panel of speakers and steering committee members reviewed the issues and discussions that they thought arose during the workshop and identified questions that, in their views, remain unresolved. The floor was then opened for questions and general discussion.
Most participants agreed that the National Health and Nutrition Examination Survey (NHANES) has great potential to help medical science, and the question is whether or not NHANES should become a 21st-century public health and medicine research study. As Les Biesecker noted, there are very good reasons to keep doing exactly the same thing over a long period of time, but hard questions should be asked about what the study is trying to accomplish, and whether a conservative approach or a more novel and creative approach is the right way to go.
Biesecker commented that only two things affect health: genes and environment. Further, he said, it may be the case that NHANES does not have, as an objective, a goal of understanding how one or the other of those things affects health. It would be scientifically justifiable to not have one or the other of them as an objective, and that justification should drive how the study is designed and carried out. Whether or not NHANES should explore what might be called this ideological terrain seems to
be, as he interpreted from some of the workshop discussion, an open question.
Gail Jarvik noted that the NHANES program espouses a dual mission: a primary mission involving participant health assessments and a secondary mission of biobanking for research. In her opinion, understanding and collecting genetic data is essential to the primary mission and to the scientific study of health. Gene-environment interactions are important in human health but very difficult to study, she said, and NHANES offers a unique opportunity to understand how these interactions impact health.
Wiley Burke and Marc Williams wondered about stopping to rethink what genomic parameters should or could be included in NHANES and what is reasonable within the context of NHANES. Williams suggested it would be useful to reevaluate why DNA collection was initiated in the survey, given the regulatory constraints that affect secondary use of the data. He said it may be helpful to restart the thought processes, better understand the landscape, and then take purposive steps forward. Burke noted that once NHANES leadership is clear on that and has carefully analyzed the ways in which genomic parameters could be added to the NHANES database, there would then be the question of what is returnable. Several workshop participants have suggested that there are not many returnable results to date. If this is so, Burke suggested, the process is to first figure out what research is worth doing, then figure out what DNA-based information will be generated, and then address which genetic information (presumably only highly actionable information) would be returned.
Ellen Clayton stated that, in her view, the first matter to be resolved is the extent to which NHANES data and samples can be used for research under the current constraints, or under any constraints that are foreseeable. She said there is a radical disconnect at this point in time between what is going on with regard to scientific data sharing generally and what is going on with regard to NHANES. The NHANES program needs to figure out what can be done with genetic data and how they can be made more readily available to investigators in a way that permits research.
Jennifer Madans stated it would be helpful for NHANES to have people who are interested in gene-environment relationships look at what is currently in the survey in terms of environmental information and the sample size, and try to simulate the array of useful information that could be derived. Many workshop participants expressed their opinions that the gene-environment data are a potentially unique source of research information. However, said Madans, NHANES has been counseled that
the sample sizes may be too small and that the statistical power to explore interactions may be lacking. It would be very useful to have a range of input on this empirical question, she suggested.
David Weir said he saw as a pressing need that the NHANES program has an ongoing system by which markers, preferably those that are attached to commercial chips, are identified and binned appropriately. Although different thresholds for when and where to report (based, for example, on how old the sample is) might be set, at least there would be an ordering that would provide a shared understanding.
As illustrated by different presenters and participants during this workshop, there is considerable ongoing debate among scientists, medical professionals, and researchers about the need to return results. Clayton argued that, to her, the answer depends on what is acceptable, rather than on individual preferences. Jarvik said that NHANES is not compelled to return results, and noted that at this point in time, there are no genetic results that are returnable. Not only are there consents saying that results will not be returned, she continued, but also some if not most of the genetic data are from non-CLIA (Clinical Laboratory Improvement Amendments) samples run in non-CLIA labs using non-CLIA procedures, and none has a high enough expected validity to compel return.
Clayton noted that the study context is particularly salient for this discussion. NHANES is a one-time, highly involved engagement with participants, but after a written report, there is no further contact. A question to consider is whether this one-time intervention leads to some long-term obligation to follow up.
This discussion led to exploration of duration of the responsibility to follow up. Susan Wolf mentioned some recent convergence around the idea that, in primary research, the duration of project funding may limit the duration of the responsibility. But with a biorepository, she said, there is no clear indicator of where responsibility ends.
With regard to the cost of returning results, and how this might detract from other NHANES or National Center for Health Statistics (NCHS) activities, Jarvik said that there are more expensive and less expensive ways to proceed. There are many already-studied roads that could be followed with reasonable efficiency to return high-impact results should they ever be found, she suggested, and there may be an unwarranted level of concern about the frequency of incidental findings. While NHANES does not have genomic expertise, the expertise does exist elsewhere. In her view, NHANES does not need to hire an entire board to
determine what to return, because many groups are working on that issue and their knowledge can be leveraged.
Wolf noted that NHANES is already returning many results and possibly also returning incidental findings or at least worrying about them. If one is scanning, she said, one is likely dealing with incidental findings. Echoing the sentiments of several workshop participants, Wolf found it implausible that NCHS would do a public study and say that it will sit on certain data and not return them to participants. In her opinion, this is a “nonstarter” in view of the fact that NHANES already touts the return of nongenetic results as a benefit of participation. She commented that some of the questions that this workshop has agonized about have been asked and answered already in the design of the study, and that it makes sense to contextualize the genomics issue in light of NHANES policies regarding nongenetic data.
Wolf suggested that NHANES could use a research mechanism to more rigorously ascertain what a population is thinking about return of results and how incidental findings are treated. There may be information to be learned about genetics from other parts of the survey: how the study is handling incidental findings revealed during scans, how participants are handling the return of STD information, and the same for pregnancy information. She further noted that there is a literature about how incidental findings can be both benefit and burden, and suggested this distinction might be highlighted in NHANES communications.
Jeffrey Botkin emphasized that, in his opinion, the NHANES-participant deal may be more complicated. There is the short term, in which well-validated, familiar clinical tests are being used to describe the participants, give a snapshot of their current physiology, and so forth, and he noted the guarantee or promise to return those results directly to people in the context of that short-term relationship around the study recruitment. Then, he continued, there is separate language around long-term research results, in which it is plainly agreed the participants will not receive those results. These two should not be conflated as the same promise. To him, both are part of the deal, and based on the explicit language of the consent form, one can make a distinction between the short-term obligations and the understanding about those results, and whatever obligation one might have longer term with experimental results that may not emerge for years or a decade or more.
The key for NHANES, according to Kelly Edwards, is adaptive governance, an approach to governance that would be dynamic and responsive
to the changing landscape of health research. Edwards stated that to get too focused on the particular—whether there are 56 or 118 actionable variants, or whether one methodology is marginally better than another—is to miss the broader picture.
She said that some changes can be made at an executive board level, some changes require consultation with other advisory boards, perhaps including community advisory boards, and other changes are at a significant enough level that each individual must be asked what they would like to do going forward, as some at the workshop proposed with the re-contact/re-consent approach to the retrospective data collection. A proactive opting into a return-of-results model is a straightforward way of reengaging participants, Edwards observed.
Virginia Cain mentioned that NHANES has considered two approaches to genetic information going forward. One would be to bank the samples and do research in the future, and the other approach is to do gene sequencing up front. She noted that the process around incidental findings involves a level of complexity that would be hard to communicate to people and wondered what participant expectations would be. Wiley Burke mentioned that up-front sequencing poses an interesting communication challenge that, in her view, probably starts by not telling people you will be doing a whole-genome sequence. Any results to report back will be a discrete set of results, which Burke suggested could be explained to people so they know exactly what to expect. Edwards commented there is always a question about the subtleties of what people understand, and one approach is to minimize or match expectations. She mentioned that some projects have developed short animated videos to explain some of these complex concepts. With simple consultation around public education and science literacy, there are ways to describe these concepts that could live on the NHANES Website as videos, and NHANES already has experience with videos for other aspects of health.
Les Biesecker mentioned the issues of familiarity and comfort, and used a doctor visit with a complete history and physical examination as an analogy. Virtually everyone has experienced a visit to a doctor, visits that are always unconsented, are exploratory, have a positive predictive value and a sensitivity for finding things, and may reveal incidental findings that were not looked for. He noted everything about searching a genome is the same as in an exam. What is different about a history and a physical exam, he said, is that people are so used to them and have been undergoing them for so long that they usually do not worry about them anymore. Genomics is new, so, as he said, the hair goes up on the back of people’s necks, but the problems and challenges are the same.
Several presenters mentioned possible changes to the NHANES consent forms. If there ever is any possibility of return of results, this possibility should be talked about at the front end of the consent process, according to Clayton. Weir suggested that consent language should probably say that reporting will occur if and when the results are actionable, and the challenge is to develop language that can explain this to people.
With regard to specimens that have already been collected, Botkin stated that the consent language is determinative. NHANES made an agreement with survey participants about not returning results, and it is perfectly appropriate to honor that consent language, given all the complexities and difficulties around the return-of-results issue. An alternative would be to go back to them and say NHANES has a new understanding and new capabilities, there are circumstances in which to consider returning results, and ask participants if they would be interested in obtaining results under those circumstances. If they were interested and opted in, this would be a perfectly acceptable alternative. These are ethically acceptable alternatives, Botkin said, although the latter is a much more burdensome alternative.
According to Botkin, once there is an opportunity to talk with people in any detail about the trade-offs involved with biobanking and use of genomic information and electronic medical records, people understand the need for a trade-off between confidentiality and utility. People expect a high level of protection, he said, but not to the extent that a resource such as NHANES, which is such a valuable one, would become substantially less useful.
Jarvik commented that what is heard most commonly in the clinical setting is that people want their data to have an impact and that they are willing to take some chances. Participant engagement does not only revolve around the return of results. Engagement can tell clinicians, researchers, and biobank curators many things about governance, what kind of consent to have, what sharing policies to have, and what kinds of research participants value most, she said.
Several workshop presenters discussed whether NHANES is going to produce data that will generate findings, and whether the science will
produce results that are valid and returnable. Benjamin Berkman posed an intermediate question: Assuming a study will produce valid findings, is there something ethically distinct about population-based biobank research? This question is pertinent not only to NHANES, but also to this class of research, he said. He posed a series of questions: Are there certain characteristics of population-based studies that distinguish them from other kinds of research where a stronger obligation is felt to return incidental findings? Can a line be drawn saying that to the extent there is an obligation to return incidental findings, this obligation only applies to some core set of clinical studies? Can a set of characteristics be defined that would, on a principled justifiable basis, argue that population-based research is just too different and does not have an obligation to report to participants, even if results are going to be generated?
One key step toward the answers, Berkman suggested, is to determine how difficult it would be to return results and whether an institutional review board (IRB) would approve a no-return approach given a high degree of difficulty. If it is true that returning findings will impact research in such a way that one will be unable to answer fundamental questions and will somehow impinge on the ability to do core parts of one’s mission, Berkman said that argument can be used to justify a no-return policy or a very-high-bar return policy. Data suggest that IRBs do consider arguments about core capacity to accomplish a scientific mission, he said, if in fact that capacity is limited by allocating resources to a return-of-results policy. Some IRBs are willing to approve a no-return paradigm. According to Berkman, the disagreement among workshop participants about return of incidental findings underlines the fact that this is still a field in flux, and there should be room for different positions with regard to research that does not look at all like clinical research.
Burke noted that there seems to be strong support within the field for the notion that secondary uses of biobank specimens may generate less concern, obligation, sense of duty, or even ability to return results, which is a very important distinction. But, as Jarvik argued, highly actionable genetics findings should be returned in keeping with the fact that other highly actionable medical findings are returned to human subjects. It is difficult for the NHANES study to step through the secondary-use-of-data exception, she said.
The return-of-results discussion has implications for the research enterprise more broadly, according to Botkin. He stated there is no doubt that if one does a research scan and sees a tumor, one has to respond to that. On the other hand, if investigators’ responsibility to return data is expanded, the line between research and clinical care may be blurred. In Botkin’s view, there is no clear upper limit in terms of what the researchers’ obligations are, both in terms of what to return and how often to
review their data to find out which data points may have changed from Bin Two to Bin One. This could enormously complicate the research enterprise in a way that may not end up benefiting the overall welfare of the public, because so many resources would be diverted.
NCHS staff raised several questions during the final discussion. The NHANES program is grappling with the issue of which guidelines it should adopt when considering the return of genetic results. When study participants need advice about glucose levels, the American Diabetes Association has guidelines, and the National Cholesterol Education Program has useful guidelines for cholesterol thresholds. In the same vein, Kathryn Porter asked, what is the best place to look for guidance on Bin One variants? Burke noted that while the American College of Medical Genetics and Genomics (ACMG) recommendations (the 56-gene list) have received the most attention, several workshop presenters identified other lists such as the one described by Jarvik (see also Fabsitz et al., 2010). But as Les Biesecker stressed, the important point is that any list is just a starting point. He said such a list is not a how-to manual for dealing with every single variant in every situation, and it is not about research; it is only a starting point for constraining one’s problem to a greater degree than it might otherwise be.
What a list does not do, Biesecker continued, is answer the pragmatic questions about how one evaluates the results, which ones should be CLIA validated, and which ones should go back to participants or patients. It will not work to have external guidelines, he said, because as this workshop has shown, there is no single commonly accepted formulation. Issues have to be worked on in-house with expertise from people who are following the debates, who know the range of recent literature, who can weigh the medical and genetic pros and cons, and who can help decide that variant X meets one’s criteria for these good reasons and variant Y does not.
A “Returning Results Board,” a third party, could be extremely useful, suggested Edwards. She pointed out that an analogy that occurs in clinical settings is end-of-life decision making, where one has a living will or an advance directive and has tried to think through every worst-case scenario but may end up in a different ambiguous scenario that the written wishes did not anticipate. A list or policy that is too specific about which results will or will not be returned is destined to miss the mark, according to Edwards, because results will often be rare and circumstances will evolve. A body of people with the requisite expertise
to help make decisions in real time seems like a necessary component of NHANES.
Another topic of discussion involved the NHANES DNA data bank, which currently is closed to new research proposals and tests. Porter explained that any past tests that have been done are available for secondary analysis, but new genetic tests are not permitted because the IRB said there is no way of determining whether results would need to be reported to participants. Madans noted that the NCHS would like to open the bank to new proposals, and asked if, leaving aside issues of re-consent and consent going forward, new proposals could be entertained given the consent currently in place.
Burke said in her view, feedback at the workshop from most participants suggested that it would be ethically acceptable to generate genomic data, even genomic data that met some criteria of reportability, and not report back for samples that were collected under a consent that said no return of results. One could also consider a re-consent process to return, Burke suggested. Several participants suggested that going forward, as future samples are collected and there is an anticipation of additional genomic research being done, it would be most appropriate to change the consent form so that return of results is possible. Accompanying that could be a very careful process that defines the criteria for the threshold of return, which likely would be very high, Burke said.
Weir said that because genetic analyses are done by third parties who come to NHANES to use the samples, the onus is on the researchers to say what they are going to look at and what the reportability is. To him, NHANES would then have the obligation to communicate the results and the information. No one who runs large population studies can possibly stay on top of all the curation that is needed to know about what markers matter and for which people, he observed.
One participant wondered whether NCHS could argue to its IRB that results of tests on already-stored samples do not have to be reported. Jarvik said she thought yes, because the samples in the DNA bank were not extracted under the standard practices that are now used when data are going to be returned. The same consideration would apply to genotypes that were generated in outside non-CLIA labs, she asserted, because the samples do not have the tracking that would generally be used for a valid test result. Madans commented that a first step for NHANES would be to reevaluate if results from existing samples would be reportable from a quality point of view. This has not yet been done.
Botkin suggested a layered argument can be made. In his view, the no-return consent is sufficient, and concerns about sample integrity can be layered on top to strengthen that argument. Another layer he identified
is the policy that NHANES has already adopted; that is, not to support or permit studies that are likely to develop incidental findings or findings on so-called Bin One conditions. The unlikelihood of finding other Bin One conditions (not currently in the bin) further reduces the probability that this would ever be an issue. Weir agreed, adding that the consent used was state of the art then and is still state of the art in much of the rest of the research community, even if standards are evolving.